Professional Documents
Culture Documents
(Premier Reference Source) Pandian Vasant - Innovation in Power, Control, and Optimization - Emerging Energy Technologies (Premier Reference Source) - IGI Global (2011) PDF
(Premier Reference Source) Pandian Vasant - Innovation in Power, Control, and Optimization - Emerging Energy Technologies (Premier Reference Source) - IGI Global (2011) PDF
(Premier Reference Source) Pandian Vasant - Innovation in Power, Control, and Optimization - Emerging Energy Technologies (Premier Reference Source) - IGI Global (2011) PDF
Nader Barsoum
Curtin University, Malaysia
Jeffrey Webb
Swinburne University of Technology, Malaysia
Senior Editorial Director: Kristin Klinger
Director of Book Publications: Julia Mosemann
Editorial Director: Lindsay Johnston
Acquisitions Editor: Erika Carter
Development Editor: Myla Harty
Production Editor: Sean Woznicki
Typesetters: Jennifer Romanchak, Chris Shearer
Print Coordinator: Jamie Snavely
Cover Design: Nick Newcomer
Copyright © 2012 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.
All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Editorial Advisory Board
Milorad Bojic, University of Kragujevac, Serbia
Gianfranco Rizzo, University of Salerno, Italy
Janos Sebestyen Janosy, KFKI Atomic Energy Research Institute, Hungary
Ivan Zelinka, Thomas Bata University in Zlin, Czech Republic
Davor Skrlec, University of Zagreb, Croatia
Nikolai I. Voropai, Energy Systems Institute, Russia
Monica Chis, Siemens Program and System Engineering, Romania
Cengiz Kahraman, İstanbul Technical University, Turkey
Valentina E. Balas, “Aurel Vlaicu” University of Arad, Romania
Arturo Suman Bretas, Universidade Federal do Rio Grande do Sul, Brazil
Etienne Kerre, Ghent University, Belgium
Radu Emil Precup, “Politehnica” University of Timisoara, Romania
Gerardo Maximiliano Mendez, Instituto Tecnologico de Nuevo Leon, Mexico
Table of Contents
Preface..................................................................................................................................................... x
Acknowledgment.................................................................................................................................xiii
Chapter 1
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems......................... 1
Nicolay Voropai, Energy Systems Institute, Russia
Irina Kolosok, Energy Systems Institute, Russia
Elena Korkina, Energy Systems Institute, Russia
Alexey Paltsev, Energy Systems Institute, Russia
Anna Glazunova, Energy Systems Institute, Russia
Victor Kurbatsky, Energy Systems Institute, Russia
Nikita Tomin, Energy Systems Institute, Russia
Alexander Gamm, Energy Systems Institute, Russia
Irina Golub, Energy Systems Institute, Russia
Roman Bershansky, Energy Systems Institute, Russia
Daniil Panasetsky, Energy Systems Institute, Russia
Dmitry Efimov, Energy Systems Institute, Russia
Dmitry Popov, Energy Systems Institute, Russia
Christian Rehtanz, University of Dortmund, Germany
Ulf Häger, University of Dortmund, Germany
Chapter 2
Hopfield Lagrange Network for Economic Load Dispatch................................................................... 57
Vo Ngoc Dieu, Asian Institute of Technology, Thailand
Weerakorn Ongsakul, Asian Institute of Technology, Thailand
Chapter 3
Renewable Energy and Sustainable Development................................................................................ 95
Abdeen Mustafa Omer, Energy Research Institute, UK
Chapter 4
Demand-Side Response Smart Grid Technique for Optimized Energy Use........................................ 137
Fouad Kamel, University of Southern Queensland, Australia
Marwan Marwan, Queensland University of Technology, Australia
Chapter 5
Soft Computing and Computational Intelligent Techniques in the Evaluation of Emerging Energy
Technologies........................................................................................................................................ 164
Selcuk Cebi, Karadeniz Technical University, Turkey
Cengiz Kahraman, Istanbul Technical University, Turkey
İhsan Kaya, Yıldız Technical University, Turkey
Chapter 6
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms:
Kurdistan Electric Network Case Study.............................................................................................. 198
Mohammad Saleh, University of Kurdistan, Iran
Hassan Bevrani, University of Kurdistan, Iran
Chapter 7
Many-to-Many Assignment Problems: Lagrangian Bounds and Heuristic......................................... 220
Igor Litvinchev, Nuevo Leon State University, Mexico
Socorro Rangel, São Paulo State University, Brazil
Chapter 8
Power Systems Investments: A Real Options Analysis....................................................................... 248
João Zambujal-Oliveira, Instituto Superior Técnico & Technical University of Lisbon, Portugal
Chapter 9
Optimal Configuration and Reconfiguration of Electric Distribution Networks................................. 268
Armin Ebrahimi Milani, Islamic Azad University, Iran
Mahmood Reza Haghifam, Tarbiat Modares University, Iran
Chapter 10
A Descriptive Approach for Power System Stability and Security Assessment.................................. 293
A. G. Tikdari, University of Kurdistan, Iran
H. Bevrani, University of Kurdistan, Iran
G. Ledwich, Queensland University of Technology, Australia
Chapter 11
Analyses and Monitoring of Power Grid............................................................................................. 315
Rana A. Jabbar, Rachna College of Engineering and Technology, Pakistan
Muhammad Junaid, Rachna College of Engineering and Technology, Pakistan
M. A. Masood, Rachna College of Engineering and Technology, Pakistan
A. Bashir, Rachna College of Engineering and Technology, Pakistan
M. Mansoor, Rachna College of Engineering and Technology, Pakistan
Chapter 12
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic
Algorithms and Hybrid Genetic Algorithms Pattern Search Approaches . ......................................... 344
Pandian Vasant, University Technology Petronas, Malaysia
Index.................................................................................................................................................... 377
vii
Foreword
Since the beginning of our civilization, the human race has had to confront numerous technological
challenges such as finding the optimal solution of various problems including control technologies,
power sources construction, and energy distribution, amongst others. These examples encompass both
ancient as well as modern technologies like automatic theatre controlled by special programmes in
ancient Greece, the first electrical energy distribution network in USA, mechanical, electronical, as well
as computational controllers, et cetera. Technology development of those and related areas has had and
continues to have a profound impact on our civilization and lifestyle.
The topics discussed in this book are of these mentioned areas and mutually joined into a compre-
hensive text, which while discussing the specific selected topics, give a deeper insight to the interdis-
ciplinary fusion of those modern and promising areas of emerging technologies. This book discusses
the mutual intersection of interesting fields of research, as hybrid renewable energy and energy saving,
solar and fuel cells, power systems, chaos and power quality, soft computing, simulators, and software
engineering, amongst others. Novel techniques are also discussed in this book, which are able to handle
tasks such as control of various technological and energetical systems, optimization by means standard,
as well as novel methods. Together with many interesting emerging technologies, a reader will also find
in the book various mathematical and algorithmical methods used for proposed technologies including
models like fuzzy logic, neural network, evolutionary algorithms, or Hybrid System Optimization.
Therefore, this book titled “Innovation in Power, Control and Optimization: Emerging Energy
Technologies,” edited by Pandian Vasant, Nader Barsoum, and Jeffrey Webb, is a timely volume to be
welcomed by the community focused on power control and optimization as well as computational intel-
ligence community and beyond. This book is devoted to the studies of common and related subjects in
intensive research fields of emerging technologies. For these reasons, I enthusiastically recommend this
book to our scientists and engineers working in the above mentioned fields of research and applications.
Ivan Zelinka
Czech Republic January 2011
Ivan Zelinka (born 1965) is currently working at the Technical University of Ostrava (VSB-TU), Faculty of Electrical Engi-
neering and Computer Science. He graduated consequently at Technical University in Brno (1995 – MSc.), UTB in Zlin (2001
– Ph.D.) and again at Technical University in Brno (2004 – assoc. prof.) and VSB-TU (2010 - professor). Before academic
career he was an employed as TELECOM technician, computer specialist (HW+SW) and Commercial Bank (computer and
LAN supervisor). During his career at UTB he proposed and opened 7 different lectures. He also has been invited for lectures
at 7 universities in different EU countries plus role of the keynote speaker at the Global Conference on Power, Control and
viii
Optimization in Bali, Indonesia (2009) and Interdisciplinary Symposium on Complex Systems (2011), Halkidiki, Greece.
He is responsible supervisor of grant research of Czech grant agency GAČR named Softcomputing methods in control, co-
supervisor of grant FRVŠ - Laboratory of parallel computing. He was also working on numerous grants and two EU project
like member of team (FP5 - RESTORM) and supervisor (FP7 - PROMOEVO) of the Czech team. Currently he is a head of
Department of Applied Informatics and in total he has been supervisor of more than 27 MSc. and 19 Bc. diploma theses. Ivan
Zelinka is also supervisor of doctoral students including students from the abroad. He was awarded by Siemens Award for his
Ph.D. thesis, as well as by journal Software news for his book about artificial intelligence. Ivan Zelinka is a member of British
Computer Socciety, Machine Intelligence Research Labs (MIR Labs - http://www.mirlabs.org/czech.php), member of expert
team of company DaySpring Global Multinational Inc., division Knowledge Management & Mining division (see also http://
www.dsgm.ca/consulting.asp), IEEE (committee of Czech section of Computational Intelligence), a few international program
committees of various conferences and three international journals (Associate Editor of MSE, Hindawi, http://www.hindawi.
com/journals/mse/editors.html, Editorial Council of Security Revue, http://www.securityrevue.com/editorial-council/, Edito-
rial board - Journal of Computer Science, Riga, Latvia). He is author of journal articles as well as of 5 books in Czech and 8
chapters in 6 books in English language.
ix
Foreword
Igor Tyukhov
Moscow State University of Environmental Engineering, Russia
Igor I. Tyukhov is Executive Director of UNESCO Chair “Ecologically clean engineering” at the Moscow State University of
Environmental Engineering (MSUEE) and Deputy Chair Holder of the UNESCO Chair “Renewable Energy and Rural Elec-
trification” (part time) at the All-Russian Research Institute for Electrification of Agriculture (VIESH), member of the Interna-
tional Solar Energy Society, and Associate Editor of Solar Energy Journal. He graduated from the Moscow Power Engineering
Institute, V. A. Fabrikant Physics Department, 1972. He got degree “kandidat technicheskikh nauk” (Ph.D.) at 1979. He has
been with the Moscow Power Engineering Institute for more than 30 years teaching various physics disciplines and conduct-
ing research work on solar energy, solar concentrators, optical metrology, and semiconductor technology . Dr. Tyukhov was
visiting scholar at George Mason University (GFDP, 1999/200 academic year), at University of Oregon (Fulbright Program,
2002/2003), and at the Oregon Institute of Technology (2003). Dr. Tyukhov is expert in the field of photovoltaics and renewable
energy. He is author more than 200 papers, more than 20 patents, several chapters in the books and coauthor of monograph.
x
Preface
Many engineering systems and science problems suffer from the issue of developing a system that can
cope with variations of system or control parameters, measurements uncertainty, and complex multi-
objective optimization criteria. The need for a priori knowledge and the inability to learn from past
experience make the design of robust, adaptive, and stable systems a difficult task. Currently, research
on energy resources is of great importance for future oil replacements, particularly in vehicles and
other transportation. Computational intelligence has been proven to provide successful solution of com-
plex optimization problems by fuzzy logic, neural networks, evolutionary algorithms, and genetic algo-
rithms. They include system identification, parameter estimation, multi-objective optimization, robust
solutions, adaptive systems, self-organization, and failure analysis.
This book aims to provide relevant theoretical frameworks and the latest empirical research findings
in these areas. It is written for professionals who want to improve their understanding of the strategic
role in the area of power, control, and optimization. Each book chapter is written by experts in their
particular field of expertise.
Chapter 1 of this text describes coordinated intelligent operation and emergency control of electric
power systems.
In Chapter 2 a Hopfield Lagrange network (HLN) is proposed for solving economic load dispatch
(ELD) problem. HLN is a combination of Lagrangian function and continuous Hopfield neural network
where the Lagrangian function is directly used as the energy function for the continuous Hopfield neural
network.
The increased availability of reliable and efficient energy services that stimulates new develop-
ment alternatives such as solar, wind, et cetera is discussed in Chapter 3. This chapter elaborates on
the potential for such integrated systems in the stationary and portable power market in response to the
critical need for a cleaner energy technology. Anticipated patterns of future energy use and consequent
environmental impacts (acid precipitation, ozone depletion, and the greenhouse effect or global warm-
ing) are comprehensively discussed in this chapter.
Chapter 4 describes a dynamic smart Grid concept which enables electricity end-users to be acting
on controlling, shifting, or curtailing own demand to avoid peak-demand conditions according to infor-
mation received about electricity market conditions over the Internet.
The global warming and energy need requires developing emerging energy technologies for the
electricity, heat, and transport markets are subject of discussion in Chapter 5. In this chapter are also
discussed in great detail the emerging energy technologies that aim at increasing efficiency of energy
utilization processes from energy sources and diminish CO2 exhalation. The main aim of the chapter
is to exhaustively present soft computing and computational intelligent techniques in the evaluation of
emerging energy technologies.
xi
Chapter 6 presents an overview of key issues and technical challenges in a regional electric net-
work, following the integration of a considerable amount of wind power. A brief survey on wind power
system, the present status of wind energy worldwide, common dynamic models, and control loops for
wind turbines is given.
Modified Lagrangian bounds and a greedy heuristic are proposed and discussed in Chapter 7 for
many-to-many assignment problems taking into account capacity limits for tasks and agents. A feasible
solution recovered by the heuristic shown to speed up the subgradient technique to solve the modified
Lagrangian dual. A numerical study is presented to compare the quality of the bounds and to demonstrate
the efficiency of the overall approach.
Energy projects with extended life cycles and initial investments can be unprofitable under discount
cash flow methods. Therefore, real options analysis has become relevant as a pricing technique for these
types of projects, with private risks and high investment levels. Following this question, the work pre-
sented in Chapter 8 analyses different real options approaches to select the most acceptable for investing
decisions in the energy sector.
Power loss reduction is considered as one of the main purposes for a distribution system’s design-
ers and operators especially for recent non-governmental networks. Moreover, the nature of power
loss challenges different methods to solve this problem, while various studies indicate effectiveness
of reconfiguration and its high portion for this case. Thus, “reconfiguration” can be introduced as an
optimization procedure to obtain economical high quality operation by changing the status of sectional-
izing switches in these networks. Some major points, such as using different switch types, considering
number of switching, and time varying loads which are almost neglected or not applied simultaneously
in most pervious essays are discussed in Chapter 9.
In Chapter 10, the power system is considered as a continuum, and the propagated electromechani-
cal waves initiated by faults and other random events are studied to provide a new scheme for stability
investigation of a large dimensional system. For this purpose, the measured electrical indices (such as
rotor angle and bus voltage) following a fault in different points among the network are used, and the
behavior of the propagated waves through the lines, nodes, and buses is analyzed. The impact of weak
transmission links on a progressive electromechanical wave using energy function concept is addressed.
In Chapter 11 analyses and monitoring of the power grid in Pakistan is presented.
Finally, in Chapter 12, a solution is proposed to a certain nonlinear programming difficulties related to
the presence of uncertain technological coefficients represented by vague numbers. Only vague numbers
with modified s-curve membership functions are considered. The proposed methodology consists of novel
genetic algorithms and a hybrid genetic algorithm pattern search (Vasant, 2008) for nonlinear program-
ming for solving problems that arise in industrial production planning in uncertain environments. Real
life application examples in production planning and their numerical solutions are analyzed in detail.
The new method suggested has produced good results in finding globally near-optimal solutions for the
objective function under consideration.
xii
The editors of this text want to thank all the contributors to this text for their time, energy and invalu-
able expertise that we believe will make this book a success and extremely valuable resource in the area
of power management, control and optimization of engineering problems.
Pandian Vasant
Petronas University of Technology, Malaysia
Nader Barsoum
Curtin University, Malaysia
Jeffrey Webb
Swinburne University of Technology, Malaysia
xiii
Acknowledgment
We would like to this golden opportunity to sincerely thank the following friends and colleagues of us
for their valuable help and strong support of book chapters of the manuscript. Their marvelous feedback,
opinion, constructive comments, and suggestions for the improvement of the overall outstanding qual-
ity of the book chapters are gratefully acknowledged.
Furthermore, we sincerely thank the group of IGI Global at Hershey PA, USA, for their great help
and excellent support on this book project. In particular, special thanks go to Ms. Jan Travers, Ms. Myla
Harty, and Mr. Dave De Ricco of IGI Global for their great help.
Last but not least, we sincerely express our sincere thanks and appreciation to members of PCO
Global for their great support.
Chapter 1
Coordinated Intelligent
Operation and Emergency
Control of Electric
Power Systems
Nicolay Voropai Irina Golub
Energy Systems Institute, Russia Energy Systems Institute, Russia
Alexander Gamm
Energy Systems Institute, Russia
ABSTRACT
This chapter presents the following approaches and developments: (1) the approach to power system
state estimation based on structural and functional decomposition. PMU measurements are used to
coordinate the solutions obtained in individual areas; (2) a non-iterative method to calculate voltage
DOI: 10.4018/978-1-61350-138-2.ch001
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
magnitude and phase at all the buses, allowing the state parameters of EPS to be obtained fast enough;
(3) the approach to the super short-term forecasting of state variables on the basis of neural network
technologies and algorithms of nonlinear optimization that is realized in the ANAPRO software; (4)
an analysis of the possibility of determining weak ties and cut sets in EPS; (5) a control system based
on the multi-agent technique; (6) the development of selective automatic systems intended to prevent
and eliminate out-of-step operation on the basis of synchronized voltage phase measurements obtained
from Phasor Measurement Units.
2
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
The chapter includes: new conditions and All these factors essentially complicate power
requirements for power system monitoring and system operation, increase its changeability and
control; wide area monitoring of power system unpredictability, raise danger of severe emergen-
state using new technologies and tools; compre- cies with undesirable development and massive
hensive emergency control systems and defense consequences for a system and consumers and,
plans for large power systems; coordinated con- therefore, call for prompter and more adequate
cepts and control systems for improving stability response of control systems. This generates the
and security of power systems; coordinated wide need to improve and develop principles and sys-
area monitoring and control of large power sys- tems of power system operation control which
tems; development of computer technologies and can be based on:
modern tools towards smart power grids
• New systems for measurement of operat-
The Problem of Monitoring, ing parameters (PMU) and their control
Forecasting and Control in (FACTS, energy storage devices etc.) that
Electric Power System significantly improve EPS observability
and controllability;
The trends in expansion of EPSs and changes • Modern communication systems, new in-
in the conditions of their operation have led to formation technologies and artificial intel-
considerable transformations in their structure and ligence methods, highly efficient comput-
operation. These transformations are conditioned ers, which totally change the processes
by the following factors: of acquisition, transmission, presentation
(visualization) and use of information on
• Increase in scale, expansion of territories power systems;
to be serviced, interconnection of dif- • Efficient mathematical control theory
ferent power systems for joint operation methods in multicriteria non-coincident
which results in creation of interregion- conditions.
al, interstate and intercontinental power
interconnections; Based on the above circumstances the so
• Decentralization of power supply due to a called Smart Grid concept was developed (Of-
wider use of distributed generation sources fice, 2003 and European, 2006). This problem
that are connected to the distribution net- is very often associated with the integration of
work nodes; renewable energy sources in EPS as well as with
• Restructuring of power industry, which of- distribution electric network based on the informa-
ten makes its structure drastically different tion technologies and artificial intelligence in the
from the technological structure of EPS as distributed control systems of power supply and
a technically single unit and from its con- power consumption (Chuang et al. 2004; Amin et
trol structure; al. 2005). The most comprehensive understanding
• Liberalization of relationships in the elec- of the Smart Grid concept to date has been given
tric power industry which leads to many by Shahidehpour (2009). In general the Smart
participants of relationships with different, Grid concept can be represented as a set of the
often opposing, interests, in the course of following components:
expansion, operation and control of power
system. • Generation (enhancement of reliability
and economic efficiency of electric ener-
3
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
gy production through the use of modern • Consumers (equipping them with highly
highly intelligent systems for monitor- intelligent systems intended for electricity
ing and control, integration of renewable control and metering, demand-side man-
sources, distributed generation and energy agement and load control in emergency
storage devices on the basis of the Internet situations).
technologies);
• Electric transmission network (wide scale In Russia power industry reforms and new
monitoring of operating conditions and methods and technical tools of control have also
their control with the help of new devices encouraged the development of principles and
and technologies (FACTS, PMU, artificial methods of EPS dispatching and emergency con-
intelligence, etc.) in order to provide pow- trol (Ayuev et al, 2008; Voropai, 2008).
er supply reliability and electric network Let us consider the problems of improving the
controllability); principles and enhancing the efficiency of systems
• Substations (automation of substations that intended for EPS operation and emergency control.
are based on the advanced electric equip- The most important directions here include an
ment by using modern systems for diag- increase in control adaptability and expansion of
nosing, monitoring and control on the basis coordination among control stages, devices and
of information and computer technologies systems. For this purpose it is necessary to develop
in order to provide reliability and control- an efficient system of wide scale monitoring and
lability of substations); forecasting of operating conditions and control of
• Distribution electric network (a radical EPS. Figure 1 shows power system states and the
increase in its controllability and reliabil- blocks of problems on monitoring, forecasting and
ity through introduction of distributed mi- control that correspond to these states.
croprocessor-based control and protection The blocks of monitoring and forecasting of
systems with the use of new information, the EPS normal, pre-emergency and post-emer-
computer and Internet technologies);
Figure 1. Time diagram of events in the system for monitoring, forecasting and control in EPS
4
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
gency operating conditions include the following The enumerated problems have certain
problems: specificity when applied to normal conditions,
on the one hand, and to pre-emergency and
• System state estimation; post-emergency conditions, on the other hand.
• Forecasting the parameters of expected Pre-emergency and post-emergency conditions
operating conditions. Forecasting is nec- (at system restoration) require higher speed of
essary because during the state estimation algorithm operation and smaller time intervals
procedure current state is estimated with between individual states for which the problems
some delay, while monitoring and control of operation monitoring and control are solved.
problems require some advance estimation A separate problem is monitoring of emergency
of system state (“to control is to foresee”); conditions. Dynamic nature of emergency condi-
let us note that for these two blocks of tions requires that the results of monitoring be
problems the advance time can vary; promptly transferred to dispatcher. They should
• Detection of weak points in the system in have an integrated form and indicate dangerous
the expected operating conditions; points in the system in terms of undesirable de-
• Determination of margins for transfer ca- velopment of emergency. The primary informa-
pabilities of ties in the expected condi- tion, however, obtained as a result of emergency
tions; This is necessary to efficiently use conditions monitoring, should arrive at the auto-
the margins in the operating conditions and matic control devices and systems with the view
automatic control through appropriate con- of their adaptation to the current power system
trol actions; emergency state.
• Visualization of the expected conditions; In addition to monitoring and forecasting of
• Determination of indices and criteria for the operating conditions the efficient adaptive
transition from normal to pre-emergency methods and algorithms for choosing the points
conditions and, vice versa, from post- of control action application and adjustments on
emergency to normal conditions. the basis of advanced achievements of the control
theory and artificial intelligence play an important
A special explanation should be given concern- part in increasing the adaptability of control. The
ing the last problem. It relates to the following methods and algorithms are implemented by dis-
basic principles of power system operation control patcher and automatic control systems and provide
in a market environment: adaptation of controls to the current system state
and possible disturbances.
• under normal operating conditions – ef- It is necessary to considerably expand and
ficient contract relationships among the improve coordination of control. This can be done
participants of the wholesale markets for in the following directions:
electricity, capacity and ancillary services,
that are coordinated on a market basis; 1. Expansion of EPS operation control
• under a threat of emergency conditions – a coordination:
transition from market criteria of operation A. In terms of time - from coordinated
control to the centralized principles of dis- design of control systems to their
patching control; implementation through dispatching
• under emergency and post-emergency con- and automatic devices;
ditions – the use of strictly centralized dis-
patching and automatic emergency control.
5
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
6
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
countries (Clements, 1972; El-Kleib, 1992; Iwa- the calculated scheme into subsystems by one or
moto, 1989 and others). another method (Gamm, 2007). The functional
Until recently state estimation in EPS was decomposition is made in accordance with the
mainly based on the SCADA measurements: volt- problems solved within the SE procedure. The
age magnitudes, branch power flow, nodal power main of them are: analysis of network topology
injections and, occasionally, current magnitudes. (formation of current calculated scheme); analysis
The advent of WAMS (Wide-Area Measurement of observability; analysis of bad data; calculation
System) that contains phasor measurement units of estimates and calculation of steady state with
(PMU) as the main measurement equipment regard to the estimates obtained.
(Phadke, 2002) makes it possible to synchro-
nously and accurately control the EPS state and Methods of Structural Decomposition
essentially improve the results of state estimation The calculated scheme can be divided into subsys-
(Gamm, 1997). The use of PMU measurements tems by the following techniques: decomposition
offers new possibilities in decomposition of the utilizing geographical characteristics (Gamm,
state estimation problem. 1995; Falcao, 1995), decomposition by boundary
The paper considers the algorithm of state esti- nodes (Gamm, 1983; Clements, 1972), by tie-lines
mation by the test equation technique (Clements, (Gamm, 1983; Gamm, 1995; Abdel-Rahman,
1972) that employs structural decomposition of 2001), based on the structure of gain matrix
state estimation problem, i.e. division of the cal- (Wallach, 1981), by Danzig-Wolf decomposition
culated scheme into subsystems, and functional algorithm (El-Kleib, 1992), and others.
decomposition of the SE problem (detection of The main algorithms of SE problem decompo-
bad data, state estimation on the basis of quadratic sition suggest dividing the calculated scheme into
and robust criteria). The two-level algorithm is subsystems whose boundaries are either nodes or
proposed to divide the calculated scheme into branches. In this case the SE problem is solved
subsystems for state estimation by the test equa- iteratively unless the boundary conditions are
tion technique. Application of the test equation met. The method of decomposition with bound-
technique that allows the values of measured ary nodes has been chosen for realization in the
variables to be fixed by setting zero variances offered algorithm.
for them, and placement of Phasor Measurements In this case the equality of voltage magnitudes
Units at boundary nodes (El-Kleib, 1992) make it and phases at the boundary nodes should be met [2]:
possible to essentially simplify the procedure of
coordinating the solutions obtained for separate U i = U j = ... = U k ; (1)
subsystems. In addition the paper presents the
algorithm of PMU placement at boundary nodes.
The example of calculation for a fragment of real δi = δ j = ... = δ k ; (2)
power system is given.
Besides the boundary balance relationships
Decomposition of State should be met. For example for boundary node l,
Estimation Problem common for the i, j, …, k-th subsystems
7
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Ql + ∑ ∑Q lm
(U l , δl , U m , δ m ) = 0 , PMU placement under SE problem decomposi-
s =i, j,..., k m∈ω s tion the algorithm based on the annealing method
(4) was developed. They will be presented in Section
4. Compared to the PMU-based decomposition
where ω s - a set of nodes of the s-th subsystem, methods proposed in (Zhao, 2005; Jiang, 2007;
that are adjacent to the l-th node. Jang, 2008) the number of PMUs in our case study
did not increase, yet made it possible to perform
The Use of PMU Data in Structural parallel state estimation by subsystems, to solve
Decomposition coordination problem without iteration and obtain
Development and improvement of software for an optimal but not pseudo-optimal, as in (Zhao,
monitoring and control of power systems at a 2005), solution that coincides with the solution
qualitatively new level have become possible ow- for the entire network.
ing to WAMS (Wide-Area Measurement System) Placement of PMUs at boundary nodes makes
that allows the EPS state to be controlled synchro- it possible to register boundary variables U and
nously and accurately. The devices for measuring δ measured highly accurately. In this case the
phasors (Phasor Measurement Units) are the basic operating conditions of some subsystems can
measurement equipment in this system. be calculated independently of one another and
The results of solving the state estimation solution of the coordinated problem consists in
problem can be essentially improved by using calculating nodal injections by (3), (4) using the
the PMU data. As compared to the standard set estimates of power flows.
of measurements received from SCADA system
PMUs placed at a node provide accurate (the error Algorithm of Structural Decomposition Using
is 0.2-0.5%) measurements of voltage magnitude Test Equations
and phase at this node as well as the magnitudes The idea of decomposing the state estimation
and phases of currents in the branches adjacent problem with PMU placement at boundary nodes
to this node. is rather attractive. In reality, however, due to
The possibilities of using synchronized pha- high cost of PMUs they can only be used when
sor measurements for distributed state estimation the number of boundary nodes is small.
were discussed in (Zhao, 2005; Jiang, 2007; Jang, To calculate large inhomogeneous schemes
2008 and others). In (Zhao, 2005) the method is the authors propose a method of dividing the
suggested to decompose the calculated scheme calculated scheme with respect to voltage levels
into the areas with PMU to be installed in each (Wallach, 1981). This method decreases essen-
area. The data from these PMU are then used to tially a negative impact of un-homogeneity of
solve a coordination problem. In (Jiang, 2007) calculated scheme and telemetric information in
the authors suggest placing PMU at a basic node calculation of subsystems of one voltage class but
of each area. The PMU measurements coordinate for the complex scheme inevitably leads to a large
the SE problem solution of each area. (Jang, number of boundary nodes. Therefore, the paper
2008) presents a diakoptic-based distributed SE proposes a two-stage algorithm to decompose the
algorithm, in which PMUs are used to coordinate calculated scheme into subsystems that combines
voltage angles of each area SE solution. the positive features of both approaches.
The chapter suggests the use of PMU measure- At the first stage the scheme is divided into
ments for distributed state estimation to coordinate rather large areas with minimum number of inter-
the solutions obtained for individual areas. For system ties and boundary nodes. This decomposi-
tion can be made on the basis of administrative
8
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
division, for example, the entire scheme of Russia’s of a certain voltage class before solving the state
UES is decomposed into regional power subsys- estimation problem (Gamm, 2007).
tems of large regions in the country that operate State estimation is made according to two
in parallel or it can be decomposed artificially criteria: the method of weighted least squares
into separate areas by special algorithms (Gamm, and the robust criterion that allows the estimates
1995). PMUs are placed at the boundary nodes of to be obtained and bad data to be suppressed
the areas. Highly accurate measurements obtained simultaneously.
from PMU make it possible to register the values Control is transferred to one or another state
of magnitudes and phases of nodal voltages at the estimation program depending on operation of
boundary nodes and make calculations for the the bad data detection program. In case of bad
areas in parallel. data detection or their absence the program for
At the second stage the calculated scheme of calculation of estimates operates on the basis of
each area in turn is divided into subsystems that the least squares method. However, if it is im-
correspond to the levels of nodal voltages. The possible to detect erroneous measurements and,
calculations start with the subsystem of the highest hence, identify bad data the program operates
voltage level (750-500 kV). Normally this part of according to the robust criterion (Gamm, 2005).
the scheme is well provided with highly accurate State estimation is made starting at the upper level
telemetry and contains a basic node. Then the of the structural decomposition.
calculations are made successively for the rest
of the subsystems. The subsystems are ranked Full Algorithm
by voltage levels (220 kV, 110 kV, etc.). Every
time the node bordering the subsystem of higher The full algorithm for solving the state estima-
voltage level is chosen as a basic one. tion problem based on structural and functional
After the calculation of the low level sub- decomposition is as follows.
systems a coordination problem is solved for
all areas. In this case boundary conditions (1), 1. The complete calculated scheme of EPS is
(2) are met automatically, and the coordination decomposed into rather large areas. Phasor
problem implies calculation of nodal injections measurement units are placed at the boundary
at the boundary nodes on the basis of power flow nodes of subsystems. In the subsystems that
estimates obtained for each area (Equations (3) have no basic node of the complete scheme
and (4)). one of the boundary nodes with PMU of the
highest voltage class is chosen as a basic
Functional Decomposition one. Measurements of nodal injections at
The functional decomposition of the SE problem boundary nodes are excluded from the vector
is performed in accordance with the problems of measurements.
solved within the state estimation procedure. The 2. At the second stage of decomposition the
main of them are: analysis of network topology; calculated scheme of each area is divided into
analysis of observability; analysis of bad data; subsystems that correspond to the levels of
calculation of estimates and steady state by the nodal voltages. The boundaries of the sub-
estimates obtained. systems are the nodes adjacent to the nodes
The current calculated scheme is formed for of the voltage class of this subsystem. For
the entire scheme. Bad data analysis and calcula- example for the 750-500 kV voltage class
tion of estimates and steady state are performed subsystem the nodes with the voltage of
by the test equation technique for each subsystem 220 kV are boundary nodes and vice-versa.
9
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
3. The calculation starts with the subsystem of all boundary nodes. Based on the measurements
the highest voltage level (750-500 kV) for to be received from the placed PMUs the voltage
each subsystem. Normally this part of the magnitude and phase at a neighboring node can be
scheme is well provided with highly accurate calculated using the electrical circuit equations. In
measurements and contains a basic node. The the paper the vector voltage measurement obtained
state estimation algorithm for subsystems by the equations is called the “calculated” PMU.
with boundary nodes is as follows: The study shows that the accuracy of parameters
A. For each subsystem that contains of the calculated PMU practically equals the ac-
boundary nodes the problem of bad curacy of measurements of the physical PMU
data detection is solved by the test (Kolosok, 2009). With an optimal combination of
equation technique. physical and calculated PMU at all boundary nodes
B. In case of bad data detection or their of subsystems it is possible to determine voltage
absence the state estimation is made magnitudes and phases required to coordinate
according to least squares method. the solutions obtained for individual subsystems.
C. In the event that erroneous measure- In order to minimize the number of PMUs we
ments cannot be detected and hence it is analyze not only the list of boundary nodes but
impossible to detect bad data, the state the list of internal lines within subsystems that
estimation is made according to the are incident to these nodes as well. The bound-
robust criterion (bad data suppression). ary nodes may happen to belong to one and the
4. The rest of the subsystems in the scheme are same subsystem and bound one and the same
successively calculated. They are ranked by line. Then it is enough to place a physical PMU
voltage level (220 kV, 110 kV, etc.). Every at one end of the line and a “calculated” PMU at
time the node bordering the subsystem of the other. To choose the optimal PMU placement
higher voltage level is chosen as a basic node. the algorithm based on the simulated annealing
The estimates of the boundary variables of method was developed.
the state vector that are obtained at the upper A fragment of a real scheme divided into 3
level of decomposition are registered. subsystems (Figure 2 and Figure 3) will be con-
5. The injections at boundary nodes between sidered for illustration of the algorithm operation.
the subsystems of different voltage class are Figures 2 and 3 show a variant of system divi-
calculated. sion into subsystems with boundary lines: between
6. After all subsystems of the first level of subsystems I and II - lines 2-5 and 4-7, between
decomposition have been calculated similar subsystems I and III – lines 3-9 and 4-9, and
problem is solved for the boundary nodes between subsystems II and III – lines 7-10.
with PMU. The objective function of the annealing method
in this problem has the form:
Placement of PMU at Decomposition
of the State Estimation Problem K1PMU + K 2var iants
min E = , (5)
K 3calculated + K 4 subsystems
For decomposition of power system state estima-
tion problem it is necessary to maintain accurate
values of voltage magnitudes and phases at where
boundary nodes of subsystems for iteration-free
solution of coordination problem. A simple but • K1PMU – the number of placed PMUs;
not an optimal solution is placement of PMUs at
10
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
11
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
The criterion of operation completion is consisting of 107 nodes and 175 branches. The
min E . calculations were based on real measurements.
The practical result of searching and obtain- The efficiency of the algorithm was assessed by
ing the optimal placement of PMU at division of comparing the results of calculations made for
system into subsystems with boundary lines is subsystems to the results of the calculation made
given in Table 1, line 1. Figures 2 and 3 show that for the entire scheme.
with an account taken of the calculated scheme At the first stage the Genetic Algorithm (Ko-
topology the optimal result is given by PMUs losok, 2003) was used to divide the entire scheme
(black squares) placed at boundary node 4 and into two subsystems containing 55 and 52 nodes
nodes 6 and 12 which are adjacent to the bound- including 6 boundary nodes in which the PMU
ary ones. The circles denote the nodes for which data (measurements of magnitudes and phases of
the “calculated” PMUs are obtained. nodal voltages) were modeled. The calculations
Figure 3 presents the variant of dividing system of these subsystems were carried out in parallel
into subsystems by boundary nodes: between which reduced the time of solving the SE problem
subsystems I and II - nodes 5 and 7, between almost twice: from 0.49 s to 0.27 s.
subsystems I and III – nodes 4 and 7, between At the second stage of decomposition each
subsystems II and III – boundary node 7. Table of the subsystems in turn was decomposed into
1, line 2, shows the optimal PMU placement three subsystems corresponding to the voltage
under the following division into subsystems: the levels of 500 kV, 220 kV, 110 kV and lower. The
optimal result Eopt = 0.14 is given by the PMU calculation of these subsystems according to the
above algorithm was made successively; therefore
(black square), installed at boundary node 7 which
the full time of solution could increase. However,
is common for all subsystems, the “calculated”
owing to the improved convergence of the iteration
PMUs provide pseudo-measurements at nodes
processes in the calculation of subsystems of the
4,5,8,10, which are adjacent to the boundary ones
same voltage class the total time of the calcula-
(denoted by circles).
tion for all the three subsystems practically did
The calculations show that the number of PMUs
not change.
placed at decomposition of the scheme is consider-
More efficient operation of bad data detection
ably lower than the number of boundary nodes.
algorithm and application of the robust criterion
of SE (Kolosok, 2003) for two of six subsystems
Calculation Example
improved considerably the results of state estima-
tion: the value of the SE objective function at the
In order to test the efficiency of the suggested
point of solution decreased almost by a factor
decomposition algorithm of state estimation
the calculations were made for a real scheme
Table 1. The practical result of searching and obtaining the optimal placement of PMU at division of
system into subsystems with boundary lines
Division into PMU placement at Variants of calculated Calculated nodes, The number of Function Е
subsist.
nodes, K1PMU PMU, K 2var iants K 3calculated subsist. K4
12
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
of 6 and the estimates at boundary nodes were PMU for Fast Calculation of Steady
noticeably improved. State in Electric Power Systems
Structural and functional decomposition of
state estimation problem is an effective method The current state of an electric power system
to solve the problems arising during calculation (EPS) can be properly determined by a minimum
of large schemes. set of state variables that will make it possible to
The proposed two-level algorithm for structural uniquely determine all the rest of state variables.
decomposition of the SE problem allows one to Such a set of variables is called a state vector. As
simultaneously process the data for local sub- a rule it is voltage magnitudes and phases
systems of considerably smaller dimensionality; x {δ, U }.
=
decrease the adverse impact of un-homogeneity of The speed of determining the current system
the calculated scheme and telemetric information state depends on the speed of calculating all state
when calculating one-voltage-class subsystems; vector components. As a rule the state vector
essentially simplify solution of the coordination components are calculated in the process of state
problem which, in this case, does not require itera- estimation problem solution by the iterative
tive calculations by subsystems; and reduce the methods. In emergency situations the speed of
time for SE problem solving for the entire scheme. achieving the result can prove to be insufficient.
Functional decomposition of the SE problem Measurement of these components by using PMU
allows one to coordinate interaction between the is the most attractive method. PMU installed at
problems solved at different levels, organize a the bus provides measurements of voltage mag-
flexible choice of the method to solve one or an- nitude U i and phase δi at this bus, current mag-
other state estimation problem for each subsystem, nitude and angle value between voltage and
integrate the methods of artificial intelligence current φij in the entire branches incident to this
and numerical methods, accelerate the process
bus or in some of them subject to transmission
of measurement processing, and, thus, reduce the
capacity of communication channels (Phadke,
time of system state estimation.
2002). In practice, however, installation of PMU
The use of measurements from PMUs placed
at each bus is impossible.
at boundary nodes of subsystems allows state esti-
The paper presents a non-iterative method to
mation to be performed for individual subsystems
calculate voltage magnitude and phase at all the
independent of one another. Solution of coordi-
buses, allowing the state parameters of EPS to be
nated problem in this case implies calculation of
obtained fast enough.
nodal injections at boundary nodes and does not
require iterative calculations.
Problem Statement
The developed algorithm of placing PMU by
the annealing method and the use of “calculated”
To calculate load flow in EPS fast, the complex
PMU make it possible to reduce the number
network configurations are reduced to a radial
of PMUs required for solving the coordination
form and the voltage magnitude and phase at all
problem which in this case is essentially lower
buses are calculated directly. For this purpose an
than the number of boundary nodes.
optimal set of measurements from the SCADA
Simulation calculations as well as the calcula-
system and PMU and the known relations between
tion of a real scheme demonstrate the efficiency of
the state variables in EPS are applied.
the suggested approach to electric power system
The calculated PMU is installed at bus j. Instal-
state estimation.
lation of the calculated PMU means calculation of
13
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
voltage magnitude and phase at the neighboring Calculation of δ, U at all EPS Buses
buses by the formulas:
Calculation of δ, U starts with the reference bus
that is determined in accordance with (Glazu-
U j = U i − I ij ( R cos φij + X sin φij ) , (6)
nova, 2009). PMU is installed at the reference
bus. The process of searching for the reference
I ij ( X cos φ − R sin φ) bus reveals buses and branches with insufficient
δ j = δi − arctg . measurements for determination of δ, U at the
U i − R cos φ − X sin φ
(7) next bus. Their calculation can be continued by
installation of additional PMUs. These buses are
The phases δ k at bus k are calculated through also called reference buses.
measurements of ( U k , Pk − j , Qk − j ). To do this, the
Algorithm for Calculation of δ, U
active and reactive voltage drops (∆U ' , ∆U '' )
are calculated by the expressions 1. Reduction of the scheme to a radial form
(removal from the network graph of buses
Pk − j R + Qk − j X forming loops).
∆U ' = (8)
Uk 2. Determination of reference buses, from
which the voltages are calculated in each
radial scheme. Installation of PMUs at the
Pk − j X − Qk − j R reference buses.
∆U '' = (9)
Uk 3. Installation of the calculated PMUs at the
buses adjacent to the reference ones.
4. Calculation of voltage magnitude and phase
then δ k is determined at all buses of each radial scheme.
5. Calculation of voltages at the removed
∆U '' buses. To do this, the state estimation of
δ k = δ j − arctg . (10)
U j − ∆U ' each scheme is performed, with the number
of buses equal to m+1, where m – number
of branches adjacent to the removed one.
With the measurements shown in Figure 3 the
6. Comparison of values of voltage phasors
voltage magnitude at the bus k is calculated by
that are obtained at point 4 with the values
the formula:
calculated at point 5. If the difference is
less than the threshold d, then go to point
U k = (U k − ∆U ' )2 + (∆U '' )2 . (11) 7, otherwise to point 11. The threshold is
determined by the formula:
δ k is calculated by expression (5), where
d = 3σ ,
(∆U ' , ∆U '' ) are determined through
Pj−k , Q j−k ,U k .
where σ - variance of measurements. For voltage
σu2 = 2 ; d u = 3 * 2 = ±4.24ΚBm ;
14
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
d δ = ±0.2o .
Where d p = 3 * 25 = ±15MW ,
15
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
The given network scheme is reduced to the Voltage magnitudes (if it is needed) and phases
radial form by removal from the graph of bus 3 for other buses are calculated by the following
as the most connected (Figure 4). The dashed lines measurements from the SCADA system:
show branches adjacent to the removed bus. Bus At buses 4, 6 – by the active and reactive power
13 is a reference bus. For the non-iterative calcu- flows at the beginning of the branch (10), (11).
lation of load flow to be made on the basis of At buses 11, 5, 7 – by the active and reactive
measurements presented in Figure 4 an addi- power flows at the end of branch and voltage in
tional PMU should be installed at bus 10. Branch the following succession:
7-8 becomes a boundary one.
As a result the transformed scheme has two 1. Calculation of voltage drop from (8), (9);
radial schemes with the vertex at bus 13 (13-1- 2. Calculation of δ k by Equations (10), where
11; 13-2; 13-12-5-4-6-7), and the vertex at bus k=5, 7, 11.
10 (10-8; 10-9), one boundary branch (7-8), four
branches adjacent to the removed bus. Voltage at the bus removed from the scheme
PMUs are installed at buses 13, 10. Voltage shown in Figure 4 is calculated by separation of
magnitudes and phases are calculated at the the 5-bus scheme (Figure 5).
neighboring buses 1, 2, 12, 8, 9 by using the data For the obtained scheme the state estimation
of PMU from Equations (6), (7). is performed based on available measurements
by the algorithm described in (Gamm, 2007).
16
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Voltages calculated by Equations (6)-(11) are The values of active and reactive power flows
applied as initial data. Here it is taken into con- in the branch 3–4 are calculated.
sideration that the calculated PMUs are installed Balances at buses 3, 4 are maintained by correc-
at buses 2, 9, 12, i.e. the voltage magnitude and tion of the injection values there. The calculation
phase at these buses are measured with a high results are given in Table 2.
degree of accuracy. Then the following condition The second column presents the values of
is analyzed: measurements. The results of SE of the 13-bus
scheme are given in column 3. The fourth column
U SE − U cal < d , (12) shows results of SE for the 5-bus scheme. The
fifth column presents voltage values calculated
by the suggested algorithm. The Table 2 indicates
where U SE – estimated voltage values, U cal – that the voltage phasor obtained by the non-iter-
voltages calculated in the radial scheme. If condi- ative method coincides with SE results.
tion (7) is true, it means that there are no errors The values of active and reactive power flows
in measurements and the obtained state is adequate. in the boundary branch 7-8 are calculated based on
This condition holds true for bus 4: the known voltages. The values of obtained flows
and the values of flows from SE, measurements
509 SE − 508cal < 4.25 . (13) of flows are shown in Table 3.
2 3 4 5
17
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Ranch P Q
Meas SE SS Meas SE SS
7–8 -266 -264 -258 83 80 79
Analysis of the obtained values reveals that 5. State estimation for the 3-bus scheme (2-
the power flows in the boundary branch are 3-4). For the schemes with removed buses
within the normal range: the problem of SE is not solved because of
installation of the calculated PMUs there.
Pmeas − Pss < d . (14) Voltage magnitude and phase at bus 3 are
calculated through SCADA measurements
and the calculated PMU data.
Hence, the steady state of the 13-bus scheme 6. Adjustment of power flows in the boundary
is calculated. branches 6-12; 6-11, 6-13, 4-3. The calcula-
tion results are shown in Table 4.
14-Bus Scheme 7. Analysis of the obtained results.
Figure 6 presents a 14-bus test IEEE scheme.
SS of the considered scheme is calculated by Table 4 shows that condition (14) for the
using the described algorithm. boundary branches is true (the difference between
the calculated flow and the measurement is less
1. For reduction of scheme to the radial form than the threshold equal to 15 MW and 30 MVAr).
3 buses (2, 4, and 6) are removed from the Hence, the steady state for the 14-bus scheme is
network graph. Two radial schemes with the calculated.
vertices at buses 9 and 5 are obtained. The results of SE and the results obtained on
2. Installation of PMUs at buses 9 and 5. the basis of the suggested technique are pre-
3. Installation of calculated PMUs at buses 1, sented in Table 5.
2, 4, 6, 7, 10, 14. Data of PMUs are shown by bold type and
4. Calculation of voltages at the buses of data of the calculated PMUs – by italic type.
each radial scheme: 14-13-12; 10-11; 7-8. From Table 5 it is seen that the voltage values
Calculation starts with the buses, where the obtained by different techniques coincide.
calculated PMUs are installed.
Branch P Q
SE SS SE SS
6 –11 6.89 7.4 11.4 9.2
6–12 8.18 8.01 9.8 2.12
6–13 17.9 21.7 22.39 10.8
3–4 -23 -27 -5 -8
18
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
19
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
3 5
Nº
Û δ̂ U cal δ cal
1 380 0 380 0
20
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
(devices and systems of diagnostics) and also a self-organizing Kohonen network in the struc-
monitoring the EPS operation conditions seems to ture of the main system of data processing of the
be highly important because of radically changed WAMS system is offered in (Handshin, 2006).
development trends and complicated operating The Kohonen network in this case is capable
conditions of large-scale IPSs. of analyzing schemes and operating conditions
It should be underlined that in these conditions “on-line” and “off-line”. For the given ANN
a system of operation monitoring is a powerful tool the problem is solved by division of the studied
to study dynamic characteristics of EPSs in real situations into clusters with close typical change
time for different system states. The system en- of processes in EPS. Hence, the whole variety of
hances control efficiency of normal and emergency schemes and operating conditions can be visual-
conditions in the current market environment. ized effectively and dangerous contingencies in
the studied EPSs can be detected timely.
Artificial Intelligence Methods The experimental studies have shown (Kur-
in Problems of Operating batsky, 2009) that the Kohonen maps applied allow
Condition Monitoring an adequate detection of inadmissible combina-
tions of the scheme and operating parameters
Effective organization of the system of IPS and a correct choice of necessary control actions.
operating condition monitoring is possible by In addition, a key monitoring problem is the
an extensive involvement of new tools for the problem of forecasting state variables of EPS.
analysis and calculations of operating conditions, In this case within the system state estimation
primarily technologies of artificial intelligence. (Glazunova, 2009; Gamm, 2007) it is supposed
It should be underlined that the methods of to forecast all the state variables for a very short
artificial intelligence have nothing to do with algo- time span. The lead time can be from several
rithmic calculations and therefore, do not require seconds (an interval of obtaining information on
complex computational mathematical models to be EPS state variables in a unit time) to one minute.
constructed for an object. The artificial intelligence Information on the system state is obtained by us-
methods reproduce (copy) individual functions of ing the SCADA system and from PMU recorders.
the creative activity of human brain, which makes Normally modern monitoring systems employ
it possible to find optimal decisions in a large set for these purposes traditional forecasting models
of possible states in the shortest time. such as ARIMA and Kalman filter. These make
The following are primarily the main technolo- it possible to obtain rather good forecasts pro-
gies of artificial intelligence: vided the dynamics of random variations of state
variables is represented as a stationary Gaussian
• Models applying an apparatus of artificial process.
neural networks (ANN) (Haykin, 2006, In the cases where state variables are rather
Ossovsky, 2004) variable and nonstationary it is more sensible to use
• Genetic Algorithms (Goldberg, 1989) neural network models. In the course of numerous
• Hybrid and fuzzy systems (Pospelov, studies and calculations the traditional forecasting
1986). methods have proved to be insufficiently accurate
in modeling state variables of power systems as
The Kohonen maps applied for visualization compared to the neural network methods.
and analysis of schemes and operating conditions In spite of this fact it is necessary to emphasize
in EPS may be given as an example of practical that the advantage of the neural network forecast
use of the artificial intelligence technologies. Use as compared, for example to ARIMA and Kalman
21
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
filter is most pronounced in the cases where the procedure is based on the criterion of minimiz-
trend of analyzed time series differs from the linear ing the total risk of ANN error (Ossovsky, 2004).
one and the initial data contain a considerable value This allows the balance between the reliability of
of irregular component (Borovikov, 2006). In the learning and the quality of model to be reached,
cases, however, where time series of the process thus making the neural network forecasting model
contains a linear trend the traditional approaches generalize better.
practically do not differ from the neural network
ones in terms of forecasting accuracy. R(w)=Ew(W)+𝜆Ec(W) (15)
22
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
algorithms as a basis for the approach. The al- trade between different regions in terms of market
gorithms are used to analyze the time series and requirements.
networks on the basis of Gaussian radial functions,
in particular, the GRNN structure. The GRNN Study of the Properties of a
networks are learned practically in a flash, which Large Electric Power System
is extremely important for on-line forecasts, and by Using Singular Analysis
are robust to the presence of bad data.
The experimental calculation results obtained Transfer capability of a tie determines the maxi-
while minutely forecasting active power flow in mum power which can be transmitted over this tie
the 500 kV intersystem transmission lines that without deteriorating system reliability. Electric
connects two large-scale interconnected power power system has so called weak ties. The transfer
systems are presented in Figure 7 and Table 7. capabilities of these ties will be the first to achieve
The calculation data show that the neural their limit if the changes occur in the system
network model allows one to decrease the mi- conditions and this limit will vary depending on
nutely forecast error almost twice as compared the system state.
to the forecast made by the ARIMA model. Transition to market leads to the power flows
As it is seen from Figure 7 the time series of close to the maximum admissible ones. Therefore
change in the active power flow on this section it is important to find the methods to detect weak
of ESR is extremely variable and non-stationary ties in electric cutsets in which stability losses are
which does not allow the traditional forecasting most probable and to organize monitoring of their
models to be used. transfer capabilities. It is also very important to
As is seen the neural network forecast on the study the factors that determine the weakeness
basis of the proposed approach allows one to of ties and their transfer capabilities and find the
decrease the forecasting error by a factor of 2 as methods to reinforce the ties. Accurate knowledge
compared to the ARIMA model. of weak ties and their transfer capabilities is very
The need for wider application of monitoring important to plan maintenance and management
systems is growing increasingly urgent in the light of wholesale electricity markets.
of the future possibility to interconnect the main With gradual increase in the power transmitted
power grids of Europe and Russia. The techno- along the tie line the maximum power will be
logical progress in the last decades has shown that reached in the end at which there is no solution
the use of perspective information technologies, to the load flow problem. This is the result of
first of all ANN, can provide reliable operation degeneration of the Jacobean matrix J. The matrix
of future interconnected power grids of EU and relates the changes of active ∆P and reactive
Russia, optimal use of energy resources on vast ∆Q powers with changes in phases ∆δ and
territories and mutually beneficial electricity
23
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Figure 7. Use different models for short-term forecast “60 second ahead” of active power plow
magnitudes ∆U of nodal voltages in the system In (Voitov, 2000) the authors analyze the pos-
of linearized equations of power balances sibility of applying σmin for estimation of static
stability with respect to voltage. In the paper the
∆P ∆δ ∂ P ∂ δ ∂ P ∂ U ∆δ singular analysis technique is used to detect the
∆Q = J ∆U = ∂ Q ∂ δ ∂ Q ∂ U ∆U ties with the parameters and/or high active
(16) power flows that lead to a considerable rise in
voltage phases and deterioration of the Jacobi
Degeneration of the Jacobi matrix in a gen- matrix conditionality. The study on interrelation
eral case, as is shown in [Venikov, 1975], is a between the Jacobi matrix conditionality and
local index of the global phenomenon related to transfer capabilities of ties and cut sets employs
the loss of static stability. The sign of degeneration the method of contribution factor (Bialek, 2000).
can be closeness to zero of the determinant
det( J ) , equality of the conditionality number Singular Decomposition of
cond ( J ) = σmax σmin to infinity, where σmax and the Jacobean Matrix
σmin are maximum and minimum singular values Singular decomposition of the n × n asymmetri-
of the Jacobi matrix. cal Jacobi matrix can be presented in the form:
The minimum singular value of the Jacobi
matrix is another index, since the closer σmin to n
24
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
where W = (w1, w2 ,..., wn ) and V = (v1, v2 ,..., vn ) by the difference of the components of the right
- orthogonal matrices of left and right singular first singular vector (v1δ k − v1δ l ) that correspond
vectors, σ1 ≤ σ2 ≤ σ3 ≤ .... ≤ σn - singular val- to the phases of nodal voltages at the k-th and l-th
ues given in an ascending order. nodes and can be applied as an index of the tie
If σ1 = σmin is considerably lower than the weakness. The larger the change in ∆δij with
rest of the singular values then, all other condi- increase in the flow along the tie the faster the
tions being equal, the least contribution to the maximum of the transmitted power will be reached
change in phases and magnitudes of nodal volt- in it and the Jacobi matrix will degenerate.
ages is made by the first term of the decomposi- Another index of tie weakness that allows
tion (17). Taking into account the inversion of one to determine the network parameters or state
matrix J this can be written as follows: variables f, that make the greatest impact on the
conditionality of the Jacobi matrix is the derivative
∆δ ∆P ∆P of the minimum singular value with respect to f
T
∆U = J ∆Q = (v1 w1 / σ1 ) ∆Q .
−1
(18) ∂σ1 ∂f = w1T (∂J ∂f ) v1 .
Introduction of scalar value For example the tie k − l can be called weak
∆P
, which is called the first if the change in its admittance ykl leads to the
∆S 1 = (w1T / σ1 )
∆Q maximum change in σ1
generalized disturbance shows that the maximum
∂2 P ∂ 2 P
changes in phases and magnitudes of voltage will
∂σ1 ∂δ ∂y ∂U ∂ykl v1δ
occur at the nodes that are called sensors in = w1T (∂J ∂∂ykl )v1 = (w w1U ) kl
∂ykl 1δ ∂ 2 Q ∂ 2Q v1U
(Golub, 1995) and correspond to the maximum
∂δ ∂ykl ∂U ∂ykl
25
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
the conditionality of the Jacobi matrix and the simple and implies that the most loaded weak
nodal admittance matrix. ties are excluded unless the studied subsystem is
divided into independent subsystems.
Contribution Factor of
Power Transmission Case Study
The contribution factor method (Bialek, 2000) Let us consider a 14-node network scheme (see
makes it possible to determine for a known load Figure 8). Nodes 1, 3, 101, 201 and 203 are gen-
flow between what ties in the electric network the erating, node 101 is slack, nodes 4, 6, 100, 202 are
power of each generator node is distributed and load ones. Directions of active power flows for the
in what proportion. If to assume that all tie lines, base network state that is far from the steady-state
along which the power is transmitted from gen- stability limit are shown in the scheme by arrows.
erator node j, have equal transfer capabilities then The plan for detection of weak ties and cut
with an increase of power at the generator node sets of the test scheme is the following:
power flow in the tie that has the maximum coef-
ficient aklj will reach its limit value earlier than 1. Determination, for the base conditions by
the flows in the other ties. The algorithm (Golub, using the singular analysis, of the nodes
1995; Gamm, 2003) can be used to determine the with the maximum response to generalized
coefficients aklj . It suggests searching for the paths disturbances, the nodes that cause maximum
from each generator on the directed graph with generalized disturbances and also the weak
the orientation of branches that coincides with cut sets, the changing conditions of which
the direction of flows along these branches. have the greatest effect on the Jacobi matrix
If power of the generator node increases by conditionality.
the value ∆ p , the power flow in the tie k - l will 2. Analysis of factors influencing the transfer
capability of weak ties.
increase by the value ∆ pkl = aklj ⋅ ∆ p , where 3. Confirmation of the fact that with the increase
aklj is a contribution factor that determines the in power transmitted on the path along weak
contribution of generator node j to the flow along ties the limit of transmitted power is reached
the tie k - l from the node k to the node l. in the most loaded weak tie.
In the subsystem transmitting power of a
specific generator, the ties ranked as weak on the Components of the right singular vector that
basis of singular analysis may have low contribu- correspond to phases of the nodal voltages are
tion factors and vice versa the ties that are not presented in Figure 9a, curve а. They determine
weak may have high contribution factors. To find nodes 200-203 as nodes with sensor phases and
the compromise solution the ties of each subsys- separate two groups of coherent generators – 201,
tem can be assigned the weight equal to the 203 and 1, 3, whose voltage phases have a similar
product of the tie weakness index and the contri- response to external disturbances.
bution factor, for example the weighting coeffi- Analysis of the components of the left singu-
cient for the tie k - l can be represented by lar vector that correspond to nodal voltage
(v1δ k − v1δ l ) ⋅ aklj . phases, Figure 9b, curve а, reveals that generation
change at nodes 201 and 203, should lead to the
The information on the maximum weighting maximum response of voltage phases.
coefficients makes it possible to detect weak cut Curves а in Figures 9c and 9d illustrate the
sets in every subsystem. This procedure is rather values of difference of the right singular vector
26
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
components that correspond to node numbers at • Slack node 101 to load nodes 4 and 6 - sub-
the tie ends and derivatives of the minimum sin- system b);
gular vector of the Jacobi matrix with respect to • Generator node 201 to load nodes 4, 6 and
tie admittances. Analysis of the curves reveals ties 202 - subsystem c) that includes weak ties
200-8, 8-5 and 202-100 as weak. The ranging of 200-8, 8-5 and 202-100;
weak ties using the first and second criteria differs. • Generator node 203 to load nodes 4, 6,
Therefore, it is possible to suppose that weakness 202 - subsystem d) that includes weak tie
of the 202-100 depends to a greater extent on 202-100.
the conditions and weakness of ties 200-8, 8-5
on their admittances, the latter is proved by the The effect of the increase in power transmitted
studies carried out in (Golub, 1995). over weak ties on the Jacobi matrix conditional-
We will show that with the increasing active ity will be illustrated by considering the simulta-
power flows on the path with loaded weak ties the neous equal increase (decrease) of active power
limit of transmitted power is found in such ties. of generation at nodes 201 and 203, whose con-
Let us determine by the method of contribution tribution to the generalized disturbance is maximal,
factor the subsystems in Figure 10 transmitting by the value ∆ p . This will lead to decrease (in-
power from: crease) of generation at base node 101 by the
value 2 ⋅ ∆ p plus the changing losses.
• Generator nodes 1 and 3 to load node 4 -
subsystem а);
27
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
28
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Figure 10. Subsystems transmitting active power from certain generator to load nodes
Figure 11a shows the change of minimum sharp decrease of conditionality number. The best
singular value of the Jacobi matrix with increase conditionality of the Jacobi matrix is determined
and decrease in the active power of generation at with the decreasing generation by the value
nodes 201 and 203. When generation at node 201 35 ⋅ ∆ p with respect to the base state denoted
or 203 increases by the value larger than by null. In Figure 11b curves b are constructed
16.75 ⋅ ∆ p , steady-state stability is violated with for the limit conditions and characterize the in-
Figure 11. Change of the minimum singular value of the Jacobean matrix and the voltage phase difference
29
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
creasing sensitivity due to deterioration of the In the initial and heavy conditions in subsystem
Jacobi matrix conditionality. с ties 200-8, 8-5 and 202-100 are the most loaded
Change in the difference of voltage phases in weak ties. With the change of power at node 201
the ties on the path of power transmission from the contribution factors of these ties vary insig-
generator nodes 201 and 203 in subsystems с and nificantly. When the power of node 201 changes
d, Figure 11b, shows that the Jacobi matrix is by the value −35 ⋅ ∆ p , at which the Jacobi
degenerated, because the phase difference in the matrix conditionality is the best, tie 202-100 is
202-100 becomes high. no longer weak and even its tripping does not lead
The weakness of the 202-100 is also confirmed to instability. The weak cutest is observed in the
by the impact of increasing admittances of weak most loaded weak ties 202-100 and 8-5 of sub-
ties on the minimum singular value of the Jacobi system с.
matrix and the limit value of generation at nodes On the basis of the singular analysis of the
201 and 203 that is shown in Table 8. With practi- Jacobi matrix and the method of contribution
cally the same impact on the conditionality number factors the index for determination of the most
the increasing admittance of tie 202—100 leads loaded weak ties is suggested. In these ties with
to much higher transfer capability of the network the increase in transmitted power the limit of
in comparison with ties 200-8, 8-5. transfer capability is reached earlier than in other
Table 9 illustrates the change in contribution ties, which leads to steady-state stability loss.
factors of ties in subsystems с and d with the Further studies on weak ties should provide for
increasing generation at nodes 201 and 203 and consideration of the constraints imposed on op-
also the values of weighting coefficients in the erating parameters.
initial and limit conditions.
Weak tie 202-100 in subsystem d is the most
heavily loaded. With the increase of generator
node power the contribution factor of this tie also
increases. Tie 202-100 in subsystem d has maxi-
mum weighting coefficients in both initial and
heavy conditions.
Table 8. Impact of increasing admittances of weak ties in per cent (Y %) on the limit power ∆ p of
generation at nodes 201 and 203 and the minimum singular value of the Jacobi matrix
Increase of
Y%
∆p σ1 = σmin
202-100 200-8 8-5 200-8 8-5
30
Table 9. Change of the contribution factors for ties in subsystems с and d with increase of generation at nodes 201 and 203 by the value ∆ p
and the weighting coefficients (v1δ k − v1δ l ) ⋅ aklj shown in bold type for the initial and limit conditions
Ties of subsystems с d
∆p
200-8 8-5 5-6 5-2 2-4 100-6 200-202 202-100 100-4 202-100 100-4 100-6
-45.0 0.999 0.985 0.896 0.076 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001
-25.0 0.950 0.940 0.810 0.111 0.111 0.002 0.041 0.011 0.001 0.260 0.024 0.053
-15.0 0.929 0.908 0.750 0.130 0.130 0.005 0.071 0.030 0.002 0.428 0.035 0.053
-5.00 0.900 0.884 0.710 0.142 0.141 0.007 0.048 0.048 0.003 0.530 0.037 0.075
0.00 0.901 0.873 0.700 0.146 0.145 0.007 0.098 0.057 0.004 0.570 0.037 0.073
0.00 0.353 0.425 0.008 0.014 0.003 0.002 0.000 0.036 0.001 0.365 0.013 0.018
4.00 0.895 0.865 0.687 0.148 0.147 0.007 0.104 0.063 0.004 0.600 0.036 0.069
8.00 0.889 0.856 0.673 0.149 0.148 0.072 0.110 0.069 0.004 0.620 0.035 0.065
12.00 0.882 0.845 0.650 0.149 0.148 0.007 0.117 0.076 0.004 0.646 0.034 0.061
16.00 0.869 0.823 0.630 0.145 0.143 0.008 0.130 0.087 0.004 0.668 0.033 0.058
16.75 0.850 0.805 0.613 0.139 0.136 0.008 0.141 0.095 0.005 0.670 0.033 0.060
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
16.75 0.336 0.387 0.003 0.002 0.000 0.014 0.001 0.073 0.001 0.513 0.004 0.006
31
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
A Multi-Agent Approach to instability that may occur any time after the first
Coordination of Different severe contingency and lead to blackout.
Emergency Control Devices Existing practice shows that if protection
Against Voltage Collapse system works correctly, most power systems have
sufficient stability to withstand the first heavy
Power Industry spends a lot of money to protect disturbance in EHV transmission system. The
a power system against different severe distur- post-disturbance phase represents a deceptively
bances. Nevertheless, large interconnected Power calm period that lasts several minutes with a normal
Systems throughout the world are frequently level of frequency and then voltage collapse that
subjected to widespread blackouts which interrupt lasts seconds (Lachs, 1992).
millions of consumers and cost billions of dollars. The first heavy disturbance leads to increase in
Analysis of the recent blackouts showed, that the reactive power losses and reactive power out-
the most severe interruptions occurred in highly put of rotating units in the vicinity of the affected
loaded interconnected power systems due to EHV region. So, the first disturbance effects influence
line disruption followed by multiple contingencies only the affected region, being initially a local
(CIGRE, 2007). These accidents highlighted the problem. But some time after, the lack of reac-
deficiency of the existing protection systems that tive power in the affected region might increase
cannot maintain the integrity of the transmission considerably, leading to voltage collapse in the
grid during multiple contingencies (Lachs, 2002). neighboring regions and even in the whole system.
Power system behavior in an emergency state is This happens because if the disturbance is not dealt
characterized by complex interaction between dis- with timely, the after-effects spread out through the
crete and continuous control devices. Continuous EHV transmission network and actuate different
control devices are automatic voltage regulators, control devices such as automatic voltage regula-
turbine governors, FACTS devices, etc. tors, automatic transformer tap changers, current
Discrete control devices are different pro- protection relays, etc. These control devices act
tection relays, under load tap changers, etc. at the different speed, respond to changes in the
Currently both continuous and discrete control immediate vicinity and act without coordination
devices substantially use local signals only and with one another. Their actions in response to the
do not coordinate their actions with each other. post-disturbance conditions are actually the main
Absence of coordination between discrete and cause of power system breakdown; consequently,
continuous control devices is the shortcoming of the timely control of the discrete and continuous
the existing protection system and it may lead to control devices under the post-disturbance condi-
blackout. This section presents a control system tions is the only means to prevent voltage collapse
based on the multi agent approach. The control of the whole system (Lachs, 2002). Undoubtedly,
system provides coordination between discrete the absence of different control actions coordina-
and continuous control devices to prevent volt- tion during the post-disturbance period can cause
age instability. different types of instability. But first of all, one
should cope with voltage instability because it
Voltage Instability Mechanism was the main cause of the recent blackouts. New
system protection system philosophy has to be
To understand the importance of the discrete and proposed to prevent voltage instability during the
continuous control devices coordination, one post-disturbance period.
should understand the mechanism of voltage
32
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
33
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
voltage levels and increase of reactive power • Countermeasure 4. Strategic load shed-
outputs on rotating units (Lachs, 2003). Reduction ding at selected transmission substations
of voltages and increase of rotating unit excitation only if voltage levels and reactive outputs
were proposed in different papers to indicate the do not meet the requirements, or some
proximity to voltage collapse. Thus, these two transmission lines are overloaded.
criteria may be used to detect the critical situation • Countermeasure 5. Rearranging genera-
appearance and activate protection system. tor MW outputs. Connecting part of the
disconnected load.
Control Actions
Power industry has already used the philosophy Countermeasures 1–3 have approximately the
of load shedding by selecting non-essential load same execution time and their main purposes are
to prevent frequency reduction. The analysis to impede the sharp increase of series reactive
of recent blackouts showed that the rapid load power losses, to increase transmission line charg-
shedding is usually the only way to prevent the ing and to inhibit tap changing on subtransmis-
collapse of the whole system (CIGRE, 2007). On sion and distribution transformers. Load is shed
the one hand, load shedding should be as fast as (Countermeasure 4) only after countermeasures
possible, on the other hand, it should be optimal. 1 – 3. This will decrease the amount of the load
The optimal load shedding scheme can be realized to be shed. Countermeasure 5 considers an opti-
by using different optimization procedures, but mization procedure. The optimization procedure
it is hard to solve optimization problem for any takes much more time in comparison with coun-
possible situation in advance, because the number termeasures 1–4 and provides post-emergency
of situations is too big. This means that some op- operation optimization.
timization computations should be made during Thereby, countermeasures 1–4 provide fast
the post-disturbance period. In spite of the fact control of the post-disturbance phenomenon to
that there is a number of optimization techniques avoid voltage collapse and countermeasure 5 pro-
that can be used to calculate emergency control vides long-time-period post-emergency operation
actions quickly, the amount of input data required optimization. The proposed control principles can
to solve the problem is usually too big. The state be applied to various parts of the grid that work
estimation alone can take from tens of seconds to independently. Briefly, the control actions aim to
minutes. However, load shedding under the post- control the capacity of the available reactive power
disturbance conditions has to work faster. Hence, resources and do not let reactive power demand of
load shedding procedure has to use less complex the affected region increase beyond their sustain-
methods to control post-disturbance phenomenon. able capacity to exclude the possibility of voltage
The following simple countermeasures to control instability (Bellifemine, 2000).
post-disturbance phenomenon were proposed in The proposed control system can be built by
(Lachs, 1992): using distributed intelligence principles. The
distributed intelligence is taken to mean the multi-
• Countermeasure 1. Fast tap changing on agent system.
transmission substation transformers.
• Countermeasure 2. Raising terminal volt- Multi-Agent Control System Structure
age on selected synchronous condensers
and hydro generators. The proposed multi-agent control system provides
• Countermeasure 3. Fast tap changing on reactive power control to prevent generator trip-
selected generator transformers.
34
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
ping and preserve load bus voltages within the MAS Ontology
normal range. Agents communicate with each other, by us-
A power system is used to illustrate the main ing some communication language. According
principles of the proposed multi-agent approach. to FIPA standards, messages exchanged by
This power system is a part of the modified 24 bus agents have a number of fields and in particular:
IEEE One Area RTS-96 system (see Figure 15). sender, receiver, communicative intention (also
It is divided into two subsystems - Subsystem A called”performative”), content, language, ontol-
and Subsystem B that correspond to transmission ogy and some fields used for control. Ontology
and subtransmission plus distribution systems is the vocabulary of symbols and their meanings.
respectively. The proposed MAS consist of two For the effective communication, both the sender
types of agents: Load Agents and Generator and the receiver must ascribe the same meaning to
Agents. Any agent at any time has the following symbols. Ontology can include different elements
set of local data: such as agent actions, terms, concepts, etc. Agent
actions indicate actions that can be performed by
• Local state variables (primary and second- some agents. Terms are expressions identifying
ary voltages, power flows, etc.). entities (abstract or concrete) that”exist” in the
• Operating characteristics of the local world. For voltage control purposes, following
equipment (generator terminal voltage, tap the simplest Voltage Control Ontology can be
range of the tap changer, excitation current proposed:
of the generator, etc.). Agent actions of the Voltage Control Ontology:
35
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
FIPA Request Interaction Protocol and include stopped increasing reactive power. Thus, Gen-
Increase Reactive Power action of the Voltage erator Agent always knows when reactive power
Control Ontology. The sequence diagram for the increase in his subsystem is stopped. If reactive
Request Interaction protocol used by the Generator power increasing is stopped, but Generator Agent
Agent is depicted in Figure 12. is still overexcited, it starts Load Shedding pro-
Before sending a message, Generator Agent cedure (see Figure 13).
could use a rule set to identify whether receiver FIPA Contract Net Interaction Protocol is used
is able to help him. In our research, we used the in Load Shedding procedure. In this protocol, the
following simple rule: Generator Agent do not initiator wishes to optimize some function that
send Request message to another agent if electric characterizes the Load Shedding Procedure. We
coupling between them has become too weak. use minimal voltage rate function, but of course,
For instance, if Bus202 – Bus203 active power it could be function, which includes some eco-
flow is equal to zero, Generator Agent at Bus 203 nomic aspects. Generator Agent sends n Call for
does not send Request message to Generator Agent Proposal messages to Load Agents and solicits
at Bus 202. from them m proposals and k refuses. The propos-
In response to his request, Generator Agent can als contain voltage rates at primary buses of the
receive either Refuse or Agree message. Agree Load Agents. After that, Generator Agent accepts
message means that Request Interaction proto- j proposals and sends j Accept-Proposal mes-
col participant starts to increase reactive power. sages to those Load Agents which have the low-
Sometime later, Generator Agent will receive est voltage rates at their primary buses. When
Inform-Done message with Stop Reactive Power Load Agent receives Accept-Proposal message it
Increase action, which means that the participant
36
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
starts to shed the load until its primary voltage values. Changing transmission transformer tap
will not increase up to the specified value. ratio, Load Bus agent must coordinate its actions
Now consider situation when Generator Agent with generators in transmission system.
receives Request message. First, it analyzes oper-
ating characteristics of the generator and if they Multi-Agent Control System
are within the normal range it starts to increase Implementation
reactive power output according to the algorithm,
presented in Figure 14. The success of multi-agent system mainly depends
on the availability of appropriate technology
Load Agent (development tools, programming languages)
Load Agent obtains local information about pri- that allows its implementation. Any kind of
mary and secondary voltages at the substation, programming language could be used for MAS
transformer tap ranges and active power flows. realization, but object-oriented languages are more
Load Bus agent takes part in Load Shedding pro- suitable, because the concept of agent is close to
cedure. It also can shed the load independently the concept of object.
in case of critical voltage drop. If it is installed at The computer model of the proposed MAS
transmission system substation, Load Agent can for power system voltage stability control was
take part in reactive power regulation. In this case, implemented in JADE. JADE has become a firm
Load Agent changes transmission transformer tap favorite with researchers in power engineering in
ratio until primary voltage will not decrease or recent years. JADE implements a famous object-
secondary voltage will not increase up to specified oriented language Java. Agents, developed for
37
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
the JADE platform consist of three basic layers: To provide communication between Matlab
a message handling layer; a behavioral layer; a and JADE, Box Agents are used. Box Agents are
functional layer. Message handling layer is respon- Java objects that contain different data structures.
sible for the sending and receiving of messages During Time Domain Simulation, information
from other agents. The behavioral layer provides about power system operating conditions at each
control of when an agent has to implement some integration step passes from Matlab environment
task. The functional layer embodies the action to JADE by means of Box Agents. After that,
the agent can perform. JADE provides program- agents inside JADE environment process this
mers with the following ready-to-use functions: information, produce control actions if needed,
full compliance with the FIPA specifications; put information about control actions inside Box
efficient transport of asynchronous messages; a Agents and pass Box Agents back to Matlab en-
simple agent life-cycle management; a library of vironment. Thus, there is no need to use computer
interaction protocols, etc. For further information hard disc during the simulation, all computations
about JADE platform, see (Bellifemine, 2007), are performed inside the main memory and simu-
(Taylor, 1997), (Milano, 2005). lation process is faster.
Necessary power flows and time domain The proposed MAS software realization allows
simulations were carried out in Matlab/PSAT one to use complex Matlab/PSAT routines and to
environment (Milano, 2008). Java capabilities of model complex behavior of the agents.
the JADE environment were used to implement
communication between Matlab/PSAT and JADE.
38
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Figure 15. A part of the modified IEEE One Area RTS-96 system
• To explore the influence of the ULTCs After modification, IEEE One Area RTS-96
actions during low voltage conditions, system contains 42 buses. Parameters of the
transformers equipped with ULTCs were unmodified 24-bus test system can be found in
installed between subtransmission system PSAT test folder (Milano, 2008). Parameters of
and distribution system loads. the modified 42-bus test system can be found
• Each load was modeled as 50% constant in [Modified]. For better understanding of the
impedance and 50% constant current for transient process, agents were installed only at
both active and reactive components. the buses depicted in Figure 15.
39
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
• 2 seconds after the simulation starts. Loss • ULTCs are installed at the subtransmission
of the generator connected to the Bus 201. substations Bus204 – Bus210. ULTC time
• 40 seconds after the simulation starts. Loss delay for the first tap movement is 20 sec-
of Bus208 – Bus207 line. onds. ULTC time delay for subsequent tap
movements is 5 seconds. ULTC tap range
Preliminary Remarks to the Simulation Process is ±12 steps.
During the simulation process, two types of au-
tomatic systems are considered: Voltage reductions at load substations during
the simulation are shown in Figure 16a. The change
• Automatic system based on conventional of rotor currents during simulation is represented
principles in Figure 16b.
• Automatic system based on multi-agent After the first disturbance, rotor current of the
principles. generator, connected to Bus203, reaches its ther-
• Both automatic systems do not provide for mal limit, and AVR reference voltage of the
decentralized Under Voltage Load Shedding generator starts to decrease. 20 seconds after the
(UVLS) scheme. Undoubtedly, decentral- first disturbance, ULTCs on all transformers at
ized ULVS scheme is an effective means of the affected subtransmission substations starts to
preventing voltage collapse and it should be work. This leads to further decrease of generator
provided for both conventional and multi- 203 AVR reference voltages. Compensating reac-
agent automatic systems. However, the tive power shortage, generator 202 increases its
main purpose of the simulation is to dem- excitation current. After the second disturbance,
onstrate the MAS advantages in relation to rotor current of generator 202 reaches its thermal
reactive power sources coordination for the limit and rotor current of generator 203 exceeds
purpose of generator tripping prevention. It its thermal limit. AVR reference voltages of both
should also be mentioned, that the proposed generators continue to decrease and after a while,
centralized multi-agent ULVS scheme dif- this will lead to generator 203 tripping and to the
fers from conventional centralized ULVS voltage collapse.
scheme, because it is actuated without time
delay in case when there is no available re- Dynamic Simulation for Automatic System
active power in a subsystem. Based on Multi-agent Principles
In addition to the set of local devices, represented
Dynamic Simulation for Automatic System for conventional automatic system, multi-agent
Based on Conventional Principles automatic system also includes ULTCs for trans-
Conventional automatic system includes the fol- mission transformers at Bus101 – Bus103. Trying
lowing set of the decentralized devices: to exclude generator tripping, multiagent automatic
system coordinates the work of local devices.
• TG and AVR at each generator. Voltage reductions at load substations during the
• OXLs at the generators, connected to simulation are shown in Figure 17a. The change
Bus201 – Bus203.
40
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
41
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
42
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
of rotor currents during simulation is presented in power system interconnections which may cause
Figure 17b. damage to equipment, interruption of power supply
After the first disturbance, rotor current of the to consumers and unwanted development of emer-
generator, connected to Bus203, reaches its thermal gency processes with severe consequences for the
limit and the generator sends request message to entire interconnection and its parts (Sovalov, 1988;
generator 202 and to the transmission transformers, Pourbeik, 2006; Mаkаrov, 2005 et al). With the
connected to Bus101 – Bus103. Transmission future possibility of an interconnection between
transformers at Bus101 – Bus103 as well as gen- UCTE and IPS/UPS power systems by use of AC
erator 202 are trying to decrease reactive power tie lines the occurrence of out-of-step operation at
shortage. the interface between these power systems can be
Their joint actions decrease generator 203 dangerous for both of them and result in undesir-
excitation current. Excitation current becomes able consequences for the systems and consumers.
lower than its thermal limit, and generator 203 Measures are therefore required to detect, prevent
AVR reference voltages starts increase. After the and eliminate out-of-step conditions.
second disturbance, rotor currents of both genera- Special automatic out-of-step protection sys-
tors reach their thermal limits and generators send tems (OSPS) have been used in electric power
request messages to each other and to transmis- systems for reliable, timely and selective detec-
sion transformers at Bus101 – Bus103, but in this tion and elimination of out-of-step conditions
case, the generators receive refuse messages and (Sovalov, 1988; Gonik, 1988; Brinkis, 1975). The
immediately start load shedding procedure. Thus, most effective system is the so called selective
during the transient process, rotor currents of the OSPS which is based on the angle measurement
generators remain within the normal range. This fact (Brinkis, 1975; Quintana, 1991). Previously the
excludes the possibility of the generator tripping. difficulties of comparing angles at different po-
The absence of the control devices coordina- sitions within the network made us use indirect
tion during the post-disturbance period is one of angle calculation. Most popular methods of this
the main causes of the voltage instability, which calculation are based on determination of current
permanently occurs in power systems all over the amplitude or complex impedance at a connection
world. The proposed multi-agent control system point of automatic system (Brinkis, 1975). In this
provides reactive power control by coordinating the case the system is represented by a two-machine
work of different discrete and continuous control equivalent with regard to cutset of the ties in
devices in a post-disturbance period. The reactive which the OSPS is installed. The parameters of
power control in a post-disturbance period prevents the two-machine equivalent are determined on
generator tripping and maintains load bus voltages the assumption that the motion of generators in
within the normal range. The efficiency of this ap- the initial system along both sides of the cutset at
proach has been proved by numerical simulations. issue is coherent (Brinkis, 1975; Narovlyansky,
2005 et al). This assumption is based on the fact
Advanced Emergency Control that kinetic energy of generators’ mutual oscilla-
System for Prevention and tions in the transient process under disturbance,
Elimination of Power System Out- in the case of out-of-step conditions, passes to
Of-Step Operation Using PMU the kinetic energy of the out-of-step motion of
two groups of generators along both sides of the
Out of-step operation in power system intercon- cutset at which the out-of-step conditions occur,
nections is one of the most severe emergency while inter-machine oscillations within these two
conditions. It is related to the loss of stability in groups of generators decrease essentially.
43
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
The easiest way to implement an indirect In (Padiyar, 2006) the measurements of voltage
method for calculating the transmission angle is the phases and differences of their first derivatives are
use of angle dependence of transmission current. used to forecast power system stability losses ac-
The disadvantage of this method is a wide scatter cording to the criterion based on energy function. In
of operating angle values of the OSPS under the (Yutian, 2008) an integrated criterion is suggested
assumed current pickup settings due to various to reveal the center of oscillations with the use of
possible compositions and structure of ties in the estimates of the voltage magnitude projection at
cutset in different schemes and conditions of power some point of the tie line between two parts of the
system. Besides, this dependence is nonlinear. system and current along this tie line when using
The lesser error is obtained by the use of the so the two-machine equivalent of power system on
called phantom scheme, i.e. by modeling of the the basis of the generators motion coherence in
voltage phasor of a point located at some distance these two parts of the system which is estimated
from the site of automatic system placement, for on the basis of currently measured angles.
example a receiving end of transmission line. The authors of this chapter suggest the prin-
Such an approach is used in the OSPS installed in ciples of creating a modified selective out-of-step
power systems in Russia. Particularly complicated protection and prevention system (SOSPPS) with
conditions for selective operation of OSPS occur in the use of PMU measurements. Its efficiency is
the multi-frequency out-of-step conditions along demonstrated on the test power system.
several cutsets (Brinkis, 1975).
Further development of selective OSPS has Principles of Designing a
resulted in the creation of a multifunctional Modified SOSPPS
device. The device makes it possible not only to
eliminate the out-of-step conditions if they have The Scheme of Interrelation Between States
occurred but also to prevent their occurrence. It and Control Actions
has two stages of control actions (Brinkis, 1975): Loss of synchronism in power system operation
the control actions of the first stage are intended at a cutset can be caused by two main reasons:
to prevent the loss of stability and for this purpose
generation is disconnected in the surplus part of • The maximum admissible transfer capabil-
the system and fast reserve is used (or secondary ity of the cutset is exceeded and, thus, the
load is shed) in the deficient part. If these control a periodic static stability of the system is
actions are insufficient and fail to prevent out-of- lost;
step operation the control actions of the second • The transient stability is lost as a result of
stage are triggered and split power interconnection. disturbance on one of the ties or near the
The use of synchronized voltage phase mea- considered cutset.
surements obtained from PMU offers principally
new capabilities of implementing the selective In both cases an indicator for loss of synchro-
OSPS and selective out-of-step protection system nism and the beginning of out-of-step conditions
(Phadke, 2008). Some OSPSs have been lately is the difference in voltage phases on the ends of
suggested on the basis of PMU. In (Centeno, the most critical tie line of the considered cutset.
1997, Bozchalui, 2006) in order to reveal tran- In other words there is some maximum value of
sient instability the equal-area criterion is used the voltage phase difference δlim as
, whose excess
when representing the system by a two-machine indicates the beginning of the out-of-step condi-
equivalent. Its parameters are determined by the tions.
complex values of power system state variables.
44
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
In order to avoid this critical situation it is the power system to return to a secure state it is
necessary to maintain the cutset load at the level necessary to perform control actions to reduce
not exceeding some admissible level correspond- the loading of the cutset by decreasing the gen-
as
ing to δ lim < δ lim . The difference between δ lim eration of power plants on the transmitting side
as and by using fast reserve (or disconnecting sec-
and δ lim should take into account irregular varia-
ondary consumers) on the receiving side. If these
tions of flows along the tie lines and the need to
control actions are sufficient the power system
ensure transient stability of power systems under
returns to the normal state. However, if the control
standard disturbances. In Russia dispatching
actions are insufficient the system passes to the
centers of power systems use the recommended
emergency state (out-of-step conditions) which
values of transmission loading margins under
is eliminated by disconnecting the cutset (by di-
normal and post emergency states. In the power
viding the power system). Should the splitting of
systems of UCTE there are no similar explicit
the power system be unsuccessfully, then an
recommendations. Nevertheless, setting the value
as
emergency situation can develop and the post-
δlim < δlim can be expedient. emergency state may turn out to be severe and
Thus, power systems can have four states even turn into a blackout. In the event of a suc-
(Figure 18): secure, dangerous, emergency (out- cessful splitting the generation and load in both
of-step conditions) and post-emergency. The se- subsystems are balanced by generation discon-
cure state of power systems is determined by the nection in the surplus subsystem and by auto-
condition δij < δlim . PMU measurements are used matic frequency load shedding in the deficient
subsystem. Post-emergency state in this case will
to trace the current value of δij (t ) . The dangerous
be less severe as compared to the previous one.
state of power system occurs at δij > δlim . For
Figure 18. A scheme of interrelation between power system states and control actions in SOSPPS
45
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
It should be noted that traditionally in Russia’s phases of a critical time line in the cutset. Decrease
power system dispatching practice automatic load in the second derivative of voltage phase differ-
shedding and OSPS are considered separately due ence indicates conservation of transient stability
to the fact that the former is adjusted and operates of EPS. System transition to an emergency state
using the power flow values as an indicator of (out-of-step condition) is revealed provided that
overload while the latter is intended for the trans- at least for three cycles of measurements by using
mission angle action, with transmission angles PMU, the second derivative value of voltage phase
being determined by indirect methods (Sovalov, difference for the critical tie line at the cutset does
1988). The use of one and the same indicator not go down below some small value dδ min . In
which is the difference between PMU-measured this case each cycle may account for several scores
voltage phases on both ends of power transmission of milliseconds. Theoretically, dδmin = 0 , how-
allows one to consider both types of automatic ever practically this value is not equal to zero
systems as a single integrated emergency control because of errors and noise in measurements and
system. also inaccurate determination of the second de-
rivative due to discrete measurements. Determina-
Criteria for Actions of SOSPPS’s Stages
tion of the acceptable value dδmin is an indepen-
As noted above the conditions for transition from
dent problem.
normal (safe) state to the emergency state is for-
Hence, for the direct power flow through the
mulated as δij > δlim . Hence the criterion for
cutset from node if to node j the criterion for action
action of the cutset unloading stage of SOSPPS of SOSPPS’s division stage subject to transient
will look as stability will have the form
ul
Cact = ( )
δij (t ) > δlim . (20)
( )
C+ = ∆δij (t ) > 0 ∧
d 2 δ (t )
ij
≥ d δ
∧ ...
dt 2 min
In the event that the action of the cutset unload- d δ (t − 2T )
2
ij ,
∧
S
ing stage is insufficient or inefficient, then the ≥ d δ min
(22)
dt 2
difference in the voltage phases along the critical
tie line of the cutset at its overload continues to
increase and reaches the value δ limas
. This indicates where TS – cycle length between the PMU mea-
the loss of aperiodic static stability of the power surements.
system along the considered cutsets and the need The corresponding criterion C− for the reverse
to split the system. The criterion for action of the power flow through the cutset (from node j to
SOSPPS’s division stage will have the form node i) is determined in a similar way.
The general criterion of transition to an
as
Cact = ( δij (t ) > δlim
as
). (21) emergency state (out-of-step condition) and the
action of SOSPPS’s division stage is written in
the following way:
To formulate the criterion for action of SOS-
PPS’s division stage according to the conditions
of transient instability of power system under
(
as
Cas = Cact ∧ (C+ ∨ C− ) . ) (23)
46
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
As a result EPS is split into two isolated thoroughly and individually. And the results of
subsystems. studies will determine specific features of the
designs of automatic systems.
Related Problems The other component of the problem is the ac-
The formulated approach for improving SOSPPS curate determination of the second derivatives of
based on PMU measurements includes only its variation of voltage phase differences. It depends
basic principles illustrated by the test example on the monotone change of phase differences and
later in Chapter III. Actually consideration should the length of intervals between measurements, as
also be given to the related problems to be solved well as on measurement errors. The problem is
in the course of approach application. Below are that the second derivatives must be calculated by
the main problems. numerical differentiation of measured parameters.
As indicated above, it is reasonable to measure This question also requires thorough additional
voltage phases on the basis of PMU on the ends studies.
of the critical tie line at the cutset. The problem The results of studies on both components
is that the increase of transmitted power along the determine efficiency of using criterion (22) and
cutset results in different loading of individual tie in particular, certainty in setting the value dδ min
lines at the cutset. Such a situation is caused by the The issue about selectivity of action of the
parameters of the tie lines and also the structure modified SOSPPS at the multi-frequency out-of-
and parameters of adjacent electric networks. This step condition for the case of stability loss at
un-homogeneity of an electric network is revealed several cutsets remains to some extent open. It
in different disturbance-sensitivity of nodes and tie seems that selectivity of work of automatic systems
lines at the loaded cutset (Voitov, 1999). In other should be sufficiently high and acceptable, since
words those elements affect operation parameters the change of voltage phases on the ends of tie
(voltage, power flows etc.) changing to variable lines at the cutset is a quite definite indicator of
extents. Hence, conditions for transition to the the beginning out-of-step condition. Here the
out-of-step operation are formed first of all in the as
value δlim is close to 90° and the value δ lim is
most sensitive tie line. The tie line turns out to be uniquely determined by the state variables of the
critical at the cutset and it is expedient to place critical tie line of the cutset and the required
PMUs on its ends. margins of its transfer capability.
It should be noted that significance of a critical
time line at the cutset requires additional studies. Test Studies
It is explained by the fact that with the start of
out-of-step condition change in the voltage angles Let the test electric power system operates in
will be observed in all tie lines of the cutset. It post-emergency conditions when the tie 8-5 is
is important to establish the extent to which the loaded at 90% of its maximal transmission capa-
change in tie line loading at the beginning of
bility, and voltage mutual angles equal δ8−5 = 36,6
out-of-step condition is significant in terms of
the efficiency of SOSPPS operation. and δ 202−100 = 50,5.
Another problem is the necessity for verifica- Let us consider as a disturbance the unsched-
tion of criterion (23). It has two components. The uled disconnecting the one of two lines of the tie
first is associated with available errors and noise 8-5. The behaviors of voltage mutual angles and
in the measurements by using PMU, delays in in- their time-derivatives without any control actions
formation transmission, measurement frequency, are shown in Figure 19.
etc. These technical properties should be studied
47
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Figure 19. Time behavior of δ ( ), dδ/dt ( ) and d2δ/dt2 ( ),) in the absence of control actions:
a) for the tie 8-5, b) for the tie 202-100
Suppose the system is equipped with SOSPPS, we suppose (only with a view to exemplify the
and a starting value for action of the cutset unload- study) the inadmissibility of further mutual asyn-
ing stage (see criterion (20)) is set as δ min(8−5) = chronous motion for the power system, then the
52,2. At the time of t = 0,4 s the control actions starting value for action of the cutset division
will be realized as partial disconnections (of stage (see criterion (21)) in accordance with Fig-
as
generation in power surplus part of the system ure 20 is to be set as δ lim( 8−5)
= 57,3. At that case
and of load in its deficient part). Suppose the the system is being divided at the time of t = 0,5
larger disconnections are to be highly undesirable s (when the first derivative of the angle reaches
through technical and/or economical limitations. d δ8−5
The behaviors of voltage mutual angles and its maximum = 44 grad/s).
dt
their time-derivatives with above mentioned After splitting the system into two separate
control actions are shown in Figure 20. subsystems (one of them with the surplus and other
Figure 20 demonstrates the inefficiency of with the lack of active power) each subsystem
undertaken control actions for providing the sys- faces the challenge of bringing the frequency to
tem stability. The non-periodic growth of mutual admissible level. The solution is in the further
angles in the cutset, which appeared virtually reducing the power (of the generation and load
straight after the disturbance, lasts also after these respectively). This reducing may ensue less than
actions (although not so fast as without them). If that would be required under saving the parallel
48
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Figure 20. Time behavior of δ ( ), dδ/dt ( ) and d2δ/dt2 ( ) when maximum permissible unload-
ing the cutset: a) for the tie 8-5, b) for the tie 202-100
operation with unified frequency. In general the for measuring operating parameters, their
system splitting involves the more beneficial transfer, processing and application of
effect the higher power of the subsystem which operating conditions control of EPS. The
separates with the power lack. suggested basic principles of the system of
monitoring and forecasting the operating
conditions and control of EPS substantially
CONCLUSION enhance efficiency and adaptability of
the coordinated operation and emergency
In conclusion, we note the following: control in EPS. The results described in the
paper illustrate efficiency of the approach
1. Essential sophistication of the operat- and applied methods and information
ing conditions of current EPSs enhances technologies.
danger of heavy system emergencies and 2. Structural and functional decomposition
requires improvement and development of of state estimation problem is an effective
the principles and control systems of EPS method to solve the problems arising during
operating conditions. For this purpose it is calculation of large schemes. The proposed
necessary to apply new methods and tools two-level algorithm for structural decom-
49
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
50
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Ayuyev, B. I., Erohin, P. M., Neuymin, V. G., CIGRE Task Force. (2007, April). C2.02.24: CI-
Mashalov, E. V., & Shubin, N. G. (2005). The GRE defense plan against extreme contingencies.
software complex of optimal power flow solu- Clements, K. A., Denison, O. J., & Ringle, R. J.
tion for United Power System of Russia in a (1972). A multy-area approach to state estimation
competitive market. In Proceedings IEEE Power in power system networks. In Proc. IEEE Power
Tech Conference, St. Petersburg. Eng. Soc. Meeting, San Francisco, CA. (Paper
Barzam, A. B. (1964). System automatic devices. C72 465-3).
Moscow, Russia: Energiya. (in Russian) El-Kleib, A. A., Nieplocha, J., Singh, H., &
Bellifemine, F., Caire, G., & Greenwood, D. (2007). Maratukulam, D. J. (1992). A decomposed state
Developing multi-agent systems with JADE. Lon- estimation technique suitable for parallel proces-
don, UK: Wiley. doi:10.1002/9780470058411 sor implementation. IEEE Transactions on Power
Systems, 7(3), 1088–1097. doi:10.1109/59.207322
Bellifemine, F., Caire, G., Trucco, T., & Rimassa,
G. (2000). JADE programmer’s guide. CSELT & ENAS. (2004). Methodological guidelines on
University of Parma. Retrieved from http://www. power system stability. (in Russian)
jade.tilab.com/doc Etingov, P. V. (2006). Application of the systems of
Bialek, W. (2000). Tracing-based unifying frame- transients monitoring to control FACTS devices.
work for transmission pricing of cross-border Proceedings of the Workshop Methodological
trades in Europe. In Proc 2000 International Problems in Reliability Studies of Bulk Energy
Conference on Electric Utility Deregulation and Systems, Kharkov National Technical University,
Restructuring, and Power Technologies, London. Kharkov, Ukraine, (p. 6). (in Russian)
51
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
European Commission. (2006). European smart Gamm, A. Z., Zaika, R. A., & Kolosok, I. N.
Grids technology platform: Vision and strategy (2005). Robust methods of EPS state estimation
for Europe’s electricity networks of the future (p. based on the test equations and their software
45). Luxemburg. support using genetic algorithms. [in Russian].
Electrichestvo, 10, 2–8.
Falcao, D. M., Wu, & Murphy L. (1995). Parallel
and distributed state estimation. IEEE Trans- Glazunova, A., Kolosok, I., & Korkina, E. (2008,
actions on Power Systems, 10(2), 724–730. April). Test equation method for state estimation
doi:10.1109/59.387909 using PMU measurements. Proc. of Conf. Moni-
toring of Power System Dynamics Performance.
FIPA. (2010). The current set of standard FIPA
Saint Petersburg.
specifications. Retrieved from http://www.fipa.
org/repository/standardspecs.html Glazunova, A., Kolosok, I., & Korkina, E. (2009).
PMU placement on the basis of SCADA measure-
Gamm, A., Grishin, Y., Glazunova, A., Kolosok,
ments for fast load flow calculation in electric
I., & Korkina, E. (2007). New EPS state estima-
power systems, Proceedings of the International
tion algorithms based on the technique of test
Conference PowerTech’2009, Bucharest.
equations and PMU measurements. In Proc of
the International Conference PowerTech’2007, Goldberg, D. (1989). Genetic algorithms in search,
Lausanne. optimization and machine learning.
Gamm, A. Z. (1983). Decomposition algorithms Gonik Ya, E., & Iglitsky, E. S. (1988). Automatic
for solution of power system state estimation elimination of out-of-step condition. Moscow,
problem. Electronnoye Modelirovanie, 3, 63–68. Russia: Energoatomizdat. (in Russian)
Gamm, A. Z., & Golub, I. I. (1995). The prob- Handshin, E. (2006). Asset management and sur-
lem of weak places in electric power systems. In vivability of electric enrgy systems during large
Proc. 1995 International Conference on Electri- disturbances. In E. Handshin, E. Hauptmeter, &
cal Power Engineering Power Tech, Stockholm, D. Hause (Eds.), Liberalization and modernization
(pp. 542-546). of power systems: risk assessment and optimiza-
tion for asset management: The 3rd International
Gamm, A. Z., & Golub, I. I. (2003). Contribution
workshop, Irkutsk Energy Systems Institute (pp.
factor of active and reactive power transmission
156-168).
in electric power system. Elektrichestvo, 3, 9–16.
Haykin, S. (2006). Neural networks: A compre-
Gamm, A. Z., & Grishin, Yu. A. (1995). Distrib-
hensive foundation (2nd ed.). Williams Publishing
uted information processing in automated power
House.
system control systems. In Proc. V Int. Workshop
Distributed information processing. Novosibirsk, Iwamoto, S., Kusano, M., & Quantana, V. H.
Russia, (pp. 243-247). (in Russian) (1989). Hierarchical state estimation using
a fast rectangular-coordinate method. IEEE
Gamm, A. Z., Kolosok, I. N., & Paltsev, A. S.
Transactions on Power Systems, 4(3), 870–879.
(2007). Methods for decomposition of EPS state
doi:10.1109/59.32574
estimation problem when solving it on the basis of
multiagent technologies. In Scientific Proceedings
of Riga Technical University Power and Electrical
Engineering, Riga, (pp. 205-214).
52
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Jang, W., Vittal, V., & Heydt, G. T. (2008, No- Lachs, W. R. (1992, May). Voltage instability in
vember). Diakoptic state estimation using Pha- interconnected power systems: A simulation ap-
sor Measurements Units. IEEE Transactions on proach. IEEE Transactions on Power Systems,
Power Systems, 23(4), 1580–1589. doi:10.1109/ 7(2), 753–761. doi:10.1109/59.141782
TPWRS.2008.2002285
Lachs, W. R. (2002, May). Controlling Grid
Jiang, W., Vittal, V., & Heydt, G. T. (2007, May). integrity after power system emergencies. IEEE
A distributed state estimator utilizing synchro- Transactions on Power Systems, 17(2), 445–450.
nized Phasor measurements. IEEE Transactions doi:10.1109/TPWRS.2002.1007916
on Power Systems, 22(2), 563–571. doi:10.1109/
Lachs, W. R. (2003, February). A new horizon
TPWRS.2007.894859
for system protection scheme. IEEE Transactions
Kolosok, I. N., Korkina, E. S., & Paltzev, A. S. on Power Systems, 18(1), 334–338. doi:10.1109/
(2009). Decomposition of the state estimation TPWRS.2002.807065
problem at calculation for the schemes of large
Li, L., Liu, Y., Mu, H., & Yu, Z. (2008). Out-of-step
dimensionality. Proceedings of International
splitting scheme based on PMUs. Proceedings of
Conference Liberalization and Modernization
DRPT’2008 Int. Conf. Nanjing, China.
of Power Systems: Coordinated Monitoring and
Control towards Smart Grids, Energy Systems Makarov Yu, V., Reshetov, V. I., Strojev, V. A., &
Institute, Irkutsk, Russia, (p. 7). (in Russian) Voropai, N. I. (2005). Blackout prevention in the
United States, Europe and Russia. Proceedings
Kolosok, I. N., & Zaika, R. A. (2003). A genetic
of the IEEE, 93(11), 1942–1955. doi:10.1109/
algorithm as a means of enhancing the efficiency
JPROC.2005.857486
of information validation methods for the power
system control. In Proc. 30 International Conf. McArthur, S. D. J., Davidson, E. M., Catterson, V.
Information Technologies in Science, Education, M., Dimeas, A. L., Hatziargyriou, N. D., Ponci, F.,
Business and Protection of Natural Resources. & Funabashi, T. (2007, November). Multi-agent
Gurzuf, Ukraine, (pp. 145–147). systems for power engineering applicationz - Part
I. IEEE Transactions on Power Systems, 22(4),
Kurbatsky, V. G., & Tomin, N. V. (2008). Ap-
1743–1752. doi:10.1109/TPWRS.2007.908471
plication of the ANAPRO software for analysis
and forecasting of state parameters and process Milano, F. (2005). An open source power system
characteristics in electric power systems. Proceed- analysis toolbox. IEEE Transactions on Power
ings of the 8th Baikal All-Russian Conf. Information Systems, 20(3), 1199–1206. doi:10.1109/TP-
and Mathematical Technologies in Science and WRS.2005.851911
Management, part 1 (pp. 91-99). Irkutsk, Russia:
Milano, F. (2008, February). Documentation for
SEI SB RAS.
PSAT, version 2.0.0.
Kurbatsky, V. G., & Tomin, N. V. (2009). Use of the
Mogilko, R. (2008, April). SMART-WAMS re-
ANAPRO software to analyze and forecast operat-
corder of transient variables: Experience in design
ing parameters and technological characteristics
and implementation experience - Prospects for
on the basis of macro applications. Proceedings
development. Proc. of Conf. Monitoring of Power
of IEEE Bucharest Power Tech (p. 7). Bucharest,
System Dynamics Performance, Saint Petersburg.
Romania: IEEE Press.
53
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Narovlyansky, V. G., & Nalevin, A. A. (2005). A Phadke, G. (2002). Synchronized Phasor measure-
method for determination of equivalent param- ments. A historical overview. In Proc. IEEE/PES
eters of power system network during out-of-step Transmission and Distribution Conference, Nº.1,
condition. [in Russian]. Elektrichestvo, 8, 15–21. (pp. 476-479).
Office of Electric Transmission and Distribution, Pospelov, D. A. (Ed.). (1986). Fuzzy sets in the
United State Department of Energy. (2003). Grid control and artificial intelligence models. Mos-
2030: A national version for electricity’s second cow, Russia: Nauka. (in Russian)
100 years (p. 52). Washington, DC.
Pourbeik, P., Kundur, P. S., & Taylor, C. W.
Ohura, Y., E. A. (1990). A predictive out-of-step (2006). The anatomy of a power grid blackout.
protection system based on observation of the IEEE Power and Energy Magazine, 4(5), 22–29.
phase difference between substations. IEEE doi:10.1109/MPAE.2006.1687814
Transactions on Power Delivery, 5(4), 661–667.
Quintana, V. H., & Mueller, H. (1991). Partitioning
doi:10.1109/61.103664
of power networks and application to security con-
Ossovsky, S. (2004). Neural networks for informa- trol. IEE Proceedings. Generation, Transmission
tion processing (trans. I. D. Rudinsky). Moscow, and Distribution, 138(6), 535–545. doi:10.1049/
Russia: Finansy i statistika. (in Russian) ip-c.1991.0067
Padiyar, K. R., & Krishna, S. (2006). Online Shahidehpour, M. (2009). Smart Grid: A new
detection of loss of synchronism using energy paradigm for power delivery. Proceedings of
function criterion. IEEE Transactions on Power IEEE Bucharest Power Tech (p. 7). Bucharest,
Delivery, 21(1), 163–171. doi:10.1109/TP- Romania: IEEE Press.
WRD.2005.848652
Sovalov, S. A., & Semenov, V. A. (1988). Emer-
Panasetsky, D. A. (2009). Multi-agent approach gency control in power systems. Moscow, Russia:
to coordination of different emergency control Energoatomizdat. (in Russian)
Devices against voltage collapse. Proceedings
Taylor, C. W., & Erickson, D. C. (1997, January).
of International Conference Liberalization and
Recording and analyzing the July 2 cascading
Modernization of Power Systems: Coordinated
outage. Computer Applications in Power Systems,
Monitoring and Control towards Smart Grids,
10(1), 26–30. doi:10.1109/67.560830
Energy Systems Institute, Irkutsk, Russia, (pp.
171-176). (in Russian) Venikov, V. A., Stroev, V. A., Idelchik, V. I.,
& Tarasov, V. I. (1975, August). Estimation of
Panasetsky. (n.d.). Modified 24 bus IEEE one
electrical power system steady-state stability in
area RTS-96 test system. Retrieved from http://
load flow calculations. IEEE Transactions on
panasetsky.net/wpcontent/uploads/FILES/d 024
Power Apparatus and Systems, 94(3), 152–159.
my.mdl
doi:10.1109/T-PAS.1975.31937
Phadke, A. G., & Thorp, J. S. (2008). Synchronized
Voitov, O. N., Voropai, N. I., Gamm, A. Z., Golub,
phasor measurements and their applications.
I. L., & Efimov, D. N. (Eds.). (1999). Analysis
Berlin, Germany: Springer. doi:10.1007/978-0-
of Inhomogenities in Electric Power Systems,
387-76537-2
Novosibirsk, (p. 302).
54
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
Voropai, N. I. (2008). Problems and directions Voropai, N. I., Kolosok, I. N., Kurbatsky, V. G.,
of control system development in electric power Etingov, P. V., Tomin, N. V., Korkina, E. S., & Palt-
systems. Proceedings of the 3rd International sev, A. S. (2010). Intelligent coordinated operation
Scientific and Technical Conference Energy and emergency control in electric power systems.
System: Control, Competition, Formation (p. Proc. Conference on Control Methodologies and
7). Ekaterinburg, Russia: Ural State Technical Technology for Energy Efficiency (CMTEE-2010),
University. (in Russian) Vilamoura, Portugal.
Voropai, N. I., & Etingov, P. V. (2006). Develop- Wallach, Y., & Handschin, E. (1981). An ef-
ment of adaptation methods of fuzzy logic power ficient parallel processing method for power
system stabilizer. Proceedings of 2006 IEEE PES system state estimation. IEEE Transactions on
General Meeting, IEEE Press, Montreal, Canada, Power Systems, 100(1), 4402–4406. doi:10.1109/
(p. 7). TPAS.1981.316852
Voropai, N. I., Etingov, P. V., Oudalov, A. S., Wang, X. M., & Vittal, V. (2004). System island-
Germond, A., & Sherkaoui, R. (2005). Congestion ing using minimal cut sets with minimum net flow.
management using coordinated control of FACTS IEEE PES General Meeting, Denver, USA.
devices and load management. Proceedings of 15th
Zhao, L., & Abur, A. (2005). Multiarea state esti-
PSCC, University of Liege, Liege, Belgium, (p. 7).
mation using synchronized phasor measurements.
IEEE Transactions on Power Systems, 20(2),
611–617. doi:10.1109/TPWRS.2005.846209
55
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
The problem of monitoring, forecasting and control in electric power system: Voropai N.I.
Decomposition of power system state estimation problem with the use of PMU data for large di-
mension schemes: Kolosok I.N., Korkina E.S., Paltsev A.S.
PMU for fast calculation of steady state in electric power systems: Glazunova A.M.
Artificial intelligence technologies for monitoring large power interconnections: Kurbatsky V.G.,
Tomin N.V.
Study of the properties of a large electric power system by using singular analysis: Gamm A.Z.,
Golub I., Bershansky R.G
Advanced Emergency Control System for Prevention and Elimination of Power System out-of-Step
Operation Using PMU: Voropai N.I, Rehtanz C., Efimov D.N., Popov D.B., Häger U.
56
57
Chapter 2
Hopfield Lagrange Network
for Economic Load Dispatch
Vo Ngoc Dieu
Asian Institute of Technology, Thailand
Weerakorn Ongsakul
Asian Institute of Technology, Thailand
ABSTRACT
In this chapter, a Hopfield Lagrange network (HLN) is proposed for solving economic load dispatch
(ELD) problems. HLN is a combination of Lagrangian function and continuous Hopfield neural network
where the Lagrangian function is directly used as the energy function for the continuous Hopfield neu-
ral network. In the HLN method, its energy function augmented by Hopfield terms from the continuous
Hopfield network could damp out oscillation of the conventional Hopfield network during the conver-
gence process. Consequently, the proposed HLN can overcome the disadvantages of the conventional
Hopfield network in solving optimization problems for its simpler implementation, better global solution,
faster convergence time, and larger scale applications. The proposed method has been tested on differ-
ent ELD problems including all thermal units, thermal units with fuel constraint, and both thermal and
hydro units. The obtained results from the test cases have shown that the proposed method is effective
and efficient for solving the ELD problems. Therefore, the HLN method is the new contribution to the
development of new methods for solving optimization problems in power systems.
DOI: 10.4018/978-1-61350-138-2.ch002
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Hopfield Lagrange Network for Economic Load Dispatch
58
Hopfield Lagrange Network for Economic Load Dispatch
programming (QP) (Irving & Sterling, 1985), inte- However, the meta-heuristic methods may suffer
rior point method (IPM) (Ponnambalam, Quintana from slow convergence and local optimum for
& Vanelli, 1992), and Lagrangian relaxation (LR) large-scale problems. A clear consensus is pres-
(Fisher, 1973). Conventional methods require the ently heading toward the hybrid models, which
models of optimal generation scheduling problems are the combinations of both conventional and
to be represented as piecewise linear or polyno- non-conventional methods and can handle the
mial approximations of monotonically increasing present day complicated problems commonly
nature. However, such an approximation may lead seen within developed countries. Moreover,
to sub-optimal solution, resulting in huge loss of parallel techniques are also designed for solution
revenue over the time. Hence, the new trend in methods to reduce computational time in large-
the recent time is to use more realistic models of scale problems. However, the development of
hydro and thermal plants. these techniques is still limited due to hardware
In recent years, meta-heuristic optimization dependence. On the other hand, developing new
techniques have attracted much attention of re- methods which can deal with complicated prob-
searchers due to their ability to seek for global lems and obtain fast solution is being continued.
optimal solution for problems with complicated In the recently new trends, hybrid systems are
constraints. These methods have proved to be considered as promising methods and widely
very efficient since they do not place any restric- used for optimal generation scheduling.
tion on the shape of the cost curves and other Many hybrid systems have been proposed for
non-linearity in model representation. Although solving the ELD problems such as fuzzy logic
these heuristic methods do not always guarantee and GA (Song, Wang, Wang & Johns, 1997),
to find the globally optimal solution, they can hybrid EP and sequential QP (Attaviriyanupap,
provide a reasonable solution (sub-optimal near Kita, Tanaka & Hasegawa, 2002), combined
globally optimal) in a sufficient computational GA and TS (Ruangpayoongsak, Ongsakul &
time. Among the many artificial intelligence Runggeratigul, 2002), hybrid PSO and sequential
based methods, some popular ones have been QP (Victoire & Jeyakumar, 2004), hybrid EP and
implemented for solving the ELD problems LP (Somasundaram, Lakshmiramanan & Kup-
consisting of Hopfield neural network (HNN) pusamy, 2005), combined DE and QP (Coelho
(Park, Kim, Eom, & Lee, 1993), simulated & Mariani, 2006), etc. Hybrid systems can be
annealing (SA) (Wong & Fung, 1993), tabu usually the combinations whether among the
search (TS) (Ongsakul, Dechanupaprittha, & conventional methods or between the conven-
Ngamroo, 2004), ant colony search algorithm tional methods and the meta-heuristic methods
(ACSA) (Song, Chou, & Stonham, 1999), genetic or among the meta-heuristic methods to utilize
algorithm (GA) (Bakirtzis, Petridis & Kazarlis, the advantages of element methods. Therefore,
1994), evolutionary programming (EP) (Wong the hybrid systems can deal with more compli-
& Yuryevich, 1998), differential evolution (DE) cated problems and obtain better solution with
(Nomana & Iba, 2008), and particle swarm shorter computational time than many single
optimization (PSO) (Jeyakumar, Jayabarathi & meta-heuristic search methods.
Raghunathan, 2006). Most of the conventional In this chapter, a new method based on a com-
methods can offer good solution in a short com- bination of continuous Hopfield neural network
putational time but they can only deal with simple and Lagrangian function is proposed for solving
and small or medium-scale problems meanwhile the ELD problems including:
most of the meta-heuristic methods can deal with Basic ELD problem: This is a very simple form
more complicated and larger scale problems. of ELD problem. The problem is to minimize an
59
Hopfield Lagrange Network for Economic Load Dispatch
objective of total cost for thermal generating units optimization problems in different fields. The
subject to power balance constraint and power continuous Hopfield neural networks are a recur-
generation limits and ramp rate constraints for rent network type that operates in an unsupervised
each unit during a schedule time horizon, usually manner. The action of a Hopfield network is based
1 hour. The objective function considered for on the minimization of its energy function which
this problem here is a sum of quadratic fuel cost is mapped from an optimization problem to the
functions of all online units and the power loss network will converge to a solution of the problem.
in transmission system is computed via Kron’s One of the advantages of the Hopfield network is
loss formula where the power loss is a function that it can efficiently handle variable limits by its
of power outputs of units. sigmoid function. However, the applications of the
Fuel constrained ELD problem: This problem Hopfield network to the optimization problems
is more complex than the basic ELD problem with are limited to simple problems with linear con-
more objective and constraints added. The objec- straints due to the complex process of mapping
tive of the problem is to minimize both total fuel from the problem to the neural network. Moreover,
cost and emission level of thermal generating units large number of iterations and oscillation during
while satisfying different constraints including convergence process are also the major concerns
power balance, fuel delivery, fuel storage, genera- that the Hopfield network can be suffered while
tor capacity limits, fuel delivery limits, and fuel solving optimization problems.
storage limits for a certain period of time. Both In the implementation of the conventional
the objectives of fuel cost and emission considered continuous Hopfield neural network (HNN) to
in this problem are modeled as quadratic func- the ELD problems (Park et al., 1993; Su & Chiou,
tions which are the function of power outputs of 1997), an energy function for the problem is pre-
generating units. defined including objective and constraints associ-
Hydrothermal ELD problem: This problem ated with weighting factors, and then mapped into
includes both thermal and hydro generating units the HNN to determine the weighting connections
but only fuel cost of thermal units is considered for the neurons. There are many drawbacks for
while neglecting the total cost of hydro units since such an application. Firstly, the solution by the
it is very small and negligible. The hydraulic HNN for the problem is sensitive to the selected
constraints for hydro units are also included in weighting factors associated with the defined
addition to the constraints for the thermal units. energy function which may lead to local optimal
The purpose of this problem is to minimize total solution if these weighting factors are not carefully
fuel cost of thermal generating units subject to tuned. Secondly, HNN is very difficult to deal
power balance, continuity of reservoir head, with the complicated problems with nonlinear
and generator capacity limit constraints. The constraints since the problem constraints have to
functions considered in this problem including be linearized before implementing in HNN. Lastly,
fuel cost of thermal units and water discharge the computational effort of HNN for solving the
and reservoir head variation for hydro units are ELD problem is high since it needs large number of
quadratic functions. iterations to obtain optimal solution, thus it is dif-
ficult for HNN to deal with large-scale problems.
Some improvements for the conventional HNN
BACKGROUND have been proposed to overcome the mentioned
drawbacks. To speed up the computational time for
Continuous Hopfield neural networks (Hopfield, HNN, the sigmoid function of continuous neurons
1982, 1984) have been widely used for solving has been linearized (Su & Chiou, 1997, 2000).
60
Hopfield Lagrange Network for Economic Load Dispatch
By doing this way, the solution for the problem Lagrange network (HLN), is proposed to over-
is directly found with a very short computational come to the difficulties of the Hopfield network.
time similar to analytical approach. However, this In the proposed HLN method, the Lagrangian
improvement can be applied to simple problems function is directly used as the energy function
with non-binding inequality constraints. Another of the Hopfield network. The advantages of the
improvement of HNN has been proposed in (Yal- proposed neural network over to the conventional
cinoz & Short, 1997) where the energy function Hopfield network are as follows:
of HNN is formulated in quadratic programming
form for identifying with quadratic programming • The proposed neural network is not nec-
problem to determine the weight connections for essary to predefine an energy function as-
the neurons. For this improvement, the improved sociated with penalty factors and map the
HNN can deal with large-scale problems and its problem into the Hopfield network for
speed of calculation has been also considerably determining the synaptic interconnections
improved. Nevertheless, this formulation can only among the neurons.
apply to quadratic programming problems with • Since the proposed neural network uses
linear constraints. An adaptive HNN has been Lagrangian function as the energy function
proposed in (Lee, Sode-Yome & Park, 1998; for the Hopfield network, it can efficiently
Lee, Nuroglu & Sode-Yome, 2000) to speed up handle constraints of the problems without
the convergence of HNN by adaptively adjusting causing constraint mismatch. Moreover,
slope, bias, and learning rates of neurons. With the proposed neural network is not limited
the new adjustment techniques, the adaptive to the simple problems with linearized con-
HNN can deal with non-convex problems and straints as the Hopfield network, especially
its computational speed has been also improved. for the time-coupling constraints.
However, the implementation of this neural net- • The proposed neural network can give a
work is similar to the conventional HNN; that is very fast convergence to the optimal solu-
the weight connections for neurons have to be tion compared to the conventional Hopfield
pre-determined and the problem constraints have network.
to be linearized. Recently, a modification for HNN • The proposed neural network can eas-
has been proposed in (da Silva, Nepomuceno & ily deal with large-scale and complicated
Bastos, 2004) by representing the energy of HNN optimization problems via Lagrangian
in two terms, a confinement term that groups the relaxation.
constraints and an optimization term that conducts
the network output to the equilibrium points. The The proposed HLN model is solved using
minimization for energy function of this modi- sub-gradient technique with updating step sizes
fied HNN is conducted in two stages for the two which will be easily tuned for each problem while
corresponding terms of the energy function. This the slope of sigmoid function for continuous
modified HNN can deal with more complicated neurons can be fixed. With the new improve-
problem compared to the conventional HNN and ments, the proposed HLN method could solve any
its ability to find global optimal solution has been optimization problems that the Hopfield network
also improved. However, for implementation this can. Moreover, the solution quality obtained by
modified neural network, the problem constraints the proposed neural network is also higher than
also need to be linearized. that from the conventional HNN. Therefore, the
In this research, a new improvement of con- newly proposed neural network could be one of
tinuous Hopfield neural network, called Hopfield
61
Hopfield Lagrange Network for Economic Load Dispatch
the new options for solving optimization problems where λi is Lagrangian multiplier associated with
in power systems. constraint i.
To apply the Lagrangian function in HLN, the
continuous and multiplier neurons corresponding
HOPFIELD LAGRANGE NETWORK to independent variables and Lagrangian multipli-
AND APPLICATIONS ers are respectively needed. In HLN, the neurons
representing continuous variables are called
Hopfield Lagrange Network continuous neurons, and the neurons representing
for Optimization Problem Lagrange multiplier are called multiplier neurons.
The energy function for HLN is formulated
Optimization Problem Formulation based on the Lagrangian function as follows:
62
Hopfield Lagrange Network for Economic Load Dispatch
Figure 1. Sigmoid function of continuous neurons with different slopes where xk,max = 1 and xk,min = 0
63
Hopfield Lagrange Network for Economic Load Dispatch
tuned based on experiments. Based on experiments the values of the updating step sizes are usually
some observations are draw as follows. smaller than 1.
If the slope σ < 1, the neural network converges
very fast but the obtained solution may be local Initialization
optimum. In contrast, if the slope σ > 1, the solu- The neural network requires initialization for each
tion from the neural network is global optimum neuron. It is observed that a good initialization for
but slightly slower convergence, the larger value multiplier neurons can speed up the convergence
of σ used the better solution obtained. Therefore, process. However, this initialization does not af-
the preferable values for σ is ranging from 10 to fect on the final solution by the neural network.
100, since higher values of σ lead to longer to In this research, the initial outputs of continu-
converge but the improvement in the obtained ous neurons are initialized based on “medium
solution is inconsiderable. start”, e.g. the outputs of continuous neurons are
The updating step sizes αx and αλ for neurons initialized at middle point of independent variable
usually depend on the problem being considered. limits as follows:
It is observed that the larger the values of the
updating step sizes, the closer the discrete system 1
behavior of the neural network, producing values
Vk(,0x) =
2
(x k ,max + x k ,min ) (12)
at the upper and lower limits of each neuron. On
the contrary, the smaller the values of updating where Vk,x(0) is the initialization of Vk,x of continu-
step sizes, the slower convergence of the neural ous neuron k.
network. To determine these parameters, a small The multiplier neurons can be initialized at
value will be chosen first and then gradually in- zero. However, to speed up the convergence pro-
crease until the network behaves like a discrete cess of the neural network, the multiplier neurons
system. The proper values will be obtained. Note is initiated by the solution of the equation in (6),
64
Hopfield Lagrange Network for Economic Load Dispatch
65
Hopfield Lagrange Network for Economic Load Dispatch
dE ∂E dVi ,λ
Substituting (6) into (18): = (20)
dt ∂Vi ,λ dt
2
dE dg (U ) dU
= − c k ,x k ,x (19) Substituting Vi,λ in (11) into (20):
dt dU i dt
dE ∂E dgm (U i ,λ ) dU i ,λ
Since gc(Uk,x) is a monotonically increasing = (21)
dt ∂Vi ,λ dU i ,λ dt
function as shown in Figure 1, the value of the
derivative dgc(Uk,x)/dUk,x is always positive. Con-
sequently, the right hand side of Equation (19) is Substituting (7) into (21):
always negative. Therefore, the energy function
(5) is always minimized when there is a change dE dU i ,λ
2
66
Hopfield Lagrange Network for Economic Load Dispatch
It is obvious that the right hand side of equa- Pi 0 − Pi ≤ DRi , if generation decreases
tion (22) is always positive. Therefore, the energy (28)
function always seeks for maximum value when
there is a status change on the multiplier neurons where ai, bi, ci fuel cost coefficients for unit i;
associated with equality constraints.
Bij, B0i, B00 transmission loss formula coeffi-
HLN for Solving Basic ELD Problem cients;
The objective of the basic ELD (BELD) problem N total number of online units;
here is to minimize total cost of thermal generating
units of a system over some appropriate period PD total load demand of the system (MW);
(one hour typically) while satisfying various
constraints including power balance, generator Pi output power of unit i (MW);
power limits, and ramp rate constraints.
Mathematically, the BELD problem is formu- Pi0 initial output power of unit i (MW);
lated as follows:
Pi,min, Pi,max lower and upper generation limits of
N unit i (MW);
Min F = ∑ (ai + bi Pi + ci Pi 2 ) (23)
i =1
PL total network loss of the system (MW);
∑P −P
i =1
i L − PD = 0 (24)
The Lagrangian function L of the problem is
formulated as follows:
N N N
PL = ∑ ∑ Pi Bij Pj + ∑ B0i Pi + B00 (25) N N
i =1 j =1 i =1 L = ∑ (ai + bi Pi + ci Pi 2 ) + λ PD + PL − ∑ Pi
i =1
i =1
(29)
Generator operating limits
To represent in HLN, N continuous neurons and
Pi ,min ≤ Pi ≤ Pi ,max ; i = 1, 2, …, N (26)
one multiplier neurons are required. The energy
function E of the problem is formulated based on
Ramp rate constraints the Lagrangian function as follows:
67
Hopfield Lagrange Network for Economic Load Dispatch
N
The outputs of continuous neurons represent-
E = ∑ (ai + bV i pi )
i pi + cV
2
i =1
ing for output power of units are calculated via
V the sigmoid function:
N N pi
+Vλ PD + PL − ∑Vpi + ∑ ∫ gc−1 (V )dV
i =1 P
i =1 − Pi ,low
Vpi = gc (U pi ) = i ,high 1 + tanh (σU pi ) + Pi ,low
0
(30) 2
(36)
where Vλ output of multiplier neuron representing
Lagrangian multiplier λ; where the new generator limits are redefined as
follows:
Vpi output of continuous neuron i representing
for output power Pi. Pi ,high = min {Pi ,max , Pi 0 + URi } (37)
dU pi ∂E ∂P
L − 1 + U
=− = − (bi + 2cV
i pi ) + Vλ Pi,high maximal possible power output of unit i;
dt ∂Vpi ∂Vpi pi
(31) Pi,low minimal possible power output of unit i.
dU λ ∂E N
=+ = PD + PL − ∑Vpi (32) The outputs of multiplier neurons are defined
dt ∂Vλ i =1 by a transfer function as follows:
where Vλ = gm (U λ ) = U λ (39)
∂PL N
The outputs of neurons are initialized by:
= 2∑ BijVpj + B0i (33)
∂Vpi j =1
Pi ,max + Pi ,min
Vpi(0) = (40)
Uλ input of multiplier neuron corresponding to 2
the output Vλ.;
(0)
1 N bi + 2cV
Upi input of continuous neurons corresponding Vλ(0) =
N
∑ 1 − ∂P i pi
∂Vpi
(41)
to the outputs Vpi. i =1 L
The algorithm for updating inputs of neurons The maximum error for the network at iteration
is defined: n is calculates as follows:
∂E (n )
Errmax = max {∆P (n ) , ∆Vpi(n ) , ∆Vλ(n ) }
U pi(n ) = U pi(n −1) − αi (34)
∂Vpi N
= max PD + PL − ∑Vpi(n ) , Vp(in ) −Vpi(n −1) , Vλ(n ) −Vλ(n −1)
i =1
(42)
∂E
U (n )
λ =U (n −1)
λ + αλ (35)
∂Vλ
68
Hopfield Lagrange Network for Economic Load Dispatch
69
Hopfield Lagrange Network for Economic Load Dispatch
Figure 4. Maximum error of HLN for three-unit system neglecting power loss
Figure 5. Energy function of HLN for three-unit system neglecting power loss
70
Hopfield Lagrange Network for Economic Load Dispatch
Figure 6. Energy production cost for three-unit system neglecting power loss
71
Hopfield Lagrange Network for Economic Load Dispatch
When power loss is included, the selected prices. Fuel contracts are generally under a take-
updating step sizes are αi = 0.02 and αλ = 0.001, or-pay agreement including both maximum and
and the outputs of neurons are initialized at Vpi(0) minimum limits on delivery of fuel to generating
= [375 250 125]T, Vλ(0) = 9.3312. Their correspond- units over life of the contract. The fuel storage is
ing inputs of neurons determined by inverse usually within a specified limit to allow for inac-
functions of sigmoid and transfer functions are curate load forecasts and the inability to deliver
Upi(0) = [0 0 0]T and Uλ(0) = 9.3312, respectively. on time of suppliers (Asgarpoor, 1994).
The total cost produced by the HLN method is Assuming that the entire schedule time hori-
8,344.21 ($/h) and the solution is found as follows: zon is divided into M subintervals each having
a constant load demand and that all generating
λ = Vλ = 9.5302 ($/MWh) units are available and remain on-line for M
subintervals. The objective is to simultaneously
P1 = Vp1 = 435.409 (MW) minimize generation cost and emission level of
generating units over the M subintervals such that
P2 = Vp2 = 299.666 (MW) the constraints for power balance, fuel delivery
and fuel storage for any given subinterval as
P3 = Vp3 = 130.746 (MW) well as maximum-minimum fuel delivery, fuel
storage, and generator operating constraints for
PL = 15.821 (MW) each generating unit are satisfied.
The problem formulation for a system hav-
In this case, the HLN method finds the optimal ing N thermal generating units scheduled in M
solution with the same number of iterations as in subintervals is as follows (Basu, 2002).
the case neglecting power loss in 0.010 seconds.
In the obtained solutions, all constraints are Min {Ffc + Fem} (43)
met; that is, generator outputs are within there
lower and upper limits and total power generation M N
from the generators totally meets the load demand Ffc = ∑ ∑ tk (a fi + bfi Pik + c fi Pik2 ) (44)
k =1 i =1
plus power loss (if included) requirement. Obvi-
ously, in the case with power loss neglected, the M N
total operation cost and energy production cost Fem = ∑ ∑ tk (aei + bei Pik + cei Pik2 ) (45)
are lower than those for the case with power loss. k =1 i =1
Problem Formulation N
∑P
i =1
ik − PLk − PDk = 0 ; k = 1,…, M (46)
Fuel constrained ELD (FELD) problem or fuel
scheduling is an important part of utility for opera- N N N
tion and planning since it is a complex problem PLk = ∑ ∑ Pik Bij Pjk + ∑ B0i Pik + B00
of very large dimensions with a wide range of i =1 j =1 i =1
72
Hopfield Lagrange Network for Economic Load Dispatch
N
FDk fuel demand of the units during subinterval
∑F
i =1
ik − FDk = 0 ; k = 1, …, M (48)
k, in tons;
The fuel storage at subinterval k in (49) can be Pi,min, Pi,max lower and upper generation limits of
rewritten in terms of initial fuel storage as follows: thermal unit i, in MW;
73
Hopfield Lagrange Network for Economic Load Dispatch
M N
L = ∑ ∑ tk (a fi + bfi Pik + c fi Pik2 ) + (aei + bei Pik + cei Pik2 )
dU pik ∂E
=−
k =1 i =1
dt ∂Vpik
M N M N
+∑ λk PLk + PDk − ∑ Pik + ∑ γk ∑ Fik − FDk
k =1
i =1
k =1
i =1
tk (bfi + 2c fiVpik ) + tk (bei + 2ceiVpik )
M N k ∂P
+∑ ∑ ηik X ik − X i 0 − ∑ (Fil + tlQil ) = − dQ
+Vλk L k
− 1 +Vηik × tk ik
+ U pik
k =1 i =1 l =1
∂Vpik dVpik
(55)
(57)
To represent in HLN, 3N×M continuous
neurons and (N+2)×M multiplier neurons are dU fik ∂E
=− = − {Vγk −Vηik + U fik }
required. The energy function E of the problem dt ∂Vfik
is formulated based on the Lagrangian function (58)
in terms of neurons as follows.
dU xik ∂E
M N =− = − {Vηik + U xik } (59)
E = ∑ ∑ tk (a fi + bfiVpik + c fiVpik
2
) + (aei + beiVpik + ceiVpik2 ) dt ∂Vxik
k =1 i =1
M N M N
+∑Vλk PLk + PDk − ∑Vpik + ∑Vγk ∑Vfik − FDk
k =1 i =1
k =1 i =1 dU λk ∂E N
M N k =+ = PDk + PLk − ∑Vpik (60)
+∑ ∑Vηik Vxik − Xi 0 − ∑ (Vfil + tlQil )
dt ∂Vλk i =1
k =1 i =1 l =1
N pik
V Vfik Vxik
M
+∑ ∑ ∫ gc−1 (V )dV + ∫ gc−1(V )dV + ∫ gc−1(V )dV dU γk
∂E N
k =1 i =1
0 0 0
=+ = ∑Vfik − FDk (61)
(56) dt ∂Vγk i =1
The dynamics of HLN for updating neuron Uλk, Uγk, Uηik inputs of multiplier neurons
inputs are defined as follows: corresponding to the outputs Vλk, Vγk and Vηik,
respectively.
74
Hopfield Lagrange Network for Economic Load Dispatch
Vγk = gm (U γk ) = U γk (75)
∂E
U (n )
xik =U (n −1)
xik − αx (67)
∂Vxik
Vηik = gm (U ηik ) = U ηik (76)
∂E
U λ(nk ) = U λ(nk −1) + αλ (68)
∂Vλk The maximum error for the neural network at
iteration n is determined as follows:
∂E
U γ(nk ) = U γ(nk −1) + αγ (69) Errmax = max {∆Pk , ∆Fik , ∆X ik , ∆Vpik , ∆Vfik , ∆Vxik }
∂Vγk
(77)
∂E where
U η(nik) = U η(nik−1) + αη (70)
∂Vηik
N
∆Pk = PDk + PLk − ∑Vpik (78)
where αp, αf, αx continuous neuron updating step i =1
sizes;
N
P − Pi ,min
Vpik = gc (U pik ) = i ,max 1 + tanh (σU pik ) + Pi,min (n )
∆Vpik = Vpik (n −1)
−Vpik (81)
2
(71)
75
Hopfield Lagrange Network for Economic Load Dispatch
Unit 1 2 3 4 5
afi ($/h) 25 60 100 120 40
bfi ($/MWh) 2.0 1.8 2.1 2.2 1.8
cfi ($/MW h) 2
0.008 0.003 0.001 0.004 0.002
aei (kg/h) 80 50 70 45 30
bfi (kg/MWh) -0.805 -0.555 -0.955 -0.600 -0.555
cei (kg/MW h) 2
0.018 0.015 0.012 0.008 0.012
di(ton/h) 0.83612 2.00669 3.34448 4.01338 1.33779
ei (ton/MWh) 0.066890 0.060200 0.070230 0.073580 0.060200
fi (ton/MW2h) 0.00026756 0.00010033 0.00004013 0.00013378 0.00005017
Pi,max (MW) 75 125 175 250 300
Pi,min (MW) 20 20 30 40 50
Fi,max (tons) 1000 1000 2000 3000 3000
Fi,min (tons) 0 0 0 0 0
Xi,max (tons) 10000 10000 20000 30000 30000
Xi,min (tons) 0 0 0 0 0
Subinterval 1 2 3
Duration tk (h) 168 168 168
Load demand PDk (MW) 700 800 650
Fuel demand FDk (tons) 7000 7000 7000
76
Hopfield Lagrange Network for Economic Load Dispatch
delivery cost, and ηik represents for fuel storage method gives a total cost of $ 1,044,738.45 and
cost for each unit for each sub-interval. The an emission of 534,478.54 kg with a computa-
computational time for this case is 0.13 seconds. tional time of 0.27 seconds. The obtained solution
Case 2: The initial storage for 5 units is Xi0 = for this case is given in Table 6. In this case, the
[2000; 5000; 5000; 500; 8000]T tons. This is also fuel delivered to unit 5 is also increased while
called “shortage case” since the initial fuel storage fuel delivered to the others is reduced.
for unit 4 can only guaranty it to operate in a very The results from the three cases show that the
short term. The solution for this case is more dif- final solutions are little affected by the initial
ficult to be found since the constraint condition conditions of fuel storage. The proposed HLN
is more restricted than the full case. In this case, could find the corresponding optimal solution for
the total cost and emission obtained by the pro- each case. In all cases, the obtained total costs
posed method are $1,044,701.09 and 534,515.92 and emissions are not much difference from each
kg, respectively. The solution for this case is other.
given in Table 5. Due to the initial fuel shortage, In this multi-objective optimization problem,
the fuel delivered to unit 4 in this case is more the single objective optimization is also considered
than that in Case 1 while fuel delivered to other as an option for decision maker. In this case, the
units is less. For this case, the computational time single objective problems are also considered for
is 0.17 seconds. purity of fuel cost dispatch and purity of emission
Case 3: The initial storage for 5 units is Xi0 = dispatch with the initial storage from Case 1. In
[2000; 2500; 2500; 8000; 500]T tons. This case these single objective cases, the optimal solution
is also a shortage case with the initial fuel short- is easier to be found than the cases with multi-
age for unit 5 and lower initial fuel storage than objectives since there is no conflict between the
that in Case 1 for units 2 and 3. This case is more objectives. The parameters of HLN for these cases
restricted than Case 2 since the initial storage is are chosen as follows: αp = 10-5, αf = αx = αλ = 0.1,
lower. Therefore, the optimal solution is also more and αγ = αη = 10-7. The remaining parameters are
difficult to be found. For this case, the proposed
Sub λk γk
interval ($/MWh) ($/ton-h) Unit 1 2 3 4 5
1 942.6 0.0012 Pik (MW) 75.00 21.26 175.00 167.07 161.67
Fik (103 tons) 0.7816 0.5425 1.5275 2.1183 2.0300
Xik (10 tons)
3
1.7950 3.9770 3.8996 7.3750 8.1688
ηik ($/ton-h) 0.0076 0.0021 0.0071 0.0056 0.0049
2 1,148.1 0.0023 Pik (MW) 75.00 125.00 175.00 218.03 206.97
Fik (103 tons) 0.7697 0.5971 1.5804 2.1278 1.9250
Xik (10 tons)
3
1.5781 2.9707 2.8520 6.1286 7.7741
ηik ($/ton-h) 0.0084 0.0043 0.0090 0.0068 0.0053
3 884.1 0.0031 Pik (MW) 75.00 111.58 162.07 152.57 148.78
Fik (103 tons) 0.7739 0.6619 1.6544 2.0978 1.8120
Xik (10 tons)
3
1.3653 2.1652 2.0312 5.6629 7.8554
ηik ($/ton-h) 0.0092 0.0064 0.0109 0.0073 0.0052
77
Hopfield Lagrange Network for Economic Load Dispatch
Sub λk γk Unit 1 2 3 4 5
interval ($/MWh) ($/ton-h)
1 942.7 0.0039 Pik (MW) 75.00 21.27 175.00 167.34 161.69
Fik (103 tons) 0.6894 0.4206 1.3338 2.8664 1.6898
Xik (10 tons)
3
1.7027 3.8549 3.7059 0.6236 7.8284
ηik ($/ton-h) 0.0079 0.0023 0.0074 0.0193 0.0052
2 1148.3 0.0057 Pik (MW) 75.00 125.00 175.00 217.98 207.02
Fik (10 tons)
3
0.6673 0.4636 1.3914 2.9414 1.5363
Xik (103 tons) 1.3834 2.7151 2.4694 0.1914 7.0445
ηik ($/ton-h) 0.0091 0.0049 0.0098 0.0252 0.0059
3 884.2 0.0067 Pik (MW) 75.00 111.59 162.08 152.53 148.80
Fik (103 tons) 0.6828 0.5441 1.5212 2.8310 1.4209
Xik (10 tons)
3
1.0795 1.7916 1.5152 0.4593 6.7345
ηik ($/ton-h) 0.0106 0.0076 0.0125 0.0208 0.0062
Sub λk γk Unit 1 2 3 4 5
interval ($/MWh) ($/ton-h)
1 942.6 0.0039 Pik (MW) 75.00 121.25 175.00 167.09 161.66
Fik (10 tons)
3
0.6141 0.6259 1.5930 1.5408 2.6262
Xik (103 tons) 1.6275 1.5604 1.4650 6.7974 1.2652
ηik ($/ton-h) 0.0082 0.0084 0.0127 0.0061 0.0156
2 1148.1 0.0057 Pik (MW) 75.00 125.00 175.00 218.04 206.96
Fik (103 tons) 0.5890 0.7311 1.7417 1.5148 2.4234
Xik (10 tons)
3
1.2298 0.6881 0.5788 4.9378 1.3690
ηik ($/ton-h) 0.0098 0.0130 0.0176 0.0081 0.0152
3 884.2 0.0067 Pik (MW) 75.00 111.57 162.04 152.60 148.79
Fik (10 tons)
3
0.5990 0.9113 1.9567 1.4531 2.0800
Xik (103 tons) 0.8421 0.1320 0.0606 3.8269 1.7181
ηik ($/ton-h) 0.0119 0.0216 0.0290 0.0096 0.0140
the same as the ones selected for the three cases the total cost is much lower than that in the case
above. of bi-objective case meanwhile the emission in
Fuel Cost Objective Only: In this case, there this case is much higher and all constraints are
only fuel cost objective is minimized while the satisfied. The obtained solution for the economic
emission objective is neglected. The obtained dispatch is given in Table 7.
total cost by HLN is $1,002,468.61 and the emis- Emission objective only: This case is opposite
sion 744,675.06 kg. It is obvious that when the to the case considering only fuel cost objective.
fuel cost objective is priority to be considered, The obtained emission level in this case is
78
Hopfield Lagrange Network for Economic Load Dispatch
Sub λk γk Unit 1 2 3 4 5
interval ($/MWh) ($/ton-h)
1 451.7 0.0011 Pik (MW) 43.00 125.00 175.00 61.03 295.97
Fik (103 tons) 0.7470 0.5495 1.5324 1.9964 2.1747
Xik (10 tons)
3
2.1214 3.9461 3.9045 8.5664 6.9541
ηik ($/ton-h) 0.0066 0.0021 0.0071 0.0046 0.0060
2 537.7 0.0022 Pik (MW) 74.98 125.00 175.00 125.02 300.00
Fik (10 tons)
3
0.7368 0.6065 1.5881 1.8905 2.1780
Xik (103 tons) 1.8718 2.9491 2.8647 8.2345 5.8707
ηik ($/ton-h) 0.0073 0.0044 0.0089 0.0049 0.0071
3 435.6 0.0029 Pik (MW) 37.00 125.00 175.00 49.04 263.96
Fik (103 tons) 0.6925 0.6888 1.6836 1.7422 2.1929
Xik (10 tons)
3
2.0064 2.0345 1.9204 8.6951 5.1670
ηik ($/ton-h) 0.0069 0.0068 0.0112 0.0045 0.0078
521,793.49 kg and the total cost $1,072,673.58. in Table 8. It can be observed from the obtained
The obtained total cost in this case is much result that the power outputs of generating units
higher than the case with only fuel cost objective in this case are different to those from the eco-
while the emission level is opposite to the previ- nomic dispatch since they are decided by the
ous case. In fact, the performance of emission different objective coefficients.
dispatch is the same manner with the economic The obtained test results from the proposed
dispatch except the different objective coefficients. HLN above are better than those from the con-
The solution for this emission dispatch is given ventional Hopfield network in (Basu, 2002) for
Sub λk γk Unit 1 2 3 4 5
interval ($/MWh) ($/ton-h)
1 468.3 0.0012 Pik (MW) 75.00 111.41 162.68 211.66 139.25
Fik (10 tons)
3
0.7809 0.5321 1.5110 2.1715 2.0046
Xik (103 tons) 0.7679 0.5853 1.5614 2.2237 1.8617
ηik ($/ton-h) 0.0076 0.0019 0.0069 0.0061 0.0047
2 612.4 0.0024 Pik (MW) 75.00 125.00 175.00 250.00 175.00
Fik (103 tons) 0.7710 0.6377 1.6159 2.2506 1.7249
Xik (10 tons)
3
1.7943 4.0664 4.0284 6.8761 8.3703
ηik ($/ton-h) 0.0084 0.0041 0.0087 0.0077 0.0049
3 421.9 0.0032 Pik (MW) 75.00 102.19 150.67 194.39 127.75
Fik (10 tons)
3
1.5755 3.0482 2.9619 5.3297 8.2359
Xik (103 tons) 1.3599 2.3136 2.2371 4.4988 8.4429
ηik ($/ton-h) 0.0092 0.0060 0.0104 0.0087 0.0047
79
Hopfield Lagrange Network for Economic Load Dispatch
all cases considered. This shows that the proposed Φ(Phk ) = a ph + bph Phk + cph Phk2 (89)
method is better than the conventional Hopfield
neural network in dealing with complicated prob- Ψ(dhk ) = adh + bdhdhk + cdhdhk2 (90)
lems.
M N1
Min F = ∑ ∑ tk (ai + bi Pik + ci Pik2 ) (84) where ai, bi, ci cost coefficients for thermal unit i;
k =1 i =1
Power balance constraints adh, bdh, cdh reservoir head variation coefficients
for hydro unit h;
N1 N2
∑P
i =1
ik + ∑ Phk − PLk − PDk = 0 ; k = 1,…, M
h =1
Bpq, B0p, B00 coefficients for the system;
(85) dh0, dhM the initial and final height of the reser-
N 1 +N 2 N 1 +N 2 N 1 +N 2
voir head of hydro unit h, in ft;
PLk = ∑ ∑
p =1 q =1
Ppk Bpq Pqk + ∑p =1
B0 p Ppk + B00
dhk the height of the reservoir head of hydro unit
(86) h in interval k, in ft;
Continuity of reservoir head constraints fh the surface of the vertical sided tank of hydro
unit h;
tk
dhk = dhk −1 + (rhk − qhk ) ; h = 1, …, N2 Pik generation output of thermal unit s during
fh
(87) sub-interval k, in MW;
qhk = Φ(Phk ) × Ψ(dhk ) (88) Phk generation output of hydro unit h during
subinterval k, in MW;
80
Hopfield Lagrange Network for Economic Load Dispatch
HLN Implemented to the Vλk, Vγh outputs of the multiplier neurons associ-
HELD Problem ated with power balance and water constraint,
respectively.
The Lagrange function L is formulated as follows:
The dynamics of Hopfield Lagrange model
for updating neuron inputs based on the previous
information are follows:
81
Hopfield Lagrange Network for Economic Load Dispatch
dU pik ∂E ∂P
Lk − 1 + U
∂E
=− = − tk (bi + 2cV
i pik ) + Vλk
pik
(n )
U pik (n −1)
= U pik − αi (103)
dt ∂Vpik
∂Vpik
∂Vpik
(96)
∂E
dU phk ∂E ∂P
Lk − 1 +V t ∂qhk + U
(n )
U phk (n −1)
= U phk − αh (104)
=− = −
Vλk phk ∂Vphk
dt ∂Vphk ∂V γh k
∂Vphk
phk
(97)
∂E
N1 N2 U λ(nk ) = U λ(nk −1) − αλ (105)
dU λk ∂E ∂Vλk
=+ = PDk + PLk − ∑Vpik − ∑Vphk
dt ∂Vλk i =1 h =1
(98)
∂E
U γ(nh ) = U γ(nh−1) − αγ (106)
∂Vγh
dU γh ∂E M
=+ = ∑ tk (qhk − rhk ) −Wh (99)
dt ∂Vγh k =1
where αi, αh updating step sizes for continuous
neurons;
where
N1 N2
αλ, αγ updating step sizes for multiplier neurons.
∂PLk
= 2∑ BipVppk + 2∑ BihVphk + B0i
∂Vpik p =1 h =1 The outputs of continuous neurons are calcu-
(100) lated by a sigmoid function:
Bip loss coefficients related to thermal plants; Since multiplier neurons are unconstrained
outputs, the outputs are defined as below:
Bhq loss coefficients related to hydro plants;
Vλk = gm(Uλk) = Uλk (109)
Bih, Bhi loss coefficients between thermal and
hydro plants, Bih = BhiT; Vγh = gm(Uγh) = Uγh (110)
Upik, Uphk inputs of the neurons ik and hk, re- The maximum error for the neural network is:
spectively;
Errmax = max {∆Pk , ∆Wh , ∆Vpik , ∆Vphk }
Uλk, Uγh inputs of the multiplier neurons. (111)
82
Hopfield Lagrange Network for Economic Load Dispatch
N1 N2
Water inflow of reservoirs during schedule
∆Pk = PDk + PLk − ∑Vpik − ∑Vphk (112)
i =1 h =1
period is supposed to be zeros. Transmission loss
coefficient matrix is given below, per MW:
M
∆Wh = ∑ t (q − rhk ) −Wh (113) 4.0 1.0 1.5 1.5
k hk
k =1 1.0 3.5 1.0 1.2
B = × 10−5
1.5 1.0 3.9
9 2.0
(n )
∆Vpik = Vpik (n −1)
−Vpik (114) 1.5 1.2 2.0 4.9
(n )
∆Vphk = Vphk (n −1)
−Vphk (115) The schedule time horizon is 48h which is
divided in four sub-periods with 12 hours for
each supplying to the load demand of [1200 1500
1400 1700] MW. The allowable volumes of water
Numerical Results
for hydro plants 3 and 4 for the whole period are
given by:
Fixed-Head HELD Problem
In this case, the reservoir head variation of hydro
W3 = 125,000 acre-ft
units is neglected; therefore water discharge of
each hydro unit in (88) is a function of only its
W4 = 286,000 acre-ft
generation output.
The test system has two thermal and two hydro
The parameters of the HLN for the problem
plants with their characteristics given as follows:
are selected after tuning as follows: σ = 100, αi
= αh = 3×10-4, αλ = 10-2 and αγ = 7.5×10-7. The
F1(P1) = 380 + 6.75P1 + 0.00225P12 $/h
maximum number of iterations and the maximum
error for the neural network are set to 2,500 and
47.5 MW ≤ P1 ≤ 450 MW
10-4, respectively.
When the system power loss is neglected,
F2(P2) = 600 + 5.28P2 + 0.0055P22 $/h
the total power generation from the thermal and
100 MW ≤ P2 ≤ 1000 MW hydro units is balanced to only load demand.
The proposed method provides a total cost of
q3(P3) = 260 + 8.5P3 + 0.00986P32 acre-ft/h $ 353,444.60 with a computational time of 0.8
seconds. The water discharge cost for each hydro
0 MW ≤ P3 ≤ 250 MW plant for the whole schedule time is [0.73 0.51]
T
$/acre-ft. Obviously, this water discharge cost
q4(P4) = 250 + 9.8P4 + 0.0114P42 acre-ft/h reflects to the practice that the marginal cost for
hydro is small and negligible in the operating
0 MW ≤ P4 ≤ 500 MW cost calculation. Therefore, the energy produc-
tion cost for the system is mainly based on the
energy production cost from thermal units. The
where F1 and F2 are fuel cost functions of thermal solution obtained by the HLN method for this case
power plants 1 and 2, respectively and q3 and q4 is given Table 9. As observed from this result, all
are water discharge functions for hydro power constraints including power balance and water
plants 3 and 4, respectively. discharge are satisfied.
83
Hopfield Lagrange Network for Economic Load Dispatch
Subinterval 1 2 3 4
Duration tk (h) 12 12 12 12
Load demand PDk (MW) 1200 1500 1400 1700
Production cost λk ($/MWh) 103.53 116.55 111.66 129.67
Thermal unit 1 P1k (MW) 417.01 450.00 450.00 450.00
Thermal unit 2 P2k (MW) 304.37 403.01 366.06 502.38
Hydro unit 3 P3k (MW) 167.84 243.09 214.92 250.00
Hydro unit 4 P4k (MW) 310.78 403.91 369.02 497.62
Water discharge q3k (acre-ft) 23572.79 34907.03 30507.11 36015.00
Water discharge q4k (acre-ft) 52760.45 72817.82 65025.60 95395.30
The obtained result by HLN method for this in the case without power loss. This is because
case is better than that from the conventional the allowable volumes of water for hydro units
Hopfield neural network in (Basu, 2003). This are fixed, thus the power outputs from thermal
confirms that the proposed HLN is more efficient units are increased to compensate to power loss
than the conventional Hopfield neural network in in the system leading to more fuel consumption.
approach to complicated problems. The final solution for this case is given in Table
When the system power loss is included, the 10. This result is also satisfies all constraints of
total power generation from both thermal and power balance and water discharge.
hydro units balances to load demand plus power
loss in the system. The HLN method obtains a Variable-Head HELD Problem
total cost of $ 375,933.65 with the water discharge This case considers all constraints as in the
cost for each hydro plant for the whole period is problem formulation. The test system consists
[0.76 0.52]T $/acre-ft for a computational time of of two thermal and two hydro plants. Their data
0.11 seconds. The total cost and energy production is given below:
cost in this case are obviously higher than those
Table 10. Solution for fixed-head HELD problem with power loss
Subinterval 1 2 3 4
Duration tk (h) 12 12 12 12
Load demand PDk (MW) 1200 1500 1400 1700
Production cost λk ($/MWh) 111.09 129.20 122.93 147.42
Power loss PLk (MW) 31.42 48.92 42.57 63.69
Thermal unit 1 P1k (MW) 442.16 450.00 450.00 450.00
Thermal unit 2 P2k (MW) 326.19 445.10 404.21 563.69
Hydro unit 3 P3k (MW) 160.56 247.12 217.39 250.00
Hydro unit 4 P4k (MW) 302.52 406.69 370.96 500.00
Water discharge q3k (acre-ft) 22547.18 35552.36 30885.46 36015.00
Water discharge q4k (acre-ft) 1095.33 73453.53 65451.15 96000.00
84
Hopfield Lagrange Network for Economic Load Dispatch
F1(P1) = 25 + 3.2P1 + 0.0025P12 $/h The transmission loss coefficient matrix is:
85
Hopfield Lagrange Network for Economic Load Dispatch
When power loss is neglected, the total power generation from hydro plant 3 during the schedule
generation from the thermal and hydro plants is time is similar to hydro plant 4. However, hydro
balanced to only load demand at each interval for plant 3 is not off at the hours of low load demand
the whole schedule time horizon. The obtained since the allowable water discharge volume for
total cost by HLN is $ 62,839.58 and the water this plant is higher than that for plant 4 which can
discharge cost of hydro plants [10.23 3.93]T $/ guarantee it to operate for the whole scheduled
MCF with a computational time of 0.13 seconds. time. The water head of the hydro plants changes
The solution by the HLN for this case is given according to the water discharge for each time-
in Table 11. As observed from the table, based interval of one hour. With the contribution of the
on the obtained schedule, hydro plant 4 is off at hydro power generation in the system, the energy
hours 2-4 for low load demand and generates high production cost is lower than the case with all
power at the high load demand hours 9-20 so as thermal unit system. The power generation outputs
its total power generation is fitted to the pre-fixed of plants for this case for the whole schedule time
allowable discharge water volume. The power are given in Figure 9.
Table 11. Solution for variable-head HELD problem neglecting power loss
86
Hopfield Lagrange Network for Economic Load Dispatch
Figure 9. Power generation of plants for variable-head HELD problem neglecting power loss
When power loss included in the problem, the Future Research Directions
total power generation from the thermal and hy-
dro plants has to satisfy load demand plus power For further research directions based on the HLN
loss in the system. The total cost is $ 67,952.42 method, more developments of HLN and its
with the water discharge cost of [10.40 3.99]T $/ broader implementations to optimization problems
MCF for each hydro plant. The total computa- in power systems will be considered as follows:
tional time of HLN for this case is 0.15 seconds.
The final solution is given in Table 12. The sched- • Adaptive updating mechanism will be con-
ule for this system in this case is also similar to sidered to replace the current updating step
the case neglecting power loss except higher sizes since they need to be tuned in the
thermal power generation to compensate the HLN method for each problem.
power loss leading to higher energy production. • Other ELD problems for thermal units such
The power generation outputs of power plants for as combined heat and power economic dis-
this case and the comparison of energy production patch, economic dispatch with piecewise
costs for both cases during the whole schedule fuel const function, economic dispatch
time are given in Figures 10 and 11, respectively. with prohibited operating zones etc will be
For the both cases, the variations of power considered and solved by implementation
generation outputs and energy production costs of the proposed HLN method.
are corresponding to the variation of the load
demand in the whole schedule time horizon.
87
Hopfield Lagrange Network for Economic Load Dispatch
Table 12. Solution for variable-head HELD problem with power loss
Hr. PD λ PL P1 P2 P3 P4 d1 d2
(MW) ($/MWh) (MW) (MW) (MW) (MW) (MW) (ft) (ft)
1 800 4.2028 22.31 151.45 364.54 276.36 29.97 299.91 249.92
2 700 4.0849 16.97 135.06 314.40 255.75 11.77 299.83 249.89
3 600 3.9627 12.40 117.71 261.49 233.19 0 299.75 249.89
4 600 3.9624 12.40 117.67 261.38 233.35 0 299.68 249.88
5 600 3.9622 12.40 117.63 261.26 233.51 0 299.60 249.88
6 650 4.0262 14.59 126.79 289.16 246.10 2.53 299.52 249.87
7 800 4.2010 22.31 151.17 363.72 277.39 30.04 299.43 249.79
8 1000 4.4428 35.37 184.06 464.41 319.54 67.36 299.32 249.62
9 1330 4.8671 64.19 239.46 633.66 390.54 130.52 299.19 249.28
10 1350 4.8922 66.24 242.61 643.27 394.87 135.49 299.05 248.92
11 1450 5.0270 77.03 259.47 694.60 416.83 156.13 298.90 248.51
12 1500 5.0946 82.77 267.76 719.81 427.84 167.37 298.75 248.07
13 1300 4.8190 61.18 233.17 614.55 383.66 129.80 298.61 247.73
14 1350 4.8841 66.24 241.42 639.73 394.51 140.58 298.47 247.37
15 1350 4.8821 66.24 241.14 638.89 394.43 141.78 298.33 247.00
16 1370 4.9072 68.33 244.27 648.45 398.73 146.89 298.19 246.62
17 1450 5.0145 77.04 257.67 689.27 416.24 163.86 298.05 246.19
18 1570 5.1813 91.20 278.02 750.99 442.93 189.26 297.89 245.69
19 1430 4.9820 74.81 253.53 676.74 411.49 163.06 297.74 245.26
20 1350 4.8712 66.26 239.54 634.14 393.80 148.77 297.60 244.88
21 1270 4.7630 58.27 225.67 591.84 376.32 134.44 297.47 244.54
22 1150 4.6058 47.34 205.20 529.34 350.44 112.36 297.36 244.25
23 1000 4.4164 35.39 180.01 452.28 318.55 84.56 297.25 244.04
24 900 4.2935 28.46 163.36 401.32 297.55 66.24 297.15 243.88
• The HLN method will be also applied for fuel constrained economic load dispatch and hy-
solving hydrothermal economic dispatch drothermal economic load dispatch. By directly
with cascaded hydro plants. using Lagrangian function as the energy function
• The implementation of the HLN for solv- of continuous Hopfield network in the HLN, it
ing large-scale optimization problems in is not necessary to pre-define an energy function
power systems will be also studied due to for the problem and map the problem into neural
its fast convergence to optimal solution. network like the conventional Hopfield network.
Moreover, the HLN method can simultaneously
process all variables and constraints, so it can
CONCLUSION quickly converge to optimal solution. The ob-
tained results from the test cases have shown that
In this chapter, the proposed HLN method has been the proposed HLN is reliable for finding optimal
efficiently implemented for solving different ELD solutions of the considered problems. The obvi-
problems including basic economic load dispatch, ous advantages of the proposed HLN method for
88
Hopfield Lagrange Network for Economic Load Dispatch
Figure 10. Power generation of plants for variable-head HELD problem with power loss
Figure 11. Energy production costs for both cases with and without power loss
89
Hopfield Lagrange Network for Economic Load Dispatch
solving the economic dispatch problems in power Attaviriyanupap, P., Kita, H., Tanaka, E., &
systems as well as optimization problems are its Hasegawa, J. (2002). A hybrid EP and SQP for
ability to efficiently deal with nonlinear objectives dynamic economic dispatch with nonsmooth
and constraints, to properly handle time-coupling fuel cost function. IEEE Transactions on Power
constraints by Lagrangian function and variable Systems, 17(2), 411–416. doi:10.1109/TP-
limits by sigmoid function of continuous neurons WRS.2002.1007911
in Hopfield neural network, to quickly find the
Bakirtzis, A., Petridis, V., & Kazarlis, S. (1994).
optimal solution for the problems, and to deal with
Genetic algorithm solution to the economic dis-
very large-scale problems with multiple schedule
patch problem. IEE Proceedings. Generation,
periods. In addition, unlike the population based
Transmission and Distribution, 141(4), 377–382.
methods such as evolutionary programming, dif-
doi:10.1049/ip-gtd:19941211
ferential evolution or particle swarm optimization,
the proposed method needs only one run to obtain Basu, M. (2002). Fuel constrained economic emis-
the optimal solution that does not depend on the sion load dispatch using Hopfield neural networks.
initially assumed solution for the algorithm. One Electric Power Systems Research, 63(1), 51–57.
more highlighted characteristic of HLN is that its doi:10.1016/S0378-7796(02)00090-1
energy function is simultaneously minimized with
Basu, M. (2003). Hopfield neural networks for op-
respect to continuous neurons and maximized
timal scheduling of fixed head hydrothermal power
with respect to multiplier neurons which satisfy
systems. Electric Power Systems Research, 64(1),
the Lagrangian function condition. Therefore,
11–15. doi:10.1016/S0378-7796(02)00118-9
this is a contribution for development of a new
computational tool for solving ELD problems Chowdhury, B. H., & Rahman, S. (1990). A review
in power systems in particular and optimization of recent advances in economic dispatch. IEEE
problems in general. On the other hand, a draw- Transactions on Power Systems, 5(4), 1248–1259.
back of the HLN is that the updating step sizes doi:10.1109/59.99376
for the continuous neurons have to be tuned by
Coelho, L. S., & Mariani, V. C. (2006). Combin-
experiments for different problems. However,
ing of chaotic differential evolution and quadratic
these parameters can be easily tuned by starting
programming for economic dispatch optimization
from small values and then gradually increasing
with valve-point effect. IEEE Transactions on
them until the continuous neurons produce the
Power Systems, 21(2), 989–996. doi:10.1109/
solution at their lower and upper limits, thus the
TPWRS.2006.873410
obtained parameter values can be used.
da Silva, I. N., Nepomuceno, L., & Bastos, T. M.
(2004). An efficient Hopfield network to solve
REFERENCES economic dispatch problems with transmission
system representation. International Journal
Asgarpoor, S. (1994). Comparison of linear, of Electrical Power & Energy Systems, 26(9),
nonlinear, and network flow programming tech- 733–738. doi:10.1016/j.ijepes.2004.05.007
niques in fuel scheduling. Electric Power Systems
Research, 30(3), 169–174. doi:10.1016/0378- Fisher, M. L. (1973). Optimal solution of sched-
7796(94)00851-5 uling problems using Lagrange multipliers:
Part I. Operations Research, 21, 1114–1127.
doi:10.1287/opre.21.5.1114
90
Hopfield Lagrange Network for Economic Load Dispatch
Happ, H. H. (1977). Optimal power dispatch – Nomana, N., & Iba, H. (2008). Differential evolu-
A comprehensive survey. IEEE Transactions tion for economic load dispatch problems. Electric
on Power Apparatus and Systems, PAS-96(3), Power Systems Research, 78(8), 1322–1331.
841–854. doi:10.1109/T-PAS.1977.32397 doi:10.1016/j.epsr.2007.11.007
Hopfield, J. J. (1982). Neural networks and Ongsakul, W., Dechanupaprittha, S., & Ngam-
physical systems with emergent collective com- roo, I. (2004). Parallel tabu search algorithm for
putational abilities. Proceedings of the National constrained economic dispatch. IEE Proceed-
Academy of Science, USA: Vol. 79 (pp. 2554-2558). ings. Generation, Transmission and Distribution,
151(2), 157–166. doi:10.1049/ip-gtd:20040460
Hopfield, J. J. (1984). Neurons with graded re-
sponse have collective, computational properties Park, J. H., Kim, Y. S., Eom, I. K., & Lee, K. Y.
like those of two-state neuron. Proceedings of (1993). Economic load dispatch for piecewise
the National Academy of Sciences of the United quadratic cost function using Hopfield neural
States of America, 81, 3088–3092. doi:10.1073/ network. IEEE Transactions on Power Systems,
pnas.81.10.3088 8(3), 1030–1038. doi:10.1109/59.260897
Irving, M. R., & Sterling, M. J. H. (1985). Eco- Ponnambalam, K., Quintana, V. H., & Vanelli,
nomic dispatch of active power by quadratic A. (1992). A fast algorithm for power system
programming using a sparse linear comple- optimization problems using an interior point
mentary algorithm. International Journal of method. IEEE Transactions on Power Systems,
Electrical Power & Energy Systems, 7(1), 2–6. 7(2), 892–899. doi:10.1109/59.141801
doi:10.1016/0142-0615(85)90002-X
Ruangpayoongsak, N., Ongsakul, W., & Rungger-
Jeyakumar, D. N., Jayabarathi, T., & Raghuna- atigul, S. (2002). Constrained economic dispatch
than, T. (2006). Particle swarm optimization for by combined genetic and simulated annealing
various types of economic dispatch problems. algorithm. Electric Power Components & Systems,
International Journal of Electrical Power & 30(9), 917–931. doi:10.1080/15325000290085235
Energy Systems, 28(1), 36–42. doi:10.1016/j.
Somasundaram, P., Lakshmiramanan, R., & Kup-
ijepes.2005.09.004
pusamy, K. (2005). Hybrid algorithm based on EP
Lee, K. Y., Nuroglu, F. M., & Sode-Yome, A. and LP for security constrained economic dispatch
(2000). Real power optimization with load flow problem. Electric Power Systems Research, 76(1-
using adaptive Hopfield neural network. Engineer- 3), 77–85. doi:10.1016/j.epsr.2005.04.005
ing Intelligent Systems, 8(1), 53–58.
Song, Y. H., Chou, C. S., & Stonham, T. J. (1999).
Lee, K. Y., Sode-Yome, A., & Park, J. H. (1998). Combined heat and power economic dispatch
Adaptive Hopfield neural networks for economic by improved ant colony search algorithm. Elec-
load dispatch. IEEE Transactions on Power Sys- tric Power Systems Research, 52(2), 115–121.
tems, 13(2), 519–526. doi:10.1109/59.667377 doi:10.1016/S0378-7796(99)00011-5
Liang, Z.-X., & Glover, J. D. (1992). A zoom Song, Y. H., Wang, G. S., Wang, P. Y., & Johns,
feature for a dynamic programming solution to A. T. (1997). Environmental/economic dispatch
economic dispatch including transmission losses. using fuzzy logic controlled genetic algorithms.
IEEE Transactions on Power Systems, 7(2), IEE Proceedings. Generation, Transmission and
544–550. doi:10.1109/59.141757 Distribution, 144(4), 377–382. doi:10.1049/ip-
gtd:19971100
91
Hopfield Lagrange Network for Economic Load Dispatch
Stahl, E. C. M. (1930). Load division in intercon- Wong, K. P., & Yuryevich, J. (1998). Evolution-
nections. Electrical World, 95, 434–438. ary-programming-based algorithm for environ-
mentally constrained economic dispatch. IEEE
Stahl, E. C. M. (1931). Economic loading of
Transactions on Power Systems, 13(2), 301–306.
generating stations. Electrical Engineering, 50,
doi:10.1109/59.667339
722–727.
Yalcinoz, T., & Short, M. J. (1997). Large-scale
Su, C.-T., & Chiou, G.-J. (1997). A fast-compu-
economic dispatch using an improved Hopfield
tation Hopfield method to economic dispatch of
neural network. IEE Proceedings. Generation,
power systems. IEEE Transactions on Power Sys-
Transmission and Distribution, 144(2), 181–185.
tems, 12(4), 1759–1764. doi:10.1109/59.627888
doi:10.1049/ip-gtd:19970866
Su, C.-T., & Chiou, G.-J. (1997). An enhanced
Hopfield model for economic dispatch consider-
ing prohibited zones. Electric Power Systems
ADDITIONAL READING
Research, 42(1), 72–76. doi:10.1016/S0378-
7796(96)01208-4 Book Chapters
Su, C.-T., & Lin, C.-T. (2000). New approach El-Hawary, M. E., & Christensen, G. S. (1979).
with a Hopfield modeling framework to economic Optimal economic operation of electric power sys-
dispatch. IEEE Transactions on Power Systems, tems. London: Academic Press, Inc. Abdelaziza,
15(2), 541–545. doi:10.1109/59.867138 Kothari, D. P., & Dhillon, J. S. (2006). Power
van den Berg, J., & Bioch, J. C. (1993). Con- system optimization. New Delhi: Prentice Hall
strained optimization with a continuous Hop- of India Private Limited.
field-Lagrange model. (Technical report EUR- Kothari, D. P., & Nagrath, I. J. (2003). Modern
CS-93-10), Erasmus University Rotterdam, Comp. power system analysis (3rd ed.). Boston: Mc
Sc. Dept., Faculty of Economics. Graw-Hill.
Victoire, T. A. A., & Jeyakumar, A. E. (2004). Hy- Saadat, H. (2002). Power system analysis (2nd
brid PSO–SQP for economic dispatch with valve- ed.). Boston: Mc Graw-Hill.
point effect. Electric Power Systems Research,
71(1), 51–59. doi:10.1016/j.epsr.2003.12.017 Wood, A. J., & Wollenberg, B. F. (1996). Power
generation operation and control (2nd ed.). John
Wells, D. W. (1968). Method for economic secure Wiley & Sons, Inc.
loading of a power system. Proceedings IEE,
115(8), 1190–1194. Wadhwa, C. L. (2006). Electrical power systems
(4th ed.). New Age International Publisher.
Wong, K. P., & Fung, C. C. (1993). Simulated
annealing based economic dispatch algorithm. Zhu, J. (2009). Optimization of power system
IEE Proceedings. Generation, Transmission operation. John Wiley & Sons, Inc., Publication.
and Distribution, 140(6), 509–515. doi:10.1049/ doi:10.1002/9780470466971
ip-c.1993.0074
92
Hopfield Lagrange Network for Economic Load Dispatch
93
Hopfield Lagrange Network for Economic Load Dispatch
Yalcinoz, T., Cory, B. J., & Short, M. J. (2001). Fuel Constrained Economic Load Dispatch:
Hopfield neural network approaches to economic The economic dispatch problem with more con-
dispatch problems. Int. J. Electrical and Energy straints of fuel delivery and fuel storage for each
Systems, 23(6), 435–442. doi:10.1016/S0142- generating unit.
0615(00)00084-3 Hopfield Lagrange Network: A continuous
Hopfield neural network with its energy func-
Yalcinoz, T., & Short, M. J. (1998). Neural net-
tion based on Lagrangian function. By using the
works approach for solving economic dispatch
Lagrangian relaxation for the energy function,
problem with transmission capacity constraints.
the Lagrange Hopfield network is simpler, easier,
IEEE Transactions on Power Systems, 13(2),
and more efficient than the conventional Hopfield
307–313. doi:10.1109/59.667341
neural network in solving optimization problems.
Yalcinoz, T., Short, M. J., & Cory, B. J. (1999). Hydrothermal Economic Load Dispatch:
Security dispatch using the Hopfield neural net- An economic dispatch problem for both thermal
work. IEE Proceedings. Generation, Transmission and hydro generating units. In this problem, the
and Distribution, 146(5), 465–470. doi:10.1049/ hydro constraints are also taken into consideration
ip-gtd:19990506 in addition to the thermal and system constraints.
This problem is more complicated than the eco-
nomic dispatch problem due to integration of
time-coupling constraints.
KEY TERMS AND DEFINITIONS
Lagrangian Function: A function used to
convert a constrained optimization problem to an
Economic Load Dispatch: One of the optimi-
unconstrained optimization problem so that the
zation problems in power systems. In this problem,
problem could be easier to be solved.
it is assumed that the generating units are online
Sigmoid Function: A nonlinear function used
during the considered schedule time. The objective
to calculate the outputs of continuous neurons
of the economic dispatch problem is usually to
based on their inputs. In the sigmoid function, its
minimize the total cost of thermal generating units
slope can be adjusted leading to the change of the
while satisfying the unit and system constraints
function shape and the outputs will be changed
such as power balance, power generation limits,
accordingly.
ramp rate constraints, etc.
Transfer Function: A linear function used to
Energy Function: A function of the outputs
calculate the outputs of multiplier neurons based
of neurons defined for Hopfield neural network.
on their inputs. The slope of this function cannot
In the Hopfield neural network, any change in
be usually adjusted and the outputs of neurons
the status of continuous neurons always leads to
are equally set to their inputs.
minimization of the energy function.
94
95
Chapter 3
Renewable Energy and
Sustainable Development
Abdeen Mustafa Omer
Energy Research Institute, UK
ABSTRACT
The increased availability of reliable and efficient energy services stimulates new development alter-
natives. This article discusses the potential for such integrated systems in the stationary and portable
power market in response to the critical need for a cleaner energy technology. Anticipated patterns of
future energy use and consequent environmental impacts (acid precipitation, ozone depletion and the
greenhouse effect or global warming) are comprehensively discussed in this chapter. Throughout the
theme several issues relating to renewable energies, environment, and sustainable development are ex-
amined from both current and future perspectives. It is concluded that green energies like wind, solar,
ground-source heat pumps, and biomass must be promoted, implemented, and demonstrated from the
economic and/or environmental point view.
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Renewable Energy and Sustainable Development
more economical in their use of energy for heat- • Cost efficiency is the financial expenditure
ing, lighting, cooling, ventilation and hot water on energy relative to the level of environ-
supply. Passive measures, particularly natural or mental comfort and productivity that the
hybrid ventilation rather than air-conditioning, building occupants attained. The overall
can dramatically reduce primary energy con- cost efficiency can be improved by im-
sumption (Omer, 2009a). However, exploitation proving the indoor environmental quality
of renewable energy in buildings and agricultural and the energy efficiency of a building.
greenhouses can, also, significantly contribute
towards reducing dependency on fossil fuels. Several definitions of sustainable development
Therefore, promoting innovative renewable ap- have been put forth, including the following com-
plications and reinforcing the renewable energy mon one: development that meets the needs of
market will contribute to preservation of the the present without compromising the ability of
ecosystem by reducing emissions at local and future generations to meet their own needs. The
global levels. This will also contribute to the World Energy Council (WEC) study found that
amelioration of environmental conditions by without any change in our current practice, the
replacing conventional fuels with renewable ener- world energy demand in 2020 would be 50-80%
gies that produce no air pollution or greenhouse higher than 1990 levels (WEC, 2009). According
gases (during their use). The provision of good to the USA Department of Energy (DoE) report,
indoor environmental quality while achieving annual energy demand will increase from a current
energy and cost efficient operation of the heating, capacity of 363 million kilowatts to 750 million
ventilating and air-conditioning (HVAC) plants kilowatts by 2020 (DoE, 2009). The world’s energy
(devices) in buildings represents a multi variant consumption today is estimated to 22 billion kWh
problem (Omer, 2009b). The comfort of building per year, 53 billion kWh by 2020 (WEC, 2009).
occupants is dependent on many environmental Such ever-increasing demand could place signifi-
parameters including air speed, temperature, rela- cant strain on the current energy infrastructure and
tive humidity and air quality in addition to lighting potentially damage world environmental health
and noise. The overall objective is to provide a by CO, CO2, SO2, NOx effluent gas emissions
high level of building performance (BP), which and global warming (ASHRAE, 2005). Achiev-
can be defined as indoor environmental quality ing solutions to environmental problems that we
(IEQ), energy efficiency (EE) cost efficiency (CE), face today requires long-term potential actions for
and environmental performance (EP). sustainable development. In this regards, renew-
able energy resources appear to be the one of the
• Indoor environmental quality is the per- most efficient and effective solutions since the
ceived condition of comfort that building intimate relationship between renewable energy
occupants experience due to the physical and sustainable development. More rational use
and psychological conditions to which of energy is an important bridge to help transition
they are exposed by their surroundings. from today’s fossil fuel dominated world to a world
The main physical parameters affecting powered by non-polluting fuels and advanced
IEQ are air speed, temperature, relative technologies such as photovoltaics (PVs) and fuel
humidity and air quality. cells (FCs) (Abdeen, 2008a).
• Energy efficiency is related to the provi- An approach is needed to integrate renewable
sion of the desired environmental condi- energies in a way to meet high building perfor-
tions while consuming the minimal quan- mance. However, because renewable energy
tity of energy. sources are stochastic and geographically diffuse,
96
Renewable Energy and Sustainable Development
their ability to match demand is determined by particular greenhouse effect, therefore, contributes
adoption of one of the following two approaches to global warming. The application of greenhouses
(EUO, 2000): the utilisation of a capture area for plants growth can be considered one of the
greater than that occupied by the community to measures in the success of solving this problem.
be supplied, or the reduction of the community’s Maximising the efficiency gained from a green-
energy demands to a level commensurate with the house can be achieved using various approaches,
locally available renewable resources. employing different techniques that could be ap-
For a northern European climate, which is plied at the design, construction and operational
characterised by an average annual solar irradi- stages. The development of greenhouses could be
ance of 150 Wm-2, the mean power production a solution to farming industry and food security
from a photovoltaic component of 13% conver- (Abdeen, 2008b).
sion efficiency is approximately 20 Wm-2 (Duffie Energy security, economic growth and envi-
and Beckman, 1980). For an average wind speed ronment protection are the national energy policy
of 5 ms-1, the power produced by a micro wind drivers of any country of the world. As world
turbine will be of a similar order of magnitude, populations grow, many faster than the growth
though with a different profile shape. In the UK, rate of 2%, the need for more and more energy
for example, a typical office building will have is exacerbated (Figure 1). Enhanced lifestyle and
a demand in the order of 300 kWhm-2yr-1 (EUO, energy demand rise together and the wealthy in-
2000). This translates into approximately 50 Wm-2 dustrialised economics, which contain 25% of the
of façade, which is twice as much as the available world’s population, consume 75% of the world’s
renewable energies (Lysen, 1983). Thus, the aim energy supply (WEC, 2009). The world’s energy
is to utilise energy efficiency measures in order to consumption today is estimated to 22 billion kWh
reduce the overall energy consumption and adjust per year (WEC, 2009). About 6.6 billion metric
the demand profiles to be met by renewable ener- tons carbon equivalent of greenhouse gas (GHG)
gies. For instance, this approach can be applied to emission are released in the atmosphere to meet
greenhouses, which use solar energy to provide this energy demand (WEC, 2009). Approximately
indoor environmental quality. The greenhouse 80% is due to carbon emissions from the com-
effect is one result of the differing properties of bustion of energy fuels (Abdeen, 2008c). At the
heat radiation when it is generated at different current rate of usage, taking into consideration
temperatures. Objects inside the greenhouse, or population increases and higher consumption of
any other building, such as plants, re-radiate the energy by developing countries, oil resources,
heat or absorb it. Because the objects inside the natural gas and uranium will be depleted within a
greenhouse are at a lower temperature than the few decades. People could depend on new nuclear
sun, the re-radiated heat is of longer wavelengths, technologies that will enable much slower uranium
and cannot penetrate the glass. This re-radiated depletion in the future. As for coal, it may take
heat is trapped and causes the temperature inside two centuries or so. Technological progress has
the greenhouse to rise. Note that the atmosphere dramatically changed the world in a variety of
surrounding the earth, also, behaves as a large ways. It has, however, also led to developments
greenhouse around the world. Changes to the e.g., environmental problems, which threaten man
gases in the atmosphere, such as increased carbon and nature. Build-up of carbon dioxide and other
dioxide content from the burning of fossil fuels, GHGs is leading to global warming with unpre-
can act like a layer of glass and reduce the quantity dictable but potentially catastrophic consequences.
of heat that the planet earth would otherwise radi- When fossil fuels burn, they emit toxic pollutants
ate back into space (Brain, and Mark, 2007). This that damage the environment and people’s health
97
Renewable Energy and Sustainable Development
with over 700,000 deaths resulting each year, ac- Hydrogen for fuel cells and the sun for PV have
cording to the World Bank review of 2000. At the been considered for many years as a likely and
current rate of usage, taking into consideration eventual substitute for oil, gas, and coal. The sun
population increases and higher consumption of is the most abundant element in the universe. The
energy by developing countries, oil resources, and use of solar thermal energy or solar photovoltaic
natural gas will be depleted within a few decades. (PVs) for the everyday electricity needs has a
A Figure 2 shows the annual and estimated world distinct advantage: avoid consuming resources
population and energy demand, and Figure 3 the and degrading the environment through pollut-
world oil productions in the next 10-20 years. As ing emissions, oil spills and toxic by-products.
for coal, it may take two centuries or so. One must A one-kilowatt PV system producing 150 kWh
therefore endeavour to take precautions today for each month prevents 75 kg of fossil fuel from
a viable world for coming generations. being mined (WEC, 2009), and 150 kg of CO2
Research into future alternatives has been and from entering the atmosphere and keeps 473
still being conducted aiming to solve the complex litres of water from being consumed (Abdeen,
problems of this recent time e.g., rising energy 2008d). Electricity from fuel cells can be used
requirements of a rapidly and constantly growing in the same way as grid power: to run appliances
world population and global environmental pol- and light bulbs and even to power cars since each
lution. Therefore, options for a long-term and gallon of gasoline produced and used in an inter-
environmentally friendly energy supply have to nal combustion engine releases roughly 12 kg of
be developed leading to the use of renewable CO2, a GHG that contributes to global warming.
sources (water, sun, wind, biomass, geothermal,
hydrogen production by electrolysis of water and People, Power and Pollution
fuel cells. Renewables could shield a nation from
the negative effect in the energy supply, price and Over millions of years ago plants covered the earth,
related environment concerns. converting the energy of sunlight into living tissue,
Figure 1. Annual and estimated world population and energy demand in [Million of barrels per day of
oil equivalent (MBDOE)] (Omer, 2008a)
98
Renewable Energy and Sustainable Development
Figure 2. World oil productions in the next 10-20 years (Omer, 2008a)
some of which was buried in the depths of the earth sources. Coal, oil and gas, which will certainly
to produce deposits of coal, oil and natural gas. be of great value to future generations, as they
The past few decades, however, have experienced are to ours, are non-renewable natural resources.
many valuable uses for these complex chemical This is particularly true now as it is, or soon will
substances, manufacturing from them plastics, be, technically and economically feasible to sup-
textiles, fertiliser and the various end products of ply all of man’s needs from the most abundant
the petrochemical industry. Indeed, each decade energy source of all, the sun. The sunlight is not
sees increasing uses for these products. Renew- only inexhaustible, but, moreover, it is the only
able energy is the term used to describe a wide energy source, which is completely non-polluting.
range of naturally occurring, replenishing energy
99
Renewable Energy and Sustainable Development
Industry’s use of fossil fuels has been blamed gaining experience in designing policies to address
for warming the climate. When coal, gas and oil climate change (Abdeen, 2009a).
are burnt, they release harmful gases, which trap The World Summit on Sustainable Develop-
heat in the atmosphere and cause global warm- ment in Johannesburg in 2002 committed itself
ing (Bos et al, 1994). However, there has been an to ‘‘encourage and promote the development of
ongoing debate on this subject, as scientists have renewable energy sources to accelerate the shift
struggled to distinguish between changes, which towards sustainable consumption and produc-
are human induced, and those, which could be tion’’. Accordingly, it aimed at breaking the link
put down to natural climate variability. Never- between resource use and productivity. This can
theless, industrialised countries have the highest be achieved by the followings:
emission levels, and must shoulder the greatest
responsibility for global warming. However, ac- • Trying to ensure economic growth does
tion must also be taken by developing countries not cause environmental pollution.
to avoid future increases in emission levels as • Improving resource efficiency.
their economies develop and populations grow, as • Examining the whole life-cycle of a
clearly captured by the Kyoto Protocol (Abdeen, product.
2008e). Notably, human activities that emit carbon • Enabling consumers to receive more infor-
dioxide (CO2), the most significant contributor to mation on products and services.
potential climate change, occur primarily from • Examining how taxes, voluntary agree-
fossil fuel production. Consequently, efforts to ments, subsidies, regulation and informa-
control CO2 emissions could have serious, negative tion campaigns, can best stimulate inno-
consequences for economic growth, employment, vation and investment to provide cleaner
investment, trade and the standard of living of technology.
individuals everywhere. Scientifically, it is dif-
ficult to predict the relationship between global The energy conservation scenarios include
temperature and GHG concentrations. The climate rational use of energy policies in all economy
system contains many processes that will change sectors and the use of combined heat and power
if warming occurs. Critical processes include heat systems, which are able to add to energy savings
transfer by winds and tides, the hydrological cycle from the autonomous power plants. Electricity
involving evaporation, precipitation, runoff and from renewable energy sources is by definition
groundwater and the formation of clouds, snow, the environmental green product. Hence, a renew-
and ice, all of which display enormous natural able energy certificate system, as recommended
variability (UNECA, 2003b). by the World Summit, is an essential basis for
The equipment and infrastructure for energy all policy systems, independent of the renewable
supply and use are designed with long lifetimes, energy support scheme. It is, therefore, important
and the premature turnover of capital stock in- that all parties involved support the renewable
volves significant costs. Economic benefits occur energy certificate system in place if it is to work
if capital stock is replaced with more efficient as planned. Moreover, existing renewable energy
equipment in step with its normal replacement technologies (RETs) could play a significant miti-
cycle. Likewise, if opportunities to reduce future gating role, but the economic and political climate
emissions are taken in a timely manner, they will have to change first. The change in climate is
should be less costly. Such a flexible approach real. It is happening now, and GHGs produced by
would allow society to take account of evolving human activities are significantly contributing to
scientific and technological knowledge, while it. The predicted global temperature increase of
100
Renewable Energy and Sustainable Development
between 1.5 and 4.5oC could lead to potentially governments could take to promote and increase
catastrophic environmental impacts (DEFRA, the use of renewables is to improve access for
2006). These include sea level rise, increased renewables to the energy market. This access to
frequency of extreme weather events, floods, the market needs to be under favourable conditions
droughts, disease migration from various places and, possibly, under favourable economic rates as
and possible stalling of the Gulf Stream. This has well. One move that could help, or at least justify,
led scientists to argue that climate change issues better market access would be to acknowledge
are not ones that politicians can afford to ignore, that there are environmental costs associated with
and policy makers tend to agree (DEFRA, 2006). other energy supply options and that these costs
However, reaching international agreements on are not currently internalised within the market
climate change policies is no trivial task as the price of electricity or fuels. This could make a
difficulty in ratifying the Kyoto Protocol has significant difference, particularly if appropriate
proved (UNECA, 2004). subsidies were applied to renewable energy in
Therefore, the use of renewable energy sources recognition of the environmental benefits it offers.
and the rational use of energy, in general, are the Similarly, cutting energy consumption through
fundamental inputs for any responsible energy end-use efficiency is absolutely essential. This
policy. However, the energy sector is encounter- suggests that issues of end-use consumption of
ing difficulties because increased production and energy will have to come into the discussion in
consumption levels entail higher levels of pollu- the foreseeable future (Levine et al, 2005).
tion and eventually climate change, with possibly However, RETs have the benefit of being
disastrous consequences. At the same time, it is environmentally benign when developed in a
important to secure energy at an acceptable cost sensitive and appropriate way with the full involve-
in order to avoid negative impacts on economic ment of local communities. In addition, they are
growth. To date, renewable energy contributes diverse, secure, locally based and abundant. In
as much as 20% of the global energy supplies spite of the enormous potential and the multiple
worldwide (Abdeen, 2009b). Over two thirds of benefits, the contribution from renewable energy
this comes from biomass use, mostly in develop- still lags behind the ambitious claims for it due
ing countries, some of it unsustainable. Yet, the to the initially high development costs, concerns
potential for energy from sustainable technologies about local impacts, lack of research funding and
is huge. On the technological side, renewables poor institutional and economic arrangements
have an obvious role to play. In general, there is (IPCC, 2001).
no problem in terms of the technical potential of Hence, an approach is needed to integrate
renewables to deliver energy. Moreover, there renewable energies in a way that meets high
are very good opportunities for RETs to play an building performance requirements. However,
important role in reducing emissions of GHGs because renewable energy sources are stochastic
into the atmosphere, certainly far more than have and geographically diffuse, their ability to match
been exploited so far. However, there are still some demand is determined by adoption of one of the
technical issues to address in order to cope with following two approaches (Parikn et al, 1999):
the intermittency of some renewables, particularly the utilisation of a capture area greater than that
wind and solar. Yet, the biggest problem with rely- occupied by the community to be supplied, or the
ing on renewables to deliver the necessary cuts in reduction of the community’s energy demands to
GHG emissions is more to do with politics and a level commensurate with the locally available
policy issues than with technical ones (DEFRA, renewable resources.
2006). For example, the single most important step
101
Renewable Energy and Sustainable Development
102
Renewable Energy and Sustainable Development
Energy and Environmental Problems individual travel by cars has led to an increase in
road traffic and hence a shift in attention paid to
Technological progress has dramatically changed the effects and sources of NOx and volatile organic
the world in a variety of ways. It has, however, also compound (VOC) emissions.
led to developments of environmental problems, Environmental problems span a continuously
which threaten man and nature (UNECA, 2003a). growing range of pollutants, hazards and ecosys-
During the past two decades the risk and reality tem degradation over wider areas. The main areas
of environmental degradation have become more of environmental problems are: major environ-
apparent. Growing evidence of environmental mental accidents, water pollution, maritime pol-
problems is due to a combination of several fac- lution, land use and sitting impact, radiation and
tors since the environmental impact of human radioactivity, solid waste disposal, hazardous air
activities has grown dramatically because of the pollutants, ambient air quality, acid rain, strato-
sheer increase of world population, consumption, spheric ozone depletion and global warming
industrial activity, etc., throughout the 1970s most (greenhouse effect, global climatic change) (Table
environmental analysis and legal control instru- 2). The four more important types of harm from
ments concentrated on conventional effluent gas man’s activities are global warming gases, ozone
pollutants such as SO2, NOx, particulates, and CO destroying gases, gaseous pollutants and micro-
(Table 1). Recently, environmental concerns has biological hazards (Table 3). Notably, human
extended to the control of micro or hazardous activities that emit carbon dioxide (CO2), the most
air pollutants, which are usually toxic chemical significant contributor to potential climate change,
substances and harmful in small doses, as well to occur primarily from fossil fuel production. Con-
that of globally significant pollutants such as CO2. sequently, CO2 emissions could have serious,
Aside from advances in environmental science, negative consequences for economic growth,
developments in industrial processes and struc- employment, investment, trade and the standard
tures have led to new environmental problems. of living of individuals everywhere. The earth is
For example, in the energy sector, major shifts warmer due to the presence of gases but the
to the road transport of industrial goods and to global temperature is rising. This could lead to
Table 1. EU criteria pollutant standards in the ambient air environment (Omer, 2008a)
Pollutant EU limit
CO 30 mg/m2; 1h
NO2 200 μg/m2; 1h
O3 235 μg/m2; 1h
SO2 250-350 μg/m2; 24 h
80-120 μg/m2; annual
PM10 250 μg/m2; 24 h
80 μg/m2; annual
SO2 + PM10 100-150 μg/m2; 24 h
40-60 μg/m2; annual
Pb 2 μg/m2; annual
Total suspended particulate (TSP 260 μg/m2; 24 h
HC 160 μg/m2; 3 h
103
Renewable Energy and Sustainable Development
Table 2. Significant EU environmental directives in water, air and land environments (Omer, 2008a)
the sea level rising at the rate of 60 mm each pollution, acid precipitation, ozone depletion,
decade with the growing risk of flooding in low- forest destruction and emission of radioactive
lying areas (Figure 4). substances (Table 4). These issues must be taken
At the United Nations Earth Summit at Rio in into consideration simultaneously if humanity is
June 1992 some 153 countries agreed to pursue to achieve a bright energy future with minimal
sustainable development (Boulet, 1987). A main environmental impacts. Much evidence exists,
aim was to reduce emission of carbon dioxide and which suggests that the future will be negatively
other GHGs. Reduction of energy use in buildings impacted if humans keep degrading the environ-
is a major role in achieving this. Carbon dioxide ment.
targets are proposed to encourage designers to During the past century, global surface tem-
look at low energy designs and energy sources. peratures have increased at a rate near 0.6oC/
Problems with energy supply and use are century and the average temperature of the At-
related not only to global warming that is taking lantic, Pacific and Indian oceans (covering 72%
place, due to effluent gas emission mainly CO2, of the earth surface) have risen by 0.06oC since
but also to such environmental concerns as air 1995. Global temperatures in 2001 were 0.52oC
104
Renewable Energy and Sustainable Development
Table 4. Global emissions of the top fourteen nations by total CO2 volume (billion of tones) (Omer, 2009)
105
Renewable Energy and Sustainable Development
Figure 5. Global mean temperature changes over the period of 1990-2100 and 1990-2030 (Omer, 2009)
106
Renewable Energy and Sustainable Development
etc.). These water resources are renewable but distributional effects would threaten the long-
potentially destructible resources. While moder- term viability of the economy. This inflation-
ate amounts of human water extractions from ary dynamics is not due to any form of market
a given regional water system can be sustained imperfection but is a natural consequence of the
for indefinite periods. Excessive extractions will competitive arbitrage behaviour of unregulated
change the geographical and climatic conditions private resource owners.
supporting the water cycle and will diminish
the regenerative capacity of the regional water Sustainability Concept
system, thereby reducing the potential for future
withdrawals. Typically, recovery from any such Absolute sustainability of electricity supply is a
resource degradation will be very slow and dif- simple concept: no depletion of world resources
ficult, if not impossible; resource degradation is and no ongoing accumulation of residues. Relative
partially irreversible (Erreygers, 1996). sustainability is a useful concept in comparing the
To make sustainable water extraction economi- sustainability of two or more generation technolo-
cally viable, the sustainable policy has to break gies. Therefore, only renewables are absolutely
even (all costs are covered by revenues) while sustainable, and nuclear is more sustainable than
unsustainable policy has to be unprofitable (costs fossil. Energy used to produce devices and plants
exceed revenues): for renewable energy is not sustainable. How-
ever, any discussion about sustainability must
(1+r) vt-1 = 5yt + vt (1) not neglect the ability or otherwise of the new
technologies to support the satisfactory opera-
Where: r is the interest rate, t=year, yt is the tion of the electricity supply infrastructure. The
revenue, vt is initial costs recovered by revenue, electricity supply system has been developed to
and vt-1 is all costs are covered by revenues. have a high degree of resilience against the loss
of transmission circuits and major generators, as
(1+r) vt-1 > 105yt (2) well as unusually large and rapid load changes.
It is unlikely that consumers would tolerate any
(1+r) vt-1 < [105/(105-5)] vt (3) reduction in the quality of the service, even if
this were the result of the adoption of otherwise
The term [105/(105-5)] is to define the natural benign generation technologies. Renewables are
productivity factor of the water resource as (1+g) generally weather-dependent and as such their
= [105/(105-5)]; g is the natural productivity rate. likely output can be predicted but not controlled.
Rate g will be close to zero if the sustainable The only control possible is to reduce the output
extraction level is much smaller than the unsustain- below that available from the resource at any given
able level. Using g, the equation can be as follows: time. Therefore, to safeguard system stability and
security, renewables must be used in conjunction
vt > (1+r)/(1+g) vt-1 (4) with other, controllable, generation and with large-
scale energy storage. There is a substantial cost
Regulatory measures that prevent resource associated with this provision (Abdeen, 2009d).
owners from adopting certain unsustainable It is useful to codify all aspects of sustain-
extraction policies are a necessary pre-condition ability, thus ensuring that all factors are taken into
for the effective operation of a privatised natural account for each and every development proposal.
resource sector. Unregulated water privatisation Therefore, with the intention of promoting debate,
would result in an inflationary dynamics whose the following considerations are proposed:
107
Renewable Energy and Sustainable Development
1. Long-term availability of the energy source arguments for incineration with energy recovery,
or fuel. rather than material recycling. Cleaner produc-
2. Price stability of energy source or fuel. tion approach and pollution control measures
3. Acceptability or otherwise of by-products are needed in the recycling sector as much as
of the generation process. in another. The industrial sector world widely is
4. Grid services, particularly controllability of responsible for about one third of anthropogenic
real and reactive power output. emissions of carbon dioxide, the most important
5. Technological stability, likelihood of rapid greenhouse gas. Industry is also an important
technical obsolescence. emitter of several other greenhouse gases. And
6. Knowledge base of applying the technology. many of industry’s products emit greenhouse gases
7. Life of the installation – a dam may last more as well, either during use or after they become
than 100 years, but a gas turbine probably waste. Opportunities exist for substantial reduc-
will not. ing industrial emissions through more efficient
8. Maintenance requirement of the plant. production and use of energy: fuel substitutions,
the use of alternative energy technologies, process
Environmental Aspects modification, and by revising materials strategies
to make use of less energy and greenhouse gas
Environmental pollution is a major problem facing intensive materials. Industry has an additional
all nations of the world. People have caused air role to play through the design of products that
pollution since they learned to how to use fire, use less energy and materials and produce lower
but man-made air pollution (anthropogenic air greenhouse gas emissions. Table 5 summarises
pollution) has rapidly increased since industriali- the classification of data requirements.
sation began. Many volatile organic compounds Development in the environmental sense is a
and trace metals are emitted into the atmosphere rather recent concern relating to the need to man-
by human activities. The pollutants emitted into age scarce natural resources in a prudent manner-
the atmosphere do not remain confined to the because human welfare ultimately depends on
area near the source of emission or to the local ecological services. The environmental interpre-
environment, and can be transported over long tation of sustainability focuses on the overall
distances, and create regional and global envi- viability and health of ecological systems- defined
ronmental problems. The privatisation and price in terms of a comprehensive, multiscale, dy-
liberalisation in energy fields has been secured namic, hierarchical measure of resilience, vigour
to some extent (but not fully). There should be and organisation. Natural resource degradation,
availability and adequate energy supplies to the pollution and loss of biodiversity are detrimental
major productive sectors. The result is that, the because they increase vulnerability, undermine
present situation of energy supplies is for better system health, and reduce resilience. The envi-
than ten years ago. ronmental issues include:
A great challenge facing the global community
today is to make the industrial economy more like • Global and transnational (climate change,
the biosphere, that is, to make it a more closed ozone layer depletion).
system. This would save energy, reduce waste • Natural habitats (forests and other
and pollution, and reduce costs. In short, it would ecosystems).
enhance sustainability. Often, it is technically • Land (agricultural zones).
feasible to recycle waste in one of several dif- • Water resources (river basin, aquifer, water
ferent ways. For some wastes there are powerful shed).
108
Renewable Energy and Sustainable Development
109
Renewable Energy and Sustainable Development
• Best practice tools- making use of free help • Product stewardship- taking the broad-
and advice from government best practice est view of ‘producer responsibility’ and
programmes (energy efficiency, environ- working to reduce all the ‘downstream’ ef-
mental technology, resource savings). fects of products after they have been sold
• Innovation and ecodesign- rethinking the on to customers.
delivery of ‘value added’ by the business, • Openness and transparency- publicly re-
so that impact reduction and resource ef- porting on environmental performance
ficiency are firmly built in at the design against meaningful targets; actively us-
stage. ing clear labels and declarations so that
• Cleaner, leaner production processes- pur- customers are fully informed; building
suing improvements and savings in waste stakeholder confidence by communicat-
minimisation, energy and water consump- ing sustainability aims to the workforce,
tion, transport and distribution, as well as the shareholders and the local community
reduced emissions. Tables (6-8) indicate (Figure 6 and Table 9).
energy conservation, sustainable develop-
ment and environment. With the debate on climate change, the prefer-
• Supply chain management- specifying ence for real measured data has been changed.
more demanding standards of sustainabil- The analyses of climate scenarios need an hourly
ity from ‘upstream’ suppliers, while sup- weather data series that allows for realistic
porting smaller firms to meet those higher changes in various weather parameters (REN21,
standards. 2007). By adapting parameters in a proper way,
Technological criteria Energy and environment criteria Social and economic criteria
Primary energy saving in regional scale Sustainability according to greenhouse Labour impact
gas pollutant emissions
Technical maturity, reliability Sustainable according to other pollutant Market maturity
emissions
Consistence of installation and maintenance re- Land requirement Compatibility with political, legislative and
quirements with local technical known-how administrative situation
Continuity and predictability of performance Sustainability according to other envi- Cost of saved primary energy
ronmental impacts
110
Renewable Energy and Sustainable Development
Table 8. Positive impact of durability, adaptability and energy conservation on economic, social and
environment systems
111
Renewable Energy and Sustainable Development
data series can be generated for the site. Weather the shorter time horizons associated with market
generators should be useful for: forces as deregulation replaces vertical integration.
Sustainable low-carbon energy scenarios for the
• Calculation of energy consumption (no ex- new century emphasise the untapped potential
treme conditions are required) of renewable resources. Rural areas can benefit
• Design purposes (extremes are essential), from this transition. The increased availability of
and reliable and efficient energy services stimulates
• Predicting the effect of climate change new development alternatives. It is concluded
such as increasing annually average of that renewable environmentally friendly energy
temperature. must be encouraged, promoted, implemented, and
demonstrated by full-scale plant especially for use
This results in the following requirements: in remote rural areas (CEC, 2000).
This is the step in a long journey to encourage a
• Relevant climate variables should be gen- progressive economy, which continues to provide
erated (solar radiation: global, diffuse, di- us with high living standards, but at the same time
rect solar direction, temperature, humidity, helps reduce pollution, waste mountains, other
wind speed and direction) according to the environmental degradation, and environmental
statistics of the real climate. rationale for future policy-making and interven-
• The average behaviour should be in accor- tion to improve market mechanisms. This vision
dance with the real climate. will be accomplished by:
• Extremes should occur in the generated
series in the way it will happen in a real • ‘Decoupling’ economic growth and envi-
warm period. This means that the gener- ronmental degradation. The basket of indi-
ated series should be long enough to assure cators illustrated shows the progress being
these extremes, and series based on aver- made. Decoupling air and water pollution
age values from nearby stations. from growth, making good headway with
CO2 emissions from energy, and transport.
Growing concerns about social and environ- The environmental impact of our own in-
mental sustainability have led to increased interest dividual behaviour is more closely linked
in planning for the energy utility sector because of to consumption expenditure than the econ-
its large resource requirements and production of omy as a whole.
emissions (Roriz, 2001). A number of conflicting • Focusing policy on the most important en-
trends combine to make the energy sector a major vironmental impacts associated with the
concern, even though a clear definition of how to use of particular resources, rather than on
measure progress toward sustainability is lacking. the total level of all resource use.
These trends include imminent competition in the • Increasing the productivity of material and
electricity industry, global climate change, expect- energy use that are economically efficient
ed long-term growth in population and pressure to by encouraging patterns of supply and de-
balance living standards (including per capital en- mand, which are more efficient in the use
ergy consumption). Designing and implementing of natural resources (the aim is to promote
a sustainable energy sector will be a key element innovation and competitiveness) and in-
of defining and creating a sustainable society. In vesting in areas such as energy efficiency,
the electricity industry, the question of strategic water efficiency and waste minimisation.
planning for sustainability seems to conflict with
112
Renewable Energy and Sustainable Development
• Encouraging and enabling active and efficient, inherently cleaner technologies that do
informed individual and corporate not rely heavily on abatement. For existing plant,
consumers. the requirements are likely to lead to improved
control of:
On some climate change issues (such as global
warming), there is no disagreement among the • NOx emissions, by the adoption of infur-
scientists. The greenhouse effect is unquestion- nace combustion control and abatement
ably real; it is essential for life on earth. Water techniques.
vapour is the most important GHG; next is carbon • Acid gases, by the adoption of abate-
dioxide (CO2). Without a natural greenhouse ef- ment techniques and optimisation of their
fect, scientists estimate that the earth’s average control.
temperature would be –18oC instead of its present • Particulate control techniques, and their
14oC. There is also no scientific debate over the fact optimisation, e.g., of bag filters and elec-
that human activity has increased the concentration trostatic precipitators.
of the GHGs in the atmosphere (especially CO2
from combustion of coal, oil and gas) (Leszek and Lifecycle analysis of several ethanol feedstocks
Jakub, 2009). The greenhouse effect is also being shows the emissions per ton of feedstock are high-
amplified by increased concentrations of other est for corn stover and switchgrass (about 0.65
gases, such as methane, nitrous oxide, and CFCs tons of CO2 per ton of feedstock) and lowest for
as a result of human emissions. Most scientists corn (about 0.5 ton). Emissions due to cultivation
predict that rising global temperatures will raise and harvesting of corn and wheat are higher than
the sea level and increase the frequency of intense those for lignocellulosics, and although the lat-
rain or snowstorms. Climate change scenarios ter have a far higher process energy requirement
sources of uncertainty and factors influencing the (Figure 8). GHG emissions are lower because this
future climate are: energy is produced from biomass residue, which
is carbon neutral.
• The future emission rates of the GHGs. The waste and resources action programme
• The effect of this increase in concentration has been working hard to reduce demand for
on the energy balance of the atmosphere. virgin aggregates and market uptake of recycled
• The effect of these emissions on GHGs and secondary alternatives (Figure 7). The pro-
concentrations in the atmosphere, and gramme targets are:
• The effect of this change in energy balance
on global and regional climate. • To deliver training and information on the
role of recycling and secondary aggregates
in sustainable construction for influences
Wastes in the supply chain, and
• To develop a promotional programme to
Waste is defined as an unwanted material that highlight the new information on websites.
is being discarded. Waste includes items being
taken for further use, recycling or reclamation. Sulphur in Fuels and Its
Waste produced at household, commercial and Environmental Consequences
industrial premises are control waste and come
under the waste regulations. Waste Incineration Organic sulphur is bonded within the organic
Directive (WID) emissions limit values will favour structure of the coal in the same way that sulphur
113
Renewable Energy and Sustainable Development
Emissions will also, of course, occur from • Internally, by baffling and restricting flow
petroleum-based or shale-based fuels, and in of the burnt gas away from the burner, re-
heavy consumption, such as in steam raising. sulting in flame re-entry of part of it.
There will frequently be a need to control SO2 • Externally, by diverting up to 10% of the
emissions. There are, broadly speaking, three flue gas back into the flame.
ways of achieving such control:
Particles
• Pre-combustion control: involves carrying
out a degree of desulphurisation of the fuel. Some of the available control procedures for
• Combustion control: incorporating into the particles are summarised in Table 12. Figure 9
combustion system something capable of shows the variation of distribution factor with
trapping SO2. particle size.
Figure 7. Comparison of thermal biomass usage options, CHP displacing natural gas as a heat source
(Omer, 2009) Billion tones (Bton) Scenarios are (1) household, (2) commercial, (3) agriculture and (4)
industrial
114
Renewable Energy and Sustainable Development
Figure 8. The lifecycle energy balance of corn and Switchgrass reveal a paradox: corn, as an ethanol
feedstock requires less energy for production, i.e., more of the original energy in starch is retained in
the ethanol fuel. Nevertheless, the Switchgrass process yields higher GHG emissions. This is because
most of the process energy for Switchgrass process is generated from biomass residue.
115
Renewable Energy and Sustainable Development
Figure 9. The variation of distribution factor against particle size for coal undersizes in a classifier. The
sizes correspond to mid-point for ranges (Omer, 2008c)
116
Renewable Energy and Sustainable Development
Ground Source Heat Pumps benefit renewable energies that are designed for
space conditioning, but also will contribute to
The ground is as universal as air and solar radiation the global mix of green power and green heat
(Mortal, 2002). Over the past twenty years, as the capacity. Earth energy (also called geothermal
hunt for natural low-carbon energy sources has or ground source heat pumps or GeoExchange),
intensified, there has been an increased endeavour which transfers absorbed solar heat from the
to investigate and develop both earth and ground ground into a building for space heating or water
water thermal energy storage and usage (Lund, heating. The same system can be reversed to reject
Freeston, and Boyd, 2005). Geothermal energy heat from the interior into the ground in order to
solutions, although well known, are another in provide cooling. A typical configuration buries
our armoury of renewable energy sources that are polyethylene pipe below the frost line to serve as
within our immediate grasp to use and integrate the head source (or sink), or it can use lake water
with an overall energy policy (Huttrer, 2001). and aquifers as the heat medium (Omer, 2008c).
For high temperature heat storage with tem-
peratures in excess of 50oC the particular concerns Effects of Urban Density
were:
Compact development patterns of buildings can
• Clogging of wells and heat exchangers due reduce infrastructure demands and the need to
to fines and precipitation of minerals. travel by car. As population density increases,
• Water treatment to avoid operational prob- transportation options multiply and dependence
lems resulting from the precipitation of areas, per capita fuel consumption is much lower
minerals. in densely populated areas because people drive
• Corrosion of components in the groundwa- much less. Few roads and commercially viable
ter system. public transport are the major merits. On the other
• Automatic control of the ground water hand, urban density is a major factor that deter-
system. mines the urban ventilation conditions, as well as
the urban temperature. Under given circumstances,
Three main techniques that are used to ex- an urban area with a high density of buildings
ploit the heat available are geothermal aquifers, can experience poor ventilation and strong heat
hot dry rocks and GSHPs. Geothermal energy is island effect. In warm-humid regions these fea-
the natural heat that exists within the earth and tures would lead to a high level of thermal stress
that can be absorbed by fluids occurring within, of the inhabitants and increased use of energy in
or introduced into the crystal rocks. Heat pump air-conditioned buildings (Reddy, Williams, and
technology can be used for heating only, or for Johansson, 2007).
cooling only, or be ‘reversible’ and used for heat- However, it is also possible that a high-density
ing and cooling depending on the demand. More urban area, obtained by a mixture of high and low
generally, there is still potential for improvement buildings, could have better ventilation condi-
in the performance of heat pumps. tions than an area with lower density but with
As consumers in less-developed countries buildings of the same height. Closely spaced or
increase their capacity of electricity and green high-rise buildings are also affected by the use
power, developed nations are starting to realise of natural lighting, natural ventilation and solar
the benefits of using low-grade thermal energy energy. If not properly planned, energy for elec-
for green heat applications that do not require tric lighting and mechanical cooling/ventilation
high-grade electricity. This shift will not only may be increased and application of solar energy
117
Renewable Energy and Sustainable Development
systems will be greatly limited. Table 13 gives a ogy in buildings has an important role to play.
summary of the positive and negative effects of A review of options indicates benefits and some
urban density. All in all, denser city models require problems (BS, 1989). There are two key elements
more careful design in order to maximise energy to the fulfilling of renewable energy technology
efficiency and satisfy other social and develop- potential within the field of building design; first
ment requirements. Low energy design should the installation of appropriate skills and attitudes
not be considered in isolation, and in fact, it is in building design professionals and second the
a measure, which should work in harmony with provision of the opportunity for such people to
other environmental objectives. Hence, building demonstrate their skills. This second element may
energy study provides opportunities not only for only be created when the population at large and
identifying energy and cost savings, but also for clients commissioning building design in particu-
examining the indoor and outdoor environment lar, become more aware of what can be achieved
(Aroyeun, et al., 2009). and what resources are required. Terms like passive
cooling or passive solar use mean that the cooling
Energy Efficiency and Architectural of a building or the exploitation of the energy of
Expression the sun is achieved not by machines but by the
building’s particular morphological organisation
The focus of the world’s attention on environmen- ((EFC, 2000). Hence, the passive approach to
tal issues in recent years has stimulated response themes of energy savings is essentially based on
in many countries, which have led to a closer the morphological articulations of the construc-
examination of energy conservation strategies tions. Passive solar design, in particular, can realize
for conventional fossil fuels. Buildings are im- significant energy and cost savings. For a design to
portant consumers of energy and thus important be successful, it is crucial for the designer to have a
contributors to emissions of greenhouse gases good understanding of the use of the building. Few
into the global atmosphere. The development and of the buildings had performed as expected by their
adoption of suitable renewable energy technol- designers. To be more precise, their performance
118
Renewable Energy and Sustainable Development
had been compromised by a variety of influences natural and artificial ingredients such as soil and
related to their design, construction and opera- vegetation, urban fabrics and pollution (Zuatori,
tion. However, there is no doubt that the passive 2005). The most dominant outside parameter is, of
energy approach is certainly the one that, being course, the sun’s irradiation, our planet’s primary
supported by the material shape of the buildings energy source. All these elements can be measured
has a direct influence on architectural language in physical terms and are therefore the subject
and most greatly influences architectural expres- of science. Within the second concept, however,
siveness (Lazzarin et al, 2002). Furthermore, form one considers the emotional and intellectual en-
is a main tool in architectural expression. To give ergies, which are the prime inexhaustible source
form to the material things that one produces is of renewable power (Anne et al, 2005). In this
an ineluctable necessity. In architecture, form, case, cultural parameters, which are not exactly
in fact, summarises and gives concreteness to measurable, are involved. However, they repre-
its every value in terms of economy, aesthetics, sent the very essence of the architectural quality.
functionality and, consequently, energy efficiency Objective scientific measurement parameters tell
(David, 2003). The target is to enrich the expres- us very little about the emotional way of perceiv-
sive message with forms producing an advantage ing, which influences the messages of human are
energy-wise. Hence, form, in its geometric and physical sensorial organs. The perceptual reality
material sense, conditions the energy efficiency arises from a multitude of sensorial components;
of a building in its interaction with the environ- visual, thermal, acoustic, olfactory and kinaesthet-
ment. It is, then, very hard to extract and separate ics. It can also arise from the organisational quality
the parameters and the elements relative to this of the space in which different parameters come
efficiency from the expressive unit to which they together, like the sense of order or of serenity.
belong. By analysing energy issues and strategies Likewise, practical evaluations, such as useful-
by means of the designs, of which they are an inte- ness, can be involved too. The evaluation is a
gral part, one will, more easily, focus the attention wholly subjective matter, but can be shared by a
on the relationship between these themes, their set of experiencing persons (Randal et al, 1998).
specific context and their architectural expres- Therefore, these cultural parameters could be dif-
siveness. Many concrete examples and a whole ferent in different contexts in spite of the inexorable
literature have recently grown up around these levelling on a planet- wide scale. However, the
subjects and the wisdom of forms and expedients parameters change in the anthropological sense,
that belong to millennia-old traditions has been not only with the cultural environment, but also in
rediscovered. Such a revisiting, however, is only, relation to function. The scientifically measurable
or most especially, conceptual, since it must be parameters can, thus, have their meanings very
filtered through today’s technology and needs; profoundly altered by the non-measurable, but
both being almost irreconcilable with those of the describable, cultural parameters.
past. Two among the historical concepts are of However, the low energy target also means to
special importance. One is rooted in the effort to eliminate any excess in the quantities of material
establish rational and friendly strategic relations and in the manufacturing process necessary for the
with the physical environment, while the other construction of our built environment. This claims
recognises the interactions between the psyche and for a more sober, elegant and essential expression,
physical perceptions in the creation of the feeling which is not jeopardising at all, but instead enhanc-
of comfort. The former, which may be defined ing, the richness and preciousness of architecture,
as an alliance with the environment deals with while contributing to a better environment from
the physical parameters involving a mixture of an aesthetic viewpoint (Yadav et al, 1997). Argu-
119
Renewable Energy and Sustainable Development
ably, the most successful designs were in fact the Energy Efficiency
simplest. Paying attention to orientation, plan
and form can have far greater impact on energy Energy efficiency is the most cost-effective way
performance than opting for elaborate solutions of cutting carbon dioxide emissions and improve-
(EIBI, 1999). However, a design strategy can fail ments to households and businesses. It can also
when those responsible for specifying materials have many other additional social, economic and
for example, do not implement the passive solar health benefits, such as warmer and healthier
strategy correctly. Similarly, cost-cutting exercises homes, lower fuel bills and company running costs
can seriously upset the effectiveness of a design and, indirectly, jobs. Britain wastes 20 per cent
strategy. Therefore, it is imperative that a designer of its fossil fuel and electricity use (Witte, et al.,
fully informs key personnel, such as the quantity 2002). This implies that it would be cost-effective
surveyor and client, about their design and be to cut £10 billion a year off the collective fuel bill
prepared to defend it. Therefore, the designer and reduce CO2 emissions by some 120 million
should have an adequate understanding of how tones. Yet, due to lack of good information and
the occupants or processes, such as ventilation, advice on energy saving, along with the capital
would function within the building (Lam, 2000). to finance energy efficiency improvements, this
Thinking through such processes in isolation huge potential for reducing energy demand is not
without reference to others can lead to conflicting being realised (Paul, 2001). Traditionally, energy
strategies, which can have a detrimental impact utilities have been essentially fuel providers and
upon performance. Likewise, if the design intent of the industry has pursued profits from increased
the building is not communicated to its occupants, volume of sales. Institutional and market arrange-
there is a risk that they will use it inappropriately, ments have favoured energy consumption rather
thus, compromising its performance. Hence, the than conservation. However, energy is at the centre
designer should communicate in simple terms of the sustainable development paradigm as few
the actions expected of the occupant to control activities affect the environment as much as the
the building. For example, occupants should be continually increasing use of energy. In addition,
well informed about how to guard against summer more than three quarters of the world’s consump-
overheating. If the designer opted for a simple, tion of these fuels is used, often inefficiently,
seasonally adjusted control; say, insulated sliding by only one quarter of the world’s population.
doors were to be used between the mass wall and Without even addressing these inequities or the
the internal space. The lesson here is that designers precious, finite nature of these resources, the scale
must be prepared to defend their design such that of environmental damage will force the reduction
others appreciate the importance and interrelation- of the usage of these fuels long before they run
ship of each component (IEA, 2008). A strategy out (WB, 2003b).
will only work if each individual component is Throughout the energy generation process
considered as part of the bigger picture. Failure to there are impacts on the environment on local,
implement a component or incorrect installation, national and international levels, from opencast
for example, can lead to failure of the strategy mining and oil exploration to emissions of the
and consequently, in some instances, the build- potent greenhouse gas carbon dioxide in ever
ing may not liked by its occupants due to its poor increasing concentration. Recently, the world’s
performance. leading climate scientists reached an agreement
that human activities, such as burning fossil
fuels for energy and transport, are causing the
temperature to rise. The Intergovernmental Panel
120
Renewable Energy and Sustainable Development
on Climate Change has concluded that ‘‘the bal- These are normally measures, which require
ance of evidence suggests a discernible human minimal investment and can be implemented
influence on global climate’’. It predicts a rate quickly. The followings are some examples of
of warming greater than any one seen in the last such measures:
10,000 years, in other words, throughout human
history. The exact impact of climate change is A. Good housekeeping, monitoring energy use
difficult to predict and will vary regionally. It and targeting waste-fuel practices.
could, however, include sea level rise, disrupted B. Adjusting controls to match requirements.
agriculture and food supplies and the possibility C. Improved greenhouse space utilisation.
of more freak weather events such as hurricanes D. Small capital item time switches, thermo-
and droughts. Indeed, people already are wak- stats, etc.
ing up to the financial and social, as well as the E. Carrying out minor maintenance and repairs.
environmental, risks of unsustainable energy F. Staff education and training.
generation methods that represent the costs of the G. Ensuring that energy is being purchased
impacts of climate change, acid rain and oil spills. through the most suitable tariff or contract
The insurance industry, for example, concerned arrangements.
about the billion dollar costs of hurricanes and
floods, has joined sides with environmentalists 2. Medium priority/medium cost:
to lobby for greenhouse gas emissions reduction.
Friends of the earth are campaigning for a more Measures, which, although involve little or no
sustainable energy policy, guided by the principle design, involve greater expenditure and can take
of environmental protection and with the objec- longer to implement. Examples of such measures
tives of sound natural resource management and are listed below:
long-term energy security. The key priorities of
such an energy policy must be to reduce fossil fuel A. New or replacement controls.
use, move away from nuclear power, improve the B. Greenhouse component alteration, e.g.,
efficiency with which energy is used and increase insulation, sealing glass joints, etc.
the amount of energy obtainable from sustainable, C. Alternative equipment components, e.g.,
renewable sources (WB, 2004). Efficient energy energy efficient lamps in light fittings, etc.
use has never been more crucial than it is today,
particularly with the prospect of the imminent 3. Long term/high cost:
introduction of the climate change levy (CCL).
Establishing an energy use action plan is the es- These measures require detailed study and design.
sential foundation to the elimination of energy They can be best represented by the followings:
waste. A logical starting point is to carry out an
energy audit that enables the assessment of the A. Replacing or upgrading of plant and
energy use and determine what actions to take. The equipment.
actions are best categorised by splitting measures B. Fundamental redesign of systems, e.g., com-
into the following three general groups: bined heat and power (CHP) installations.
121
Renewable Energy and Sustainable Development
122
Renewable Energy and Sustainable Development
Figure 10. Supply side and demand side management approach for energy
a supply of energy resources that are sustainably Hence, as much as environmental impact is as-
available at a reasonable cost and can cause no sociated with energy, sustainable development
negative societal impacts. Energy resources such requires the use of energy resources, which cause
as fossil fuels are finite and lack sustainability, as little environmental impact as possible. One
while renewable energy sources are sustainable way to reduce the resource depletion associated
over a relatively longer term. Environmental with cycling is to reduce the losses that accompany
concerns are also a major factor in sustainable the transfer of exergy to consume resources by in-
development, as activities, which degrade the creasing the efficiency of exergy transfer between
environment, are not sustainable (Jeremy, 2005). resources i.e. increasing the fraction of exergy
123
Renewable Energy and Sustainable Development
removed from one resource that is transferred to Carnot Quality Factor (CQF) = (1-To/Ts) (5)
another (Erlich, 1991).
As explained above, exergy efficiency may Exergy = Energy (transferred) x CQF (6)
be thought of as a more accurate measure of
energy efficiency that accounts for quantity and Where To is the environment temperature (K)
quality aspects of energy flows. Improved exergy and Ts is the temperature of the stream (K).
efficiency leads to reduced exergy losses (WB, Various parameters are essential to achieving
2006). Most efficiency improvements produce sustainable development in a society. Some of
direct environmental benefits in two ways. First, them are as follows:
operating energy input requirements are reduced
per unit output, and pollutants generated are • Public awareness
correspondingly reduced. Second, consideration • Information
of the entire life cycle for energy resources and • Environmental education and training
technologies suggests that improved efficiency • Innovative energy strategies
reduces environmental impact during most stages • Renewable energy sources and cleaner
of the life cycle (Dragana, 2008). Quite often, the technologies
main concept of sustainability, which often inspires • Financing
local and national authorities to incorporate en- • Monitoring and evaluation tools
vironmental consideration into setting up energy
programmes have different meanings in different The development of a renewable energy in a
contexts though it usually embodies a long-term country depends on many factors. Those important
perspective (WB, 2007). Future energy systems to success are listed below:
will largely be shaped by broad and powerful
trends that have their roots in basic human needs. 1. Motivation of the population
Combined with increasing world population, the
need will become more apparent for success- The population should be motivated towards
ful implementation of sustainable development awareness of high environmental issues, rational
(White, and Robinson, 2008). use of energy in order to reduce cost. Subsidy
Heat has a lower exergy, or quality of energy, programme should be implemented as incentives
compared with work. Therefore, heat cannot be to install renewable energy plants. In addition,
converted into work by 100% efficiency. Some image campaigns to raise awareness of renew-
examples of the difference between energy and able technology.
exergy are shown in Table 14.
The terms used in Table 14 have the following 2. Technical product development
meanings:
124
Renewable Energy and Sustainable Development
125
Renewable Energy and Sustainable Development
126
Renewable Energy and Sustainable Development
Sustainable energy is the provision of energy a result of the increased concentration of carbon
such that it meets the needs of the present without dioxide and other GHGs in the atmosphere as re-
compromising the ability of future generations to leased by burning fossil fuels. This global warming
meet their needs. A broader interpretation may al- will eventually lead to substantial changes in the
low inclusion of fossil fuels and nuclear fission as world’s climate, which will, in turn, have a major
transitional sources while technology develops, as impact on human life and the built environment.
long as new sources are developed for future gen- Therefore, effort has to be made to reduce fossil
erations to use. A narrower interpretation includes energy use and to promote green energies, particu-
only energy sources, which are not expected to be larly in the building sector. Energy use reductions
depleted in a time frame relevant to the human race. can be achieved by minimising the energy demand,
Sustainable energy sources are most often regarded by rational energy use, by recovering heat and the
as including all renewable sources, such as biofuels, use of more green energies. This article was a step
solar power, wind power, wave power, geothermal towards achieving that goal. The adoption of green
power and tidal power. It usually also includes tech- or sustainable approaches to the way in which
nologies that improve energy efficiency. This new society is run is seen as an important strategy in
and important handbook gathers the latest research finding a solution to the energy problem. The key
from around the globe in the study of sustainable factors to reducing and controlling CO2, which is
energy and highlights such topics as: monitoring the major contributor to global warming, are the
sustainable energy development; methane; energy use of alternative approaches to energy generation
and territory; biodiesel production; electrochemi- and the exploration of how these alternatives are
cal hydrogen storage; environmental policies in an used today and may be used in the future as green
electricity sector and others). energy sources. Even with modest assumptions
The move towards a de-carbonised world, about the availability of land, comprehensive
driven partly by climate science and partly by the fuel-wood farming programmes offer significant
business opportunities it offers, will need the pro- energy, economic and environmental benefits.
motion of environmentally friendly alternatives, These benefits would be dispersed in rural areas
if an acceptable stabilisation level of atmospheric where they are greatly needed and can serve as
carbon dioxide is to be achieved. This requires the linkages for further rural economic development.
harnessing and use of natural resources that produce The nations as a whole would benefit from savings
no air pollution or greenhouse gases and provides in foreign exchange, improved energy security, and
comfortable coexistence of human, livestock, socio-economic improvements. With a nine-fold
and plants. This study reviews the energy-using increase in forest – plantation cover, a nation’s
technologies based on natural resources, which are resource base would be greatly improved. The
available to and applicable in the farming industry. international community would benefit from
Integral concept for buildings with both excel- pollution reduction, climate mitigation, and the
lent indoor environment control and sustainable increased trading opportunities that arise from
environmental impact are reported in the present new income sources. The non-technical issues,
communication. which have recently gained attention, include: (1)
Environmental and ecological factors e.g., carbon
sequestration, reforestation and revegetation. (2)
CONCLUSION Renewables as a CO2 neutral replacement for fossil
fuels. (3) Greater recognition of the importance
There is strong scientific evidence that the average of renewable energy, particularly modern bio-
temperature of the earth’s surface is rising. This is mass energy carriers, at the policy and planning
127
Renewable Energy and Sustainable Development
levels. (4) Greater recognition of the difficulties Abdeen, M. O. (2009a). Environmental and
of gathering good and reliable renewable energy socio-economic aspect of possible development
data, and efforts to improve it. (5) Studies on in renewable energy use. In Proceedings of the
the detrimental health efforts of biomass energy 4th International Symposium on Environment,
particularly from traditional energy users. Athens, Greece, 21-24 May 2009.
Abdeen, M. O. (2009b). Energy use, environment
and sustainable development. In Proceedings of
ACKNOWLEDGMENT
the 3rd International Conference on Sustainable
Energy and Environmental Protection (SEEP
A special thanks to my spouse Kawthar Abdelhai
2009), Paper No.1011, Dublin, Republic of Ire-
Ali for her support and her unwavering faith in me.
land, 12-15 August 2009.
Her intelligence, humour, spontaneity, curiosity
and wisdom added to this article. Abdeen, M. O. (2009c). Energy use and environ-
mental: Impacts: A general review. Journal of
Renewable and Sustainable Energy, 1(5), 1–29.
REFERENCES
Abdeen, M. O. (2009d). Energy use, environ-
Abdeen, M. O. (2008a). Renewable building ment and sustainable development. In Mancuso,
energy systems and passive human comfort solu- R. T. (Ed.), Environmental cost management
tions. Renewable & Sustainable Energy Reviews, (pp. 129–166). New York, NY: NOVA Science
12(6), 1562–1587. doi:10.1016/j.rser.2006.07.010 Publishers.
Abdeen, M. O. (2008b). People, power and pollu- Aroyeun, S. O. (2009). Reduction of aflatoxin
tion. Renewable & Sustainable Energy Reviews, B1 and Ochratoxin A in cocoa beans infected
12(7), 1864–1889. doi:10.1016/j.rser.2006.10.004 with Aspergillus via Ergosterol Value. World
Review of Science. Technology and Sustain-
Abdeen, M. O. (2008c). Energy, environment able Development, 6(1), 75–90. doi:10.1504/
and sustainable development. Renewable & WRSTSD.2009.022459
Sustainable Energy Reviews, 12(9), 2265–2300.
doi:10.1016/j.rser.2007.05.001 ASHRAE. (2005). Commercial/institutional
ground source heat pump engineering manual.
Abdeen, M. O. (2008d). Focus on low carbon Atlanta, GA: American Society of Heating, Re-
technologies: The positive solution. Renewable & frigeration and Air-conditioning Engineers, Inc.
Sustainable Energy Reviews, 12(9), 2331–2357.
doi:10.1016/j.rser.2007.04.015 Barton, A. L. (2007). Focus on sustainable devel-
opment research advances (pp. 189–205). New
Abdeen, M. O. (2008e). Development of inte- York, NY: NOVA Science Publishers, Inc.
grated bioenergy for improvement of quality of
life of poor people in developing countries. In Bos, E., My, T., Vu, E., & Bulatao, R. (1994). World
Magnusson, F. L., & Bengtsson, O. W. (Eds.), population projection: 1994-95. Baltimore, MD
Energy in Europe: Economics, policy and strategy & London, UK: John Hopkins University Press,
(pp. 341–373). New York, NY: NOVA Science World Bank edition.
Publishers. Boulet, T. (1987). Controlling air movement: A
manual for architects and builders (pp. 85–138).
New York, NY: McGraw-Hill.
128
Renewable Energy and Sustainable Development
Brain, G., & Mark, S. (2007). Garbage in, energy Energy use in offices (EUO). (2000). Energy
out: Landfill gas opportunities for CHP projects. consumption guides 19 (ECG019) Energy ef-
Cogeneration and On-Site Power, 8(5), 37–45. ficiency best practice programme. London, UK:
UK Government.
BS 5454. (1989). Storage and exhibition archive
documents. British Standard Institute. London. Erlich, P. (1991). Forward facing up to climate
change. In Wyman, R. C. (Ed.), Global climate
Commission of the European Communities (CEC).
change and life on Earth. London, UK: Chapman
(2000). Towards a European strategy for the se-
and Hall.
curity of energy supply. (Green Paper, Brussels,
29 November 2000 COM 769). Erreygers, G. (1996). Sustainability and stability
in a classical model of production. In Faucheux,
Davenport, A., Galbraith, A., Stockdale, M., Wil-
S., Pearce, D., & Proops, J. (Eds.), Models of
son, S., Mitchell, R., & Hewitson, R. … Woodley,
sustainable development. Cheltenham.
M. (2005). Building and land management for
law students, 5th edition. Oxford: UK. Farm Energy Centre (EFC). (2000). Helping
agriculture and horticulture through technology,
David, E. (2003). Sustainable energy: Choices,
energy efficiency and environmental protection.
problems and opportunities. The Royal Society
Warwickshire.
of Chemistry, 19, 19–47.
Hindsworth, M. F., & Lang, T. B. (2009). Com-
DeCarlo, F., & Bassano, A. (2010). Freshwa-
munity participation and empowerment (pp.
ter ecosystems and aquaculture research (pp.
235–261). New York, NY: NOVA Science Pub-
63–105). New York, NY: Nova Science Publish-
lishers, Inc.
ers, Inc.
Huttrer, G. (2001). The status of world geothermal
DEFRA. (2006). Energy resources. Doncaster,
power generation 1995-2000. Geothermics, 30,
UK: Sustainable Development and Environment.
1–27. doi:10.1016/S0375-6505(00)00042-0
Department of Energy (DoE). (2009). Annual
IEA. (2008). Combined heat and power: Evalu-
energy demand. USA.
ating the benefits of greater global investment.
DETR. (1994). Best practice programme-intro-
IPCC. (2001). Climate change 2001 (3 volumes).
duction to energy efficiency in buildings. Don-
United Nations International Panel on Climate
caster, UK. UK: Department of the Environment,
Change. UK: Cambridge University Press.
Transport and the Regions.
Jeremy, L. (2005). The energy crisis, global
Duffie, J. A., & Beckman, W. A. (1980). Solar
warming and the role of renewables. Renewable
engineering of thermal processes. New York, NY:
Energy World, 8(2).
J. Wiley and Sons.
Kowalczyk, L., & Piotrowski, J. (2009). Energy
EIBI (Energy in Building and Industry). (1999).
costs, international developments and new direc-
Constructive thoughts on efficiency, building
tions (pp. 1–37). New York, NY: NOVA Science
regulations, inside committee limited. Inside
Publishers, Inc.
Energy: Magazine for Energy Professional, 13-
14. UK: KOPASS. Lam, J. C. (2000). Shading effects due to nearby
buildings and energy implications. Energy Con-
version and Management, 47(7), 47–59.
129
Renewable Energy and Sustainable Development
Lazzarin, R., D’Ascanio, A., & Gaspaella, A. Omer, A. M. (2009b, October-November). Prin-
(2002). Utilisation of a green roof in reducing ciple of low energy building design: Heating,
the cooling load of a new industrial building. In ventilation and air conditioning. [Mumbai, India.].
Proceedings of the 1st International Conference Cooling India: India’s Premier Magazine on the
on Sustainable Energy Technologies (SET), (pp. Cooling Industry, 5(4), 26–46.
32-37). Porto, Portugal. 12-14 June 2002.
Omer, A. M. (2010a). Development of sustainable
Levine, M., & Hirose, M. (2005). Energy efficiency energy research and applications. In Lee, W. H.,
improvement utilising high technology: An as- & Cho, V. G. (Eds.), Handbook of sustainable
sessment of energy use in industry and buildings. energy (pp. 385–418). New York, NY: NOVA
Report and Case Studies. London, UK: World Science Publishers, Inc.
Energy Council.
Omer, A. M. (2010b). The crux of matter: Water in
Lund, J. W., Freeston, D. H., & Boyd, T. L. (2005). the Republic of the Sudan (pp. 1–50). New York,
Direct application of geothermal energy: 2005 NY: NOVA Science Publishers, Inc.
Worldwide Review. Geothermics, 34, 691–727.
Parikn, J., Smith, K., & Laxmi, V. (1999). In-
doi:10.1016/j.geothermics.2005.09.003
doors air pollution: A reflection on gender bias.
Lysen, E. H. (1983). Introduction to wind energy Economic and Political Weekly, 34(9).
(pp. 15–50). The Netherlands: CWD.
Paul, F. (2001). Indoor hydroponics: A guide to
Meffe, S., Perkson, A., & Trass, O. (1996). Coal understanding and maintaining a hydroponic
beneficiation and organic sulphur removal. Fuel, nutrient solution. UK. 2001.
75, 25–30. doi:10.1016/0016-2361(95)00171-9
Randal, G., & Goyal, R. (1998). Greenhouse
Mortal, A. (2002). Study of solar powered heat technology. New Delhi, India: Narosa Publish-
pump for small spaces. Portugal. ing House.
Omer, A. M. (2008a). Green energies and the REN21. (2007). Renewables global status report.
environment. Renewable & Sustainable En- Retrieved from www.ren21.net
ergy Reviews, 12, 1789–1821. doi:10.1016/j.
Reddy, A., Williams, R., & Johansson, T.
rser.2006.05.009
(2007). Energy after Rio: Prospects and chal-
Omer, A. M. (2008b). On the wind energy re- lenges. United Nations Development Programme
sources of Sudan. Renewable & Sustainable (UNDP). Retrieved from http://www.undp.org/
Energy Reviews, 12(8), 2117–2139. doi:10.1016/j. seed/energy/-exec-en.html
rser.2006.10.010
Roriz, L. (2001). Determining the potential energy
Omer, A. M. (2008c). Energy demand for heating and environmental effects reduction of air con-
and cooling equipment systems and technology ditioning systems. Commission of the European
advancements. In White, J. R., & Robinson, W. H. Communities DG TREN.
(Eds.), Natural resources: Economics, manage-
Shao, S. (2002). Thermodynamic analysis on heat
ment and policy (pp. 131–165).
pumps with economiser for cold regions. China.
Omer, A. M. (2009a). Energy use and environ-
mental impacts: A general review. Renewable and
Sustainable Energy, 1, 1–29.
130
Renewable Energy and Sustainable Development
UNEP. (2003). Handbook for the international United Nations Economic Commission for Africa
treaties for the protection of the ozone layer. (UNECA). (2003). Making science and technol-
Nairobi, Kenya: United Nations Environment ogy work for the poor and for SD in Africa. Paper
Programme. prepared by the SD Division with the assistance
of a senior international consultant, Akin Adubifa,
UNIDO. (2007). Changing courses sustainable
January.
industrial development, as a response to agenda
21. Vienna. United Nations Economic Commission for Africa
(UNECA). (2003a). The state of food security in
United Nations. (UN). (2001). World urbanisation
Africa. Progress Report of the 3rd Meeting of the
prospect: The 1999 revision. New York, NY: The
Committee on Sustainable Development, 7–10
United Nations Population Division.
October, Addis Ababa, Ethiopia.
United Nations. (UN). (2002a). Science and
United Nations Economic Commission for Europe
technology as a foundation for SD. Summary by
(UNECE). (2004). Note by the ECE Secretariat,
the Scientific and Technological Community for
Steering Group on Sustainable Development.
the Multi-Stakeholder Dialogue Segment of the
Second Meeting of the 2003/2004 Bureaus,
fourth session of the Commission on SD acting as
Conference of European Statisticians, Statistical
the preparatory committee for the World Summit
Commission, Geneva, Switzerland.
on SD. Note by the Secretary-General. Commis-
sion on SD acting as the preparatory committee United Nations Under-Secretary General and the
for the World Summit on SD Fourth Preparatory United Nations Environment Programme (UNEP).
Session, 27 May–7 June, 2002. (2000). Overview: Outlook and recommendations.
Global Environment Outlook. Retrieved from
United Nations. (UN). (2002b). Global challenge
http://grid.cr.usgs.gov/geo2000/ov-e/0012.htm
global opportunity: Trends in sustainable devel-
opment. Department of Economics and Social, Vilinac, D. (2008). Plant medicines: An herbalist’s
World Summit on Sustainable Development, perspective. World Review of Science. Technology
Johannesburg, SA. and Sustainable Development, 3(2), 140–151.
doi:10.1504/WRSTSD.2008.018556
United Nations. (UN). (2002c). Implementation of
the United Nations millennium declaration. Report White, J. R., & Robinson, W. H. (2008). Natural
of the Secretary-General, United Nations General resources: Economics, management and policy
Assembly. Retrieved from http://www.un.org (pp. 1–49). New York, NY: NOVA Science Pub-
lishers, Inc.
United Nations Economic Commission for Africa
(UNECA). (2002). Address by Josué Dioné: Witte, H. (2002). Comparison of design and op-
Science and technology policies for sustainable eration of a commercial UK ground source heat
development and Africa’s global inclusion. Sus- pump project. Groenholland BV.
tainable Development Division ATPS Conference,
World Bank. (WB). (2003a). World development
11 November, Abuja, Nigeria.
report. New York, NY: Oxford University Press.
131
Renewable Energy and Sustainable Development
World Bank. (WB). (2003b). Global economic Abdeen, M. O. (2008b). People, power and pollu-
prospects: Realizing the development promise of tion, Renewable and Sustainable Energy Reviews,
the Doha agenda. Washington, DC: The World Vol.12 No.7, p.1864-1889, United Kingdom,
Bank, World Bank, The International Bank for September 2008.
Reconstruction and Development.
Abdeen, M. O. (2008c). Energy, environment and
World Bank. (WB). (2004). World developments sustainable development, Renewable and Sustain-
report 2004: Making services work for poor able Energy Reviews, Vol.12, No.9, p.2265-2300,
people. Washington, DC: World Bank. United Kingdom, December 2008.
World Bank. (WB). (2006). Sustainable de- Abdeen, M. O. (2008d). Focus on low carbon
velopment in the 21st century. Retrieved from technologies: the positive solution, Renewable
http://lnweb18.worldbank.org/ESSD/sdvext. and Sustainable Energy Reviews, Vol.12, No.9,
nsf/43ByDocName/SustainableDevelopmentint p.2331-2357, United Kingdom, December 2008.
World Bank. (WB). (2007). World Bank sustain- Abdeen, M. O. (2008e). Development of integrated
able development reference guide. Retrieved from bioenergy for improvement of quality of life of
http://www.WorldBankSustainableDevelopmen- poor people in developing countries. In Magnus-
tReferenceGuide son, F. L., & Bengtsson, O. W. (Eds.), Energy
in Europe: Economics, Policy and Strategy- IB
World Energy Council (WEC). (2009). The world
(pp. 341–373). New York, USA: NOVA Science
energy demand in 2020.
Publishers, Inc.
World Energy Outlook (WEO). (1995). Inter-
Abdeen, M. O. (2009a). Environmental and
national Energy Agency. Paris, France: OECD
socio-economic aspect of possible development
Publications.
in renewable energy use, In: Proceedings of the
WRI (World Resource Institute). (2004). World 4th International Symposium on Environment,
resources: A guide to the global environment- Athens, Greece, 21-24 May 2009.
People and the environment. Washington, DC.
Abdeen, M. O. (2009b). Energy use, environment
Yadav, I., & Chauadhari, M. (1997). Progressive and sustainable development, In: Proceedings of
floriculture (pp. 1–5). Bangalore, India: The the 3rd International Conference on Sustainable
House of Sarpan. Energy and Environmental Protection (SEEP
2009), Paper No.1011, Dublin, Republic of Ire-
Zuatori, A. (2005). An overview on the national
land, 12-15 August 2009.
strategy for improving the efficiency of energy
use. Jordanian Energy Abstracts, 9(1), 31–32. Abdeen, M. O. (2009c). Energy use and envi-
ronmental: impacts: a general review, Journal
of Renewable and Sustainable Energy, Vol.1,
No.053101, p.1-29, United State of America,
ADDITIONAL READING
September 2009.
Abdeen, M. O. (2008a). Renewable building en- Abdeen, M. O. (2009d). Energy use, environ-
ergy systems and passive human comfort solutions, ment and sustainable development. In Mancuso,
Renewable and Sustainable Energy Reviews, R. T. (Ed.), Environmental Cost Management
Vol.12, No.6, p.1562-1587, United Kingdom, (pp. 129–166). New York, USA: NOVA Science
August 2008. Publishers, Inc.
132
Renewable Energy and Sustainable Development
Aroyeun, S. O. (2009). Reduction of aflatoxin Leszek Kowalczyk and Jakub Piotrowski. (2009).
B1 and Ochratoxin A in cocoa beans infected Energy Costs, International Developments and
with Aspergillus via Ergosterol Value. World New Directions (pp. 1–37). New York, USA:
Review of Science. Technology and Sustain- NOVA Science Publishers, Inc.
able Development, 6(1), 75–90. doi:10.1504/
Lund, J. W., Freeston, D. H., & Boyd, T. L. (2005).
WRSTSD.2009.022459
Direct application of geothermal energy: 2005
Barton, A. L. (2007). Focus on Sustainable De- Worldwide Review. Geothermics, 34, 691–727.
velopment Research Advances (pp. 189–205). doi:10.1016/j.geothermics.2005.09.003
New York, USA: NOVA Science Publishers, Inc.
Mildred, F., & Trevor, B. (2009). Lang (pp.
Brain, G., & Mark, S. (2007). Garbage in, energy 235–261). New York, USA: Community Par-
out: landfill gas opportunities for CHP projects. ticipation and Empowerment. NOVA Science
Cogeneration and On-Site Power, 8(5), 37–45. Publishers, Inc.
Commission of the European Communities (CEC). Omer, A. M. (2010). The crux of matter: water in
(2000). Towards a European strategy for the se- the Republic of the Sudan. 2010 NOVA Science
curity of energy supply. Green Paper, Brussels, Publishers, Inc., p.1-50, New York, USA.
29 November 2000 COM (2000) 769.
Paul, F. (2001). Indoor hydroponics: A guide
Farm Energy Centre (EFC). (2000). Helping to understanding and maintaining a hydroponic
agriculture and horticulture through technology, nutrient solution. UK. 2001.
energy efficiency and environmental protection.
REN21. (2007). Renewables (2007) global status
Warwickshire. 2000.
report. www.ren21.net.
Felice DeCarlo and Alessio Bassano. (2010).
Reddy, A., Williams, R., & Johansson, T. (2007).
Freshwater Ecosystems and Aquaculture Research
Energy after Rio: prospects and challenges. United
(pp. 63–105). New York, USA: Nova Science
Nations Development Programme (UNDP). http://
Publishers, Inc.
www.undp.org/seed/energy/-exec-en.html. 2007.
Huttrer, G. (2001). The status of world geothermal
Roriz, L. (2001). Determining the potential energy
power generation 1995-2000. Geothermics, 30,
and environmental effects reduction of air con-
1–27. doi:10.1016/S0375-6505(00)00042-0
ditioning systems. Commission of the European
IEA. (2008). Combined heat and power: evaluating Communities DG TREN.
the benefits of greater global investment. 2008.
Shao, S. (2002). Thermodynamic analysis on heat
Jeremy, L. (2005). The energy crisis, global warm- pumps with economiser for cold regions. China.
ing and the role of renewables. Renewable Energy
UNEP. (2003). Handbook for the International
World 2005; 8 (2).
Treaties for the Protection of the Ozone Layer.
Lam, J. C. (2000). Shading effects due to nearby Nairobi, Kenya: United Nations Environment
buildings and energy implications. Energy Con- Programme.
version and Management, 47(7), 47–59.
United Nations. (UN). (2001). World Urbanisa-
tion Prospect: The 1999 Revision. New York.
The United Nations Population Division. 2001.
133
Renewable Energy and Sustainable Development
United Nations. (UN). (2002a) Science and Tech- United Nations Under-Secretary General and the
nology as a Foundation for SD. Summary by the United Nations Environment Programme (UNEP).
Scientific and Technological Community for the (2000) ‘Overview: outlook and recommenda-
Multi-Stakeholder Dialogue Segment of the fourth tions’, Global Environment Outlook, http://grid.
session of the Commission on SD acting as the cr.usgs.gov/geo2000/ov-e/0012.htm, Earthscan,
preparatory committee for the World Summit on 1999, London.
SD. Note by the Secretary-General. Commission
Vilinac, D. (2008). Plant medicines: an herbalist’s
on SD acting as the preparatory committee for the
perspective. World Review of Science. Technology
World Summit on SD Fourth Preparatory Session,
and Sustainable Development, 5(2), 140–151.
27 May–7 June, 2002.
doi:10.1504/WRSTSD.2008.018556
United Nations. (UN) (2002b) ‘Global challenge
White, J. R., & Robinson, W. H. (2008). Natural
global opportunity: trends in sustainable devel-
Resources: Economics, Management and Policy
opment’, Department of Economics and Social,
(pp. 1–49). New York, USA: NOVA Science
World Summit on Sustainable Development,
Publishers, Inc.
Johannesburg, SA.
Witte, H. (2002). Comparison of design and op-
United Nations Economic Commission for Africa
eration of a commercial UK ground source heat
(UNECA). (2002) ‘Address by Josué Dioné:
pump project. Groenholland BV.
science and technology policies for sustainable
development and Africa’s global inclusion’, Sus- World Bank. (WB) (2003a) World Development
tainable Development Division ATPS Conference, Report, Oxford University Press, New York.
11 November, Abuja, Nigeria.
World Bank. (WB). (2003b) ‘Global economic
United Nations Economic Commission for Africa prospects: realizing the development promise of
(UNECA) (2003) Making Science and Technology the Doha agenda’, The International Bank for
Work for the Poor and for SD in Africa, Paper Reconstruction and Development, The World
prepared by the SD Division with the assistance Bank, World Bank: Washington, DC.
of a senior international consultant, Akin Adubifa,
World Bank. (WB). (2004) World Development
January.
Report 2004: Making Services Work for Poor
United Nations Economic Commission for Africa People, World Bank: Washington, DC.
(UNECA). (2003a) ‘The state of food security in
World Bank. (WB). (2006). Sustainable De-
Africa’, Progress Report of the 3rd Meeting of the
velopment in the 21st Century.http://lnweb18.
Committee on Sustainable Development, 7–10
worldbank.org/ESSD/sdvext.nsf/43ByDocName/
October, Addis Ababa, Ethiopia.
SustainableDevelopmentint
United Nations Economic Commission for Europe
World Bank. (WB). (2007). World Bank Sus-
(UNECE). (2004) ‘Note by the ECE Secretariat,
tainable Development Reference Guide, http://
Steering Group on Sustainable Development.
www. WorldBankSustainableDevelopmentRef-
Second Meeting of the 2003/2004 Bureau’,
erenceGuide.
Conference of European Statisticians, Statistical
Commission, Geneva, Switzerland.
134
Renewable Energy and Sustainable Development
KEY TERMS AND DEFINITIONS wind turbines to make electricity, wind mills for
mechanical power, wind pumps for pumping water
Renewable Energy: Renewable energy is or drainage, or sails to propel ships.
energy generated from natural resources such as Hydropower: Hydropower, hydraulic power
sunlight, wind, rain, tides, and geothermal heat, or waterpower is power that is derived from the
which are renewable (naturally replenished). force or energy of moving water, which may be
Energy obtained from sources that are essentially harnessed for useful purposes. Hydropower is
inexhaustible (unlike, for example the fossil fuels, using water to power machinery or make elec-
of which there is a finite supply). Energy sources tricity. Water constantly moves through a vast
that are, within a short time frame relative to the global cycle, evaporating from lakes and oceans,
Earth’s natural cycles, sustainable, and include forming clouds, precipitating as rain or snow, and
non-carbon technologies such as solar energy, then flowing back down to the ocean.
hydropower, and wind, as well as carbon-neutral Geothermal Energy: Geothermal power
technologies. (from the Greek roots geo, meaning earth, and
Solar Energy: Energy from the sun that is thermos, meaning heat) is power extracted from
converted into thermal or electrical energy; “the heat stored in the earth. This geothermal energy
amount of energy falling on the earth is given by originates from the original formation of the
the solar constant, but very little use has been made planet, from radioactive decay of minerals, and
of solar energy”. Energy derived ultimately from from solar energy absorbed at the surface. Heat
the sun. It can be divided into direct and indirect transferred from the earth’s molten core to under-
categories. Most energy sources on Earth are forms ground deposits of dry steam (steam with no water
of indirect solar energy, although we usually do droplets), wet steam (a mixture of steam and water
not think of them in that way. Solar energy uses droplets), hot water, or rocks lying fairly close to
semiconductor material to convert sunlight into the earth’s surface.
electric currents. Although solar energy only pro- Resource Management: Efficient incident
vides 0.15% of the world’s power and less than management requires a system for identifying
1% of US energy, experts believe that sunlight available resources at all jurisdictional levels to
has the potential to supply 5,000 times, as much enable timely and unimpeded access to resources
energy as the world currently consumes. needed to prepare for, respond to, or recover from
Biomass Energy: The energy embodied in an incident. Resource management is the efficient
organic matter (“biomass”) that is released when and effective deployment for an organization’s
chemical bonds are broken by microbial digestion, resources when they are needed. Such resources
combustion, or decomposition. Biofuels are a wide may include financial resources, inventory, hu-
range of fuels, which are in some way derived man skills, production resources, or information
from biomass. The term covers solid biomass, technology (IT).
liquid fuels and various biogases. Biofuels are Sustainable Development: Development,
gaining increased public and scientific attention, which seeks to produce sustainable economic
driven by factors such as oil price spikes and the growth while ensuring future generations’ ability
need for increased energy security. to do the same by not exceeding the regenerative
Wind Energy: Kinetic energy present in wind capacity of the nature. In other words, it’s trying
motion that can be converted to mechanical en- to protect the environment. A process of change
ergy for driving pumps, mills, and electric power in which the resources consumed (both social and
generators. Wind power is the conversion of wind ecological) are not depleted to the extent that they
energy into a useful form of energy, such as using cannot be replicated. Environmentally friendly
135
Renewable Energy and Sustainable Development
forms of economic growth activities (agriculture, life forms within the environment, and includes
logging, manufacturing, etc.) that allow the con- all variables that comprise the Earth’s biosphere.
tinued production of a commodity without damage Greenhouse Gases: Greenhouse gases are
to the ecosystem (soil, water supplies, biodiversity gases in an atmosphere that absorb and emit
or other surrounding resources). radiation within the thermal infrared range. This
Environment: The natural environment, process is the fundamental cause of the greenhouse
commonly referred to simply as the environment, effect. The main greenhouse gases in the Earth’s
encompasses all living and non-living things oc- atmosphere are water vapour, carbon dioxide,
curring naturally on Earth or some region thereof. methane, nitrous oxide, and ozone. Changes in
The biophysical environment is the symbiosis be- the concentration of certain greenhouse gases,
tween the physical environment and the biological due to human activity such as fossil fuel burning,
increase the risk of global climate change.
136
137
Chapter 4
Demand-Side Response
Smart Grid Technique for
Optimized Energy Use
Fouad Kamel
University of Southern Queensland, Australia
Marwan Marwan
Queensland University of Technology, Australia
ABSTRACT
The chapter describes a dynamic smart grid concept that enables electricity end-users to be acting on
controlling, shifting, or curtailing own demand to avoid peak-demand conditions according to informa-
tion received about electricity market conditions over the Internet. Computer-controlled switches are
used to give users the ability to control and curtail demand on a user’s premises as necessary, follow-
ing a preset user’s preferences. The computerized switching provides the ability to accommodate local
renewable energy sources as available. The concept offers further the ability to integrate charging
electrical vehicles during off-peak periods, helping thus substantially improving the utilization of the
whole electricity system. The approach is pursuing improved use of electrical energy associated with
improved energy management, reduced electricity prices and reduced pollution caused by excessive
use of combustion engine in transport. The technique is inherently restricted to take effect in frame of
energy tariff regimes based on real-time price made to encourage and reward conscious users being
proactively participating in holistic energy management strategies.
DOI: 10.4018/978-1-61350-138-2.ch004
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Demand-Side Response Smart Grid Technique for Optimized Energy Use
138
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 1. Impact of the plant capacity factor on the cost of the produced energy (Kamel, 2009b)
MWh, while operated continuously for 24 hour tomers from their normal consumption patterns
will produce energy at a cost of $25/MWh. in response to changes in the price of electricity
Figure 1 depicts the importance of operating over time. Parvania & Fotuhi-Firuzabad (2010)
power plants and electrical network infrastructure describe DSR as tariff or program established
at elevated plant capacity factor, close to the to motivate change in electric consumption by
unity, in order to verify best economic perfor- end-use customers in response to change in the
mance. Various efforts on avoiding peak demands price of electricity over time. Further on, DSR
on the electrical network are mainly aiming at programs provide means for utilities to reduce the
leveling demand throughout the year in order to power consumption and save energy, maximize
achieve as high plant capacity factors as possible utilizing the current capacity of the distribution
for all electrical power components. Additionally, system infrastructure, reducing or eliminating
leveling demands is pursued to avert or delay the the need for building new lines and expanding
urgency to expand generation capacity and net- the system as described by Dam, Mohagheghi &
work infrastructure to cope with consistently Stoupis (2008).
rising peak demands. Important methods to avert Vos (2009) described demand response as an
or delay peak demands are represented in demand- integral part of the smart grid, is a cost effective,
side response activities, both on the supplier and rapidly deployed resource that provides benefits
on the user’s side. to utilities and customers. Some advantages of
DSR according to Greening (2010) are: increased
Demand Side Response and economic efficiency of the electricity infrastruc-
Smart Grid Technologies ture, enhanced reliability of the system, relief of
power congestions and transmission constraints,
Demand Side Response (DSR) reduced energy prices and mitigated potential
market power. Further on, demand response can
Demand Side Response (DSR), as described by help reduce peak demands and therefore reduce
Albadi & El Saadany (2007) can be defined as spot price volatility as illustrated by Nguyen
the changes in electricity usage by end-use cus- (2010). In addition, Hyung Seon & Thomas (2008)
139
Demand-Side Response Smart Grid Technique for Optimized Energy Use
140
Demand-Side Response Smart Grid Technique for Optimized Energy Use
141
Demand-Side Response Smart Grid Technique for Optimized Energy Use
electricity bill from the utility. In California the Covino (2003) emergency demand response pro-
incentive of I/C program was $700/MWh/month gram provide participant with significant intensive
in 2001 as reported in (Aalami H.A, Moghaddam to reduce load. To participate on this program, all
M.P & Yousefi G.R, 2009). customers are expected to reduce their energy
consumption during the events. Tyagi & Black
Emergency Demand (2010) described this program will determine
Response Program which houses must be included in the event to
minimize cost and disruption, while alleviating
Emergency demand response program is energy- the overload condition. When asked to curtail,
efficient program that provides incentives to and verified to have performed, the resource is
customers who can reduce electricity usage for a paid the higher of $500/MWh (Rahimi & Ipakchi,
certain time; this is usually conducted at the time 2010). In New York, emergency demand response
of limited availability of electricity. According to program allows participant to be paid for reducing
142
Demand-Side Response Smart Grid Technique for Optimized Energy Use
their energy consumption upon notice from the incentives and costumer benefits associated with
New York Independent System Operator that an the DSR program used.
operating reserves deficiency or major emergency The change in energy consumption ∆d(t) at
exists (Lawrence & Neenan, 2003). the time t when the user changes demand from
d(t) to do(t) is:
Time of Use Program
∆d(t)= d(t)- do(t) (2)
According to Na & Ji-Lai (2006) Time of Use is
one of the important demand side management For participating in the DSR program, the total
methods, TOU demand side will response to the incentives P∆d(t), when A(t) is paid as incentive
price and will change the shape of the demand to the costumer at the time t for each kWh load
curve. Further on, Time of Use rate is the most reduction, can be calculated as the following:
obvious strategy developed for the management of
the peak demand in the world, which is designed P(∆d(t))=A(t).∆d(t) (3)
to encourage the consumer to modify the pattern
of electricity usage (Wen-Chen, Yi-Ping & Tzu- The total penalty PEN·Δd(t) when the consumer
Hao, 2007). For applying this program, the utility does not commit to the obligations as a member of
does not provide reward or penalty to customer. To the DSR program, pen(t) is the penalty per kWh
participate, all customers are required to remove at the time t and D(t) is the DSR program contract
their energy consumption during peak session to level of consumption in kWh, will be accounted
off-peak session as soon as their receipt informa- as the following:
tion from the utility. Kirschen suggested (2003)
in this type of contract, the rate is fixed for the PEN·(Δd(t))=pen(t).{D(t)-[d(t)-do(t)]} (4)
duration of the contract but depends on the time
of the day. As compared to the flat rate contract, The total customer’s benefit S a member of
some of the risk is shifted from the retailer to the the DSR program can make at a certain time t is
consumer because the consumer has an incentive to made up of income and expenditures. The contract
consume during periods when the rates are lower. could be mentioning a benefit from joining the
program so that e.g.B(d(t)) is the customer income
Calculation of Energy Consumption during that period from the use of d(t) kWh, at
for Maximized Benefit the same time the customer could be receiving
additional incentives P(∆d(t)) as described in
Different models are used in Demand Side Re- Equation(2). The cost of the consumed electricity
sponse program planning as described by the and any penalty for not following the program,
Federal Energy Regulatory Commission (2006). In if applicable, will be deducted from the income
the report of the strategic plan of the International as the following:
Energy Agency (2010), DSR is divided into two
basic categories, namely, the time based program S=B(d(t))+P(Δd(t))-d(t).r(t)-PEN(Δd(t))
and the incentives based program. In Aalami H.A, (5)
Moghaddam M.P & Yousefi G.R (2008) the fol-
lowing approach is used to calculate the savings where,r(t) is the rate the customer pays per kWh
in electricity expenditure based on price, demand, electricity at that time.
143
Demand-Side Response Smart Grid Technique for Optimized Energy Use
144
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 3. Smart grid network diagram (Energy Efficiency and Conservation Authority, 2009)
grid. In addition, there are social, environmental assists electricity end-users to be shifting loads
and economic advantages in switching to elec- around the clock averting peak-demand periods
tricity vehicles as described by Anna Cain, Iain and making use of on-site renewable energy
MacGill & Bruce (2010). sources as appropriate. This shall help users to be
engaged in mitigating peak demands on the elec-
The User’s DSR Concept tricity network. The proposed concept comprises
a technical set-up of a programmable internet
This concept helps developing the technical tools relay, a router, solid state switches in addition to
to be independently implemented and managed by the suitable software to control electricity demand,
end-users to enable control energy consumption Figure 5. The software’s on appropriate multime-
accustomed to user’s preferences. The proposed dia tool (CD Rom) offers users optimized control
DSR Smart-Grid scheme is targeting flattening of energy consumption.
load profiles by averting periods of peak de- The concept enables commercial and indus-
mands helping thus redistributing the load profile trial customers on fluctuating energy prices to be
throughout the day/year. The scheme helps electri- achieving immediate financial savings. For resi-
cal generating plants and electrical infrastructure dential customers on flat-rate tariffs, in contrast,
to be operated 24 hour/day achieving thus higher users are gaining financial benefits from reducing
utilization factor, enhanced efficiency and con- energy consumptions at peak-demand periods.
siderably reduced energy price. The presented Residential customers on different tariffs, where
user’s side DSR concept is providing the needed energy price differs with day time and network
balance in the electrical system to complement conditions (e.g. night tariffs), they are gaining
the efforts undertaken by suppliers to mitigate financial benefits also from shifting loads from
peak demands and improve supply reliability and day- to night-times, when electricity is cheaper.
stability, Figure 4. The scheme uses a router and a programmable
The concept is presenting a low-cost DSR internet relay and solid-state switches to control
technique implemented at user’s premises, which electrical demand at the user’s premises. The relay
145
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 4. The proposed user’s smart grid demand-side-response scheme (DSR) balancing conventional
supplier’s smart grid DSR programs
146
Demand-Side Response Smart Grid Technique for Optimized Energy Use
147
Demand-Side Response Smart Grid Technique for Optimized Energy Use
major peak demands in the morning and evening. the year corresponding to 8760 h/year, a utiliza-
It is to be noted that electricity prices in Figure 8 tion factor of 1. Those power stations are provid-
are wholesale regional references prices (RRP) ing the most economic operation at the least
i.e. prices traded to electricity suppliers. Electrical possible energy price. Any loads exceeding the
energy at this stage still needs to be transmitted base load are usually covered by other power
to the different localities then distributed to end- plants operated for shorter periods of time at
users. End-users could be industrial, commercial utilization factor lesser than 1, thus generating
or residential. Usually industrial and commercial energy at high prices. Accordingly, the intermittent
end-users are contracting electricity suppliers on operation of the expensive to-run power plants
special agreements to provide the service satis- makes their operation even more expensive.
fying their requirements. For most residential Figure 10 illustrates the occurrence of the
electricity customers, electricity pricing typically regional reference wholesale price RRP in
follow one set price “flat-rate”. Residential users Queensland during the year 2008; extracted from
are often also offered a night tariff, where elec- the Australia Energy Market Operator (2009). The
tricity prices are substantially reduced. The night figure indicates mainly that low-priced supplies
tariff corresponds off-peak times. are taking place at very high occurrences of more
Figure 9 illustrates the frequency of electric- than 80% a year, while high prices happen at
ity demand supplied in Queensland during the lower occurrences. For instance, prices around
year 2008 as extracted from data of the Australian AUD $20/MWh are occurring at frequencies of
Energy Market Operator (2009). The figure indi- about 80%, while prices of over $50/MWh have
cates mainly the fact that the higher the load above occurrences of less than 10%.
the base load the lesser likely the extent of their Based on data of 31 December 2008,
duration will be. Base load power stations are Queensland total electricity generating capacity
those operated twenty four hours/day throughout was 12487 MW; coal-fired power stations pro-
148
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 8. Wholesale electricity price in AUD $/MWh and demand in MW for a typical day in Queensland
on 5th May 2009(Australia Energy Market Operator, 2009)
Figure 9. Occurrence of electrical energy demand Queensland during 2008. Peak demand 8413 MW,
base-load 4100 MW and total supplied electrical energy 52.18 TWh (Kamel, 2009b)
149
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 10. Electricity wholesale price RRP in Queensland in 2008 (Kamel, 2009b).
vided 70% of this total capacity, while gas-fired network is carrying electricity from the transmis-
electricity accounted for 17% and renewable sion system to consumers. In Queensland, EN-
energy accounted for around 5% as stated by the ERGEX and ERGON energy are purchasing
Department of Employment Economic Develop- electrical energy from the Energy Market and
ment and Innovation (2010). These power gen- distributing it to the customer. ERGON e.g. pro-
erations are used to provide electrical energy for vides energy at several tariff options to end users.
all consumers in the Queensland area: residential, For example, Tariff 11 for all domestic consump-
commercial and industrial. However, the amounts tion 18.84 ¢/kWh, while the night rate Tariff 31
of energy produced from various generators de- for all consumption 7.7 ¢/kWh and the economy
pend on market demand, price and availability of Tariff 33 for all consumption 11.32 ¢/kWh (Ergon
sources. Figure 11 illustrates electricity generation Energy, 2010).
in Queensland according to the Department of According to Queensland Competition Au-
Employment Economic Development and Innova- thority (2000), the total energy consumption in
tion (2010). Australian grew at an annual rate of 2.6% for
Most of the power stations are directly con- the 25 years to 1997/1998. In the 2007- 2008
nected to the transmission system. The Queensland period according to Department of Employment
electricity transmission system is provided by Economic Development and Innovation (2009a)
Powerlink, licensed to operate more than 12,000 the annual electricity consumption in Queensland
kilometres of Queensland high voltage transmis- has grown by over 29% or approximately 10500
sion network, transporting electricity from the GWh, making Queensland the second highest
generators to the distribution networks as by consumer of electricity in Australia. This indicated
(Department of Employment Economic Develop- that Queensland has a significantly greater number
ment and Innovation, 2009). The distribution
150
Demand-Side Response Smart Grid Technique for Optimized Energy Use
of high energy users than any other state, most of system, the Australian Government has created a
these in regional Queensland. collaborative electricity and gas industry in the
Narayan & Smith (2005) describes since the form of the Australian Electricity Market Opera-
beginning of the 1990s, Australia’s electric power tor (Australian Energy Market Operator, 2010),
industry has undergone a series of structural which commenced operation on 1 July 2009.
reforms. In Queensland, the electricity industry The AEMO is managing power flows across the
was restructured on 1 July 1998 to prepare the Australian Capital Territory, New South Wales,
industry for participating in the competitive Na- Queensland, South Australia, Victoria and Tasma-
tional Electricity Market, which is responsible for nia. Western Australia and the Northern Territory
structure, rules and regulations in the delivery of are not currently connected to this market primarily
energy to customers (Department of Employment because of their geographic distance from the rest
Economic Development and Innovation, 2009b). of the market. AEMO’s responsibilities include
The National Electricity Market Management wholesale and retail energy market operation,
Company (NEMMCO) Limited was the Whole- infrastructure and long term market planning
sale Market and Power System Operator for the demand forecasting data and scenario analysis as
Australian NEM. NEMMCO was established in described in (Australian Energy Market Opera-
1996 to administer and manage the NEM, develop tor, 2010). The electricity market comprises of a
the market and continually improve its efficiency wholesale sector and a competitive retail sector.
and as of 1 July 2009 was replaced by the Aus- All electricity dispatched in the market must be
tralian Energy Market Operator. traded through the central spot market. The Market
To improve governance, and enhance the reli- structure of NEMMCO / AEMO can be presented
ability and sustainability of the State’s electricity as in Figure 12.
151
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 12. The Market structure of NEMMCO/AEMO (Department of Resources Energy and Tourism,
2009)
Figure 13. Fluctuation of electricity price in Queensland (Marwan & Kamel, 2010a)
152
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Marwan & Kamel (2010a) summarized in ten scenarios have been formulated to demonstrate
Figure 13 an example of classic fluctuations in the results as presented in Figure 14 and sum-
electricity price in Queensland, from 22 May 2008 marized in Table 2.
to 22 May 2009. This graph illustrates that the Scenario 1, In this scenario users are shifting
average price during that time was in the range 375 MWh peak electricity usage occurring be-
of $50/MWh (¢5/kWh) Regional Reference tween 17:00 pm-19:00 pm towards the time pe-
Wholesale Price (RRP), however, extreme prices riod 19:00 pm-21:30 pm when energy demand
occurred exceeding $500/MWh (¢50/kWh). The and prices are low. All participants are suggested
graph indicates also that excessive demands are to set-up the electricity profile to stop chosen
occurring regularly in all states on the intercon- appliance from running during that time. For
nected power network. Fraser (2005) stated that example, kettle, washing machine and air condi-
customers, even those bound by flat-rate contracts, tion could be effectively operated at optional times
must bear the additional cost for managing the of the day. No savings in energy cost due to ap-
corresponding extreme prices. plicable day-time tariffs. However, the scheme
was still able to remove congestions out of peak
Controlled Scenario Using demand areas.
the User’s DSR Scheme Scenarios 2, Users are shifting peak demand
of 375 MWh occurring between 17:00 pm-19:00
In order to evaluate the effect of the proposed pm to the period between 21:30 pm to 23:30 pm.
scheme on electricity energy saving the electricity Achievable savings $21149 per day.
price/demand in Queensland for the period 10th- Scenario 3, Users are shifting peak demand
12nd May 2010 has been used. In the following, of 375 MWh occurring between 17:00 pm-19:00
153
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Scenario NR Time to Curtail Time to Reconnect Load to Day Tarrif Night Tarrif Saving
Load Load Curtail 18.84 ¢kWh 11.32 ¢kWh ($)
1 17.00 pm 19:00 pm to 21:30 pm 375 70650 NA NA
to
2 21:30 pm to 23:30 pm 375 70650 49501 21149
19.00 pm
3 23:30 pm to 1:00 am 375 70650 42450 28200
4 1:00 am to 3:00 am 375 70650 42450 28200
5 3:00 am to 5:30 am 375 70650 42450 28200
6 5:30 am to7:00 am 375 70650 42450 28200
7 6:30 am to 10:30 am 375 70650 67125 3525
8 10.30 am 19:30 am to 23:30 am 1730 325923 277147 48785
to
9 23:30 pm to 01:30 am 1730 325923 195836 130096
19.30 pm
10 1.30 am to 4:00 am 1730 325923 195836 130096
pm to the period between 23:30 pm to 01:00 am. pm to the period between 23:30 pm to 01:30 am.
Achievable savings $28200 per day. Achievable savings $130096 per day.
Scenario 4, Users are shifting peak demand Scenario 10, Users are shifting peak demand
of 375 MWh occurring between 17:00 pm-19:00 of 1730 MWh occurring between 10:30 am-19:30
pm to the period between 01:00 am to 03:00 am. pm to the period between 01:30 am to 04:00 am.
Achievable savings $28200 per day Achievable savings $130096 per day.
Scenario 5, Users are shifting peak demand
of 375 MWh occurring between 17:00 pm-19:00 Economic Model
pm to the period between 03:00 am to 05:30 am.
Achievable savings $28200 per day While the concept is designed to be targeting flat-
Scenario 6, Users are shifting peak demand tening the national electrical demand throughout
of 375 MWh occurring between 17:00 pm-19:00 the year the concept involves an economic model
pm to the period between 05:30 am to 07:00 am. based on the maximization of financial benefits
Achievable savings $28200 per day to electricity users. The scheme is applicable in
Scenario 7, Users are shifting peak demand regions managed by the Australian Energy Man-
of 375 MWh occurring between 17:00 pm-19:00 agement Operator (AEMO) and other regions
pm to the period between 06:30 am to 10:30 am. under similar conditions. Usually the electricity
Achievable savings $3525 per day price is high during peak demands and low at
Scenario 8, Users are shifting peak demand of off-peak periods. The concept allows customers
1730 MWh occurring between 10:30 am-19:30 controlling consumption to avoid peak demands
pm. All participants are suggested to set-up the periods. In case the user is on other DSR program
electricity profile to stop some appliance to run with the supplier, the scheme is still allowing ad-
during this time. User can run chosen appliances ditional savings besides the benefits and saving
between 19:30 pm to 23:30 pm. Achievable sav- already achievable through the DSR agreement.
ings in energy cost $48785. For commercial and industrial customers on
Scenario 9, Users are shifting peak demand of fluctuating energy prices implementing the scheme
1730 MWh occurring between 10:30 am-19:30 will enable achieving immediate financial savings.
For residential customers on flat-rate tariffs, in
154
Demand-Side Response Smart Grid Technique for Optimized Energy Use
contrast, users are gaining financial benefits from savings in Queensland by curtailing energy de-
reducing energy consumptions at certain times a mand over a certain energy prices.
day; mainly averting peak-load periods or using Referring to Figure 16, the technique is able
off-peak night tariffs. to remove 6.1 TWh/year from a total of 52.18
TWh/year if users are setting switches to curtail
Multimedia Tool own loads at any regional reference prices above
$50/MWh; this is a percentage of 11.7% of the
The concept includes a multimedia tool (CD- total demand. In case users chose to curtail loads
Rom) run on local computers at user’s premises at $40/MWh, the savings will be 11.1 TWh/year;
to allow implementing and operating the DSR a percentage of 21.2%, and 24.8 TWh/year at a
model continuously. The CD-Rom is containing $30/MWh curtailment; a 47.5%.
an introductory part to address the electricity Figure 17 depicts the case where coordinated
peak demand issue to the user, a programming strategies are able to lead customers to defer
part, where the user will be recording operation loads from times of peak-demand to times of low-
profiles for the different appliances and an execu- demands. Such a procedure shall help flatten the
tive part, where the information is transferred to total energy demand to meet a constant average of
the programmable relay, which on turn sending 5941 MW for Queensland, achieving considerable
signals to electronic switches to operate the dif- improvement in the system utilization and thus in
ferent appliances. the whole system economics. In such a procedure
the technique enables deferring 3.26 TWh/year
Significance from peak to off-peak times.
Queensland Government (2009) is describing:
Figure 15 shows customers fully curtailing energy transport is the fourth largest source of
withdrawals at any energy price above $55/MWh Queensland’s greenhouse gas emissions, contrib-
as example. Figure 16 shows achievable energy uting 10.4 per cent to Queensland’s total emission
profile and the Queensland Government will invest
Figure 15. Day electricity demand curtailed for wholesale regional reference price not to exceed AUD
$55/MWh in Queensland in 2008 (Kamel 2009a)
155
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 16. Achievable energy savings by curtailing energy demand over a certain energy prices in
Queensland (Kamel, 2009a)
Figure 17. Occurrence and average electrical energy demand Queensland (Kamel, 2009a)
$1.4 million to undertake a trial of low-emission The technique offers further the present elec-
diesel-electric buses in the public transport fleet. trical supply system to be supplying the transport
The Department of Resources Energy and Tourism sector for the use of charging electric vehicles at
(2009) is indicating 1359 PJ (377.5 TWh) were low-demand times (at night). Figure 19 depicts
used for transport in Australia in 2006-07 while the possibility to utilize 21.52 TWh/year of elec-
1695 PJ (460.83 TWh) for electricity generation trical capacity, mainly of peak-load power stations,
(1TWh = 3.6 PJ) as illustrated in Figure 18. otherwise not used. The procedure helps enhanc-
156
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Figure 18. Energy consumption in Australia (Department of Resources Energy and Tourism, 2009)
Figure 19. Electrical energy demand Queensland Vs maximum utilization of generating capacity (Ka-
mel, 2009a)
157
Demand-Side Response Smart Grid Technique for Optimized Energy Use
158
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Clement-Nyns, K., Haesen, E., & Driesen, J. Department of Employment Economic De-
(2010). The impact of vehicle-to-grid on the dis- velopment and Innovation. (2010). Electricity
tribution grid. Electric Power Systems Research, generation in Queensland. Brisbane, Australia:
81(1), 185–192. doi:10.1016/j.epsr.2010.08.007 Queensland Government. Retrieved from http://
www.energyfutures.qld.gov.au/electricity_gen-
Clorefeine, A. S. (1980). Economic feasibility of
eration_in_queensland.cfm
photovoltaic energy system. Paper presented at
the 14th IEEE Photovoltaic Specialist Conf., San Department of Employment Economic De-
Diego, California. velopment and Innovation. (2010). Electricity
generation in Queensland. Brisbane, Australia:
Covino, S. (2003). Demand side response 21st
Queensland Government. Retrieved from http://
century style. Paper presented at the Power En-
www.energyfutures.qld.gov.au/electricity_gen-
gineering Society General Meeting, 2003, IEEE.
eration_in_queensland.cfm
Dam, Q. B., Mohagheghi, S., & Stoupis, J.
Department of Resources Energy and Tourism.
(2008). Intelligent demand response scheme for
(2009). Energy in Australia. Canberra, Australia:
customer side load management. Paper presented
Abare. Retrieved from http://www.ret.gov.au/
at the Energy 2030 Conference, 2008. ENERGY
energy/Documents/facts%20statistics%20publi-
2008. IEEE.
cations/energy_in_aus_2009.pdf
De-Meo, E. A. (1978). Prespectives on util-
Energy Efficiency and Conservation Authority.
ity central station photovoltaic applications.
(2009). Domestic-scale distributed generation
Solar Energy, 21, 177–192. doi:10.1016/0038-
guidance for local government. Wellington.
092X(78)90020-8
Energy User Association of Australia Document.
Department of Employment Economic Develop-
(2010). Energy User Association of Australia
ment and Innovation. (2009). Transmission and
document. Australian Government. Retrieved
distribution. Brisbane, Australia: Queensland
from http://www.euaa.com.au/index.htm
Government. Retrieved from http://www.dme.
qld.gov.au/Energy/transmission_and_distribu- Ergon Energy. (2010). Domestic tariff. Retrieved
tion.cfm from http://www.ergon.com.au/home/electric-
ity_for_your_home/ep_domestic.asp
Department of Employment Economic Devel-
opment and Innovation. (2009a). Electricity. Federal Energy Regulatory Commission. (2006).
Brisbane, Australia: Queensland Government. Assesment of demand response and advanced
Retrieved from http://www.dme.qld.gov.au/En- metering. Washington, DC. Retrieved from
ergy/electricity.cfm http://www.ferc.gov/legal/staff-reports/demand-
response.pdf
Department of Employment Economic Develop-
ment and Innovation. (2009b). National electricity Fraser,R. S.(2005). Demand side response in the
market. Brisbane, Australia: Queensland Govern- national electricity market case studies: End use
ment. Retrieved from http://www.dme.qld.gov.au/ customer awareness program.
Energy/national_electricity_market.cfm
Greening, L. A. (2010). Demand response re-
sources: Who is responsible for implementation
in a deregulated market? Energy, 35, 1518–1525.
doi:10.1016/j.energy.2009.12.013
159
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Hamidi, V., Li, F., & Robinson, F. (2009). Demand Leonard, S. L. (1977). Mission analysis of photo-
response in the UK’s domestic sector. Electric voltaic-major mission for the mid-term 1988-2000.
Power Systems Research, 79(12), 1722–1726. (Solar Energy Conversion-San 1101/PA 8 - 1/3, 3).
doi:10.1016/j.epsr.2009.07.013
Leonard, S. L. (1978). Central station power plant
Hammons, T. J. (2006). Integrating renewable application for photovoltaic. Paper presented at the
energy sources into European Grids. Proceed- Solar Energy Conversion 13th IEEE Photovoltaic
ings of the 41st International Universities Power Sp Conference.
Engineering Conference, 2006. UPEC ‘06.
Li, Z., & Yao, T. (2010). Renewable energy bas-
Hyung Seon, O., & Thomas, R. J. (2008). Demand- ing on Smart Grid. Paper presented at the Wire-
side bidding agents: Modeling and simulation. less Communications Networking and Mobile
IEEE Transactions on Power Systems, 23(3), Computing (WiCOM), 2010 6th International
1050–1056. doi:10.1109/TPWRS.2008.922537 Conference on IEEE.
International Energy Agency. (2010). Strategic Marwan, M., & Kamel, F. (2010a). Demand-side
plan for the International Energy Agency demand- response load management modelling encoun-
side management program 2004-2009. tering electrical peak demands in Eastern and
Southern Australia - Smart Grid tools. Paper
Jin-Ho, K., Tae-Kyung, H., & Kwang-Seok, Y.
presented at the Australasian Universities Power
(2009). Roadmap for demand response in the
Engineering Conference AUPEC 2010, Christ-
Korean electricity market. Paper presented at the
church New Zealand.
Power & Energy Society General Meeting, 2009.
PES ‘09. IEEE. Marwan, M., & Kamel, F. (2010b). User-con-
trolled electrical energy consumption towards
Jones, T. E. (2004). Australian example of demand
optimized usage of electricity infrastructure. Paper
side management actions. Australia.
presented at the Southern Region Engineering
Kamel, F. (2009a). Sharing communication Conference, Toowoomba Australia.
network resources for a user-controlled electri-
Na, Y., & Ji-Lai, Y. (2006). Optimal TOU deci-
cal energy consumption. Paper presented at the
sion considering demand response model. Paper
QUESTnet, Gold Coast Australia.
presented at the Power System Technology, 2006.
Kamel, F. (2009b). User-controlled energy con- PowerCon 2006. International Conference on
sumption in a transparent electricity system. Paper IEEE.
presented at the 47th Annual Conference of the
Narayan, P. K., & Smyth, R. (2005). Electricity
Australian and New Zealand Solar Energy Society,
consumption, employment and real income in
Townsville Queensland Australia.
Australia evidence from multivariate Granger
Kirschen, D. S. (2003). Demand-side view causality tests. Energy Policy, 33(9), 1109–1116.
of electricity markets. IEEE Transactions on doi:10.1016/j.enpol.2003.11.010
Power Systems, 18(2), 520–527. doi:10.1109/
Nguyen, D. T. (2010). Demand response for
TPWRS.2003.810692
domestic and small business consumers: A new
Lawrence, D. J., & Neenan, B. F. (2003). The status challenge. Paper presented at the Transmission
of demand response in New York. Paper presented and Distribution Conference and Exposition,
at the Power Engineering Society General Meet- 2010 IEEE PES.
ing, 2003, IEEE.
160
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Parvania, M., & Fotuhi-Firuzabad, M. (2010). Walawalkar, R., Blumsack, S., Apt, J., & Fernands,
Demand response scheduling by stochastic SCUC. S. (2008). Analyzing PJMs economic demand
IEEE Transactions on Smart Grid, 1(1), 89–98. response program. Paper presented at the Power
doi:10.1109/TSG.2010.2046430 and Energy Society General Meeting - Conver-
sion and Delivery of Electrical Energy in the 21st
Queensland Competition Authority. (2000). Elec-
Century, 2008 IEEE.
tricity demand forecast. NSW, Australia. Retrieved
from http://www.qca.org.au/files/QLDElectrici- Wen-Chen, C., Yi-Ping, C., & Tzu-Hao, L. (2007).
tyDemandForecast.pdf The competitive model based on the demand re-
sponse in the off-peak period for Taipower System.
Queensland Government. (2009). ClimateQ:
Paper presented at the Industrial & Commercial
Toward a greener Queensland -Transport moving
Power Systems Technical Conference, 2007. ICPS
towards a low carbon future. Retrieved from http://
2007. IEEE/IAS.
www.climatechange.qld.gov.au/__data/assets/
pdf_file/0013/24061/ClimateQ_Report_chap- Yu, C. W., Zhang, S., Chung, T. S., & Wong, K. P.
ter15.pdf (2005). Modelling and evaluation of interruptible-
load programmes in electricity markets. IEEE
Rahimi, F., & Ipakchi, A. (2010). Overview of
Proceedings Generation. Transmission and
demand response under the smart Grid and mar-
Distribution, 152(5), 581–588. doi:10.1049/ip-
ket paradigms. Paper presented at the Innovative
gtd:20045138
Smart Grid Technologies (ISGT) 2010, IEEE.
Zhong, J., Kang, C., & Liu, K. (2010). Demand
Schweppe, F. C., Caramanis, M. C., Tabors, R.
side management in China. Paper presented at
D., & Bohn, R. E. (1988). Spot electricity price.
the Power and Energy Society General Meeting,
Boston, MA: Kluwer Academic.
2010 IEEE.
Stam, E. (2005). Demand response activities in
Finlad. Paper presented at the A Nordic Confer-
ence on Enhancing and Developing Demand
ADDITIONAL READING
Response in the Energy Market, Copenhagen.
Torriti, J., Hassan, M. G., & Leach, M. (2009). Almeida, A., & Rosenfeld, A. (1988). Demand-
Demand response experience in Europe: Policies, side management and electricity end-use effi-
programmes and implementation. Energy, 35(4), ciency. Kluwer Academic Publishers.
1575–1583. doi:10.1016/j.energy.2009.05.021 America’s Energy Future Panel on Electricity
Tyagi, R., & Black, J. W. (2010). Emergency from Renewable Resources, & Council, N. R.
demand response for distribution system contin- (2010). Electricity from Renewable Resources:
gencies. Paper presented at the Transmission and Status, Prospects, and Impediments: National
Distribution Conference and Exposition, 2010 Academies Press.
IEEE PES. Barnett, D., & Bjornsgaard, K. (2000). Electric
Vos, A. (2009). Effective business models for de- power generation: a nontechnical guide. Pen-
mand response under the Smart Grid paradigm. nWell Books.
Paper presented at the Power Systems Conference
and Exposition, 2009. PSCE ‘09. IEEE/PES.
161
Demand-Side Response Smart Grid Technique for Optimized Energy Use
Barsali, S., Ceraolo, M., & Possenti, A. (2002). Garg, H. (1987). Advances in Solar Energy
Techniques to control the electricity generation Technology: Collection and storage systems. D.
in a series hybrid electrical vehicle. Energy Con- Reidel Pub. Co.
version. IEEE Transactions on, 17(2), 260–266.
Goswami, D. (2007). Advances in Solar Energy:
Books An Annual Review of Research And Development:
Earthscan in association with The American Solar
Budde, P., & Whittle, R. (2007). Smart grid:
Energy Society.
energy management and broadband. Paul Budde
Communication. Hajimiragha, A., Caizares, C. A., Fowler, M.
W., & Elkamel, A. (2010). Optimal Transition
Caves, D. W., Herriges, J. A., Hanser, P., &
to Plug-In Hybrid Electric Vehicles in Ontario,
Windle, R. J. (1988). Load impact of interruptible
Canada, Considering the Electricity-Grid Limita-
and curtailable rate programs: evidence from ten
tions. Industrial Electronics. IEEE Transactions
utilities [tariff incentives]. Power Systems. IEEE
on, 57(2), 690–701.
Transactions on, 3(4), 1757–1763.
Hamilton, K., & Gulhar, N. (2010). Taking De-
Chen, C. S., & Leu, J. T. (1990). Interruptible
mand Response to the Next Level. Power and
load control for Taiwan Power Company. Power
Energy Magazine, IEEE, 8(3), 60–65. doi:10.1109/
Systems. IEEE Transactions on, 5(2), 460–465.
MPE.2010.936352
Chua-Liang, S., & Kirschen, D. (2009). Quantify-
Hori, Y. (2004). Future vehicle driven by electric-
ing the Effect of Demand Response on Electricity
ity and Control-research on four-wheel-motored
Markets. Power Systems. IEEE Transactions on,
“UOT electric march II”. Industrial Electronics.
24(3), 1199–1207.
IEEE Transactions on, 51(5), 954–962.
Collins, M. M., & Mader, G. H. (1983). The tim-
Ibitoye, F. I. & Adenikinju, A. (2007) Future
ing of EV recharging and its effect on utilities.
demand for electricity in Nigeria. Applied En-
Vehicular Technology. IEEE Transactions on,
ergy, 84(5), 492-504. doi: DOI: 10.1016/j.apen-
32(1), 90–97.
ergy.2006.09.011
Daryanian, B., Bohn, R. E., & Tabors, R. D. (1989).
Jazayeri, P., Schellenberg, A., Rosehart, W. D.,
Optimal demand-side response to electricity spot
Doudna, J., Widergren, S., & Lawrence, D. (2005).
prices for storage-type customers. Power Systems.
A Survey of Load Control Programs for Price and
IEEE Transactions on, 4(3), 897–903.
System Stability. Power Systems. IEEE Transac-
Deep, U. D., Petersen, B. R., & Meng, J. (2009). tions on, 20(3), 1504–1509.
A Smart Microcontroller-Based Iridium Satellite-
Journal Articles
Communication Architecture for a Remote Re-
newable Energy Source. Power Delivery. IEEE Kreith, F., & Goswami, D. (2007). Handbook of
Transactions on, 24(4), 1869–1875. energy efficiency and renewable energy. CRC
Press.
Fox-Penner, P. (2010). Smart Power: Climate
Change, the Smart Grid, and the Future of Electric Masiello, R. (2010). Demand Response the other
Utilities. Island Press. side of the curve [Guest Editorial]. Power and
Energy Magazine, IEEE, 8(3), 18–18. doi:10.1109/
MPE.2010.936206
162
Demand-Side Response Smart Grid Technique for Optimized Energy Use
163
164
Chapter 5
Soft Computing and
Computational Intelligent
Techniques in the Evaluation of
Emerging Energy Technologies
Selcuk Cebi
Karadeniz Technical University, Turkey
Cengiz Kahraman
Istanbul Technical University, Turkey
İhsan Kaya
Yıldız Technical University, Turkey
ABSTRACT
The global warming and energy need requires developing emerging energy technologies for the electric-
ity, heat, and transport markets. The emerging energy technologies aim at increasing efficiency of energy
utilization processes from energy sources and diminish CO2 exhalation. The main aim of the chapter
is to exhaustively present soft computing and computational intelligent techniques in the evaluation of
emerging energy technologies. In the scope of the chapter, classification of emerging energy technolo-
gies, their application trends in the literature, a brief explanation for soft computing and computational
intelligent techniques, and literature survey of related techniques on both emerging energy technologies
and energy planning are included. Moreover, technique for order performance by similarity to ideal
solution, analytic hierarchy processes, and their fuzzy structures are introduced.
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Soft Computing and Computational Intelligent Techniques
plants and cleaner heating of homes. In addition, that it is an urgent need to develop highly ef-
crude oil made possible new transportation options ficient energy utilization processes from energy
such as road vehicle and aircraft by invention of the sources effectively and substitute energy sources
internal combustion engine (Vanek and Albright, since the emerging energy technologies are still
2008). Nowadays, fossil fuels play an important in an early phase of development (Jacobsson and
role in the transport and stationary. However, it is Bergek, 2004; Kajikawa et al., 2007). Therefore,
thought that current energy systems are not sustain- recent budgets for governmental research and
able since most of the world primary energy use development (R&D) for energy technologies
is from fossil fuels (Kajikawa et al., 2007). There have increased to support emerging energy re-
are two serious hazards with fossil fuels; the first searches (Hultman and Koomey, 2007). Moreover,
one is that the production of fossil fuel has been European Union is committed in supporting the
predicted to diminish at the middle of this century development of emerging energy technologies,
(Kajikawa et al., 2007; Jefferson, 2006) and the in improving the use of renewable energy, and in
other is that fossil fuels have caused emission of increasing the energy efficiency, to reach global
greenhouse gases into the atmosphere and also objectives of sustainability, competitiveness, and
global warming. Global warming and fossil fuel security of energy supply (Segurado et al., 2009).
depletion are two of the most important issues of When any investment or design decision about
this century. The considerations of energy security energy systems is required, a number of goals or
and climate change force increased societal interest criteria that are local, regional, or global must be
in technologies that enable a reduction in the use taken into account. It is possible to classify these
of fossil fuels. It has been well-recognized that into three categories; (1) Physical goals which meet
an effective solution to these issues is to develop physical requirements that make it possible for the
non-carbon-dioxide-emitting and inexhaustible system to operate. (2) Financial goals which are
energy resources and energy technologies (Chen et monetary objectives related to the energy system.
al., 2009). Recently, discovering of nuclear power (3) Environmental goals which are the objectives
have provide both to diminish our dependence on related to the way in which the energy system im-
fossil fuel resources, and also to provide electricity pacts the natural environment. Regional or global
without any emissions of harmful air pollutants. impacts include the emissions of greenhouse gases
Although nuclear power is cleaner than many that contribute to climate change, air pollutants
other forms of energy production and although that degrade air quality and physical effects from
nuclear energy can be produced in large quanti- extracting resources used either for materials or
ties over short periods of time, nuclear power energy (Vanek and Albright, 2008). Therefore,
generates harmful radiation and throwing out of the emerging technologies have high degree of
nuclear waste which is produced by nuclear power uncertainty and it represents the large variety of
plants is difficult and expensive. Negative effects opportunities that a new technology has to offer.
of both fossil and nuclear technology, renewable This uncertainty creates opportunities for inves-
energy technologies became more advanced and tors to engage in emerging technologies. Thus, the
the range of their applications became broader relation between uncertainty and the decision of
(Vanek and Albright, 2008). Therefore, sustainable investors to engage in emerging technologies is
and renewable energy sources such as sun, wind, very complex (Mejer et al., 2007). Furthermore,
geothermal, biomass, wave etc. and emerging traditional techniques or conventional (hard)
energy technologies have been attracted greater computing models may not present an effective
interest as an important concept while energy solution dealing with the problems in which the
planning of a country. In addition, it is thought dependencies between variables are complex or
165
Soft Computing and Computational Intelligent Techniques
ill-defined. And, these solutions may not satisfy problems in emerging energy technologies and
the decision-makers’ expectations. The selection energy science that cannot be satisfactorily solved
of the most suitable sustainable energy technol- using conventional crisp computing techniques. In
ogy for implementation is a complex problem that particular, multicriteria decision making methods
includes multiple conflicting goals or criteria. In dealing with decision making problems in emerg-
particular, the difficulties dealing with information ing energy technologies and energy science have
about qualitative criteria such as social, cultural been emphasized.
etc. during the evaluation of many conflicting cri- The organization of the chapter is as follows:
teria makes the problem complex. For example, a Section 2 presents the classification of emerging
new energy technology may need to provide good energy technologies. Section 3 introduces the
value for money, low maintenance costs, while most known MCDM techniques in the literature.
at the same time having a large and stable energy Section 4 includes the most used MCDM methods
output and positive social and environmental (AHP and TOPSIS) and information axiom, newly
effects. Multicriteria decision making methods presented to literature. Section 5 presents the lit-
(MCDM) among the soft computing methods are erature survey for emerging energy technologies.
useful and effective tools in order to take into ac- The trends for the emerging energy technologies
count simultaneously all the basic aspects of the and SC and CI techniques are given. Finally,
decision making problems while other decision- concluding remarks are given in Section 5.
support tools, such as cost–benefit analysis are
not well effective (Burton and Hubacek, 2007;
Karakosta et al., 2010). Therefore, soft computing EMERGING ENERGY
(SC) and computational intelligent (CI) techniques TECHNOLOGIES
have been widely used in the literature to solve
complex problems. The emerging energy technologies (EETs) are
SC and CI techniques which are based on classified into six main groups in terms of usage
copying ability of human mind under uncer- area. These are building technology, industry
tainty and imprecision are emerging approaches technology, transportation technology, coal
(Konar, 2007). The main characteristics of SC technology, non-conventional fuel technology,
are representation and processing of human mind and biomass technology. The classification is
and knowledge, qualitative and approximate presented in Figure 1(Lee et al., 2009). Emerg-
reasoning, computing with words, and biologi- ing energy technologies for building includes
cal models of problem solving and optimization, lighting, air conditioning, building envelop, and
and are directly related to intelligent systems and building system technologies. Industrial emerg-
applications (Karray and Silva, 2004). In this re- ing energy technologies are classified into waste
spect, these techniques differ from the respective heat technologies, common technologies, and
conventional computing techniques in that they are petroleum refinery and fine chemical technolo-
tolerant of imprecision, uncertainty, partialtruth, gies. Transportation technologies involve fuel
and approximation. The soft computing techniques efficiency improvement technologies, hybrid
comprises of fuzzy logic, artificial neural net- electronic technologies, electric, hydrogen fuel
works, probabilistic reasoning and meta-heuristic cell, and biodiesel technologies. Coal emerging
techniques such as genetic algorithm, tabu search, energy technologies are classified into direct uti-
etc. (Altun and Yalcinoz, 2008). The main aim of lization technology and conversion technologies.
this section is to introduce the most known and Non-conventional fuel technologies are oil shale/
the most used SC and CI techniques for complex oil sand technologies and gasification technolo-
166
Soft Computing and Computational Intelligent Techniques
gies of waste. Biomass technologies are direct of biomass has continues, the popularity of non-
utilization and conversion technologies. conventional fuel, transportation, and industry
Figure 2 and Figure 3 presents the trends for technologies have increased. Finally, the popular-
emerging energy technologies based on years and ity of coal and transportation technologies have
application areas of EETs, respectively. The years dramatically increases at the last period.
between 1990 and 2010 are divided into four According to the Figure 3, popularity of
periods as 1990-1995, 1996-2000, 2001-2005, the each emerging energy technology has been
and 2006-2010 and numbers of publications in clearly increasing year by year except for non-
SCI index are analyzed. Figure 2 presents the conventional fuel technology.
attractiveness of emerging energy technologies
in the literature while Figure 3 presents the most
attractive years for each technology. SOFT COMPUTING
According to Figure 2, biomass technology is AND COMPUTATIONAL
the most attractive topic for researchers between INTELLIGENT TECHNIQUES
the years 1990-1995. In the same period, building
technology is the second attractive topic. Although Soft computing (SC) and computational intelligent
the attractiveness of biomass technology is the (CI) techniques are used to obtain the closest solu-
first at the second period, its popularity decreas- tions to computationally-hard problems. The basis
es and the popularity of building and coal tech- of the SC and CI techniques are founded on bio-
nologies have increased according to previous logical or behavioral phenomena related to humans
period. At the third period, while the popularity or animals, and analogues of these technologies
167
Soft Computing and Computational Intelligent Techniques
exist in many human and animal systems (Uhrig the hardware technology advances, SC and CI
and Tsoukalas, 1999). SC and CI techniques are techniques have been intensely studied and im-
vitally practical tools for many complex problems proved in the last years, and nowadays practical
since they can tolerant of imprecision, uncertainty, applications become a reality. Such techniques
partial truth, and approximation. However, tra- present several advantages when compared to
ditional hard computing methods are often too traditional ones, such as: (i) acquisition of better
cumbersome for complex problems. They need results in the optimization processes when no
a precisely stated analytical model and often a lot prior knowledge is available, (ii) possibility of
of computational time (Zadeh, 1965). Following application to problems to which the conventional
168
Soft Computing and Computational Intelligent Techniques
methods are not suitable, (iii) simulation of hu- and it is inspired of biological neural networks.
man cognition processes, instead of trying to The ANNs are structured with a set of inter con-
solve deterministically what is not deterministic nected layers, each of them composed of nodes;
(Schirru et al, 1999). The term “soft computing the typology of connections and nodes (called
and computational intelligence” in its broadest neurons) characterizes the different typologies of
sense, encompasses a number of technologies neural networks (Bertini et al. 2010). Outputs of
that includes, but is not limited to, artificial neural neurons in a given layer are the inputs of neurons
networks (ANN), genetic algorithms (GA), fuzzy in the next layer. First and last layers are called
logic models, ant colony techniques (AC), tabu input and output layers, respectively, while those in
search (TS), expert systems (ExS), multicriteria between are hidden layers. Neurons are character-
decision making methods etc. These are the ized by a transfer function, which is applied to an
well known soft computing and computational appropriate function of inputs. The main advantage
intelligence techniques in the literature. Beside of ANN is that it does not need any mathemati-
of these techniques, genetic programming (GP), cal model, since it learns from historical data to
artificial immune system (AIS), harmony search recognize non-evident relations and patterns in a
(HS), scatter search (SC), variable neighborhood set of input–output variables, without any prior
search (VNS), pattern search (PS), differential assumption about their nature (Pena et al., 2010).
evolution (DE), evolutionary programming (EP), The structure of ANN is given in Figure 4.
evolutionary strategies (ES), simulated anneal- Genetic algorithms (GAs) are global search
ing (SA), particle swarm optimization (PSO), and optimization techniques motivated by the
swarm intelligence (SI) etc. are the relatively new process of natural selection in biological system
developed tools. However, these techniques are (Gen and Cheng, 2000; Kaya, 2009). GA as a
currently not as popular especially with regard to field of study was initiated and developed in the
the emerging energy technology. Thus, only the early 1970’s by John Holland (Holland, 1975,
main tools are briefly highlighted in this section. 1992) and his students, but its applications to
The artificial neural networks (ANN) were real-world practical problems was almost three
first introduced by McCulloch et al. (1943), who decades in developing. GA approaches are good
suggested that the biological function of the at solving the ill-posed problems such as non-
human brain could be emulated by a simplified convex functions, non-differentiable functions,
computational model (Saridakis and Dentsoras, domains not connected, badly behaved functions,
2008). The technique is a computational model multiple local optima, and multiple objectives
169
Soft Computing and Computational Intelligent Techniques
(Miranda et al., 1998). The main advantages of involves nothing more than processing highly fit
GA is that it present an approximate solution in individuals in order to produce better individuals
relatively short time compared with other random as the search progresses. A typical genetic algo-
searching methods, such as simulated annealing rithm cycle involves four major processes of
or dynamic programming (Won and Park, 2003). fitness evaluation, selection, recombination and
The first population of GA is a randomly selected creation of a new population. Based on fitness
initial solution set. To obtain an optimum solution, criterion, poorer performing individuals are
a search is conducted by moving from the initial gradually taken out, and better individuals have
population of individuals to a new population a greater possibility of conveying genetic informa-
using genetics-like operators such as selection, tion to the next generation (Altun and Yalcinoz,
crossover and mutation, which are inspired from 2008). The cycle of GA is illustrated in Figure 5.
the mechanics of natural selection and genetics Ant colony (ACO) which is used for solving
encountered in natural life. Each individual rep- combinatorial optimization problems is a coop-
resents a candidate to the optimization solution erative search algorithm inspired by the behavior
and is modeled by a value called chromosome. of ants in finding paths from the nest to food (Yang
The GA operators perform task on the chromo- and Zhuang, 2010). In the early 1992, ACO was
some, in the reproduction process, in order to proposed to literature as a metaheuristic optimiza-
produce new generations so that solution at the tion tool as by Dorigo (1992). High concentrations
global optimum may be obtained. The operation of pheromones indicate more favorable paths that
is based on a selective nature, i.e. the best candi- other members should follow in order to reach
dates in terms of fitness are chosen as parent so the optimal solution (Al-Rashidi and El-Hawary,
that the new generation holds best genetic heritage. 2009). ACO is based on the indirect communica-
For this purpose, a fitness function assigns a fit- tion of a colony of simple agents, called ants,
ness value to each individual within the population. mediated by (artificial) pheromone trails. The
This fitness value is the measure for the quality pheromone trails in ACO serve as distributed,
of an individual. The basic optimization procedure numerical information which the ants use to
170
Soft Computing and Computational Intelligent Techniques
probabilistically construct solutions to the problem alternatives based on prioritized attributes of the
being solved and which the ants adapt during the alternatives.
algorithm’s execution to reflect their search ex-
perience (Dorigo and T. Stützle, 2003). MCDM Techniques
Tabu search (TS) is another algorithm which is
used for the solution of combinatorial optimization The well-known and the most used MADM tech-
problems like the traveling salesman problem. niques in the literature are Analytic Hierarchical
TS method originally proposed by Fred Glover Process (AHP) (Saaty, 1980), Technique for
(Glover, 1986) is based on neighborhood search Order Performance by Similarity to Ideal Solu-
procedure such that the algorithm iteratively tion (TOPSIS) (Hwang and Yoon, 1981), Simple
moves from a solution to another solution in the Additive Weighting method (SAW) (Hwang and
related neighborhood, until it reaches any stopping Yoon, 1981), Elimination By Aspects (EBA)
criterion. The basic principle of TS is to pursue lo- (Tversky, 1972), ELimination and Choice Ex-
cal search whenever it encounters a local optimum pressing REality (ELECTRE) (Bernard, 1968),
by allowing non-improving moves; cycling back Preference Ranking Organisation METHod for
to previously visited solutions is prevented by the Enrichment Evaluations (PROMETHEE) (Brans
use of memories, called tabu lists, which record and Vincke, 1985). Besides these methodologies,
the recent history of the search (Grandeu, 2003). relatively new decision making methodologies
TS is a powerful algorithmic approach that has such as information axiom method (Suh, 1990),
been applied with great success to many difficult VlseKriterijumska Optimizacija I Kompromisno
combinatorial problems. The best feature of TS Resenje (VIKOR) (Opricovic and Tzeng, 2004)
is that it easily handle complicating constraints. etc. or integrated methods like Hierarchical TOP-
Thus, TS heuristics generally find good solutions SIS (Kahraman et al, 2007), have been proposed
very early in the search. Both depth and breadth to solve complex decision making problems (Cebi
need to be achieved in the searching process. and Kahraman, 2010a). In the literature, there
Although depth is usually not a problem for TS, are also fuzzy MCDM techniques such as fuzzy
breadth can be a critical issue (Gendreau and TOPSIS (Chen, 2000), fuzzy AHP (Buckley, 1985,
Potvin, 2010). Chang, 1996; Laarhoven and Pedrycz, 1983),
MCDM techniques are also known as SC and fuzzy VIKOR (Opricovic and Tzeng, 2004), fuzzy
CI techniques. MCDM techniques are classified information axiom (Kulak and Kahraman, 2005a;
into two groups as Multiple Objective Decision 2005b; Kulak et al., 2005; Kahraman and Cebi,
Making (MODM) and Multiple Attribute Deci- 2009) etc. in order to make decision makers cope
sion Making (MADM). The difference between with incomplete and vague information.
MADM and MODM is that MADM is associated In this chapter AHP, TOPSIS, Fuzzy AHP,
with problems of which numbers of alternatives Fuzzy TOPSIS and information axiom among
have been predetermined. The Decision Maker the MCDM techniques are presented.
(DM) is to select/rank a finite number of courses of
action. On the other hand, MODM is not associated Analytic Hierarchy Process
with the problems in which alternatives have been
predetermined (Lai and Hwang, 1994). In other The analytic hierarchy process (AHP) is based on
words, MODM techniques present optimization pairwise comparisons. This provides an advantage
of an alternative or alternatives on the bases of when there is no any quantitative information
prioritized objectives while MADM techniques about the problem. The mains steps of the AHP
present selection of an alternative from a set of is as follows (Önüt and Soner, 2008);
171
Soft Computing and Computational Intelligent Techniques
1 a12 a1n
C1 TOPSIS Method
1 1 a 2n
C2 a12
A = aij (1) In TOPSIS method, the distance to both positive
. and negative ideal solution is calculated and an
Cn 1 1 1 alternative is the best if the alternative has the
a1n a 2n
shortest distance from the positive ideal-solution
and the longest distance from the negative-ideal
where aij=1 when i=j, and aij=1 / aij for i,j=1,2,…,n. solution (Yoon and Hwang, 1981). The TOPSIS
Step 2. The comparison matrix is normalized method can be summarized as follows (Önüt and
and weights are obtained. Soner, 2008):
Let A1,A2,…,Aj be the j different alternatives.
a12 a1n For alternative Aj the rating of the ith aspect is
1
t2 tn denoted by fij which is the value of the ith criterion
a21 a function for the alternative Aj, and n is the number
1 2n
A = t1 tn (2) of the criteria. Then
Step 1. The normalized decision matrix is
a an 2 obtained by;
n1 1
t1 t2
fij
rij = , j=1,2,…,J and i=1,2,…,n
n n
t j = ∑ aij (3) ∑f
j =1
2
ij
i =1
(7)
where i and j represents row and column number,
Step 2. The weighted normalized value vij is
respectively. Then, a set of numerical weights,
calculated as:
w1, w2, …, wn are obtained by averaging of rows.
Step 3. The consistency analysis is done;
vij = wi ∗ rij j=1,2,…,J and i=1,2,…,n (8)
A * wi = λmax * wi (4)
where wi is the weight of the ith attribute or cri-
and the consistency index is obtained as follows: terion, and
172
Soft Computing and Computational Intelligent Techniques
n
(1983), which compared fuzzy ratios described by
∑w
i =1
i = 1. (9)
triangular membership functions. Then, Buckley
(1985) presented fuzzy priorities of comparison
Step 3. The positive ideal solution A*and the ratios whose membership functions trapezoidal.
negative ideal solution A-are determined as: And the last one is proposed by Chang (1996) with
the use of triangular fuzzy numbers for pairwise
comparison scale of fuzzy AHP, and the use of
{
A∗ = {v1∗ , ..., vi∗ } = (max vij i ∈ I ' ),(min vij i ∈ I '' )}
j j the extent analysis method for the synthetic ex-
(10) tent values of the pairwise comparisons. In this
chapter, Buckley’s (1985) fuzzy AHP approach
{
A− = {v1−, ..., vi− } = (min vij i ∈ I ' ),(max vij i ∈ I '' )}
j j
is presented in detail.
(11) Step 1. Pairwise comparison matrices are
constructed. Each element( cij ) of the pairwise
where I’ is a set of benefit criteria and I’ is a set comparison matrix (C) is a linguistic terms pre-
of cost criteria. senting which is the more important of two cri-
Step 4. The distance to positive ideal solution teria. The pairwise comparison matrix is given
is calculated by; by;
n 1 c12 … c1n
D j∗ = ∑ (v ij − vi∗ )2 , j=1,2,…,J (12)
c21 1 … c2n
Ck =
i =1
, k=1,2,3,…,K
Similarly, the distance to the negative-ideal cn 1 cn 2 … 1
solution is obtained by; (15)
n
where Ck is a pairwise comparison matrix belongs
D j− = ∑ (vij − vi− )2 , j=1,2,…,J
i =1
(13)
to kth expert for FRm. For the evaluation procedure,
the linguistic terms given in Table 1a are used.
Arithmetic mean is used to aggregate expert
Step 5. The relative closeness coefficient of
opinions.
the alternative Aj is defined by;
Step 2. Weights are calculated. At first, the
fuzzy weight matrix is calculated by Buckley’s
D j−
C j∗ = , j=1,2,…,J (14) Method as follows (Hsieh et al., 2004);
D j∗ + D j−
ri = (ci1 ⊗ ci 2 ⊗ ... ⊗ cin )1/n (16)
Step 6. The alternatives are ranked with respect
to closeness coefficients. wi = ri ⊗ (r1 + r2 + ... + rn )−1 (17)
173
Soft Computing and Computational Intelligent Techniques
Table 1a. Linguistic scale for weight matrix (Hsieh et al., 2004)
Step 3. After the fuzzy relative weight matrix is where c *j = max cij if criterion is benefit. Other-
i
obtained, defuzzification process which converts wise, if criterion is cost, following equation is
a fuzzy number into a crisp value is utilized. At used.
first, fuzzy numbers will be defuzzified into crisp
values and then normalization procedure will be a − a − a −
applied. For the defuzzification process, centroid rij = j , j , j where a −j = min aij (20)
method, which provides a crisp value based on cij bij aij i
= [r ]
R ⇒
a b c
rij = ij* , ij* , ij* (19)
di* = ∑ d(v , v )
j =1
ij
*
j
i = 1, 2, , m
c j c j c j
ij mxn
n
di− = ∑ d(v , vij
−
j
) i = 1, 2, , m
j =1
(23)
174
Soft Computing and Computational Intelligent Techniques
175
Soft Computing and Computational Intelligent Techniques
between the conventional information axiom and literature to take into account EETs based on SC
the fuzzy information axiom is that the fuzzy in- and CI techniques. For instance, ANN and GA
formation axiom uses fuzzy numbers. Kulak et al. techniques are generally used for sizing photo-
(2005) developed unweighted and weighted multi voltaic (PV) technology, since these methods
attribute axiomatic design approaches includ- present a good solution (Table 2). The performance
ing both crisp and fuzzy criteria and applied the of the PV systems depends upon several factors
methodology to an equipment selection problem. such as solar radiation, ambient temperature and
Kulak (2005) developed a decision support system wind speed. In order to size a PV system so that
for the selection of a material handling system. it can work properly, efficiently and economi-
Then, Kahraman and Cebi (2009) extended the cally to meet the desired load requirements under
usability of the fuzzy information axiom for vari- the local meteorological conditions, the charac-
ous decision making problems. teristic performance of each component in the PV
system is required (Mellit et al. 2008). According
to Table 2, the most used technique is ANN too
LITERATURE SURVEY for sizing problem of PV system. Table 2 indicates
FOR EMERGING ENERGY that SC and CI techniques have an increasing
TECHNOLOGIES popularity on the sizing of PV-systems. In par-
ticular, the numbers of the applications using SC
In this section, a literature survey has been pre- and CI technologies are
sented. At first, the general applications of SC and In the literature, MCDM methods have become
CI techniques and trends on energy problems in increasingly popular in decision-making for en-
last five years are presented. Then, the applications ergy planning. Table 3 illustrates applications of
on EETs have been given. The literature survey MCDM methods on the solution of energy plan-
is divided into two categories since the applica- ning in last five years.
tions of SC and CI techniques on the evaluation Furthermore, the applications of MCDM
of emerging energy technologies do not yield a methods on emerging energy technology are more
huge material in the literature; one of them is the than other SC and CI techniques. These studies
studies related to MCDM and the other is related are summarized as follows;
to rest of SC and CI techniques. Some studies for Beccali et al. (2003) used ELECTRE III method
the last five years deal with the applications of SC for the selection of suitable energy technologies
and CI techniques on energy planning have been in renewable energy technology diffusion plan.
given in Table 1. Table 2 presents the applications In the study, technological alternatives given in
of SC and CI techniques on EETs. The number of Table 4 are evaluated under the criteria which
presented publications on energy problems does are classified into four main groups such as
not include all in the literature. Since we aimed technological criteria, energy and environmental
to demonstrate the usefulness and possible ap- criteria, social and economic criteria (Table 5).
plications of SC and CI techniques, the papers The evaluations are implemented in three differ-
published in last five years are presented. ent scenarios; environmental–oriented scenario,
Except MCDM methods, the other SC and CI economy-oriented scenario, and energy saving
techniques are widely used in the literature to and rational use scenario.
forecast energy demand or energy supply. In Pohekar and Ramachandran (2006a, 2006b)
particular, ANN is the most widely used technique assessed the utility of parabolic solar cooker (PSC)
in all to predict energy demand or supply (Table under the criteria techno-economic, social, be-
1b). In addition, there are a few studies in the havioral and commercial comparing with other
176
Soft Computing and Computational Intelligent Techniques
Table 1b. The applications of SC and CI techniques related to energy planning for last five years
contemporary cooking energy devices. In the In the study, the perceived social, economic, and
paper, the alternatives; chulha, improved chulha, environmental cost of the small-scale energy were
kerosene stove, biogas stove, lpg stove, micro compared with the large-scale alternatives. In
wave oven, electric oven, solar box cooker, para- order to investigate whether the energy could have
bolic solar cooker are evaluated under thirty been generated at a lower social, economic, and
criteria. The criteria are categorized under five environmental cost if large-scale projects had
main criteria such as technical, economical, social, been available, a multi-criteria decision making
behavioral, and commercial (Table 6). (MCDM) methodology, MACBETH, was used
Burton and Hubacek (2007) investigated a to compare the advantages and disadvantages of
local case study of different scales of renewable a number of different renewable energy tech-
energy provision for local government in the UK. nologies. MACBETH method proposed by (Bana
177
Soft Computing and Computational Intelligent Techniques
e Costa and Vansnick, 1997). In the basis of the and environmental issues (Burton & Hubacek,
MACBETH, a series of pairwise comparisons, 2007).
where a decision-maker is asked to specify the Doukas et al. (2007) presented a direct and
difference in attractiveness between all of the transparent MCDM approach, using linguistic
alternatives is included. In the study eight renew- variables, to assist policy makers in formulating
able energy technologies of differing scales are technological energy priorities towards a sustain-
considered such as solar photovoltaic, micro-wind, able energy system. In the paper, technologies
micro-hydro, large-scale wind, large-scale hydro, given in Table 7 are handled under following
energy from waste, landfill gas and biomass (wood criteria; economic (including investment cost
chippings) based on the definition of renewable criterion and economic visibility using payback
energy used by the UK government under eight period criterion), environmental (including con-
criteria such as; capital cost, operation and main- tribution to confrontation of the climate change
tenance cost, generation capacity, lifespan, carbon phenomenon criterion and effects on natural
emissions, noise, impact upon the natural environ- environment criterion), technological (includ-
ment and social effects. These criteria were se- ing efficiency rate criterion and knowledge of
lected in order to consider the viability of renew- the innovative technology criterion), and social
able energy developments and the need to have (including contribution to employment oppor-
a breadth of criteria covering social, economic tunities’ creation criterion and contribution to
178
Soft Computing and Computational Intelligent Techniques
179
Soft Computing and Computational Intelligent Techniques
Table 5. Main and sub criteria used for evaluation of energy technologies (Beccali et al., 2003)
180
Soft Computing and Computational Intelligent Techniques
regional development criterion). In the paper a Nations framework convention on climate change,
two-staged method was developed using linguistic economic spin-off, technical spin-off, urgency of
ordered weighted averaging (LOWA) and ordered technology development, and quantity of energy
weighted maximum (OWMAX) operators for the use. The sub-criteria are possibility of developing
technologies assessment. The energy technologies technologies (domestic technical level and pos-
are firstly identified and the most promising is sibility of commercialization), potential quantity
chosen based on the country’s specific priorities of energy saving (quantity of energy saving and
and objectives. quantity CO2 saving), market size (domestic
Lee et al. (2007) determined the priorities for market size, potential export market size, and
energy technology development in the sectors of effect of generating hiring), investment cost, and
energy efficiency improvement and greenhouse ease of energy use (ease of product, applicable
gas (GHG) control plans (EGCP) for a new na- area of other technologies).
tional energy and resource technology R&D plan In another study, Lee et al. (2009) prioritized
(NERP)by using the analytic hierarchy process the energy technologies against high oil prices in
(AHP). They focused on the areas of energy ef- the energy technology roadmap (ETRM) in order
ficiency improvement and GHG control. In the to allocate R&D budget strategically. The fuzzy
paper, in order to determine priorities of energy analytic hierarchy process, which integrates the
technologies, 9 energy technologies from 3 sectors fuzzy theory into the classical AHP approach, is
given in Table 8 in terms of GHG control and 34 utilized to generate the weights of energy technol-
energy technologies from six sectors given in ogy against high oil prices of the ETRM. In the
Table 9 in terms of energy efficiency improvement paper, four criteria which are economical spin-off,
are evaluated under a set of criteria. The hierarchy possibility of commercialization, inner capacity,
of the criteria consists of five main criteria and and technical spinoff are handled. Energy tech-
five sub criteria. The main criteria are United nologies against high oil prices such as building
181
Soft Computing and Computational Intelligent Techniques
Sectors Technologies
GHG tech CO2 capture storage and conversion tech
Non-CO2 gas tech
Clean fossil tech Advanced combustion tech
Next-generation clean coal tech
Clean petroleum and conversion tech
DME tech
GTL tech
Gas hydrate
GHG policy GHG mitigation policy
technology, industry technology, transportation economic viewpoints. In the paper, the following
technology, coal technology, non-conventional criteria was used for the assessment; technological
fuel technology, and biomass technology are as- status, hydrogen technology infrastructure, R&D
sessed. human resources, R&D budgets
Lee et al. (2008; 2010a) developed an algorithm Lee et al. (2010b) determined the priorities
by integrating fuzzy analytic hierarchy process for hydrogen energy technologies by using fuzzy
(Fuzzy AHP) and the data envelopment analysis AHP. In the study, four criteria which are economic
(DEA) in order to measure the relative efficiency impact, commercial potential, inner capacity, and
of the R&D performance in the national hydrogen technical spin-off are handled to evaluate and
energy technology development. On the first stage, determine the weights of five hydrogen energy
the fuzzy AHP was used to reflect the vagueness of technologies which are hydrogen production,
human thought. On the second stage, the DEA ap- hydrogen separation and storage, polymer elec-
proach was used to measure the relative efficiency trolyte membrane fuel cell, direct ethanol fuel
of the national R&D performance in the sector of cell, and solid oxide fuel cell.
hydrogen energy technology development with
182
Soft Computing and Computational Intelligent Techniques
Sectors Technologies
Industry High-efficiency drying tech
Fine chemical processing
Energy conversion tech
Unutilized energy tech
Energy material tech
High-efficiency dying tech
Cold storage and freezing tech
Process automation and intelligence tech
Supercritical fluid process tech
Evaporation and distillation tech
Adsorption separation tech
Membrane separation tech
Crystallization tech
Building Green building tech
Building renovation tech
High-efficiency HVAC tech
CHP tech
Efficiency policy Energy efficiency improvement policy
183
Soft Computing and Computational Intelligent Techniques
Klemes et al. (2009) developed software NENT is the only energy technology assessment
named Early Market Introduction of New En- tool that targets early stage energy technology.
ergy Technologies (EMINENT). The software Karakosta et al. (2010) presented the priorities
was developed to analyze the potential impact of sustainable electricity generation technologies
of new and underdeveloped energy technolo- for five developing countries, namely Chile, China,
gies in different sectors emerging from different Israel, Kenya and Thailand by using ELECTRE III.
countries. The software involved two databases; In the paper, energy technologies given in Table
new technologies (which contains renewable 10 were analyzed under accordance with strate-
electricity generation, renewable heating and gic/developmental planning, local and regional
cooling technologies, production and distribution economic development, co2 emissions reduction,
of liquid and gaseous bio-fuels, eco-buildings, minimization of the negative effects on the natural
poly-generation, energy demand management environment at national–regional level, contribu-
and renewable energy supply in high performance tion to the employment, and contribution to energy
communities, and alternative motor fuels) and independence developing country criteria.
sectoral energy supplies and demands (which Sadegzadeh and Salehi (2010) used TOPSIS
contains information of the number of consum- methodology to rank the attractiveness and im-
ers per sector, type of demand, typical quality portance of the stack of fuel cells as a sub-system.
of the energy required and the consumption and The technologies which are taken into consider-
installed capacity per end-user). The main aim ation as follows; the situation of professional
of the program is to evaluate the market potential manpower on the industrial and semi-industrial
of energy-related early stage energy technolo- scales, the situation of professional manpower on
gies in various energy supply chains, and their the laboratory scale, the situation of know-how
performance in terms of CO2 emissions, costs of on the industrial and semi-industrial scales, the
energy supply, use of primary fossil energy, and situation of know-how on the laboratory scales,
in different subsectors of society. the situation of hardware on the industrial and
Segurado et al. (2009) compared EMINENT semi-industrial scales, the situation of hardware
with other tools which are carbon dioxide tech- on the laboratory scale. The criteria used in the
nology database (CO2DB), MARket Allocation evaluation are as follow; production of platinum
(MARKAL), IKARUS, and energy emission catalyst powder-carbon, placing catalyst on the
economy database (E3database) already on the carbon base, production of gas penetration layer,
market for energy technology assessment. The production of polymer membrane of ion exchange,
main conclusion of the comparison is that EMI- construction of membrane collection-electrode
184
Soft Computing and Computational Intelligent Techniques
with low heat, plates of current field, technology coal, and industry keep going on and these topics
of waterproofing stack of fuel cell with low heat, will be highlighted in near future.
and technology of engineering designing of the In addition, the utilization of SC and CI tech-
collection (stack) of fuel cell. Finally, the priorities niques on development of EET technologies has
of allocating power and capital for the develop- been grown up. And application of SC and CI
ment of technology of fuel cells in automotive techniques on energy problems has been also
industry are as follows; hardware on the labora- increasing fast. In particular, the application of
tory scale, know-how on the laboratory scale, MCDM on EETs and Energy problems are the
know-how on the industrial scale, professional most attractive topics in all and it is clearly un-
manpower on the laboratory scale of laboratory, derstood that their popularity will be increase in
and professional manpower on the industrial scale. near future. The percentiles of publications with
respect to last five years have been presented in
Figure 8.
FUTURE RESEARCH DIRECTIONS
185
Soft Computing and Computational Intelligent Techniques
Figure 8. The percentiles of application of both SC and CI techniques and MCDM methods on energy
and EET field
Then, SC and CI techniques have been briefly used in decision making of both emerging
explained and various numbers of publications energy technology and energy planning
related to their applications on energy planning because of the complexity of the problem.
and emerging energy technologies in the literature • Although, there are a few studies deal with
have been reviewed. From the literature, following applications of SC and CI techniques on
conclusions are obtained; emerging energy technologies, ANN, GA
and AHP techniques are widely used meth-
• The interest of researchers in emerging en- odologies among the published articles in
ergy technologies has increased in recent the literature.
years. In particular, the biggest increase • ANN and GA is the most used techniques
has been in transportation and coal tech- for sizing problem of PV systems while
nologies. Although, the interest in biomass AHP is used to obtain a decision related to
technology is the top until the last period, EET.
the interest in transportation and coal tech- • In MCDM applications, social, technical,
nologies has left behind the interest in bio- economical, and commercial criteria are
mass technology. the most used criteria during evaluation of
• For energy planning studies, ANN and emerging energy technologies.
AHP are the most used techniques in the
literature. Except for MCDM methods, the Although the published literature on the
other SC and CI techniques are usually EETs based on SC and CI techniques indicates
used to forecast energy demand/supply. that the popularity of SC and CI techniques has
• Although applications of SC and CI tech- been increasing day by day, there are two gabs
niques on EETs are narrow, the numbers of in the literature; the first one, there are not any
publications have been increasing in recent application of other SC and CI techniques such as
years. Hence, SC and CI techniques have TS, ACO, SI. The second one is that none of the
become increasingly popular in application MCDM studies in the literature takes into account
of emerging energy technologies. In par- the interdependencies among the criteria related
ticular, MCDM methods have been widely to EETs. For further research, a decision making
186
Soft Computing and Computational Intelligent Techniques
tool such as analytic network process, Choquet Azadeh, A., Ghaderi, S. F., & Sohrabkhani, S.
Integral, etc. may be used for the evaluation of (2008a). Annual electricity consumption forecast-
emerging energy technologies to handle interde- ing by neural network in high energy consum-
pendencies among the criteria. ing industrial sectors. Energy Conversion and
It should be noted that the findings given in Management, 49(8), 2272–2278. doi:10.1016/j.
this chapter are based on the data collected from enconman.2008.01.035
articles published in scientific journals, which
Azadeh, A., Saberi, M., Ghaderi, S. F., Gitiforouz,
do not include conference proceeding papers,
A., & Ebrahimipour, V. (2008b). Improved estima-
master’s theses, doctoral dissertations, textbooks
tion of electricity demand function by integration
deal with the literature. It is possible to extend
of fuzzy system and data mining approach. Energy
this study by including these sources.
Conversion and Management, 49(8), 2165–2177.
doi:10.1016/j.enconman.2008.02.021
REFERENCES Azadeh, A., & Tarverdian, S. (2007). Integration of
genetic algorithm, computer simulation and design
Abdel-Aal, R. E. (2008). Univariate modeling and of experiments for forecasting electrical energy
forecasting of monthly energy demand time series consumption. Energy Policy, 35(10), 5229–5241.
using abductive and neural networks. Comput- doi:10.1016/j.enpol.2007.04.020
ers & Industrial Engineering, 54(4), 903–917.
doi:10.1016/j.cie.2007.10.020 Bahgat, A. B. G., Helwa, N. H., Ahamd, G. E.,
& El Shenawy, E. T. (2004). Estimation of the
Al-Rashidi, M. R., & El-Hawary, M. E. (2009). maximum power and normal operating power
Applications of computational intelligence tech- of a photovoltaic module by neural networks.
niques for solving the revived optimal power flow Renewable Energy, 29(3), 443–457. doi:10.1016/
problem. Electric Power Systems Research, 79(4), S0960-1481(03)00126-5
694–702. doi:10.1016/j.epsr.2008.10.004
Bana e Costa, C. A., & Vansnick, J. C. (1997).
Altun, H., & Yalcinoz, T. (2008). Implementing Applications of the MACBETH approach
soft computing techniques to solve economic in the framework of an additive aggrega-
dispatch problem in power systems. Expert tion model. Journal of Multi-Criteria Deci-
Systems with Applications, 35(4), 1668–1678. sion Analysis, 6(2), 107–114. doi:10.1002/
doi:10.1016/j.eswa.2007.08.066 (SICI)1099-1360(199703)6:2<107::AID-
Ashhab, M. S. S. (2008). Optimization and mod- MCDA147>3.0.CO;2-1
eling of a photovoltaic solar integrated system Beccali, M., Cellura, M., & Mistretta, M. (2003).
by neural networks. Energy Conversion and Decision-making in energy planning: Application
Management, 49(11), 3349–3355. doi:10.1016/j. of the Electre method at regional level for the
enconman.2007.10.036 diffusion of renewable energy technology. Renew-
Azadeh, A., Asadzadeh, S. M., & Ghanbari, A. able Energy, 28(13), 2063–2087. doi:10.1016/
(2010). An adaptive network-based fuzzy infer- S0960-1481(03)00102-2
ence system for short-term natural gas demand
estimation: Uncertain and complex environments.
Energy Policy, 38(3), 1529–1536. doi:10.1016/j.
enpol.2009.11.036
187
Soft Computing and Computational Intelligent Techniques
Ben Salah, C., & Ouali, M. (2011). Comparison Cavallaro, F. (2009). Multi-criteria decision
of fuzzy logic and neural network in maximum aid to assess concentrated solar thermal tech-
power point tracker for PV systems. Electric Power nologies. Renewable Energy, 34(7), 1678–1685.
Systems Research, 81(1), 43–50. doi:10.1016/j. doi:10.1016/j.renene.2008.12.034
epsr.2010.07.005
Cavallaro, F. (2010a). A comparative assessment
Benlarbi, K., Mokrani, L., & Nait-Said, M. S. of thin-film photovoltaic production processes
(2004). A fuzzy global efficiency optimization using the electre iii method. Energy Policy, 38(1),
of a photovoltaic water pumping system. Solar 463–474. doi:10.1016/j.enpol.2009.09.037
Energy, 77(2), 203–216. doi:10.1016/j.sole-
Cavallaro, F. (2010b). Fuzzy topsis approach for
ner.2004.03.025
assessing thermal-energy storage in concentrated
Bernard, R. (1968). Classement et choix en solar power (CSP) systems. Applied Energy, 87(2),
présence de points de vue multiples (la méthode 496–503. doi:10.1016/j.apenergy.2009.07.009
ELECTRE). la Revue d’Informatique et de Re-
Cebi, S., & Kahraman, C. (2010a). Developing
cherche Opérationelle, 8, 57–75.
a group decision support system based on fuzzy
Bertini, I., Ceravolo, F., Citterio, M., De Felice, information axiom. Knowledge-Based Systems,
M., Di Pietra, B., & Margiotta, F. (2010). Ambi- 23(1), 3–16. doi:10.1016/j.knosys.2009.07.005
ent temperature modelling with soft computing
Ceylan, H., Ozturk, H. K., Hepbasli, A., & Utlu,
techniques. Solar Energy, 84(7), 1264–1272.
Z. (2005). Estimating energy and exergy produc-
doi:10.1016/j.solener.2010.04.003
tion and consumption values using three different
Brans, J. P., & Vincke, P. (1985). A preference genetic algorithm approaches. Part 2: Application
ranking organization method: The PROMETHEE and scenarios. Energy Sources, 27, 629–639.
method for MCDM. Management Science, 31(6), doi:10.1080/00908310490448631
647–656. doi:10.1287/mnsc.31.6.647
Chang, D. Y. (1996). Applications of the ex-
Buchholz, T., Rametsteiner, E., Volk, T. A., & tent analysis method on fuzzy-AHP. European
Luzadis, V. A. (2009). Multi criteria analysis for Journal of Operational Research, 95, 649–655.
bioenergy systems assessments. Energy Policy, doi:10.1016/0377-2217(95)00300-2
37(2), 484–495. doi:10.1016/j.enpol.2008.09.054
Chang, Y.-P. (2009). Optimal design of discrete-
Buckley, J. J. (1985). Fuzzy hierarchical value tilt angle of PV using sequential neural-net-
analysis. Fuzzy Sets and Systems, 17, 233–247. work approximation and orthogonal array. Expert
doi:10.1016/0165-0114(85)90090-9 Systems with Applications, 36(3), 6010–6018.
doi:10.1016/j.eswa.2008.06.105
Burton, J., & Hubacek, K. (2007). Is small beauti-
ful? A multicriteria assessment of small-scale en- Chaouachi, A., Kamel, R. M., & Nagasaka, K. (in
ergy technology applications in local governments. press). A novel multi-model neuro-fuzzy-based
Energy Policy, 35(12), 6402–6412. doi:10.1016/j. mppt for three-phase grid-connected photovoltaic
enpol.2007.08.002 system. [forthcoming]. Solar Energy.
188
Soft Computing and Computational Intelligent Techniques
Chen, T.-C. (2000). Extensions of the TOPSIS for Dufo-Lopez, R., Bernal-Agustin, J. L., & Contre-
group decision-making under fuzzy environment. ras, J. (2007). Optimization of control strategies
Fuzzy Sets and Systems, 114, 1–9. doi:10.1016/ for stand-alone renewable energy systems with
S0165-0114(97)00377-1 hydrogen storage. Renewable Energy, 32(7),
1102–1126. doi:10.1016/j.renene.2006.04.013
Chen, T. Y., Yu, O. S., Hsu, G. J. Y., Hsu, F. M.,
& Sung, W. N. (2009). Renewable energy tech- Ediger, V. S., & Akar, S. (2007). Arima forecast-
nology portfolio planning with scenario analysis: ing of primary energy demand by fuel in Turkey.
A case study for Taiwan. Energy Policy, 37(8), Energy Policy, 35(3), 1701–1708. doi:10.1016/j.
2900–2906. doi:10.1016/j.enpol.2009.03.028 enpol.2006.05.009
Cinar, D., Kayakutlu, G., & Daim, T. (2010). Eynard, J., Grieu, S., & Polit, M. (in press). Wave-
Development of future energy scenarios with let-based multi-resolution analysis and artificial
intelligent algorithms: Case of hydro in Turkey. neural networks for forecasting temperature and
Energy, 35(4), 1724–1729. doi:10.1016/j.en- thermal power consumption. [forthcoming]. En-
ergy.2009.12.025 gineering Applications of Artificial Intelligence.
Dong, B., Cao, C., & Lee, S. E. (2005). Apply- Gareta, R., Romeo, L. M., & Gil, A. (2006).
ing support vector machines to predict building Forecasting of electricity prices with neural
energy consumption in tropical region. Energy networks. Energy Conversion and Management,
and Building, 37(5), 545–553. doi:10.1016/j. 47(13-14), 1770–1778. doi:10.1016/j.encon-
enbuild.2004.09.009 man.2005.10.010
Dorigo, M. (1992). Optimization, learning and Gen, M., & Cheng, R. (2000). Genetic algorithms
natural algorithms. PhD thesis, Dipartimento di and engineering optimization. New York, NY:
Elettronica, Politecnico di Milano, Italy. (in Italian) John Wiley and Sons.
Dorigo, M., & Stützle, T. (2004). Ant colony Gendreau, M. (2003). Handbook of
optimization. Cambridge, MA & London, UK: metaheurıstıcs:An introductıon to tabu search
MIT Press. (Glover, F., & Kochenberger, G. A., Eds.). New
York, NY/ Boston, MA/ Dordrecht, The Neth-
Doukas, H. C., Andreas, B. M., & Psarras, J.
erlands/ London, UK/ Moscow, Russia: Kluwer
E. (2007). Multi-criteria decision aid for the
Academic Publishers.
formulation of sustainable technological energy
priorities using linguistic variables. European Gendreau, M., & Potvin, J.-Y. (2010). Handbook
Journal of Operational Research, 182, 844–855. of metaheuristics. Springer Science+Business
doi:10.1016/j.ejor.2006.08.037 Media. LLC.
Dufo-Lopez, R., & Bernal-Agustin, J. L. (2005). Georgiou, P., Tourkolias, C., & Diakoulaki, D.
Design and control strategies of PVdiesel systems (2008). A roadmap for selecting host countries
using genetic algorithms. Solar Energy, 79, 33–46. of wind energy projects in the framework of the
doi:10.1016/j.solener.2004.10.004 clean development mechanism. Renewable &
Sustainable Energy Reviews, 12(3), 712–731.
doi:10.1016/j.rser.2006.11.001
189
Soft Computing and Computational Intelligent Techniques
Ghafghazi, S., Sowlati, T., Sokhansanj, S., & Hontoria, L., Aguilera, J., & Zufiria, P. (2005).
Melin, S. (2010). A multicriteria approach to A new approach for sizing stand alone photo-
evaluate district heating system options. Applied voltaic systems based in neural networks. Solar
Energy, 87(4), 1134–1140. doi:10.1016/j.apen- Energy, 78(2), 313–319. doi:10.1016/j.sole-
ergy.2009.06.021 ner.2004.08.018
Glover, F. (1986). Future paths for integer pro- Hsieh, T. Y., Lu, S. T., & Tzeng, G. T. (2004). Fuzzy
gramming and links to artificial intelligence. MCDM approach for planning and design tenders
Computational Opeation. Research, 13, 533–549. selection in public office buildings. International
Journal of Project Management, 22, 573–584.
González, P. A., & Zamarreño, J. M. (2005).
doi:10.1016/j.ijproman.2004.01.002
Prediction of hourly energy consumption in
buildings based on a feedback artificial neural Hultman, N. E., & Koomey, J. G. (2007). The
network. Energy and Building, 37(6), 595–601. risk of surprise in energy technology costs.
doi:10.1016/j.enbuild.2004.09.006 Environmental Research Letters, 2(3), 1–6.
doi:10.1088/1748-9326/2/3/034002
González-Romera, E., Jaramillo-Morán, M. A., &
Carmona-Fernández, D. (2008). Monthly electric Hussein, A., Hirasawa, K., & Hu, J. (2004).
energy demand forecasting with neural networks A robust control method for a pv-supplied dc
and Fourier series. Energy Conversion and motor using universal learning networks. Solar
Management, 49(11), 3135–3142. doi:10.1016/j. Energy, 76(6), 771–780. doi:10.1016/j.sole-
enconman.2008.06.004 ner.2003.12.011
Haldenbilen, S., & Ceylan, H. (2005). Genetic Hwang, C. L., & Yoon, K. (1981). Multiple at-
algorithm approach to estimate transport energy tribute decision making: Methods and applica-
demand in Turkey. Energy Policy, 33, 89–98. tions. Berlin/Heidelberg, Germany/New York,
doi:10.1016/S0301-4215(03)00202-7 NY: Springer-Verlag.
Hamzaçebi, C. (2007). Forecasting of Turkey’s net Jacobsson, S., & Bergek, A. (2004). Transforming
electricity energy consumption on sectoral bases. the energy sector: The evolution of technological
Energy Policy, 35(3), 2009–2016. doi:10.1016/j. systems in renewable energy technology. Indus-
enpol.2006.03.014 trial and Corporate Change, 13(5), 815–849.
doi:10.1093/icc/dth032
Heo, E., Kim, J., & Boo, K.-J. (2010). Analysis
of the assessment factors for renewable energy Jefferson, M. (2006). Sustainable energy de-
dissemination program evaluation using fuzzy velopment: Performance and prospects. Re-
AHP. Renewable & Sustainable Energy Reviews, newable Energy, 31, 571–582. doi:10.1016/j.
14(8), 2214–2220. doi:10.1016/j.rser.2010.01.020 renene.2005.09.002
Holland, J. H. (1975). Adaptation in natural and Kahraman, C., & Cebi, S. (2009). A new multi-
artificial systems. Ann Arbor, MI: University of attribute decision making method: Hierarchical
Michigan Press. fuzzy axiomatic design. Expert Systems with
Applications, 36(3), 4848–4861. doi:10.1016/j.
Holland, J. H. (1992). Genetic algorithms.
eswa.2008.05.041
Scientific American, 267, 66–72. doi:10.1038/
scientificamerican0792-66
190
Soft Computing and Computational Intelligent Techniques
Kahraman, C., Çevik, S., Ates, N. Y., & Gülbay, Kaya, T., & Kahraman, C. (2010). Multicriteria
M. (2007). Fuzzy multi-criteria evaluation of renewable energy planning using an integrated
industrial robotic systems. Computers & Indus- fuzzy VIKOR & AHP methodology: The case of Is-
trial Engineering, 52(4), 414–433. doi:10.1016/j. tanbul. Energy, 35(6), 2517–2527. doi:10.1016/j.
cie.2007.01.005 energy.2010.02.051
Kahraman, C., & Kaya, I. (2010). A fuzzy Klemes, J., Bulatov, I., & Koppejan, J. (2009).
multicriteria methodology for selection among Novel energy saving technologies evaluation tool.
energy alternatives. Expert Systems with Ap- Computers & Chemical Engineering, 33(3), 751–
plications, 37(9), 6270–6281. doi:10.1016/j. 758. doi:10.1016/j.compchemeng.2008.07.005
eswa.2010.02.095
Konar, A. (2007). Artificial intelligence and soft
Kahraman, C., Kaya, I., & Cebi, S. (2009). A computing: Behavioral and cognitive modeling
comparative analysis for multiattribute selection of the human brain. New York, NY: CRC Press.
among renewable energy alternatives using fuzzy
Kowalski, K., Stagl, S., Madlener, R., & Omann, I.
axiomatic design and fuzzy analytic hierarchy pro-
(2009). Sustainable energy futures: Methodologi-
cess. Energy, 34(10), 1603–1616. doi:10.1016/j.
cal challenges in combining scenarios and partici-
energy.2009.07.008
patory multi-criteria analysis. European Journal
Kajikawa, Y., Yoshikawa, J., Takeda, Y., & Mat- of Operational Research, 197(3), 1063–1074.
sushima, K. (2008). Tracking emerging technolo- doi:10.1016/j.ejor.2007.12.049
gies in energy research: Toward a roadmap for
Kulak, O. (2005). A decision support system for
sustainable energy. Technological Forecasting
fuzzy multi-attribute selection of material handling
and Social Change, 75, 771–782. doi:10.1016/j.
equipments. Expert Systems with Applications,
techfore.2007.05.005
29(2), 310–319. doi:10.1016/j.eswa.2005.04.004
Karakosta, C., Doukas, H., & Psarras, J. (2010).
Kulak, O., Durmusoglu, M. B., & Kahraman,
EU-MENA energy technology transfer under the
C. (2005). Fuzzy multi-attribute equipment se-
CDM: Israel as a frontrunner? Energy Policy, 38(5),
lection based on information axiom. Journal of
2455–2462. doi:10.1016/j.enpol.2009.12.039
Materials Processing Technology, 169, 337–345.
Karatepe, E., Boztepe, M., & Colak, M. (2006). doi:10.1016/j.jmatprotec.2005.03.030
Neural network based solar cell model. Energy
Kulak, O., & Kahraman, C. (2005a). Fuzzy
Conversion and Management, 47(9-10), 1159–
multi-attribute selection among transportation
1178. doi:10.1016/j.enconman.2005.07.007
companies using axiomatic design and analytic
Karray, F. O., & Silva, C. (2004). Soft computing hierarchy process. Information Sciences, 170,
and intelligent systems design. New York, NY: 191–210. doi:10.1016/j.ins.2004.02.021
Pearson Education Limited.
Kulak, O., & Kahraman, C. (2005b). Multi-
Kaya, I. (2009). A genetic algorithm approach attribute comparison of advanced manufacturing
to determine the sample size for control charts systems using fuzzy vs. crisp axiomatic design
with variables and attributes. Expert Systems with approach. International Journal of Produc-
Applications, 36(5), 8719–8734. doi:10.1016/j. tion Economics, 95, 415–424. doi:10.1016/j.
eswa.2008.12.011 ijpe.2004.02.009
191
Soft Computing and Computational Intelligent Techniques
Laarhoven, P. J. M., & Pedrycz, W. (1983). A Lee, S. K., Mogi, G., Lee, S. K., Hui, K. S., &
fuzzy extension of Saaty’s priority theory. Fuzzy Kim, J. W. (2010). Econometric analysis of the
Sets and Systems, 11, 229–241. R&D performance in the national hydrogen en-
ergy technology development for measuring rela-
Lai, Y.-J., & Hwang, C.-L. (1994). Fuzzy multi
tive efficiency: The fuzzy AHP/DEA integrated
objective decision making methods and applica-
model approach. International Journal of Hy-
tions. Berlin, Germany: Springer-Verlag.
drogen Energy, 35(6), 2236–2246. doi:10.1016/j.
Lee, S. K., Mogi, G., & Kim, J. W. (2008). The ijhydene.2010.01.009
competitiveness of Korea as a developer of hy-
Lee, S. K., Mogi, G., Lee, S. K., & Kim, J. W.
drogen energy technology: The AHP approach.
(in press). Prioritizing the weights of hydrogen
Energy Policy, 36(4), 1284–1291. doi:10.1016/j.
energy technologies in the sector of the hydro-
enpol.2007.12.003
gen economy by using a fuzzy AHP approach.
Lee, S. K., Mogi, G., & Kim, J. W. (2009). Deci- International Journal of Hydrogen Energy. doi:.
sion support for prioritizing energy technologies doi:10.1016/j.ijhydene.2010.01.035
against high oil prices: A fuzzy analytic hierar-
Lee, S. K., Yoon, Y. J., & Kim, J. W. (2007). A
chy process approach. Journal of Loss Preven-
study on making a long-term improvement in
tion in the Process Industries, 22(6), 915–920.
the national energy efficiency and GHG control
doi:10.1016/j.jlp.2009.07.001
plans by the AHP approach. Energy Policy, 35(5),
Lee, S. K., Mogi, G., & Kim, J. W. (2009). Deci- 2862–2868. doi:10.1016/j.enpol.2006.09.019
sion support for prioritizing energy technologies
Li, G., & Shi, J. (2010). On comparing three arti-
against high oil prices: A fuzzy analytic hierar-
ficial neural networks for wind speed forecasting.
chy process approach. Journal of Loss Preven-
Applied Energy, 87(7), 2313–2320. doi:10.1016/j.
tion in the Process Industries, 22(6), 915–920.
apenergy.2009.12.013
doi:10.1016/j.jlp.2009.07.001
Li, K., & Su, H. (2010). Forecasting building
Lee, S. K., Mogi, G., Kim, J. W., & Gim, B. J.
energy consumption with hybrid genetic algo-
(2008). A fuzzy analytic hierarchy process ap-
rithm-hierarchical adaptive network-based fuzzy
proach for assessing national competitiveness
inference system. Energy and Building, 42(11),
in the hydrogen technology sector. International
2070–2076. doi:10.1016/j.enbuild.2010.06.016
Journal of Hydrogen Energy, 33(23), 6840–6848.
doi:10.1016/j.ijhydene.2008.09.028 Liao, C.-C. (2010). Genetic k-means algorithm
based RBF network for photovoltaic MPP pre-
Lee, S. K., Mogi, G., Lee, S. K., Hui, K. S., &
diction. Energy, 35(2), 529–536. doi:10.1016/j.
Kim, J. W. (2010). Econometric analysis of the
energy.2009.10.021
R&D performance in the national hydrogen en-
ergy technology development for measuring rela- Ma, J., Scott, N. R., Degloria, S. D., & Lembo, A.
tive efficiency: The fuzzy AHP/DEA integrated J. (2005). Siting analysis of farm-based centralized
model approach. International Journal of Hy- anaerobic digester systems for distributed genera-
drogen Energy, 35(6), 2236–2246. doi:10.1016/j. tion using GIS. Biomass and Bioenergy, 28(6),
ijhydene.2010.01.009 591–600. doi:10.1016/j.biombioe.2004.12.003
192
Soft Computing and Computational Intelligent Techniques
Madlener, R., Antunes, C. H., & Dias, L. C. (2009). Mellit, A., Benghanem, M., & Kalogirou, S. A.
Assessing the performance of biogas plants with (2007). Modeling and simulation of a stand-alone
multi-criteria and data envelopment analysis. Eu- photovoltaic system using an adaptive artificial
ropean Journal of Operational Research, 197(3), neural network: Proposition for a new sizing
1084–1094. doi:10.1016/j.ejor.2007.12.051 procedure. Renewable Energy, 32(2), 285–313.
doi:10.1016/j.renene.2006.01.002
Madlener, R., Kowalski, K., & Stagl, S. (2007).
New ways for the integrated appraisal of national Mellit, A., Kalogirou, S. A., & Drif, M. (2010).
energy scenarios: The case of renewable energy Application of neural networks and genetic algo-
use in Austria. Energy Policy, 35(12), 6060–6074. rithms for sizing of photovoltaic systems. Renew-
doi:10.1016/j.enpol.2007.08.015 able Energy, 35(12), 2881–2893. doi:10.1016/j.
renene.2010.04.017
Madlener, R., & Stagl, S. (2005). Sustainability-
guided promotion of renewable electricity gen- Mellit, A., Kalogirou, S. A., Hontoria, L., & Shaari,
eration. Ecological Economics, 53(2), 147–167. S. (2009). Artificial intelligence techniques for
doi:10.1016/j.ecolecon.2004.12.016 sizing photovoltaic systems: A review. Renewable
& Sustainable Energy Reviews, 13(2), 406–419.
McCulloch, W. S., & Pitts, W. A. (1943). A logical
doi:10.1016/j.rser.2008.01.006
calculus of the ideas immanent in neural nets. The
Bulletin of Mathematical Biophysics, 5, 115–133. Miranda, V., Srinivasan, D., & Proenc¸, A. M.
doi:10.1007/BF02478259 (1998). Evolutionary computation in power sys-
tems. Electric Power Systems Research, 20(2),
Meijer, I. S. M., Hekkert, M. P., & Koppenjan, J. F.
89–98. doi:10.1016/S0142-0615(97)00040-9
M. (2007). The influence of perceived uncertainty
on entrepreneurial action in emerging renewable Mróz, T. M. (2008). Planning of community
energy technology; biomass gasification projects heating systems modernization and development.
in the Netherlands. Energy Policy, 35, 5836–5854. Applied Thermal Engineering, 28(14-15), 1844–
doi:10.1016/j.enpol.2007.07.009 1852. doi:10.1016/j.applthermaleng.2007.11.020
Mellit, A. (2010). Ann-based GA for generating Murat, Y. S., & Ceylan, H. (2006). Use of artifi-
the sizing curve of stand-alone photovoltaic sys- cial neural networks for transport energy demand
tems. Advances in Engineering Software, 41(5), modeling. Energy Policy, 34(17), 3165–3172.
687–693. doi:10.1016/j.advengsoft.2009.12.008 doi:10.1016/j.enpol.2005.02.010
Mellit, A., & Benghanem, M. (2007). Sizing of Neto, A. H., & Fiorelli, F. A. S. (2008). Com-
stand-alone photovoltaic systems using neural parison between detailed model simulation and
network adaptive model. Desalination, 209(1-3), artificial neural network for forecasting building
64–72. doi:10.1016/j.desal.2007.04.010 energy consumption. Energy and Building, 40(12),
2169–2176. doi:10.1016/j.enbuild.2008.06.013
Mellit, A., Benghanem, M., Arab, A. H., & Gues-
soum, A. (2005). An adaptive artificial neural Nixon, J. D., Dey, P. K., & Davies, P. A. (in
network model for sizing stand-alone photovol- press). Which is the best solar thermal collection
taic systems: Application for isolated sites in technology for electricity generation in north-west
Algeria. Renewable Energy, 30(10), 1501–1524. India? Evaluation of options using the analytical
doi:10.1016/j.renene.2004.11.012 hierarchy process. [forthcoming]. Energy.
193
Soft Computing and Computational Intelligent Techniques
Önüt, S., & Soner, S. (2008). Transshipment site Pohekar, S. D., & Ramachandran, M. (2006a).
selection using the AHP and TOPSIS approaches Multi-criteria evaluation of cooking devices with
under fuzzy environment. Waste Management special reference to utility of parabolic solar cooker
(New York, N.Y.), 28(9), 1552–1559. doi:10.1016/j. (PSC) in India. Energy, 31(8-9), 1215–1227.
wasman.2007.05.019 doi:10.1016/j.energy.2005.04.012
Opricovic, S., & Tzeng, G. H. (2004). Compro- Pohekar, S. D., & Ramachandran, M. (2006b).
mise solution by MCDM methods: a comparative Utility assessment of parabolic solar cooker as
analysis of VIKOR and TOPSIS. European Jour- a domestic cooking device in India. Renewable
nal of Operational Research, 156(2), 445–455. Energy, 31(11), 1827–1838. doi:10.1016/j.re-
doi:10.1016/S0377-2217(03)00020-1 nene.2005.09.014
Ozturk, H. K., Ceylan, H., Canyurt, O. E., & Rivière, C., & Marlair, G. (2009). Biosafuel, a
Hepbasli, A. (2005). Electricity estimation us- pre-diagnosis tool of risks pertaining to biofuels
ing genetic algorithm approach: a case study of chains. Journal of Loss Prevention in the Pro-
Turkey. Energy, 30, 1003–1012. doi:10.1016/j. cess Industries, 22(2), 228–236. doi:10.1016/j.
energy.2004.08.008 jlp.2008.09.014
Pao, H. T. (2007). Forecasting electricity market Rovere, E. L. L., Soares, J. B., Oliveira, L. B.,
pricing using artificial neural networks. Energy & Lauria, T. (2010). Sustainable expansion of
Conversion and Management, 48(3), 907–912. electricity sector: Sustainability indicators as an
doi:10.1016/j.enconman.2006.08.016 instrument to support decision making. Renewable
& Sustainable Energy Reviews, 14(1), 422–429.
Pao, H. T. (2009). Forecasting energy consump-
doi:10.1016/j.rser.2009.07.033
tion in Taiwan using hybrid nonlinear models.
Energy, 34(10), 1438–1446. doi:10.1016/j.en- Saaty, T. L. (1980). The analytic hierarchy process.
ergy.2009.04.026 New York, NY: McGraw Hill.
Paoli, C., Voyant, C., Muselli, M., & Nivet, M.- Sadeghzadeh, K., & Salehi, M. B. (in press). Math-
L. (in press). Forecasting of preprocessed daily ematical analysis of fuel cell strategic technologies
solar radiation time series using neural networks. development solutions in the automotive industry
[forthcoming]. Solar Energy. by the TOPSIS multi-criteria decision making
method. [forthcoming]. International Journal of
Papadopoulos, A., & Karagiannidis, A. (2008).
Hydrogen Energy.
Application of the multi-criteria analysis method
ELECTRE III for the optimisation of decen- Saridakis, K. M., & Dentsoras, A. J. (2008). Soft
tralised energy systems. Omega, 36(5), 766–776. computing in engineering design - A review. Ad-
doi:10.1016/j.omega.2006.01.004 vanced Engineering Informatics, 22(2), 202–221.
doi:10.1016/j.aei.2007.10.001
Peña, B., Teruel, E., & Díez, L. I. (in press). Soft-
computing models for soot-blowing optimization Schirru, R., Martinez, A. S., Pereira, C. M. N. A.,
in coal-fired utility boilers. [forthcoming]. Applied Domingos, R. P., Machado, M. D., & Machado,
Soft Computing. L. (1999). Intelligent soft computing in nuclear
engineering in Brazil. Progress in Nuclear
Energy, 35(3-4), 367–391. doi:10.1016/S0149-
1970(99)00022-0
194
Soft Computing and Computational Intelligent Techniques
Segurado, R., Pereira, S., Pipio, A., & Alves, Thiaux, Y., Seigneurbieux, J., Multon, B., & Ben
L. (2009). Comparison between EMINENT Ahmed, H. (2010). Load profile impact on the
and other energy technology assessment tools. gross energy requirement of stand-alone photovol-
Journal of Cleaner Production, 17(10), 907–910. taic systems. Renewable Energy, 35(3), 602–613.
doi:10.1016/j.jclepro.2009.02.002 doi:10.1016/j.renene.2009.08.005
Senjyua, T., Hayashia, D., Yonaa, A., Urasakia, N., Toksari, M. D. (2007). Ant colony optimization
& Funabashib, T. (2007). Optimal configuration approach to estimate energy demand in Turkey.
of power generating systems in isolated island Energy Policy, 35, 3984–3990. doi:10.1016/j.
with renewable energy. Renewable Energy, 32, enpol.2007.01.028
1917–1933. doi:10.1016/j.renene.2006.09.003
Tversky, A. (1972). Elimination by aspects: A
Sozen, A., & Arcaklioglu, E. (2007). Prospects theory of choice. Psychological Review, 79,
for future projections of the basic energy sources 281–299. doi:10.1037/h0032955
in Turkey. Energy Sources, Part B. Economics,
Uhrig, R. E., & Tsoukalas, L. H. (1999). Soft
Planning, and Policy, 2(2), 183–201.
computing technologies in nuclear engineering
Sozen, A., Arcaklioglu, E., Özalp, M., & Çaglar, applications. Progress in Nuclear Energy, 34(1),
N. (2005a). Forecasting based on neural network 13–75. doi:10.1016/S0149-1970(97)00109-1
approach of solar potential in Turkey. Renew-
Ünler, A. (2008). Improvement of energy demand
able Energy, 30(7), 1075–1090. doi:10.1016/j.
forecasts using swarm intelligence: The case of tur-
renene.2004.09.020
key with projections to 2025. Energy Policy, 36(6),
Sozen, A., Arcaklioglu, E., & Ozkaymak, M. 1937–1944. doi:10.1016/j.enpol.2008.02.018
(2005b). Modelling of the Turkey’s net energy
Vanek, F. M., & Albright, L. D. (2008). Energy
consumption using artificial neural network.
systems engineering: Evaluation and implementa-
International Journal of Computer Applications
tion. New York, NY: McGraw Hill.
in Technology, 22(2/3), 130–136. doi:10.1504/
IJCAT.2005.006944 Venayagamoorthy, G. K., & Welch, R. L. (2010).
Energy dispatch controllers for a photovoltaic
Suh, N. P. (1990). The principles of design. New
system. Engineering Applications of Artificial
York, NY: Oxford University Press Inc.
Intelligence, 23(2), 249–261. doi:10.1016/j.en-
Suh, N. P. (2001). Axiomatic design: Advances gappai.2009.11.001
and applications. New York, NY: Oxford Uni-
Won, J. R., & Park, Y. M. (2003). Economic
versity Press.
dispatch solutions with piecewise quadratic cost
Thakker, A., Jarvis, J., Buggy, M., & Sahed, A. functions using improved genetic algorithm.
(2008). A novel approach to materials selection International Journal of Electrical Power &
strategy case study: Wave energy extraction im- Energy Systems, 25(5), 355–361. doi:10.1016/
pulse turbine blade. Materials & Design, 29(10), S0142-0615(02)00098-4
1973–1980. doi:10.1016/j.matdes.2008.04.022
Yang, J., & Zhuang, Y. (2010). An improved ant
colony optimization algorithm for solving a com-
plex combinatorial optimization problem. Applied
Soft Computing, 10(2), 653–660. doi:10.1016/j.
asoc.2009.08.040
195
Soft Computing and Computational Intelligent Techniques
Zadeh, L. A. (1965). Fuzzy sets. Information Celik, M., Kahraman, C., Cebi, S., & Er, I. D.
and Control, 8, 338–353. doi:10.1016/S0019- (2009c). Fuzzy axiomatic design-based perfor-
9958(65)90241-X mance evaluation model for docking facilities in
shipbuilding industry: the case of Turkish ship-
Zhang, L., & Fei Bai, Y. (2005). Genetic algorithm-
yards. Expert Systems with Applications, 36(1),
trained radial basis function neural networks for
599–615. doi:10.1016/j.eswa.2007.09.055
modelling photovoltaic panels. Engineering Appli-
cations of Artificial Intelligence, 18(7), 833–844. Engin, O., Çelik, A., & Kaya, I. (2008). A fuzzy
doi:10.1016/j.engappai.2005.02.004 approach to define sample size for attributes con-
trol chart in multistage processes: An application
in engine valve manufacturing process. Applied
Soft Computing, 8(4), 1654–1663. doi:10.1016/j.
ADDITIONAL READING
asoc.2008.01.005
Aliev, R. A., & Aliev, R. (2001). Soft Computing Kahraman, C. (2008). Fuzzy Multi-Criteria Deci-
& Its Applications. Singapore: World Scientific. sion Making: Theory and Applications with Recent
Castillo, O., Melin, P., Ross, O. M., Cruz, R. S., Developments. Berlin: Springer.
& Pedrycz, W. (2007). Theoretical Advances and Kahraman, C., Ruan, D., & Dogan, I. (2003).
Applications of Fuzzy Logic and Soft Computing. Fuzzy group decision-making for facility location
Berlin: Springer. doi:10.1007/978-3-540-72434-6 selection. Information Sciences, 157, 135–153.
Cebi, S., & Kahraman, C. (2010). Developing a doi:10.1016/S0020-0255(03)00183-X
group decision support system based on fuzzy Kaliszewski, I. (2006). Soft Computing for Com-
information axiom. Knowledge-Based Systems, plex Multiple Criteria Decision Making. USA:
23(1), 3–16. doi:10.1016/j.knosys.2009.07.005 Springer.
Cebi, S., & Kahraman, C. (2010). Developing a Kaya, I. (2009). A genetic algorithm approach to
group decision support system based on fuzzy determine the sample size for attribute control
information axiom. Knowledge-Based Systems.. charts. Information Sciences, 179(10), 1552–1566.
doi:10.1016/j.knosys.2009.07.005 doi:10.1016/j.ins.2008.09.024
Celik, M., Cebi, S., Kahraman, C., & Er, D. Kaya, I., & Engin, O. (2007). A new approach to
(2009a). Application of axiomatic design and define sample size at attributes control chart in
TOPSIS methodologies under fuzzy environment multistage processes: An application in engine
for proposing competitive strategies on Turkish piston manufacturing process. Journal of Ma-
container ports in maritime transportation net- terials Processing Technology, 183(1), 38–48.
work. Expert Systems with Applications, 36(3), doi:10.1016/j.jmatprotec.2006.09.022
1, 4541–4557. doi:10.1016/j.eswa.2008.05.033
Konar, A. (1999). Artificial Intelligence and Soft
Celik, M., Cebi, S., Kahraman, C., & Er, I. D. Computing: Behavioral and Cognitive Model-
(2009b). An integrated fuzzy QFD model pro- ing of the Human Brain. Boca Raton/ London/
posal on routing of shipping investment decisions New York/ Washington, D.C.: CRC Press.
in crude oil tanker market. Expert Systems with doi:10.1201/9781420049138
Applications, 36(3), 2, 6227–6235. doi:10.1016/j.
eswa.2008.07.031
196
Soft Computing and Computational Intelligent Techniques
Rayward-Smith, V. J., Osman, I. H., Reeves, C. R., Multi Attribute Decision Making (MADM):
& Smith, G. D. (1996). Modern Heuristic Search MADM is to select the best alternative under
Methods. New York, Toronto: Wiley. multiple criteria.
Multi Objective Decision Making (MODM):
Schniederjans M. J., Hamaker J.L., Schnieder-
MODM is to present the best opportunity by
jans A. M., Information Technology Investment:
satisfying multiple objectives.
Decision-Making Methodology,
Analytic Hierarchy Process (AHP): AHP
Tettamanzi, A., & Tomassini, M. (2001). Soft is a MCDM tool which is based on pairwise
Computing: Integrating Evolutionary, Neural, comparison when there is not any quantitative
and Fuzzy Systems. Berlin: Springer. information.
Technique for Order Performance by
Triantaphyllou, E. (2000) Multi-Criteria Decision
similarity to Ideal Solution (TOPSIS): TOPSIS
Making Methods: A comparative Study, Kluwer
is another MCDM method which is based on
Academic Publishers, Baton Rogue/Louisiana.
measuring the distance to positive and negative
ideal solutions.
Information Axiom: Information axiom
KEY TERMS AND DEFINITIONS which is based on satisfying level of functional
requirements is used to determine the best alterna-
Emerging Energy Technology (EET): EET tive under multiple criteria.
is the technology that improves the utilization Fuzzy Sets: Fuzzy is based on the membership
performance of energy sources. function definition of an element. In classical sets,
Multi Criteria Decision Making (MCDM): either an element is a member of the defined set or
MCDM is to give the best decision under the multi not while each member is defined by a member-
objective or multi attribute. ship value in a fuzzy sets.
197
198
Chapter 6
Dynamic Analysis and Stability
Improvement Concerning the
Integration of Wind Farms:
Kurdistan Electric Network Case Study
Mohammad Saleh
University of Kurdistan, Iran
Hassan Bevrani
University of Kurdistan, Iran
ABSTRACT
This chapter presents an overview of key issues and technical challenges in a regional electric net-
work, following the integration of a considerable amount of wind power. A brief survey on wind power
system, the present status of wind energy worldwide, common dynamic models, and control loops for
wind turbines are given. In this chapter, the Kurdistan electric network in the Northwest part of Iran is
introduced as a case study system, and an analytical approach is conducted to evaluate the potential of
wind power installation, overall capacity estimation, and economic issues, based on the practical data.
Then, the impact of high penetration wind power on the system dynamic and performance for various
wind turbine technologies is presented. The stability of integrated system is analyzed, and the need for
revising of conventional controls and performance standards is emphasized. Finally, a STATCOM-based
control approach is addressed to improve the system stability.
DOI: 10.4018/978-1-61350-138-2.ch006
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
The recent investigation studies indicate that In the next section, a background with a brief
relatively large scale wind generation affects the literature review is presented. In section 3, an over-
power system frequency and voltage regulation, view of wind energy status around the world and
as well as other control and operation issues. This Iran is provided. Section 4 presents a discussion
impact may increase at the penetration rates that about wind power systems and the main control
are expected to be high in the next several years. schemes. Section 5 determines the potential of
On the other hand, most of existing wind turbine Kurdistan province for wind power generation.
technologies cannot provide necessary control In section 6, a preliminary study on wind energy
capabilities for the regulation issue. The power costs in Kurdistan is performed. Section 7 pres-
system control of the future will require a high ents a dynamic analysis on the impact of a high
degree of flexibility and intelligence to ensure that wind power penetration on the Kurdistan electric
it can continuously balance fluctuating power and network and introduces an appropriate control
regulate frequency/voltage deviation caused by solution for its stability improvement. Finally,
renewable energy sources such as wind (Bevrani, conclusion and future research directions are
et al., 2011). presented in sections 8 and 9, respectively.
This chapter presents an overview of new dy-
namical challenges in regional electric networks,
following a high penetration of wind power. The BACKGROUND
Kurdistan electric network in Iran is considered
as a case study. Mountainous environment, costly In order to clarify the interaction behavior between
process for electricity production from conven- wind farm(s) and the power system, building of an
tional sources, and numerous windy areas make effective dynamic model for wind power systems
Kurdistan as an appropriate region for installa- (WPSs) is needed. Model simplifications and some
tion of wind farms. In this work, an analytical comparisons between different types of WPSs
approach is conducted to evaluate the potential and wind farm equivalent models are presented
of wind power installation and overall capacity in recent performed research works (Mansouri,
estimation, and to study economic issues based et al., 2004; Ekanayake, et al., 2003; Slootweg,
on the practical data. et al., 2003; Akhmatov, et al., 2006; Fernandeza,
The impact of high penetration wind power on et al., 2006; Ledesma & Usaola, 2005).
the system dynamic and performance for different The role of WPS control strategy to qualify
wind turbine technologies including fixed-speed system output and stability augmentation is stud-
induction generator (FSIG), doubly-fed induction ied in many papers. Optimization control, power
generator (DFIG) and permanent magnet syn- smoothing and voltage control of WPSs are most
chronous generator (PMSG) is presented. Using important topics of related new research areas
DIgSILENT simulation software, the stability of (Senjyu, et al., 2006; Wang & Chang, 2004).
the integrated system is re-analyzed, and the need Increasing the penetration of wind turbine gen-
for revising of conventional controls and perfor- erators in a power system may affect the system
mance standards is emphasized. Finally, a control security/stability limits, frequency, voltage and
approach to improve the system stability using dynamic behavior (Muyeen, et al., 2009; Bev-
static synchronous compensator (STATCOM) and rani, 2009; Slootweg, 2003; Bevrani, Tikdari, &
energy storage devices is addressed. This work Hiyama, 2010). This effect can be mostly caused
is supplemented by some nonlinear simulations by fluctuation of wind power. The impacts of
on the Kurdistan power system case study using wind turbines on the power system frequency
real data and parameters. and voltage have been studied in many research
199
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
works (Jowder, 2009; Radics & Bartholy, 2008; The revised operating performance standards
Bevrani & Tikdari, 2010). Power system frequency require that most type of power plants support
response model in the presence of high wind power the electricity network throughout their opera-
penetration, frequency control issue, and a com- tion. Important key issues can be considered as
prehensive survey with some new perspectives steady state and dynamic active/reactive power
are already well addressed (Bevrani & Hiyama, capability, continuously acting frequency/voltage
2011; Bevrani, Ghosh, & Ledwich, 2010; Bevrani, control and fault ride through behavior. Some
Daneshfar, & Daneshmand, 2010). commonly used turbine designs have some limits
The effects of DFIG and induction generator in terms of achieving grid code compliance in
type of WPSs on the voltage transient behaviors several countries. For the wind farms containing
are explained and the disadvantages of the induc- these turbines, additional equipments are needed
tion generator type are shown in (Nunes, et al., (Maibach, et al., 2007).
2004). The loadability of various types of WPSs Variable generation technologies generally
is studied and it is shown that the DFIG has larger refer to generating technologies whose primary
loadability than induction generators (Bevrani & energy source varies over time and cannot
Tikdari, 2010). Frequency nadir in the presence reasonably be stored to address such variation.
of different types of the WPSs has been also Uncertainty and variability are two major factors
compared in (Erlich, et al., 2006; Gillian, et al., of a variable generator that distinguish it in con-
2005). As argued in the mentioned references, wind ventional forms of generation and may impact the
turbines affect frequency behavior because they overall system planning and operations (Bevrani
add amount of inertia to the power system. Both & Hiyama, 2011).
stator and rotor windings of induction generator In order to specify wind power potential in
type of WPSs are directly connected to the power a particular site, a long-term record of wind
grid, but in DFIG type, only stator is directly con- speed has to be statistically analyzed. There are
nected and the rotor is linked through a power several studies related to the determination of
electronic type converter. The induction generator wind characteristics and wind power potential in
WPS in turn adds much inertia than DFIG in the many countries over the whole world (Radics &
power system; and in conclusion, the induction Bartholy, 2008; Elamouri, Ben, & Amar, 2008;
generator WPS frequency response is better than Al-Abbadi, 2005; Jowder, 2009; Ucar & Balo,
systems with DFIG type in the same conditions. 2009; Weigt, 2009)
Continuous increase of installed wind power
during recent years has forced the system opera- State of Wind Power Generation
tors and responsible organizations to tighten the
performance standards and connection rules – At present, wind power has effective impact on
known as grid code - in order to limit the effects energy markets. In 2009, more than 38.3 GW of
of wind power penetration on the power system new wind power capacity was installed around the
performance and stability. world that bringing the total installed capacity up to
Interconnection procedures and standards 158.5 GW. The main markets driving this growth
need to be reviewed to ensure that the new op- rate are Asia, North America and Europe. The top
erating control schemes and their responses are five countries in wind energy installed capacity
in a consistent manner to all power generation in 2009 were US with 35064 MW (22.1%), PR
technologies, including wind generating units as China with 25805 MW (16.3%), Germany with
variable generation technologies. 25777 MW (16.3%), Spain with 19149 MW
(12.1%), and India with 10926 MW (6.9%); while
200
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
the total installed wind power in the rest of world 1. Fixed speed wind turbine (FSWT)
was 41784 MW (26.4%) (GWEC, 2010). 2. Variable speed wind turbine (VSWT)
Similar to the most countries in the Middle
East and Africa, contribution of the wind energy Major characteristics of the FSWT are brush-
for production of electricity in Iran is relatively less and rugged construction, low cost and simplic-
low. The total installed capacity of wind turbine ity. The main advantage of the VSWT is that more
in Iran by 2010 is less than 200 MW. There are energy can be extracted for a specific wind speed
many suitable areas for installing wind turbines regime. In addition, the mechanical stress is less;
in Iran; however as depicted in Table 1, the major because the rotor acts as a flywheel (Slootweg,
installed wind power is centralized in Manjil and 2003). Common VSWT structures are known as
Binalud areas. DFIG and the PMSG. A FSWT is usually directly
Because of existing high sources of oil and equipped with a grid coupled squirrel cage induc-
gas in a relatively low price in Iran, most of elec- tion generator whose speed variations are limited.
tricity has been produced by fossil fuel in the past. The power extracted from the wind energy by
However, nowadays for many reasons, as well as a wind turbine can be expressed as follows (Heier,
other countries there is a great concern towards 1998; Bansal, et al., 2002):
renewable energies like wind and solar. The gov-
ernment and other responsible organizations have 1
Pm = ρAVw 3C p (λ, β) (1)
put some efforts to expand wind and solar farms 2
in different parts of country. Currently, the poten-
tial for wind power generation is estimated to be where, Pm is the power extracted from the wind,
more than 6500MW. p is the air density (Kg/m3), A is the rotor disc
area (m2), Vw is the wind speed (m/s), and Cp is
Wind Power Controls a power coefficient which is a function of the tip
speed ratio λ and the pitch angle of rotor blades
A WPS transforms the energy presented in the β. The tip speed ratio λ is defined by
belonging wind into electrical energy. A general
scheme of this system is shown in Figure 1a. Wind ωr R
energy is transformed into mechanical energy by λ= . (2)
Vw
wind turbine units. Based on rotational speed, the
wind turbines can be split into two types:
201
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Figure 1. Wind energy conversion system; a) General scheme of WPS, and b) Drive train model
where, ωr is mechanical angular velocity of the the generator electrical torque. Finally, Kmcand
turbine rotor and R is the blade radius of the wind Dmc are the stiffness and damping of mechanichal
turbine. coupling, repectively.
As shown in Fig. 1b, the drive train model of Power extracted from a wind turbine can be
wind turbines is usually represented by two mass controlled in two states, in above and below rated
models (Slootweg, Haan, Polinder, & Kling, 2003; wind speed of wind turbine. In the above rated
Slootweg, Polinder, & Kling, 2003): wind speed, a blade pitch angle controller reduces
the power coefficient and thus the power extracted
d ωr from the wind. The pitch controller limits the
Tw − Tm = J r (3)
dt generator’s speed to a rated value (ωgen, rated) by
adjusting the pitch angle (β).
Second control state (below rated wind speed)
Tm = Dmc (ωr − ωg ) + K mc ∫ (ωr − ωg )dt
exists only for the VSWT generator type. The aim
(4) is to control the rotational speed to follow the maxi-
mum power point trajectory (MPPT), when wind
d ωg speed is in change. Since, precise measurement
Tm − Tg = J g (5)
dt of wind speed is difficult, for maximum power
point tracking operation, it is better to use the rotor
where, Jr, and Jg are inertia of wind turbine and speed as a control input instead of wind speed.
generator, ωg is the rotor speed, Tm is the me-
chanical torque from the generator shaft, Tg is
202
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Modeling and Control of DFIG active and reactive powers. The MPPT unit pro-
vides the reference signal PGrid,ref for the active
The control strategy that generally applied to power, while the reactive power QGrid,ref is typi-
control of VSWT is based on vector control tech- cally fixed at zero.
niques. An overview of dynamic model for DFIG The grid-side converter controller operates in
wind turbine and the associated control system is a grid side converter voltage oriented reference
shown in Figure 2a (Hansen, et al., 2003; Hansen, frame (DIgSILENT GmbH, 2003). Active and
et al., 2004). reactive components of the grid-side converter
The rotor-side converter operates in a stator currents are controlled by the fast inner control
flux reference frame that decomposes the rotor loop. The slower outer control loop determines the
current into active power (q-axis) and reactive q-current set point, which regulates the DC-voltage
power (d-axis) components. A fast inner current to a pre-defined value. For achieving unity power
control loop controls rotor current in d- and q- factor operation of converter, it is sufficient that
axis and a slower outer control loop regulates q-current to be regulated to zero.
Figure 2. Overall structure and main control loops of VSWT systems a) DFIG, and b) PMSG
203
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Modeling and Control of PMSG districts and farm land with many windbreaks
(Class 3), and dense urban or forest (Class 4) are
An overview for dynamic model of the PMSG wind determined as 0.0002 m, 0.03 m, 0.1 m, 0.4 m,
turbine and its control system is shown in Figure and 1.6 m, respectively. In this work, since the
2b. The control structures and related concepts studied stations are located in near cities or even
are well discussed in the recent published works inside cities and due to mountainous environment
(Conroy & Watson, 2008). of Kurdistan, the roughness lengths for most sta-
At the generator-side converter, AC-voltage tions are fixed at 1 (Class 1).
and active power are regulated (reactive power
regulation is optional). The grid-side converter Site Selection
operates in a stator voltage oriented reference
frame. A fast inner control loop regulates the d- Gillbert (2004) presented a standard for using
and q-axis current components of the grid-side in site selection in a wind farm installation pro-
PWM-converter. Current references are defined by cedure. Based on this standard and according
a slower outer control-loop regulating DC-voltage to the annual mean wind speed from 2004 to
of the intermediate DC-circuit and reactive power. 2008, among considered six cities of Kurdistan
(Bijar, Qorveh, Marivan, Saqez, Sanandaj, and
Wind Energy Potential Divandarreh), regions of Bijar and Divandarreh
Assessment in Kurdistan are determined as more fair places for wind farm
installation (Table 2).
In this section, to determine the potential of wind
power generation the hourly measured wind speed Monthly Variation of Mean Wind Speed
data over a period of almost 5 years between 2004
and 2008 from 6 stations in Kurdistan, at 10 m Figure 3 shows monthly variation of mean wind
height that obtained from Kurdistan Meteorologi- speed for selected two sites (Divandarreh and
cal Organization are statically analyzed. Extrapo- Bijar). In both stations, the highest monthly wind
lation of the 10 m data, using the power law, is speed occurs in March. In Bijar, the lowest wind
used to determine the wind data at upper heights. speed happens in January; while for Divandarreh
The power law used in this study is as follow: it happens in December.
204
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Table 2. Annual mean speed at 50 m height from the ground for different locations
Wind power class Annual mean wind Height from sea Longitude Latitude Location
speed (m/s) level (m)
Min Deg Min Deg
Class 3 6.73 1883.4 37 47 53 35 Bijar
Class 1 5.58 1906.0 48 47 10 35 Qorveh
Class 1 3.06 1286.8 12 46 31 35 Marivan
Class 1 4.25 1522.8 16 46 15 36 Saqez
Class 1 3.40 1373.4 0 47 20 35 Sanandaj
Class 3 6.72 2142.6 55 46 4 36 Divandarreh
Wind Power Installation and estimated using the levelised cost of electricity
Economic Issues (LCOE) method. For this purpose, the weibul
distribution is obtained for these sites. Wind fre-
In this section, the economic evaluation for quency distributions for Bijar and Divandarreh at
installing four wind turbines in capacity of 0.8, 60 m height are shown in Figure 5.
1.5, 2 and 3 MW, for Bijar and Divandarreh are
Figure 3. Average wind speeds for different months based on the recorded data from 2004 to 2008, for
a) Divandarreh, and b) Bijar
205
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Figure 4. Wind rose diagrams based on the recorded data from 1992 to 2006, for a) Bijar, and b) Di-
vandarreh
206
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Figure 5. Frequency distributions of wind speed at 60 m height for a) Bijar, and b) Divandarreh
1
VR V F where, γ is the incomplete gamma function (Jan-
∫V f (V )dV + ∫ f (V )dV (8)
3
CF = 3 gamshetti & Rau, 1999; Suresh, et al., 2001).
VR VC VR
207
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
LCOE =
C wt Rwt + C bb Rbb + C ci Rci + C in Rin + C misc Rmisc + C om The estimation of the KWh cost of energy
EP
delivered by the WPS operating at the given sites
$/KWh (10)
has been done under the following assumptions:
Here, the Ep is annual energy production by
1. The lifetime of the WPS (n) is assumed to
delivered WPS, Cwt is cost of wind turbine, Cbb
be 25 years.
is cost of battery bank, Cci is the civil work and
2. The discount rate (r) is taken as 12%.
installation cost, Cin is cost of the inverter, Cmisc
3. Operation and maintenance cost (Com) is
is miscellaneous costs such as connecting cables,
considered to be 2% of initial capital cost
control panel and other components; and Com is
of the WPS project (Nouni, et al., 2006)
annual operation and maintenance cost. The Rwt,
4. Useful lifetime for the battery bank and
Rbb, Rci, Rin and Rmisc present the capital recovery
inverter are assumed to be 7 and 10 years,
factors (R) for wind turbine, battery bank, civil
respectively (Nouni, et al., 2006).
work and installation, inverter and other miscel-
5. It is assumed that the WPS production is equal
laneous components, respectively.
to the amount of energy output in each year
For a given discount rate (r) and useful system
during its useful lifetime (Türksoy, 1995).
lifetime (n), the capital recovery factor can be
defined as follows:
The results of cost analysis performed in this
study for the WPS with different size ranges are
(1 + r )n r
R= (11) presented in Table 3. From this table, it is seen
(1 + r )n − 1 that the predicted maximum and minimum values
regarding electricity cost per kWh for each WPS
A break-up of relative costs for different com- are calculated by taking into account the limit val-
ponents of a typical WPS can be easily obtained ues of the band interval of WPS specific cost. The
(Nouni, et al., 2006). The cost evaluation is made minimum levelised cost of electricity is calculated
by means of this cost break-up for all WPSs. A that WPS- Vestas V112 (3 MW) is 0.074 $/kWh,
typical cost table for different wind power tech- while its maximum value is 0.118 $/kWh. These
nologies is presented (Sathyajith, 2006). A specific values are the predicted lowest values for WPS
cost of WPS can be calculated as follows, in both cases of Divandarreh and Bijar.
According to the all band intervals, the highest
CWPS=IWPS PR [$] (12) electricity costs are calculated in the case of WPS-
Gamesa G90 (2 MW) in Bijar, as 0.116 $/kWh
Where, the IWPS is the specified cost of the WPS. for lower-limit and as 0.186 $/kWh for upper-
Table 3. Cost analysis per kwh for WPS in Bijar and Divandarreh
208
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
limit. As seen from the calculation of annual Detailed system information and power system
energy production, the WPS of 3 MW rated parameters are given in (Saleh, 2010).
power among the WPS considered in the study is
most attractive in terms of the levelised unit cost. Simulation Results
Dynamic Impacts Analysis In this study, the power system simulation pro-
and Stability Improvement gram, Power Factory (DIgSILENT), is used as
a suitable tool for power system modeling and
Case Study simulation. In the simulation environment, the
conventional power plant exciter is represented
In this section, transient stability of Kurdistan using the standard model EXST1; the power
electric network in the presence of two wind system stabilizer is represented using a dual-input
farms in Bijar and Divandarreh are analyzed. power system stabilizer model (PSS2A), and
A combination of FSIG and DFIG turbines are governor-turbine is represented by the standard
used in the mentioned wind farms. Single line model GAST. Load model is represented using a
diagram of Kurdistan network with wind farms static model. The voltage dependencies on active
is shown in Figure 6. Two 50-MW wind power and reactive powers are considered as 1 and 2,
plants are added to 63 KV bus, near to the cities respectively.
of Divandarreh and Bijar. It is noteworthy that in For transient stability investigation, a three
Kurdistan, only there is one conventional power phase short circuit with duration of 0.34 sec is
plant (Sanandaj Power Plant) with above 200 MW. considered in an important location. The fault
location is shown in Figure 6. Figure 7 depicts
Figure 6. Single-line diagram of the Kurdistan electric network with two wind farms
209
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
system response including voltages at Divandarreh decrease in related bus voltages with a permanent
and Bijar buses (connected to the wind farms), oscillation in active power and speed of aggre-
rotor angle of Sanandaj power plant, WPS’s speed, gated generators are indicated. It is shown that in
and active and reactive powers. view point of reactive power compensation, the
As discussed, the WPSs commonly use the Kurdistan grid is much weaker in Bijar than Di-
induction generators to convert the wind energy vandarreh area.
into electrical energy. The induction generators
act as reactive power consumers. Therefore, the Stability Improvement
system voltage would be affected in the presence
of wind turbines, especially in the case of fixed- For the sake of system frequency and real power
speed type of WPSs. This issue can be also seen compensation in the presence of WPSs, several
from simulation results shown in Figure 7. A control approaches are already presented (Bev-
210
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
rani & Hiyama, 2011; Bevrani, Daneshfar, & STATCOM provides effective voltage support
Daneshmand, 2010). In order to reactive power at the bus which is connected to the Bijar wind
compensation, which is the main control issue in farm. That is why, for stability improvement of the
the present work, traditionally the capacitor banks example at hand, a 30-Mvar STATCOM is con-
are suggested to use. However, it is noteworthy nected to 63 KV bus, near to the Bijar wind farm.
that capacitor banks cannot provide dynamic Another reason for choosing the mentioned
compensation for events such as the sudden drop place is that the location of the reactive power
of voltage. In response to above challenge, to support should be as close as possible to a point
improve stability after grid disturbances such as at which the support is more needed. Furthermore,
short circuit faults, the STATCOM technology as in the present case study, in addition to the losses
a powerful control tool is examined. reduction and increase of power transfer capabil-
The STATCOM is extensively being used in ity, the location of the STATCOM to the center
power systems because of their ability to provide of averaged load is more appropriate because the
flexible power flow control (Muyeen, et al., 2005). impact of voltage change is more significant at
The main motivation for choosing STATCOM this point.
in wind farms is their ability to provide bus bar But it is notable that the shipping of reactive
system voltage support either by supplying and/ power at low voltages in the system running close
or absorbing reactive power into the system. The to its stability margin is not very efficient. Also,
applicability of a STATCOM in wind farms has the total amount of reactive power transfer avail-
been investigated and the results from early studies able will be influenced by the transmission line
indicate that it is able to supply reactive power power factor. Hence, the compensation devices
requirements of the wind farm under various oper- are always kept as close as possible to the center
ating conditions, for improving transient stability of equivalent load as the ratio ΔV/Vnominal will
(Chun, et al., 2000), as well as enhancement of be higher for the load bus under fault conditions
the steady-state stability margin (Saad-Saoud, et (Prabhakar, 2008).
al., 1998). Regarding the grid codes mentioned The system response in the presence of STAT-
in Section 2, it is also investigated that the me- COM for the accrued fault is shown in the Figure
dium voltage STATCOM technology which adds 8. Results show a considerable improvement in
the missing functionality to wind farms in order transient stability. The transient behavior of wind
to become grid code compliant. Especially, the farms are also improved by injecting large amounts
voltage control and the fast dynamic behavior of reactive power during the fault recovery.
during balanced as well as unbalanced grid faults Flexibility in voltage control for power quality
(fault ride-through) are highlighted (Maibach, et improvement, fast response, and applicability for
al., 2007). use with high power/load fluctuation are the main
An appropriately sized STATCOM can provide advantages of the proposed STATCOM-based
the necessary reactive power compensation when control strategy.
connected to a weak grid. Also, a higher rating
STATCOM can be used for efficient voltage con-
trol and improved reliability in the interconnected FUTURE RESEARCH DIRECTIONS
grid with wind farms. However, it is noteworthy
that the STATCOM rating is limited by economic Some important research needs in future can be
issues. The location of STATCOM is generally summarized as follows:
chosen to be a point in the system which needs
reactive power. Simulation results show that
211
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
212
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
the regulation capabilities of the active/re- • The wind turbine units must meet technical
active power of WPSs. requirements with respect to the voltage,
• Control performance standards compliance frequency, ability to rapidly isolate faulty
verification remains a major open issue for parts from the rest to the network, and have
wind power units. This concerns specific a reasonable ability to withstand abnormal
WPS capabilities and will require the de- system operating conditions. They could
velopment of additional standards for test- be able to function effectively as part of
ing, from the level of the component up to the existing electricity industry particular-
the entire WPS. ly during abnormal power system operat-
• Advanced computing algorithm and fast ing conditions when power system security
hardware measurement devices are also may be at risk.
needed to realize more effective optimal/
adaptive control schemes for the power High wind power penetration, particularly in
systems with a high penetration of WPSs. the locations far away from major load centers and
• Since, naturally the wind power is stochas- existing conventional generation units increases
tic, still it is difficult to straightly use wind the risk of tie-line overloading, and may require
turbine kinetic energy storage in the regu- network augmentation, and possibly additional
lation tasks such as frequency control. The interconnections to avoid flow constraints. With
contribution of WPSs in active power and increasing wind power penetration, the grid codes
frequency regulation refers to the ability for the connection high wind turbines capacity
of these units to regulate their power out- should be also updated (Bevrani, et al., 2011).
put (Bevrani, et al., 2010). More effective
practical algorithms and control method- • Furthermore, the updating of existing
ologies are needed to perform these issues. emergency frequency control schemes for
Further studies are needed to coordinate N-1 contingency, economic assessment/
the timing and the size of the kinetic en- analysis of the frequency regulation prices,
ergy discharge with the characteristics of further study on frequency and voltage sta-
conventional plants. bility using dynamic demand control and
• To allow the increase of wind power pen- ratios of wind turbine technologies, and
etration, a change in regulation reserve quantification of reserve margin due to in-
policy may be required. In this direction, creasing wind power penetration (Bevrani,
in addition to the deregulation policies, the et al., 2010) can be considered as other im-
amount and location of wind turbines, gen- portant research needs in future.
eration technology, and the size and char-
acteristics of the electricity system must be
considered as important technical aspects. CONCLUSION
• Continuous development of communica-
tions and information technology, as well In this chapter an intensive overview of wind en-
as market and regulatory frameworks for ergy status around the world and Iran is presented.
generation and consumption is necessary The dynamic model and the main control loops
for a power system with intelligent elec- of wind turbine technologies are explained. As
tricity meters and intelligent communica- a practical case study, the wind power potential,
tions (Bevrani, et al., 2011). economic issues, and technical challenges for
a high penetration of wind power in Kurdistan
213
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
electric network are discussed. The possibility Bevrani, H. (2009). Robust power system
of connecting a STATCOM to the wind power frequency control. New York, NY: Springer.
system in order to provide an efficient control doi:10.1007/978-0-387-84878-5
method is explored. The STATCOM as a pure
Bevrani, H., Daneshfar, F., & Daneshmand,
static device with no switched passive components,
P. (2010). Intelligent power system frequency
which provides outstanding performance for both
regulation concerning the integration of wind
steady state and dynamic operation is used as a
power units. In Wang, L. F., Singh, C., & Kusiak,
suitable control solution to decrease the undesir-
A. (Eds.), Wind power systems: Applications of
able impact of wind power plants on the transient
computational intelligence (pp. 407–437). Hei-
stability and to improve the system performance.
delberg, Germany: Springer-Verlag.
It is shown that the proposed STATCOM based
design strategy provides dynamic voltage control Bevrani, H., Ghosh, A., & Ledwich, G. (2010).
and power oscillation damping, and improves the Renewable energy sources and frequency regula-
Kurdistan network transient stability. tion: Survey and new perspectives. IET Renewable
Power Generation, 4(5), 438–457. doi:10.1049/
iet-rpg.2009.0049
ACKNOWLEDGMENT
Bevrani, H., & Hiyama, T. (2011). Intelligent
automatic generation control. CRC Press.
This work is supported by Department of Electri-
cal Engineering at University of Kurdistan. The Bevrani, H., & Tikdari, A. (2010). An ANN-based
authors would like to thank Mr. Naji Ghaderne- power system emergency control scheme in the
zhad from West Regional Electric Co., and Mr. presence of high wind power penetration. In
Chareh-Khah from Kurdistan Meteorological Wang, L. F., Singh, C., & Kusiak, A. (Eds.), Wind
Organization for their help and useful comments. power systems: Applications of computational
intelligence (pp. 215–254). Heidelberg, Germany:
Springer-Verlag.
REFERENCES
Bevrani, H., & Tikdari, A. (2010). On the neces-
Akhmatov, V., Knudsen, H., Nielsen, A. H., sity of considering both voltage and frequency in
Pedersen, J. K., & Poulsen, N. K. (2006). Mod- effective load shedding schemes. In Proceedings of
eling and transient stability of large wind farms. IEEJ Technical Meeting, (pp. 7-11). Fukui, Japan.
International Journal of Electrical Power & Bevrani, H., Tikdari, A., & Hiyama, T. (2010).
Energy Systems, 25, 123–144. doi:10.1016/S0142- An intelligent based power system load shedding
0615(02)00017-0 design using voltage and frequency information.
Al-Abbadi, N. M. (2005). Wind energy resource Int. Conf on Modeling, Identification and Control,
assessment for five locations in Saudi Arabia. (pp. 545-549). Okayama, Japan.
Renewable Energy, 30, 1489–1499. doi:10.1016/j. Chun, L., Qirong, J., & Jianxin, X. (2000). Inves-
renene.2004.11.013 tigation of voltage regulation stability of static
Bansal, R., Bhatti, T., & Kothari, D. (2002). synchronous compensator in power system. IEEE
On some of the design aspects of wind energy Power Engineering Society Winter Meeting, (pp.
conversion systems. Energy Conversion and 4, 2642-47).
Management, 43, 2175–2187. doi:10.1016/S0196-
8904(01)00166-2
214
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Conroy, J., & Watson, R. (2008). Aggregate Hansen, A., Jauch, C., Sørensen, P., Iov, F., &
modeling of wind farms containing full-converter Blaabjerg, F. (2003). Dynamic wind turbine models
wind turbine generators with permanent magnet in power system simulation tool DIgSILENT. Risoe
synchronous machines: Transient stability studies. Report R-1400(EN), Risø National Laboratory,
IET Renewable Power Generation, 3(1), 39–52. Denmark.
doi:10.1049/iet-rpg:20070091
Hansen, A., Sørensen, P., Iov, F., & Blaabjerg, F.
DIgSILENT GmbH. (2003). Dynamical modeling (2004). Control of variable speed wind turbines
of doubly-fed induction machine wind-generators. with doubly-fed induction generators. Wind Engi-
DIgSILENT Technical Documentation. Retrieved neering, 28(4). doi:10.1260/0309524042886441
from www.digsilent.com
Heier, S. (1998). Grid integration of wind energy
Ekanayake, J., Holdsworth, L., & Jenkins, N. conversion systems. Chicester, UK: John Wiley
(2003). Comparison of 5th order and 3rd order & Sons.
machine models for doubly fed induction gen-
Jangamshetti, S. H., & Rau, V. G. (1999). Site
erator (DFIG) wind turbines. Electric Power
matching of wind turbines generator: A case study.
Systems Research, 67, 207–215. doi:10.1016/
IEEE Transactions on Energy Conversion, 14(4),
S0378-7796(03)00109-3
1537–1543. doi:10.1109/60.815102
Elamouri, M., & Ben Amar, F. (2008). Wind
Jowder, F. (2009). Wind power analysis and site
energy potential in Tunisia. Renewable Energy,
matching of wind turbine generators in King-
33, 758–768. doi:10.1016/j.renene.2007.04.005
dom of Bahrain. Applied Energy, 86, 538–545.
Erlich, I., Rensch, K., & Shewarega, F. (2006). Im- doi:10.1016/j.apenergy.2008.08.006
pact of large wind power generation on frequency
Lalor, G., Mullane, A., & O’Malley, M. (2005).
stability. In Proc. IEEE Power Engineering Society
Frequency control and wind turbine technologies.
General Meeting. Montereal, Que., Canada.
IEEE Transactions on Power Systems, 20(4),
Fernandeza, L. M., Saenza, J. R., & Jurado, F. 1905–1913. doi:10.1109/TPWRS.2005.857393
(2006). Dynamic models of wind farms with fixed
Ledesma, P., & Usaola, J. (2005). Doubly fed
speed wind turbines. Renewable Energy, 31(8),
induction generator model for transient stability
1203–1230. doi:10.1016/j.renene.2005.06.011
analysis. IEEE Transactions on Energy Conver-
Gilbert, M. M. (2004). Renewable and efficient sion, 20, 388–397. doi:10.1109/TEC.2005.845523
electric power systems. Hoboken, NJ: John Wiley
Maibach, P., Wernli, J., Jones, P., & Obad, M.
& Sons.
(2007). STATCOM technology for wind parks
Gokcek, M., & Genc, M. G. (2009). Evaluation to meet grid code requirements. European Wind
of electricity generation and energy cost of wind Energy Conference-EWEC. Milan.
energy conversion Evaluation of electricity gen-
Mansouri, M. N., Mimouni, M. F., Benghanem,
eration and energy cost of wind energy conversion.
B., & Annabi, M. (2004). Simulation model
Applied Energy, 86, 2731–2739.
for wind turbine with asynchronous generator
GWEC. (2010). Global wind 2009 report. Re- interconnected to the electric network. Renew-
trieved from http://www.gwec.net able Energy, 29, 421–431. doi:10.1016/S0960-
1481(03)00225-8
215
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Muyeen, S., Mannan, M., Ali, M., Takahashi, R., Senjyu, T., Sakamoto, R., Urasaki, N., Funabashi,
Murata, T., & Tamura, J. (2005). Stabilization of T., Fujita, H., & Sekine, H. (2006). Output power
Grid connected wind generator by STATCOM. leveling of wind turbine generator for all operating
IEEE Power Electronics and Drives Systems, 2. regions by pitch angle control. IEEE Transactions
on Energy Conversion, 21, 467–457. doi:10.1109/
Muyeen, S., Tamura, J., & Murata, T. (2009).
TEC.2006.874253
Stability augmentation of a Grid-connected wind
farm. London, UK: Springer-Verlag. Slootweg, J., Polinder, H., & Kling, W. (2003).
Representing wind turbine electrical generating
Nouni, M. R., Mullick, S. C., & Kandpal, T. C.
systems in fundamental frequency simulation.
(2006). Techno-economics of small wind elec-
IEEE Transactions on Energy Conversion, 18(4),
tric generator projects for decentralized power
516–524. doi:10.1109/TEC.2003.816593
supply in India. Energy Policy, 35, 2491–2506.
doi:10.1016/j.enpol.2006.08.011 Slootweg, J. G. (2003). Wind power: Modelling
and impact on power system dynamics. PhD The-
Nunes, M. V., Lopes, J. A., Zürn, H. H., Bezerra,
sis, Delft University of Technology, Netherlands.
U. H., & Almeida, R. G. (2004). Influence of
the variable-speed wind generators in transient Slootweg, J. G., Haan, S. W., Polinder, H., &
stability margin of the conventional generators Kling, W. L. (2003). General model for repre-
integrated in electrical grids. IEEE Transactions on senting variable speed wind turbines in power
Energy Conversion, 19(4), 692–701. doi:10.1109/ system dynamics simulations. IEEE Transactions
TEC.2004.832078 on Power Systems, 18, 144–151. doi:10.1109/
TPWRS.2002.807113
Prabhakar, A. (2008). Application of STATCOM for
improved dynamic performance of wind farms in Suresh, H., Janagmshetti, & Guruprasadu, R. V.
a power grid. Master Thesis, Missouri University (2001). Normalized power curves as a tool for
of Science and Technology. identification of optimum wind turbine genera-
tor parameters. IEEE Trans Energy Conv., 16(3),
Radics, K., & Bartholy, J. (2008). Estimating and
283-8.
modelling the wind resource of Hungary. Renew-
able & Sustainable Energy Reviews, 12, 874–882. Türksoy, F. (1995). Investigation of wind power
doi:10.1016/j.rser.2006.10.009 potential at Bozcaada, Turkey. Renewable
Energy, 6(8), 917–923. doi:10.1016/0960-
Saad-Saoud, Z., Lisboa, M., Ekanayake, J.,
1481(95)00091-7
Jenkins, N., & Strba, G. (1998). Application of
STATCOMs to wind farms. IEE Proceedings. Ucar, A., & Balo, F. (2009). Evaluation of wind
Generation, Transmission and Distribution, 145, energy potential and electricity generation at
1584–1589. doi:10.1049/ip-gtd:19982178 six locations in Turkey. Applied Energy, 86,
1864–1872. doi:10.1016/j.apenergy.2008.12.016
Saleh, M. (2010). Dynamic analysis of Kurdistan
electric network in the presence of high penetra- Wang, Q., & Chang, L. (2004). An intelligent
tion of wind power and determine an appropriate maximum power extraction algorithm for inverter-
control solution. Master Thesis, University of based variale speed wind turbine systems. IEEE
Kurdistan, Iran. Transactions on Power Electronics, 19, 1242–
1249. doi:10.1109/TPEL.2004.833459
Sathyajith, M. (2006). Wind energy fundamentals,
resource analysis and economics. Berlin/Heidel-
berg, Germany: Springer-Verlag.
216
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Weigt, H. (2009). Germany’s wind energy: The Chen, Z., Guerrero, J. M., & Blaabjerg, F. (2009). A
potential for fossil capacity replacement and Review of the State of the Art of Power Electronics
cost saving. Applied Energy, 86(10), 1857–1863. for Wind Turbines. IEEE Transactions on Power
doi:10.1016/j.apenergy.2008.11.031 Electronics, 24(8), 1859–1875. doi:10.1109/
TPEL.2009.2017082
Cong, L., & Wang, Y. (2002). Co-ordinated control
ADDITIONAL READING of generator excitation and STATCOM for rotor
angle stability and voltage regulation enhancement
Akhmatov, V., & Eriksen, P. B. (2007). A Large of power systems. IET Gener. Transm. Distrib.,
Wind Power System in Almost Island Operation— 149(6), 659–666. doi:10.1049/ip-gtd:20020651
A Danish Case Study. IEEE Transactions on
Power Systems, 22(3), 937–943. doi:10.1109/ EL-Helw, H., & Tennakoon, S. B. (2008). Evalu-
TPWRS.2007.901283 ation of the suitability of a fixed speed wind
turbine for large scale wind farms considering the
Ali, M. H., & Wu, B. (2010). Comparison of new UK grid code. Renewable Energy, 33, 1–12.
Stabilization Methods for Fixed-Speed Wind doi:10.1016/j.renene.2007.08.010
Generator Systems. IEEE Transactions on
Power Delivery, 25(1), 323–331. doi:10.1109/ Fang, D., Yuan, S., Wang, Y., & Chung, T. (2007).
TPWRD.2009.2035423 Coordinated parameter design of STATCOM stabi-
liser and PSS using MSSA algorithm. IET Gener.
AouzellagLahaçani, N., D.Aouzellag, & B.Mendil. Transm. Distrib., 1(4), 670–678. doi:10.1049/
(2010). Static compensator for maintaining voltage iet-gtd:20060205
stability of wind farm integration to a distribution
network. Renewable Energy, 35, 2476–2482. Gautam, D., Vittal, V., & Harbour, T. (2009).
doi:10.1016/j.renene.2010.03.010 Impact of Increased Penetration of DFIG-Based
Wind Turbine Generators on Transient and Small
Arulampalam, A., Barnes, M., Jenkins, N., & Signal Stability of Power Systems. IEEE Trans-
Ekanayake, J. (2006). Power quality and stability actions on Power Systems, 24(3), 1426–1434.
improvement of a wind farm using STATCOM doi:10.1109/TPWRS.2009.2021234
supported with hybrid battery energy storage.
IEE Proceedings. Generation, Transmission and Gaztañaga, H., Etxeberria-Otadui, I., Ocnasu,
Distribution, 153(6), 701–710. doi:10.1049/ip- D., & Bacha, S. (2007). Real-Time Analysis of
gtd:20045269 the Transient Response Improvement of Fixed-
Speed Wind Farms by Using a Reduced-Scale
Arulampalam, A., Ekanayake, J., & Jenkins, N. STATCOM Prototype. IEEE Transactions on
(2003). Application study of a STATCOM with Power Systems, 22(2), 658–666. doi:10.1109/
energy storage. IET Gener. Transm. Distrib., TPWRS.2007.895153
150(3), 373–384. doi:10.1049/ip-gtd:20030232
Han, C., Huang, A. Q., Baran, M. E., Bhattacha-
Chang-Chien, L.-R., & Yin, Y.-C. (2009). Strate- rya, S., Litzenberger, W., & Anderson, L. (2008).
gies for Operating Wind Power in a Similar Manner STATCOM Impact Study on the Integration of a
of Conventional Power Plant. IEEE Transac- Large Wind Farm into a Weak Loop Power Sys-
tions on Energy Conversion, 24(4), 926–934. tem. IEEE Transactions on Energy Conversion,
doi:10.1109/TEC.2009.2026609 23(1), 226–233. doi:10.1109/TEC.2006.888031
217
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
Jangamshetti, S. H., & Rau, V. G. (2001). Normal- Prasad, R. D., Bansal, R. C., & Sauturaga, M.
ized Power Curves as a Tool for Identification of (2009). Wind Energy Analysis for Vadravadra
Optimum Wind Turbine Generator Parameters. Site in Fiji Islands: A Case Study. IEEE Trans-
IEEE Transactions on Energy Conversion, 16(3), actions on Energy Conversion, 24(3), 750–757.
283–288. doi:10.1109/60.937209 doi:10.1109/TEC.2009.2025326
Li, H., & Chen, Z. (2008). Overview of different Qiao, W., Harley, R. G., & Venayagamoorthy, G.
wind generator systems and their comparisons. IET K. (2009). --Coordinated Reactive Power Control
Renew. Power Gener., 2(2), 123–138. doi:10.1049/ of a Large Wind Farm and a STATCOM Using
iet-rpg:20070044 Heuristic Dynamic Programming. IEEE Trans-
actions on Energy Conversion, 24(2), 493–503.
Li, P., Keung, P.-K., & Ooi, B.-T. (2009). Devel-
doi:10.1109/TEC.2008.2001456
opment and simulation of dynamic. IET Renew.
Power Gener., 3(2), 180–189. doi:10.1049/iet- Rodríguez, J. M., Fernández, J. L., Beato, D.,
rpg:20070093 Iturbe, R., Usaola, J., & Ledesma, P. (2002).
Incidence on Power System Dynamics of High
Lu, M.-S., Chang, C.-L., Lee, W.-J., & Wang, L.
Penetration of Fixed Speed and Doubly Fed Wind
(2009). Combining the Wind Power Generation
Energy Systems: Study of the Spanish Case.
System With Energy Storage Equipment. IEEE
IEEE Transactions on Power Systems, 17(4),
Transactions on Industry Applications, 45(6),
1085–1095. doi:10.1109/TPWRS.2002.804971
2109–2115. doi:10.1109/TIA.2009.2031937
Tamrakar, I., Shilpakar, L., Fernandes, B., &
Mohod, S. W., & Aware, M. V. (2010). A
Nilsen, R. (2007). Voltage and frequency control
STATCOM-Control Scheme for Grid Connected
of parallel operated synchronous generator and
Wind Energy System for Power Quality Improve-
induction generator with STATCOM in micro
ment. IEEE Systems Journal, 4(3), 346–352.
hydro scheme. IET Gener. Transm. Distrib., 1(5),
doi:10.1109/JSYST.2010.2052943
743–750. doi:10.1049/iet-gtd:20060385
Muyeen, S. M., Takahashi, R., Ali, M. H., Mu-
Yang, Z., Shen, C., Zhang, L., Crow, M. L., & Atcit-
rata, T., & Tamura, J. (2008). Transient Stability
ty, S. (2001). Integration of a StatCom and Battery
Augmentation of Power System Including Wind
Energy Storage. IEEE Transactions on Power
Farms by Using ECS. IEEE Transactions on
Systems, 16(2), 254–260. doi:10.1109/59.918295
Power Systems, 23(3), 1179–1187. doi:10.1109/
TPWRS.2008.920082
Muyeen, S. M., Takahashi, R., Murata, T., & Tamu-
KEY TERMS AND DEFINITIONS
ra, J. (2009). Integration of an Energy Capacitor
System With a Variable-Speed Wind Generator. Power System Stability: The ability of a power
IEEE Transactions on Energy Conversion, 24(3), system to regain a state of operating equilibrium
740–749. doi:10.1109/TEC.2009.2025323 after being subjected to a physical disturbance,
Muyeen, S. M., Takahashi, R., Murata, T., & with most systems indices (voltage, angle, and
Tamura, J. (2010). A Variable Speed Wind frequency) bounded. Power system stability can
Turbine Control Strategy to Meet Wind Farm take three different forms of rotor angle, voltage,
Grid Code Requirements. IEEE Transactions on and frequency stabilities.
Power Systems, 25(1), 331–340. doi:10.1109/ Power System Control: This term is used to
TPWRS.2009.2030421 define the application of control theory and tech-
218
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
nology, optimization methodologies, and expert/ grid through a power electronic converter and the
intelligent systems to improve the performance rotor acts as a flywheel.
and functions of power systems during normal Static Synchronous Compensator (STAT-
and abnormal operations. COM): A technology being extensively used as
Capacity Factor: Capacity factor is defined as dynamic shunt compensator for reactive power
the ratio of the average power output to the rated control in transmission and distribution system.
output power of the wind energy converter system. Levelised Cost of Electricity (LCOE): The
Wind Rose Diagram: The wind rose diagram LCOE for WPSs can be described as the ratio of
illustrates the wind direction in a given site. the total annualized cost to the annual electricity
Fixed Speed Wind Turbine: Wind turbine produced by the system.
that is directly connected to the grid with a small Weibull Distribution: The probability dis-
speed variation of its rotor. tribution, which is widely used to describe the
Variable Speed Wind Turbine (VSWT): long-term records of wind speeds.
This type of wind turbine is decoupled from the
219
220
Chapter 7
Many-to-Many
Assignment Problems:
Lagrangian Bounds and Heuristic
Igor Litvinchev
Nuevo Leon State University, Mexico
Socorro Rangel
São Paulo State University, Brazil
ABSTRACT
Modified Lagrangian bounds and a greedy heuristic are proposed for many-to-many assignment problems
taking into account capacity limits for tasks and agents. A feasible solution recovered by the heuristic is
used to speed up the subgradient technique to solve the modified Lagrangian dual. A numerical study is
presented to compare the quality of the bounds and to demonstrate the efficiency of the overall approach.
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Many-to-Many Assignment Problems
ment problem for this type of problem was first to the objective function to discourage their viola-
introduced by Ross and Soland (1975). tion. The optimal value of the Lagrangian problem,
Cattrysse and Van Wassenhove (1992), Mo- considered for fixed multipliers, provides an upper
rales and Romeijn (2004) and Pentico (2007) in bound (for maximization problem) for the original
their survey papers identify a variety of applica- optimal objective. The problem of finding the
tions in which GAP has been used either directly best, i.e. bound minimizing Lagrange multipliers,
or as a sub-problem within a broader model type. is called the Lagrangian dual. Lagrangian bounds
When the tasks are jobs to be performed, and are widely used as a core of many numerical tech-
the agents are computer networks, we obtain niques for integer and combinatorial problems,
the problem described by Balachandran (1976). as well as to measure the progress of the main
Another example is the fixed-charge plant loca- algorithm and derive stopping criteria. In many
tion problem, where the agents are capacitated approximate and heuristic approaches Lagrang-
plants and the tasks are customers, where each ian solution is used as a starting or a reference
of the customer demands must be supplied from point to construct the algorithm. The literature
a single plant (Geoffrion and Graves (1974)). on Lagrangian relaxation is quite extensive, see,
Other applications that have been studied are the e.g.Lemarechal (2007), Frangioni (2005) and the
location problem (Ross and Soland (1977)), the references therein.
maximal covering location problem (Klastorin Frequently a complex system can be rep-
(1979)), various routing problems (Fisher and resented as a number of coupled subsystems.
Jaikumar (1981), Bookbinder and Reece (1988)), Accordingly, all constraints can be divided into
assignment in parallel and distributed computing binding and block ones, such that dualizing binding
(Pirkul (1986), Bokhari (1987)), R & D planning constraints results in a decomposable Lagrangian
problems (Zimokha and Rubinshtein (1988)), problem. In many cases there are different ways
loading problem in flexible manufacturing systems to specify subsystems thus resulting in different
(Kuhn (1995)), production planning (LeBlanc, decomposable Lagrangian problems for the same
Shtub, Anandalingam (1999)). original problem. For example, in different varia-
The real life applications of the assignment tions on the assignment problem we may consider
models frequently involve large number of tasks either tasks or agent’s constraints as binding ones.
and/or agents thus resulting in large-scale opti- Similar properties have routing problems, produc-
mization problems. However, most large-scale tion scheduling, location problems, to mention a
optimization problems exhibit a structure that can few (Lasdon, 2002).
be exploited to construct efficient solution tech- An approach to improve classical Lagrangian
niques. In one of the most general and common bound was proposed in Litvinchev (2007) and fur-
forms of a structure the constraints of the problem ther developed in Litvinchev, Rangel & Saucedo
can be divided into “easy” and “complicated”. (2010). The main idea of this approach is to get a
In other words, the problem would be an “easy” tighter estimation of the penalty (complementar-
problem if the complicating constraints could be ity) term arising in the classical Lagrangian func-
removed. For example, removing agent constraints tion. It is well known that under certain convexity
in the assignment problems results in independent and regularity conditions the penalty turns to zero
subproblems corresponding to tasks. for the optimal primal-dual pair (complementar-
A well-known way to exploit this structure is ity condition). However, for nonconvex (integer)
to form a Lagrangian relaxation with respect to problems the complementarity condition is not
complicating constraints. That is, the complicating necessarily fulfilled. An auxiliary optimization
constraints are relaxed and a penalty term is added problem is used to estimate the penalty term and to
221
Many-to-Many Assignment Problems
construct the modified Lagrangian bound and the Numerical results are given in Section 5, future
corresponding modified dual problem. Suppose research directions are presented in Section 6 and
the original problem has two interesting subset Section 7 concludes.
of constraints, i.e. both subsets considered sepa-
rately have attractive, while different structures.
Then we may relax one subset of constraints in BASIC CONSTRUCTIONS
the standard Lagrangian fashion, while using the TO DERIVE THE MODIFIED
other subset in the auxiliary problem to estimate LAGRANGIAN BOUNDS
the penalty term.
In this chapter we specify this approach for Consider the problem:
the many-to-many assignment problem recog-
nizing capacity limits for tasks and agents. From z * = max {cx | Dx ≤ d , Ax ≤ b, x ∈ U } ,
Lagrangian point of view it is important to note (1)
that relaxing task constraints gives independent
subproblems corresponding to agents, while relax- where x∈Rn and the set U can be of general
ing agent constraints results in tasks subproblems. structure and may contain, for example, sign con-
Thus both subproblems arising in the modified straints on x and integrality constraints on some
bound can be solved in a decomposable fashion. or all components of x. The constraints Dx≤d are
The modified Lagrangian bounds are numeri- m “complicating constraints”, while constraints
cally compared with classical ones and a greedy Ax≤b are considered “nice” in the sense that the
heuristic is applied to the (unfeasible) Lagrangian optimization problem formed with only these
solution to get an approximate solution to the constraints, together with x∈U, is easier than the
original problem. Combining the modified La- original problem. Denote by x* an optimal solu-
grangian bound with a greedy heuristic provides tion of (1) and let X={ x∈U | Ax≤b}.
high quality feasible solutions typically within Relaxing the “complicating constraints” and
0.5% of relative difference between primal and using u, an m-vector of Lagrange multipliers, we
dual bounds. If the subgradient scheme is used to can define the standard Lagrangian problem as:
solve the dual problem, we may restore feasibility
by the (computationally cheap) greedy algorithm z (u ) = max {cx + u(d − Dx ), x ∈ X , u ≥ 0} .
in each iteration of the subgradient technique and (2)
then use this feasible solution to update parameters
of the iterative scheme. Incorporating a feasible We assume for simplicity that it has an optimal
greedy solution into the subgradient scheme results solution for all u≥0and therefore we get the well
in a significant decrease in the number of itera- known Lagrangian bound:
tions without dropping the quality of the bounds.
The reminder of the chapter is organized as z * ≤ z (u ), for any u ≥ 0 . (3)
follows. In Section 2 we present the basic construc-
tions to derive the modified Lagrangian bounds
and give an illustrative example. The modified The best Lagrangian bound and the associated
bounds for many-to-many assignment problem are Lagrange multipliers u* are obtained by solving
specified in Section 3. Benders and the subgradient the Lagrangian dual problem:
techniques to solve the dual problem are presented
in Section 4, together with a greedy algorithm z (u * ) = min z (u ) ≡ w D . (4)
u ≥0
to restore feasibility of the Lagrangian solution.
222
Many-to-Many Assignment Problems
Suppose now that the original problem (1) has We will refer to W as the localization of x*, or
two interesting subproblems, i.e., both constraints simply the localization. The set W can be defined
Dx≤d and Ax≤b considered separately, have attrac- by manipulating the constraints of the original
tive, while different structures. In this case it may problem, by querying a decision maker, etc. We
be useful to reformulate the original problem prior will distinguish the case of a priori localization,
to relaxation. Introducing copy constraints,x=y, when the set W is defined before the Lagrangian
in the original problem it is possible to build a problem has been solved for some u≥0 and the
Lagrangian decomposition (Guignard and Kim, associated bound z(u) has been calculated, and
1987) by dualizing them with n multipliers λ∈Rn a posteriori localization, when W is defined or
thus obtaining an x-problem and an y-problem: corrected after the Lagrangian problem has been
solved.
z (λ) = max {(c − λ)x | x ∈ X } + max {λy | Dy ≤ d , y ∈ U } Consider a further modification of the original
x y
z * ≤ z (λ* ) = min z (λ) ≡ wLD (6) Since a pair (x,y)=(x*,x*) is feasible to (7),
λ
it is a relaxation of the original problem (1) and
hence z*=zM. Note also that if the localization W
and can strictly dominate the standard Lagrang- is such that y∈W implies Dy≤d (this is the case,
ian bound obtained by dualizing either set of for example, when constraints Dy≤d are included
constraints, Dx≤d or Ax≤b. explicitly in the definition of W) then an optimal
Instead of creating a copy y of the variable x x-solution to (7) is feasible to (1) and hence z*≥zM.
and imposing the constraint x=y, it is possible to Thus for such a localization we have z*=zM.
transform the original problem by introducing an Dualizing constraints Dx≤Dy with multipliers
aggregated copy constraint Dx=Cy, where Cis u≥0 we get the modified Lagrangian bound:
a matrix of suitable dimension. The aggregate
copy constraint is then dualized providing an z * ≤ z M ≤ z M (u ) = max {cx − uDx } + max {uDy }
aggregated Lagrangian decomposition bound x ∈X y ∈W
(8)
(Maculan & Reinoso, 1992), which, in general,
is weaker than wLD.
The modified bound zM(u) obtained for the
In what follows we consider constructions
localization W={y|Dy≤d,y∈U} coincides with
to improve the standard Lagrangian bound and
the aggregated Lagrangian decomposition bound
provide two interpretations of the approach: one
calculated for C=D.
using a reformulation of the original problem, and
To simplify further notations, we use η(u)for the
another one in terms of a subproblem to estimate
optimal objective value of the first maximization
the complementarity term.
problem in (8), and ξ(u) for the optimal objective
We assume certain information about an op-
of the second. The modified dual problem cor-
timal solution to (P), x*:
responding to the bound (MLB) is then stated as:
Assumption. A set W⊆Rn is known, such that
x*∈W.
wMD = min {η(u ) + ξ(u )} . (9)
u ≥0
223
Many-to-Many Assignment Problems
It is possible to derive (8) using other ar- Estimating the extended complementarity term
guments. The feasibility of x* to the standard
Lagrangian problem implies that cx*+u(d- u(d − Dx * ) + πUB (UB − cx * ) + πLB (LB − cx * )
Dx*)≤z(u). Since x* is feasible to (1) and u≥0, (12)
then u(d-Dx*)≥0 yields immediately z*≤z(u) as in
(3). The complementarity condition u*(d-Dx*)=0 and using the localization W as before, gives the
is fulfilled for convex problems (1) which fulfill bound
certain regularity assumptions. But even for the
convex case the term u(d-Dx*) can be strictly z * ≤ z M ≤ z M (u, π) = max {(1 − π)cx − uDx } + max {πcy + uDy )} .
positive for u≠u*. For the nonconvex case the
x ∈X y ∈W
(13)
complementarity term u(d-Dx*) can be strictly
positive for u=u*. Thus we may try to strengthen The unrestricted dual variable π stands
the standard Lagrangian bound z(u) using more for πLB−πUB. The bound zM(u,π) in (13) uses
tight estimations of the complementarity term m+1multipliers, an m-vector u and a scalar π,
instead of u(d-Dx*)≥0. while the Lagrangian decomposition bound, z(λ),
Since x*∈W, then we have for any localiza- uses an n-vector of multipliers, λ.
tion W The modified dual problem corresponding to
the bound (LMBπ) is stated as:
z (u ) ≥ cx * + u(d − Dx * ) ≥ cx * + min{u(d − Dy )}
y ∈W
such that:
where η(u,π)is used to denote the optimal objective
z ≤ max{cx + u(d − Dx )} − min{u(d − Dy )}
* value of the first maximization problem in zM(u,π),
x ∈X y ∈W
while ξ(u,π) stands for the optimal objective of
(11)
the second.
It is not hard to verify that aggregating to the
This bound, after elementary algebraic trans-
problem (7) the valid constraint cx=cyand dual-
formations, coincides with zM(u) defined in (8).
izing it together with constraints Dx≤Dy we get
We may expect that for those integer programs,
exactly the same expression for the bound as in
where the constraints Dx≤d are not active for all
(13).
feasible solutions, a reasonable choice of W may
result in θ(u ) = min {u(d − Dy )} > 0 , thus
y ∈W Defining Localizations
improving the standard Lagrangian bound z(u). and Estimating ξ(u,π)
Calculating θ(u) so far was solely intended
to estimate the complementarity term associated The critical issue in using the modified Lagrangian
with Dx≤d without taking into account the original bound (13) is constructing a suitable localization
objective function. We may “balance” z(u) and θ(u) W. From the definition of zM(u,π) in (13) it follows
by introducing some information of the original that, in general, the tighter the localization W is,
objective in the modified problem. Let LB and the smaller is ξ(u,π) and the better is the modified
UB be (unknown) lower and upper bounds for upper bound zM(u,π). From this point of view, it
the optimal objective of (1). Then we may add to is worth to retain in the definition of W as many
(1) the constraint LB≤cx≤UB and dualize it with original constraints as possible. However, the
multipliers πLB−πUB ≥0. localization should be simple enough to guarantee
224
Many-to-Many Assignment Problems
and we may possibly strengthen the standard dual ξ(u, π) ≡ max {πcy + uDy }
y ∈W1
225
Many-to-Many Assignment Problems
max {cy + v(b − Ay ) | Dy ≤ d , y ∈ U } ≡ wD2 nal constraints into a single one using two multi-
(20) pliers µ = (1, 0.5) gives the bound ws = 3 .
To calculate the modified bounds, let Ax≤b be
and defined by constraint (24) while constraint (23)
stands for Dx≤d and is dualized with the multi-
{ } and suppose
L L
wMD π (W1 ) ≤ η(u, 0) + ξ (u, 0, 0)
plier u≥0. Let U = x ∈ {1, 0}
4
= max {cx − uDx } + max {uDy | Dy ≤ d , y ∈ U }
x ∈X
≤ max {cx + u(d − Dx )} ≡ wD1 . that the localization W1 is defined by the two
x ∈X
226
Many-to-Many Assignment Problems
L
w MD (W1 ) = min η + ξ
L
wMD (W1 ) = η + ξ = 693 247 ≈ 2.8057 . u ,v ≥0,π
s.t .
2 −2π −10u
π +8u +15v
Instead of using the Lagrangian relaxation
4 − 4 π − 29u
π +16u +12v
2
ξ (u, v, π) to estimate ξ(u, π) we may use a sur-
L η ≥
, ξ ≥
3 −3π −24u
3π +18u −v
rogate relaxation. Let
0
18v
(26)
{ 4
Ws = x ∈ {1, 0} | 10.5x 1 +21x 2 +7x 3 }
+ 14x 4 ≤ 27.5
Its optimal solution is π = − 4 15 , u = 2 15 ,
be the surrogate localization obtained by mul- v = 0 , η = 1.2 , ξ = 1.6 giving wMD
L
π (W1 ) = 2.8
tiplying constraint (23) by µ1 = 1 , constraint (24) Fixing π =0 in the master problem (and thus
by µ2 = 0.5 and summing them up. These mul- computing the bound without the copy constraint
tipliers are optimal to the surrogate relaxation of (cx=cy), we get the bound η + ξ = 2.826 . Setting
the original problem and result in the surrogate π=v=0 (i.e. calculating the aggregated Lagrangian
bound ws = 3 . For W=Ws the modified dual decomposition bound) yields η + ξ = 2.842 .
bound wMD π (Ws ) is calculated using the linear
master problem. Note that the constraints corre-
sponding to the variable η in the master problem MODIFIED LAGRANGIAN
remain the same as before. The optimal solution BOUND FOR MANY-TO-MANY
to the new master problem is π = 2 15 , u = 115 , ASSIGNMENT PROBLEM
η = ξ = 4 3 giving wMD π (Ws ) = 2 23 ≈ 2.667 ,
which coincides with the bound wLD obtained by The assignment problems (AP) involve optimally
Lagrangian decomposition. If we use a localiza- matching the elements of two or more sets. When
there are only two sets, they may be referred as
227
Many-to-Many Assignment Problems
“tasks” and “agents”. For example, “tasks” may the available capacity of task j. Note that (MMAP)
be jobs to be done and “agents” may be the people has a double-decomposable structure: if we dual-
or machines that can do them. In its original ver- ize the first m restrictions (28), then the relaxed
sion, the AP involves assigning each task to a problem decomposes into n independent subprob-
different agent, with each agent being assigned lems, each having a single knapsack-type restric-
to at most one task (a one-to-one assignment). In m
m
The original constraints included in the set X
∑d x ij ij ≤ dj , j = 1...n . (29) will be considered as “easy”, while those in Y will
i =1 be treated as “complicating”. Localization W1 will
be used in the modified Lagrangian dual to cal-
In what follows we will refer to the problem culate the value wMD
L
(W1 ) . We will handle con-
(27)-(29) as (MMAP). Here xij=1 if agent i is straints y∈Y explicitly, while restrictions Py≤p
assigned to task j, and 0 otherwise, cij is the will be dualized in the estimation of ξ(u, v, π) .
profit (utility) of assigning agent i to task j, aij is Let u = {u j , j = 1 n } ≥ 0 and
the amount of agent i’s capacity used to execute
v = {vi , i = 1 m } ≥ 0 be the Lagrangian mul-
task j, and bi is the available capacity of agent i.
tipliers. Then the modified Lagrangian dual for
It is assumed that each task has its own capacity
(MMAP) is:
(time) limit, such that dij is the amount of task j’s
capacity used when executed by agent i, and dj is
228
Many-to-Many Assignment Problems
229
Many-to-Many Assignment Problems
or equivalently: LB k = (η k − ξ k ) ≤ L
wMD (W1 ) ≤ min {z (u s ) − θ(u s , v s )} = UB k ,
s =1...k
(46)
n m n
η − ∑ u d j ≥ max ∑ ∑ (cij − u d )x ij ≡ ∑ max ∑ (cij − u d )x ij
k k k
m n
k
,
j j ij j ij
x ∈X x ∈X i
(42)
iterations and for the iteration s we have:
where the maximization over X is reduced to n m n
independent maximizations over Xi due to decom- z (u s ) = ∑ u sj d j + ∑ ∑ (cij − u sj dij )x ijs
posable structure of X. That is to verify the fea- j =1 i =1 j =1
Similarly, to check the feasibility with respect θ(u s , v s ) = ∑ u sj d j − ∑ visbi + ∑ ∑ (visaij − u sj dij )yijs
j =1 i =1 i =1 j =1
to constraints (40) we need to verify: (48)
(43)
given threshold.
which results in solving n independent subprob- Assuming the objective coefficients in
lems with a single knapsack constraint and m (MMAP) are nonnegative we have z MMAP ≥ 0 .
*
binary variables. Let {yijk } be their optimal solu- The optimal solution (u, v, η, ξ ) to the problem
tion. (MP) then satisfies:
If (42) and (43) are fulfilled, stop with
k
(u, v, η, ξ ) optimal to (MP). If (42) fails, add:
L
0 ≤ z MMAP ≤ w MD (W1 ) = η * − ξ * ,
0 ≤ z MMAP ≤ z (u * ) = η * . (49)
n m n
η ≥ ∑ u jd j + ∑ ∑ (cij − u jdij )x ijk (44)
j =1 i =1 j =1 Thus we can add to the master problem (MP)
the trivial restrictions η − ξ ≥ 0 and η ≥ 0 at
beginning of the constraint generation process in
to the restricted master problem. If (43) fails, add:
order to prevent the objective function from de-
n m m n creasing unboundly at early iterations.
ξ ≤ ∑ u jd j − ∑ vibi + ∑ ∑ (viaij − u jdij )yijk
j =1 i =1 i =1 j =1
Restoring Primal Feasibility
(45)
by a Greedy Heuristic
to the restricted master problem. So in each itera-
The modified Lagrangian dual (18) provided high-
tion we add at most two constraints to (RMPk) to
quality bounds for the (MMAP) for the instances
get the next restricted master problem (RMPk+1).
tested in (Litvinchev & Rangel, 2008; Litvinchev,
On the k-th iteration of Benders technique we
Rangel & Saucedo, 2010). It turned out that the
have a lower and an upper bound for the optimal
corresponding integer Lagrangian solutions (x,y)
value w MD
L
(W1 ) of (MP):
had a higher degree of primal feasibility and sub-
230
Many-to-Many Assignment Problems
optimality than the standard Lagrangian solution. indicator (e.g., maximal cost component) while
In this section we consider a greedy algorithm to maintaining feasibility.
recover primal feasibility. The feasible solution Let x0 be a current binary point not necessary
is also used in a subgradient algorithm to obtain feasible to (MMAP). Let Ω1 be a set of all pairs
the modified bounds (Section 4.3). (i,j) with x ij0 = 0 and Ω1 be a set of all pairs (i,j)
To get a feasible Lagrangian based solution with x ij0 = 1 . D e n o t e δi = bi − ∑ j aij x ij0 ,
we use a simple greedy approach. First we try to
decrease to zero some components currently equal σ j = d j − ∑ i dij x ij0 . If minij {δi , σ j } ≥ 0 then
to 1 to obtain a feasible solution. The choice of x0 is feasible to (MMAP). Otherwise, we first
the candidate component is based on the smallest decrease, in a greedy manner, some positive x ij0
decrease of a rounding indicator (e.g. minimal cost to 0 to get a feasible solution (xgr). Then we try
component). After a feasible solution is obtained to improve this feasible solution by increasing,
we try to increase to 1 some zero components in a greedy fashion, some zero components to 1.
based on the largest increase of another rounding
Algorithm 1
gr 0
1. Let x0 be a Lagrangian solution ( x or y ). Set x = x .
2. Set tij , rij as the rounding indicators (e.g. tij = cij , rij = cij see Comment
1).
3. Feasibility test.
3.1. For x0 compute δi , σ j .
3.2. While minij δi , σ j { } ≥ 0 do (rounding down)
3.2.1. Compute: minij ∈Ω {tij } . Let this minimum be
1
attained for (i,j)’.
3.2.2. Set:
gr
3.2.3. x ij = 0 for (i,j)=(i,j)’,
3.2.4. Ω1 = Ω1 \ {(i, j )'} , Ω0 = Ω0 {(i, j ) '}
3.2.5. δi = δi + a(i , j )' , σ j = σ j + d(i , j )' ,
3.3. end_while
gr
4. Let x be a feasible solution obtained in the rounding down step,
and S 0 ⊆ Ω0 be a set of (i, j ) ∈ Ω0 with both aij ≤ δi and dij ≤ σ j
5. while S 0 ≠ 0 do (rounding up)
5.1. Compute maxij ∈S
0
{rij } . Let this maximum be attained for (i,j)’.
5.2. Set
gr
5.2.1. x ij = 1 for (i,j)=(i,j)’,
5.2.2. Ω1 = Ω1 {(i, j )'} , Ω0 = Ω0 \ {(i, j )'}
5.2.3. δi = δi − a(i , j )' , σ j = σ j − d(i , j )' ,
5.2.4. update S0
5.3. end_while
gr
6. Return greedy solution x .
7. End Algorithm 1.
231
Many-to-Many Assignment Problems
Algorithm 1 gives a summary of the Greedy ity). Similarly, we can try tij=cij/max{aij /bi,dij/dj}
Heuristic, its flowchart can be seen in Figure 1. since the relative values aij /bi,dij/dj also give a
Comment 1. The rounding down part of the measure of feasibility. In the rounding up part of
Algorithm 1 may be based on pure cost criterion the Algorithm 1 we may use rij = tij. Alternatively,
tij=cij. It is also possible to use another indicator, we can use rij = cij /min{ aij, dij }. Small values of
setting for example, tij=cij /max{aij,dij}. In this aij,dij help to obtain a small degradation in the
way we can take into account the impact of the solution feasibility. Another possibility is to set
component (i,j) in violating the constraints (the rij=cij/min{aij /bi,dij/dj}.
larger the values of aij,dij the faster we get feasibil-
232
Many-to-Many Assignment Problems
m n
The Lagrangian solution is always feasible αik = [∂φ / ∂vi ] = bi + ∑ ∑ aij yijk ,
k
(53)
either to the first or to the second group of con- i =1 j =1
straints of the problem (MMAP). So we can
simplify Algorithm 1 by considering only δi (σ j ) k
m n m n
when rounding down, depending on whether x or β jk = ∂φ / ∂u j = −∑ ∑ dij x ijk + ∑ ∑ dij yijk
i =1 j =1 i =1 j =1
y is used for rounding the modified Lagrangian (54)
solution.
Denote by sk a vector composed of all
Solving the Dual Problem by {γ , αk , β k } , let λk = {u k , πk , v k } and set:
k
by the Algorithm 1 (Litvinchev et al., 2010). A may set φLB equal to the objective function value
more detailed discussion of subgradient optimiza- of (MMAP) associated to a given feasible solu-
tion can be found in Martin (1999) and Wolsey tion. In what follows we will apply the greedy
(1998). algorithm in each iteration to get a feasible solu-
k
Let (u, π, v ) be the values of the Lagrangian tion and update φLB accordingly.
multipliers for the iteration k, φk = φ(u k , π k , v k ) Since u,v≥0, the multipliers for the next it-
and x ijk , yijk be the associated subproblems solu- eration are defined as the projection of uk and vk
tions: onto the nonnegative orthant, while π has no sign
restrictions:
n
x ijk = arg max ∑ (1 − π k )cij − u kj dij x ij , u k +1 = max {0, u k } , v k +1 = max {0, v k } , πk +1 = πk .
x ∈X i
j =1 (56)
(50)
A summary of the subgradient algorithm to
m
y = arg max ∑ (πkcij + uijkdij − vikaij x ij .
k compute the modified bound for the (MMAP)
y ∈Yi
ij
i =1 is given by Algorithm 2 and its flowchart can be
(51) seen in Figure 2.
The subgradient method is not monotone, that
After solving the subproblems, a subgradient is, it is not necessary that φk ≥ φk +1 . In practice,
is directly identified as: the parameter εk is varying in (0,2], beginning
m n m n
with εk =2. If after K consecutive iterations with
γ k = [∂φ / ∂π ] = −∑ ∑ cij x ijk + ∑ ∑ cij yijk
k
a fixed value for εk the function φ does not improve
i =1 j =1 i =1 j =1 “sufficiently’”, then a smaller value of εk is used,
(52) say, a half of εk. The stopping criteria used in
Algorithm 2 are: a) maximum iteration number
is reached; b) εk is already small enough; or c) the
relative difference between the best integer fea-
233
Many-to-Many Assignment Problems
sible solution found so far and the Lagrangian with sizes m×n for m=5,8,10 and n=50 and Set2
bound is within a given threshold. instances with m=5,10,20 and n=100. The data
were random integers generated as follows:
COMPUTATIONAL TESTS cij ∈ U [10, 5 0], aij ∈ U [5, 25], dij ∈ U [3, 20], bi
234
Many-to-Many Assignment Problems
Algorithm 2
0 0 0
1. Given initial values for u , π , v .
0 0
2. Set φUB = ∞, φLB = −∞ .
3. While (not stop) do
k k k k k k
3.1. Compute x ij , yij by (50), (51) and let φ = φ(u , π , v ) .
3.2. Use Algorithm 1 to obtain feasible solutions xgr and ygr with objec-
tive values cxgr and cygr respectively.
k +1
3.3. Let φLB = max φLB , cx , cy{ k gr gr
}, k +1
φUB = max {φUB
k
, φk }
3.4. Compute s
k
= {γ k
, αk , β k } by (52)-(54).
k
3.5. Update λk+1 by (55) with φLB = φLB .
3.6. Project uk and vk onto the nonnegative orthant
u k +1 = max {0, u k } , v k +1 = max {0, v k } .
3.7. Make stop tests.
4. end_while
5. end Algorithm 2
and divided in three classes (a, b, and c) with Algorithm 2 with the following
respect to the values of α: α(α=1), b(α =0.9) and modifications:
c(α=0.8). More details of the data generation can Replace step 2 by: Set
be found in Litvinchev et al. (2010a). 0
φUB 0
= ∞, φLB =0
We divided the computational study for prob- k +1
Replace step 3.3 by: Let φLB = φLB and
k
235
Many-to-Many Assignment Problems
236
Many-to-Many Assignment Problems
c is insignificant. When zLP is close to zIP (small depends weakly on the type of the Lagrangian
values of gap3) for both Lagrangian bounds, few bound and tends to increase with dimension.
possibilities are available for improvements. That
is the case for the instance 20×100b which has Results for MMAP Using a
gap3=0.78 and gap1=0.72 (See Table 2). Overall, Subgradient Algorithm
the results suggest that the difference between the
modified and classical bounds is larger for larger All Lagrangian-type bounds were calculated by
values of gap3. In other words, the worse is the the subgradient method. First we applied Algo-
continuous relaxation, the more possibilities we rithm 2a to compute the modified bound, with
have for improving the classical Lagrangian and without the copy constraints (zMD𝜋, zMD). Then
bound. For all the problems except for the last we applied Algorithm 2 to study how the greedy
two in Table 1, we have rel1<rel3; i.e., assuming solution given by Algorithm 1 can improve the
that zlag sufficiently improves zLP for these prob- bound quality. The latter is also used to speed
lems, the same can be concluded about the relation up the subgradient scheme to solve the modified
between zMD and zlag. The number of iterations Lagrangian dual problem. For both Algorithms 2
and 2a we used K=5, and if
237
Many-to-Many Assignment Problems
((φk − φk +1 ) / φUB
k +1
) ≤ 0.005 360 s respectively. The values of the indicators
were calculated based on the best values of the
for 5 consecutive iterations with fixed εk, this bounds obtained up to the final iteration. As can
parameter was updated to εk+1= εk/2. The stopping be seen from Tables 3 and 4, for all problem in-
criteria were specified as follows: a) at most 250 stances we have rel0 < 100%, that is, the use of
iterations were permitted; b) the runs stop if the objective copy constraint (cx = cy) has a
εk+1≤0.005; or c) ((φUBk +1 k +1
− φLB k +1
) / φLB ) ≤ 0.0001 . beneficial effect in the computation of the modi-
The numerical results obtained for the problems fied bound.
in Set1 and Set 2 using Algorithm 2a are given in Moreover, gap0 is significantly smaller than
Tables 3 and 4, respectively. For each problem, min{gap1,gap2,gap3} for all instances (except
the tables give the instance dimension (m,n), its for 20 × 100c), which means that zMD𝜋 is superior
class (a,b,c), the relative quality of the bounds to all other bounds within the computational time
(rel0,rel1,rel2,rel3), the proximity to the original limit. In most cases gap0 < 0.5% and for all
optimal value (gap0,gap1,gap2,gap3). problems of the class a the exact bound was ob-
The results in Table 3 (Set 1) and Table 4 (Set tained, gap0 = 0. Note that for problem instances
2) were obtained within a time limit of 240 s and with gap0 = 0 the proposed bound zMD𝜋 cannot
Table 3. Relative quality of the bounds; results by Algorithm 2a (the subgradient method); set1 n=50
Table 4. Relative quality of the bounds; results by Algortihm 2a (the subgradient method); set2 n=100
238
Many-to-Many Assignment Problems
be outperformed by any other bound. For most (gap0gr,gap0,gap2gr,gap2) and the number of it-
instances of class c we have gap3=min{gap1,gap2}, erations (iter).
so the LP bound is better than zlog, zmd obtained For all instances with gap0gr=gap0=0.00 the
within the time limit. In Section 5.1 the results of stopping criterion (c) was fulfilled. For all other
the Benders method showed that instances the runs were terminated by the stopping
gap1<gap2<gap3for all classes. However, for the criterion (b). As can be seen from Tables 5 and 6,
class c the difference in these indicators was very incorporating a greedy solution in the subgradient
modest: for small values of gap3 there is a little algorithm slightly improves the quality (of the
scope for bound improvements. As a byproduct approximate values) of the bounds obtained by
of the bound computation ( wMD L
π ) an integer the subgradient method, but the effect is rather
Lagrangian solution y was obtained having a much modest. For all problem instances the modified
higher degree of primal feasibility and subopti- bound (gap0gr) is significantly tighter than the
mality than the standard Lagrangian solution. (see classical Lagrangian bounds (gap2g gap2). More-
Litvinchev, Rangel & Saucedo, 2010) for more over, the number of iterations of the subgradient
details on the constraint violation and suboptimal- method reduces significantly by using a greedy
ity of the Lagrangian solutions (x,y)). solution. This takes place for all problem in-
To study the effect of introducing a heuristic stances. Note that the computational cost for one
solution on the calculation and quality of the dual iteration of the subgradient technique (solving
bounds we applied Algorithm 2 to compute the integer Lagrangian problems) is much higher than
standard ( zlag
gr
) and the modified ( z MD
gr
π ) Lagrang-
the one for obtaining a greedy solution (simply
ian bounds. The numerical results obtained for reordering data). Thus combining the subgradient
the problems in Set1 and Set 2 are given in Tables method with a “cheap feasibility recovering”
5 and 6, respectively. For each problem, the tables technique is favorable for this class of problem.
give the instance dimension (m,n), its class (a,b,c), The greedy solutions derived by the modified
the proximity to the original optimal value bound are better than the ones resulting from the
standard Lagrangian bound (see Litvinchev et al.,
2010b for the details).
Table 5. Relative quality of the bounds; results by the subgradient method (Algorithm 2 and 2a); set1 n=50
239
Many-to-Many Assignment Problems
Table 6. Relative quality of the bounds; results by the subgradient method (Algorithm 2 and 2a); set2
n=100
a 0 0 1.61 1.65 5 85 49 90
240
Many-to-Many Assignment Problems
present the proximity indicator gap0 for the bound solution of the MMAP. It takes a few iterations to
wMDL
π , obtained by using the objective copy con- get an approximate feasible solution within 0.1%
straint (cx=cy), and the proximity indicator gap2 of the relative suboptimality. The main compu-
for the classical Lagrangian bound. As can be tational cost is associated to solution of the n+m
seen from the Tables, gap0 is significantly independent one-dimensional knapsack problems
smaller than gap2. That is the approximation of ((50), (51)) in each iteration. Recovering feasibility
the bound wMD L
π obtained by the subgradient tech-
of the modified Lagrangian solution by the greedy
nique is much closer to the optimal objective algorithm is significantly cheaper. The core of the
comparing to the approximation of the classical Algorithm 2 is the modified Lagrangian bound.
bound. Moreover, gap0 is almost 10 times It provides tight approximation of the optimal
smaller than gap1 obtained by the Benders tech- objective and results in an integer Lagrangian
nique and presented in Tables 1, 2. Thus, we may solution having a high degree of primal feasibil-
conclude that introducing the objective copy ity and suboptimality. The use of this Lagrangian
constraint has a beneficial effect in the computa- solution in a simple greedy algorithm results in a
tion of the modified bound (see the end of the high quality approximate feasible solution to the
Section 2.1 for theoretical discussion of this sub- original problem.
ject).
Computing the classical Lagrangian bound
results in a solution of Lagrangian problem which FUTURE RESEARCH DIRECTIONS
is frequently used as a starting or reference point
in heuristic techniques. In the case of the modified A procedure for tightening Lagrangian bounds
bound we have two Lagrangian solutions, x and was proposed. The main idea of the approach
y, both different from the classical Lagrangian is to estimate the complementarity term in the
solution. To recover a feasible solution from an Lagrangian function. For convex problems under
infeasible Lagrangian one we used a simple greedy certain regularity conditions the complementarity
heuristic (Algorithm 1). The results presented in term turns to zero for the optimal primal-dual pair.
the Tables 5 and 6 show that: a) combining the However for integer problems the complemen-
subgradient scheme with the best feasible greedy tarity term can be different from zero even for
solution (Algorithm 2) improves significantly the optimal Lagrange multipliers since we typically
convergence of the subgradient method and b) this have positive optimal multipliers for non-active
improvement was obtained without dropping the constraints. We may expect that this approach can
quality of the bounds and feasible solutions. We be especially useful for discrete problems having a
note that to get the modified bound the Algorithm large number of nonactive constraints for feasible
2 either stops after 5-8 iterations or achieves the solutions. Here we used the approach for assign-
main improvement (in terms of the quality of the ment problems with knapsack-type inequality
modified bound and the corresponding feasible constraints, which frequently are non-active for
solution) within the same range of iterations. feasible solutions. An interesting area for future
From practical point of view, it is sufficient to research is an application of the proposed ap-
execute at most 10 iterations of the Algorithm proach to packing and cutting problems where,
2, extra iterations will only slightly change the by the physical nature of the problem, available
overall picture. resources can not be utilized completely, e.g.,
Thus we may conclude that the proposed ap- packing circles into rectangular or circular areas.
proach (Algorithm 2) can be useful for practical Another direction for future research is the use
of a posteriori localization, possibly adjusted in
241
Many-to-Many Assignment Problems
the course of solving the dual problem. It can be FINAL REMARKS AND CONCLUSION
constructed in a surrogate way, similar to local-
ization Wsmentioned in Section 2.1 or using valid In this chapter we presented a Lagrangian heuristic
inequalities as in Litvinchev and Rangel (1999) and applied it to a class of the generalized assign-
(see also Litvinchev and Tsurkov (2003)). ment problems. Our contribution is threefold. First,
The proposed method results in two subprob- we proposed the modified Lagrangian bound based
lems - one is similar to the classical Lagrangian on a more tight estimation of the complementarity
problem and the other is used to estimate the term in the Lagrangian function. For the original
complementarity term. Thus we have two La- problem having two interesting sets of constraints
grangian solutions both different from the clas- we proved that the new bound is at lest as good as
sical one. In our computational experiment with the best of the corresponding classical Lagrang-
many-to-many assignment problems we notice ian bounds. Second, we studied numerically the
that although solutions to modified Lagrangian properties of the two Lagrangian-like solutions
dual are unfeasible, constraints are typically much arising in the modified bound computations and
less violated comparing with the classical Lagrang- showed that for the modified Lagrangian solution
ian solution. This is the reason why our simple the original constraints are much less violated than
greedy algorithm aimed to restore feasibility of for the classical Lagrangian solution. Third, we
the modified Lagrangian solution provides very merged a greedy technique used to recover the
high quality approximated solutions to the original feasibility of the modified Lagrangian solution
problem. It is interesting to see if the same behavior with the subgradient algorithm thus obtaining
of the modified Lagrangian solutions takes place an approximate iterative scheme producing high
for other classes of combinatorial problems. quality feasible solutions.
It was demonstrated by the computational The usefulness of the approach was demon-
experiment, that incorporating the greedy feasible strated for a class of the generalized assignment
solution in the subgradient scheme improves problems, though the basic scheme can be applied
the convergence of the subgradient method. An for a much broader class of problems. In this
interesting area for a future research is to use chapter we focused on the general algorithmic
the modified bounds in combination with other aspects of the proposed Lagrangian based heu-
greedy approaches, e.g., using different choices ristic. We sincerely hope that this new approach
for rounding up/down components, as well as with will be useful for the researches and practitioners
more sophisticated heuristic techniques (see, e.g., working with specific (and not only assignment)
Jeet and Kutanoglu (2007)). problems in power optimization.
Another direction for future research is to im-
prove the behavior of the subgradient technique
by a more suitable choice of the initial values for ACKNOWLEDGMENT
the Lagrangian multipliers, for example, using
optimal duals for the copy constraints Dx ≤ Dy in This work was partially supported by RFBR,
the LP-relaxation of the problem (7). An alterna- Russia (09-01-00592), Mexican foundations
tive to the subgradient method used in this paper CONACyT (61343), PAICYT (CE008-09), and
would be using more stable approaches such as PROMEP (103.5/09/3905), and Brazilian agencies
center-based or bundle algorithms (e.g. Bahience CNPq and FAPESP.
et al. (2002)).
242
Many-to-Many Assignment Problems
243
Many-to-Many Assignment Problems
Litvinchev, I., Mata, M., Rangel, S., & Saucedo, Pirkul, H. (1986). An integer programming model
J. (2010). Lagrangian heuristic for a class of for the allocation of databases in a distributed com-
the generalized assignment problems. Comput- puter system. European Journal of Operational
ers & Mathematics with Applications (Oxford, Research, 26(3), 401–411. doi:10.1016/0377-
England), 60(4), 1115–1123. doi:10.1016/j. 2217(86)90142-6
camwa.2010.03.070
Ross, G. T., & Soland, R. M. (1975). A branch and
Litvinchev, I., & Rangel, S. (1999). Localization bound algorithm for the generalized assignment
of optimal solution and a posteriori bounds for ag- problem. Mathematical Programming, 8, 91–103.
gregation. Computers & Operations Research, 26, doi:10.1007/BF01580430
967–988. doi:10.1016/S0305-0548(99)00027-1
Ross, G. T., & Soland, R. M. (1977). Modeling
Litvinchev, I., & Rangel, S. (2008). Comparing facility location problems as generalized assign-
Lagrangian bounds for a class of generalized ment problems. Management Science, 24(3),
assignment problems. Computational Mathemat- 345–357. doi:10.1287/mnsc.24.3.345
ics and Mathematical Physics, 48(5), 739–746.
Srinivasan, V., & Thompson, G. L. (1972). An
doi:10.1134/S0965542508050047
algorithm for assigning uses to sources in a special
Litvinchev, I., Rangel, S., & Saucedo, J. (2010). A class of transportation problems. Operations Re-
Lagrangian bound for many-to-many assignment search, 21, 284–295. doi:10.1287/opre.21.1.284
problems. Journal of Combinatorial Optimization,
Wolsey, L. A. (1998). Integer programming. New
19(3), 241–257. doi:10.1007/s10878-008-9196-3
York, NY: John Wiley & Sons.
Litvinchev, I., & Tsurkov, V. (2003). Aggregation
Zimokha, V. A., & Rubinshtein, M. I. (1988). R
in large scale optimization. Boston, MA: Kluwer.
& D planning and the generalized assignment
Maculan, N., & Reinoso, H. (1992). Lagrangean problem. Automation and Remote Control, 49,
decomposition in integer linear programming: A 484–492.
new scheme. INFOR, 30(1), 1–5.
Martello, S., & Toth, P. (1990). Knapsack prob-
lems: Algorithms and computer implementations. ADDITIONAL READING
New York, NY: Wiley.
Bahiense, L., Maculan, N., & Sagastizabal, C.
Martin, R. K. (1999). Large scale linear and in- (2002). The volume algorithm revisited: relation
teger programming: A unified approach. Boston, with bundle methods. Mathematical Program-
MA: Kluwer. ming, 94, 41–69. doi:10.1007/s10107-002-0357-3
Morales, D. R., & Romeijn, H. E. (2004). The Barahona, F., & Anbil, R. (2000). The volume
generalized assignment problem and extensions. In algorithm: producing primal solutions with a
Du, D.-Z., & Pardalos, P. M. (Eds.), Handbook of subgradient method. Mathematical Programming,
combinatorial optimization (Vol. 5, pp. 259–311). 87, 385–399. doi:10.1007/s101070050002
Kluwer Academic Publishers.
Beasley, J. E. (1993). Lagrangean relaxation. In
Pentico, D. W. (2007). Assignment problems: Reeves, C. R. (Ed.), Modern heuristic techniques
A golden anniversary survey. European Jour- for combinatorial problems (pp. 243–303). Black-
nal of Operational Research, 176, 774–793. well Scientific Publications.
doi:10.1016/j.ejor.2005.09.014
244
Many-to-Many Assignment Problems
Boschetti, M., & Maniezzo, V. (2009). Benders Jornsten, K., & Nasberg, M. (1986). A new La-
decomposition, Lagrangian relaxation and me- grangian relaxation approach to the generalized
taheuristic design. Journal of Heuristics, 15(3), assignment problem. European Journal of Opera-
283–312. doi:10.1007/s10732-007-9064-9 tional Research, 27, 313–323. doi:10.1016/0377-
2217(86)90328-0
Burkard, R., Dell’Amico, M., & Martello, S.
(2009). Assignment problems. Philadelphia: Kleeman, M., & Lamont, G. (2008). Evolution-
SIAM. doi:10.1137/1.9780898717754 ary multi-objective optimization for assignment
problems. In Bui, L. T., & Alam, S. (Eds.), Multi-
Everett, H. III. (1963). Generalized Lagrange mul-
Objective Optimization in Computational Intel-
tiplier method for solving problems of optimum
ligence: Theory and Practice (pp. 364–387). IGI
allocation of resources. Operations Research, 11,
Global. doi:10.4018/978-1-59904-498-9.ch013
399–417. doi:10.1287/opre.11.3.399
Lemarechal, C. (2001). Lagrangian relaxation.
Fisher, M. L. (1985). An application oriented guide
In Junger, M., & Naddef, D. (Eds.), Computa-
to Lagrangian relaxation. Interfaces, 15, 10–21.
tional combinatorial optimization (pp. 115–160).
doi:10.1287/inte.15.2.10
Springer Verlag. doi:10.1007/3-540-45586-8_4
Geoffrion, A. M. (1974). Lagrangian relaxation
Li, D., & Sun, X. (2006). Nonlinear integer pro-
and its uses in integer programming. Mathematical
gramming. Springer.
Programming Study, 2, 82–114.
Lorena, L. A. N., & Narciso, M. J. (1996). Relax-
Guignard, M. (2003). Lagrangian relaxation.
ation heuristics for generalized assignment prob-
Top (Madrid), 11(2), 151–228. doi:10.1007/
lem. European Journal of Operational Research,
BF02579036
91, 600–610. doi:10.1016/0377-2217(95)00041-0
Haidar, M., Al-Rizzo, H., Chan, Y., & Akl, R.
Narciso, M. J., & Lorena, L. A. N. (1999). La-
(2009). User-based channel assignment algo-
grangean/surrogate relaxation for generalized
rithm in a load-balanced IEEE 802.11 WLAN.
assignment problems. European Journal of Op-
International Journal of Interdisciplinary Tele-
erational Research, 114(1), 165–177. doi:10.1016/
communications and Networking, 1(2), 66–81.
S0377-2217(98)00038-1
doi:10.4018/jitn.2009040105
Pentico, D. W. (2007). Assignment problems:
Held, M., & Karp, R. (1970). The traveling sales-
A golden anniversary survey. European Jour-
man problem and minimum spanning trees. Op-
nal of Operational Research, 176, 774–793.
erations Research, 18, 1138–1162. doi:10.1287/
doi:10.1016/j.ejor.2005.09.014
opre.18.6.1138
Shapiro, J. F. (1974). A survey of Lagrangean
Jeet, V., & Kutanoglu, E. (2007). Lagrangian re-
techniques for discrete optimization. Annals of
laxation guided problem space search heuristics
Discrete Mathematics, 5, 113–138. doi:10.1016/
for generalized assignment problems. European
S0167-5060(08)70346-7
Journal of Operational Research, 182, 1039–
1056. doi:10.1016/j.ejor.2006.09.060
245
Many-to-Many Assignment Problems
246
Many-to-Many Assignment Problems
APPENDIX
The literature on Lagrangian relaxation is quite extensive. Here we refer the reader to a few pioneer
and/or review papers providing a clear exposition of the subject: Everett (1963), Held & Karp (1970),
Geoffrion (1974), Shapiro (1974), Fisher (1985), Beasley (1993), Lemarechal (2001), Guignard (2003).
To our knowledge, the latter paper presents the most comprehensive and complete revision of the basic
ideas used in Lagrangian relaxation and its applications.
Numerical techniques of non-differentiable optimization used to solve Lagrangian dual problem can
be found in the references cited above. More recent approaches are presented, e.g., in Barahona & Anbil
(2000), Bahiense, Maculan & Sagastizabal (2002). Various relaxations in nonlinear integer programming
are considered in Li & Sun (2006).
The ideas to combine Lagrangian and Benders decomposition techniques with heuristic approaches
are presented in Boschetti & Maniezzo (2009).
A very complete source on modeling and solution techniques for assignment problems are provided
in the recent book of Burkard, Dell’Amico & Martello (2009) and in the survey paper Pentico (2007).
More specific Lagrangian-based solution approaches can be found in Jornsten & Nasberg (1986), Lorena
& Narciso (1996), Narciso & Lorena (1999), Jeet & Kutanoglu (2007) and the references therein. Multi-
objective assignment models and solution techniques are presented in Kleeman & Lamont (2008), while
models and algorithms for channel assignment are considered in Haidar et al. (2009) and the references
therein. See also the additional reading section.
247
248
Chapter 8
Power Systems Investments:
A Real Options Analysis
João Zambujal-Oliveira
Instituto Superior Técnico & Technical University of Lisbon, Portugal
ABSTRACT
Energy projects with extended life cycles and initial investments can be non-profitable under discount
cash flow methods. Therefore, real options analysis has become relevant as a pricing technique for
these types of projects, with private risks and high investment levels. Following this question, this study
analyses different real options approaches to select the most acceptable for investing decisions in the
energy sector. Combined cycle natural gas-fired plants constitute relevant generation assets that build-
ing decisions can mostly be studied by real options tools. Because traditional pricing approaches fail
to consider the worth of flexibility, conditions for creating a significantly large options-based value
can be found. Being unable to capture the value associated with the decision maker’s ability to react
dynamically to changing market conditions, these assets constitute a fine example of flexibility, which
contributes to increasing its intrinsic value.
The study employs a real options approach that doesn’t need to capture all the uncertainty and proposes
a process that directly determines the uncertainty associated with the first period. The results support
that its use can be considered fair. However, it shows that long periods of operation and poor adhesion
to the geometric Brownian motion by the project returns might call into question its use in the energy
market. The values for option pricing have remained inside acceptable ranges, but some shortfalls could
be found. First, the study employs Monte Carlo simulations, which can be viewed as forward-looking
processes, and option pricing problems need backward recursive solutions. Second, the study shows that
its simplicity produces results as accurate as those gathered from approaches with added complexity
and computational needs.
DOI: 10.4018/978-1-61350-138-2.ch008
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Power Systems Investments
249
Power Systems Investments
approach framing of ROA in the energy market. spread is positive. This temporarily shutdown
After a suitable framework definition, the study option highlights the importance of developing
describes the underlying methodology behind the a dynamic management attitude.
research question. It aims to assess the relevance A peaking unit power plant can also be used
of calculating the project’s volatility using two to illustrate a type of growth real option (intensity
different approaches. Then, it describes technol- and scale). This kind of option formerly quantifies
ogy used in the investment project. The choice the flexibility to expand or contract the scale of a
of this investment project results from supported project. Such units have high variable costs and
reasons of environmental efficiency improvement. can produce additional power if demand and prices
At this point, the chapter provides in detail the are high. The plant may have option to change the
uncertainty modelling. output rate per unit of time or to change the total
length of production run time. This option lets
the plant to increase the capacity of an existing
PROJECT APPRAISAL AND generation line.
VOLATILITY FUNDAMENTALS If the market conditions are weaker than origi-
nally expected, the plant should have a contract
The dynamics of investment in the power genera- option to operate below capacity or even lessen
tion industry is characterized by the existence of the scale, saving part of the planned investment
different decisions about assets used in the energy outlays. This flexibility allows mitigating losses.
generation. These decisions can be undertaken Using Monte Carlo simulation, Tseng and Barz
based on the formalization of the investment events (2002) assess a power plant capable of switching
in accordance with the real options approach. The between two modes.
following paragraphs illustrate the diverse types When a power generation facility can restart
of real options and its connection with power operations for a previous shut down project in a
generation industry. dormancy state, there are present switching options
Traditional capital budgeting analysis as- that permit the facility to expand or contract. The
sumes that projects perform every year of their option to switch refers to the feasibility of choosing
lifetime. Power plants should have an option to among alternative operating modes. Basically, this
cease the projects during their lives. This can be happens when demand is changing, the manage-
seen as an abandonment real option. This option ment can change the output mix or inputs.
gives the owner the right to sell the cash flows The deferment option is valuable when market
over the remainder of the project’s life. When the demand is uncertain and interest rate is volatile.
present value of the remaining cash flows (sal- The option to choose when to start a project is a
vage value) falls below the liquidation value, the deferment option. It is an extended concept of
power facility should be sold. During the project’s opportunity cost, which emphasizes that the deci-
evaluation phase, it is necessary to recognize the sion to invest immediately entails renouncing the
value represented by the opportunity, to recoup undertaking a similar investment and postponing
part of the investment, in case the project should the implementation of this particular investment.
be abandoned for scrap value. Fleten and Nasakkala (2010) analyzed a license to
When electricity price falls below the cost build a power plant, evaluating the optimal timing
of production, it may be most advantageous to in function of stochastic carbon emission costs.
temporarily shutdown until the electricity price The option to expand allows increase in the
recovers. Deng et al. (1998) valuate thermal scale of a project by investing more follow-on
power plants that operate only when the spark capital. When the market conditions turn out to
250
Power Systems Investments
251
Power Systems Investments
V CF1 + PV1 (CFi +1, ..,CFn ) DCF analysis, which discounts future cash flows
ϕC = ln 1 = i ∈ [1, n − 1]
V0 V0 to estimate the net present value and thereby its
(4) financial viability. The present value assumes a
full knowledge of each variable’s evolution (in-
Smith (2005) suggested that CF1 should be cluding future energy prices and carbon trading
stochastic and the other cash flows (CF2,...,CFn) prices) and discounts the future cash flows using
should be expected values conditional on the re- a risk-adjusted discount rate.
sults of the first cash flow. The problem in decision Using expression (3), the process begins with
making is finding out the best moment to invest the definition of a project value at moment 0 (V0)
(now or the next period). So, there is no need and moment 1 (V1). Afterwards, using Equation
to capture the uncertainty of all the cash flows (5) and the market data, it is possible to obtain an
(Brandão et al., 2005). This improvement permits initial value for the random variable φB To perform
to detain directly the uncertainty associated with the Monte Carlo simulation, V0is fixed for all it-
the first period, rationalising the estimate and erations and V1 is incorporated into the expression
providing a fair estimate of the project’s volatility. (4). This expression represents the project returns
and follows a GBM with a drift rate α that repre-
V CF1 +V1 (E1 (CFi +1 ) CF1 ) sents the mean of the project returns distribution
ϕB = ln 1 =
and volatility s obtained from the standard de-
V0 V0
viation of the returns variation. Therefore, the
(5) project volatility at a time interval Δt corresponds
to s / ∆t and represents the annualised per-
Brandão et al. (2005) compared this GBM ap-
centage of the standard deviation. Because the
proximation with Smith and Nau (1995) binomial
period of cash flow generation is annual, the
approximations with two individual uncertain-
process directly obtains the annual volatility from
ties in a risk-neutral environment, concluding
the standard deviation because σ equals s. Implic-
the results are not deeply different. Considering
itly, there is a postulation that project values are
that a binomial tree increases its complexity
log-normally distributed and follow a GBM with
with the number of periods and uncertainties,
a constant volatility. The process to estimate
the motivation to make use of a single variable
volatilities employs ordinary econometric ap-
approximation of the stochastic process becomes
proaches, using similar regression analyses to
clear. However, the matching with GBM should
isolate price trends. It runs the regressions with
be verified, particularly the lognormal distribution
the logarithm of values and with its residuals and
and the standard deviation homogeneity of the
observes the first-order auto-regressions. The
returns. About the adequacy of Brandão et al.’s
volatility estimates result from the standard error
(2005) approach to our study, Smith (2005) stated
of these auto-regressions.
that it should only be applicable for scale options
Baltazar (2009) considered revenues as a
and that deferment options cannot be considered
function of the electricity price (see expression
scale options. Brandão et al.’s (2005) allowed this
(1)). So, revenues follow a GBM formalised by
vision but called for some flexibility.
the expression R(𝜑c)=Ke𝜑c with a constant K. The
Following the framework of Yang and Blyth
variable 𝜑c takes values between 0 and 1 and only
(2007), the study first calculates the present value
varies from the initial period to the next, holding
of the project without considering any option.
the same value for the following periods. After
Using traditional methods, the assessment of the
generating project revenues, expression (5) per-
costs and revenues of the project would involve a
252
Power Systems Investments
mits to obtain the random variable. After running economic growth. In 2007, the electricity produc-
the simulation, it’s available the standard deviation tion was 47,253 GWh, with 28% generated from
that corresponds to the volatility of project returns. natural gas resources. The use of natural gas to
produce electricity grew 11% between 2003 and
2007 (Eurostat, 2009). A combined cycle natural
COMBINED CYCLE NATURAL gas facility is characterised by advanced technol-
GAS-FIRED PLANT ogy with a low environmental impact and high
performance compared with coal or fuel oil or
Technology Description natural gas coupled to an alternator. A high volt-
age transformer connects the plant to the national
The company’s problem evaluates the project’s distribution network. This combined cycle power
financial viability. Also suggests the best time to plant has efficiency between 55% and 56%, much
start the project, taking into account the maximiza- higher than other power plants. Such efficiency
tion of present value of the project. The action of permits consumed energy to be considered a clean
the electric market is characterized by its uncer- source. Its high efficiency also leads to a lower
tainty nature regarding the price of electricity and fuel consumption and has contributed to further
the fuel cost associated with energy production. improvements in environmental impact.
This section begins with the case study analysis The combined cycle natural gas-fired plant
of a nuclear combined cycle power plant describ- has an installed capacity of a 400 MW-type gas
ing technological advances that have enabled a turbine combined with a steam cycle. The exhaust
new generation of power plants with high income gases expelled from the gas turbine are used to
levels and low costs. In particular, the natural gas generate steam that drives a steam condensation
combined cycle plants can combine a gas cycle turbine. The two turbines are linked by a single
with a steam cycle, which offers a higher efficiency shaft and the quantity of natural gas must be 1.8
than other power plants. As gas-fired facilities, times the energy generated. The low emissions
these plants contribute to the diversification of of SO2, NOx, CO2 and particulates in combined
primary energy sources and take advantage of cycle plants have a positive effect on environ-
issuing low emissions compared with other fuels. mental performance shows the carbon emissions
This relevant characteristic constitutes a key factor for different power plants.
to meet the allowed levels of pollution. Considering the technological superiority of
Making use of an ROA, this study models a a combined cycle natural gas-fired plant with
problem of a power-generating plant investment regard to CO2 emissions, the following section
problem as a deferment option associated with describes the uncertainties in market production
the spread between electricity prices and variable and how the evaluation model considers them.
costs (e.g., fuel costs). The option value rises when The definition of parameters associated with
interest rates and the time to maturity increase. uncertainties comes from market studies, namely
Ceteris paribus, if the strike price increases, the the electricity and fuel markets.
option price declines. In our real options model,
the plant manager has the right to defer the plant Uncertainty Modelling
building. The power plant’s investment decision
process can be defined by call options. Some uncertain variables, such as electricity price,
The consumption of electricity in Portugal natural gas price, the price for carbon emissions
between 2003 and 2007 grew on average 2.6% and project value, should be modelled (Reedman
per year, which represented an increase above et al., 2006). The modulation process implies
253
Power Systems Investments
Table 1. Pollutant emissions for different types of power plants (g/kWh), Source: Baltazar (2009).
studying the electricity market and retrieving some tors (OMEL, 2009; OMIP, 2009), it is possible to
historical price data. In this context, Frayer and extract a historical volatility of electricity price of
Uludere (2001) considered that it is appropriate to 54%. Assuming this level of annual volatility can
use contracts with maturities similar to the plant’s provide a mid-to-long-term standard deviation in
estimated life cycle. electricity prices, 10,000 iterations of Monte Carlo
The Iberian Electricity Market (MIBEL) re- simulations (are carried out for a time framework
sulted from a partnership between Portugal and of 25 years (Baltazar, 2009).
Spain. Its aim is to integrate the two electricity Algeria provides a large percentage of the
systems. The formation of a regional market in natural gas consumed by the Iberian Peninsula.
the Iberian Peninsula constitutes a step further Several port terminals for natural gas in Spain and
for the construction of an internal market in the one in Portugal receive it through the Maghreb–
European Union. As a consequence of the daily Europe pipeline. The price of Brent constitutes the
transactions of electricity contracts, the price var- pattern for the price of natural gas (Monaghan,
ies every day. This procedure helps corroborate 2009). Thus, the natural gas price is considered
that futures markets, for electricity and natural a market uncertainty in this plant project. The
gas, have inadequate liquidity to provide reliable estimation of gas price volatility results primar-
market-based forecasts. Because the futures mar- ily from historical market data between 1987 and
kets represent a small portion of the MIBEL and 2009 (EIA, 2009). From the analysis of daily spot
the contracts are not traded very far in advance. prices, Baltazar (2009) obtained a Brent histori-
Therefore, projections are made to simulate cal price volatility of 26%, which serves as the
how prices will change stochastically over time initial value of the simulation process. Following a
using Monte Carlo techniques. Brandão et al. similar process to the one of the electricity prices,
(2005) only considered the uncertainties with a the study uses Monte Carlo simulations (Spinney
market equivalent. Although the project contains and Watkins, 1996).
three market uncertainties (electricity price, natu- Today, carbon emission licenses are daily
ral gas price and the price for carbon emissions), traded in the carbon exchange market (European
only the first two are modulated as stochastic Climate Exchange). The launch of new licenses is
variables. Because electricity demand has no carried out in stages and depends on an administra-
market equivalent, it is analysed in the sensitivity tive grant. After 2012, the changes in the carbon
analysis section. In the current context, the study exchange market will invalidate using current
chooses to model electricity prices using a GBM, prices as references for the future. Uncertainty
for which it is necessary to estimate volatility. in the price of carbon emissions results from a
Using historical market data since January biased market and depends on the release of new
2000, which is available from the market opera- regulations. Thus, the allowances of carbon don’t
254
Power Systems Investments
assume a nature of market uncertainty (EUETS, and Uludere (2001), who considered that DCF
2009). According to the study of other combined models, based on production cost-based simula-
cycle natural gas-fired plants with similar profiles, tions (like this one), attribute short cumulative
the study considers an average annual decrease values to typical gas-fired peaking facilities.
of carbon emissions of 1.1%. The DCF analysis includes a deterministic risk
appraisal and applies a risk premium over the
Project Value Modelling risk-free interest rate. Therefore, the stochastic
nature of price variations is not taken into account.
The project is yet to start and because of economic That assumption cannot be accepted in a market
conditions and adverse financial conjuncture, with fast technological development, characterised
the company is considering postponing the in- by uncertainties about energy prices, prices for
vestment. The combined cycle natural gas-fired carbon dioxide emissions, raw materials and O
plant has two stages: investment and production. & M costs. Although the simplicity of integrating
These stages are later used to price outlooks and the risk analysis into a unique risk-adjusted dis-
forward cost simulations. The investment stage count rate is tempting, the disadvantages resulting
involves the construction and reliability testing from a poor understanding of new risks outweigh
phases. Table 2 describes the base parameters of the benefits.
the investment project. From here, the analysis evolves from a de-
The production phase starts four years after terministic DCF model into two directions. One
the beginning of the project. This means that considers the electricity and natural gas prices
building phase takes four years and the reliabil- per se. The other reflects cash flows composed
ity tests are performed during the first two years by the difference between the functions of the
of production. The present cost of the building electricity price and the fuel price. The project
phase is €205 million (Table 3). The production income depends primarily on electricity demand
phase lasts for 25 years with a 10-year guaranteed and its growth over the years, as well as the elec-
power purchase agreement (Baltazar, 2009). Us- tricity price charged in the Iberian market. The
ing historical market data since January 2000, it electricity cost of production essentially depends
is possible to extract trends from the electricity on fuel prices such as the natural gas bought in
and natural gas prices. A DCF analysis showed producer countries and distributed through the
the discounted 29-year net present value is a national natural gas network. Because no reliable
negative value of €57 million (Baltazar, 2009), long-term market-based predictions for forward
considering a risk-adjusted discount rate of 8% fuel prices are available, the method uses the
and a tax rate of 26.5% (Table 5). Under these company production cost-based technique. This
conditions, the project should be rejected. This method retrieves the cost according to the market
level of project value seems to confirm Frayer prices, adding up the other costs associated with
255
Power Systems Investments
Δ(σ2) (1) Δ(πΔp) (2) (2)/(1) Δ(δ) (1) Δ(πΔp) (2) (2)/(1)
-50.00% -51.67% 0.033 10.00% -23.24% -1.324
-58.33% -62.79% 0.076 20.00% -46.89% -1.344
-66.67% -74.60% 0.119 30.00% -71.72% -1.391
-75.00% -85.56% 0.141
Δ(ψ) (1) Δ(πΔp) (2) (2)/(1) Δ(ρ) (1) Δ(πΔp) (2) (2)/(1)
0.00% 0.00% NA 10.00% -10.87% -0.087
-33.33% -23.58% -0.293 20.00% -21.62% -0.081
-66.67% -34.34% -0.485 30.00% -32.50% -0.083
carbon dioxide emissions and operations and power plant operation with carbon capture and
maintenance. storage would not be cost effective. However,
The next phase combines two uncertainties because technology is improving at a fast pace,
(electricity price and natural gas price) in the the investment might turn out to be cost effective
cash flow statement and obtains the initial project in the near future.
volatility. This initial value for volatility is dis- The study integrates two uncertainties (elec-
similar to individual uncertainties and depends tricity price and natural gas price) to evaluate the
on the structure of the cash flows. Brandão et al. model value over time. It assumes that the two
(2005) attributed much of this difference to effects variables follow a GBM with a volatility of 54%,
related to operational leverage. In our case, as the for the electricity price process and a volatility
difference between costs and revenues is small, of 26% for the natural gas price process. The
a tiny change in revenues or costs has a relevant procedure excludes any variable correlation val-
impact on cash flows. Volatility changes its value ues. When compared with Frayer and Uludere’s
according these changes. (2001) peak (15%) and off-peak values (22%),
Considering this framework, the ROA can the volatility estimates seem to be overvalued,
show the project creates value under different which confirms Smith (2005).
economic and technological conditions. A DCF After considering a two-variable approach,
analysis of deterministic net present value ignores the process of Brandão et al. (2005) implies a
the flexibility to start the project at the right time. combination of the volatility forecasts associated
When the project value varies stochastically over with the two random variables. Although there
time, the possibility of waiting and seeing is are alternative ways of achieving it (Smith, 2005;
clearly an asset for managers (Trigeorgis, 1996). Amendola and Storti, 2008), the process produces
Assuming the existing technical conditions, a a volatility level of 120%. When the uncertainties
256
Power Systems Investments
are somewhat correlated, the project volatility 40% of the appliances, the study avoids the use
should be higher than the sum of the individual of binomial lattices to make it safer to examine
volatilities. In the absence of additional elements, the results. The approximation of project uncer-
Lima and Suslick (2006) confirmed empirically tainty applies a risk-neutral valuation and includes
these considerations in similar projects. In sum- options in the decision nodes of the tree. As in
mary, Table 3 shows the historical volatility of Copeland and Antikarov (2001), real option value
uncertainty variables of the model. results from the binomial tree filled with project
Using the DCF method, the plant investment volatility replicated from forward simulations.
shows not to be cost effective (negative project The model approach integrates two uncertain-
value). Although this method does not take into ac- ties (electricity and natural gas price) and one
count price uncertainties and can produce incorrect option (to invest or abandon the project during
information to support decisions, modelling these the time of license). The payoff of each period
uncertainties allows us to catch the investment reflects the cash flows that result from the ran-
opportunity when it becomes cost effective. The dom variables. Baltazar (2009) studied an initial
following section discusses the pricing of the de- investment value of €205 million and a risk-free
ferment option using the approaches suggested by rate of 2.16% (MTSP, 2009).
Copeland and Antikarov (2001), Smith (2005) and Using simulated volatilities of 113% (electric-
Brandão et al. (2005). This section also contains ity price) and 84% (natural gas price) obtains,
a sensitivity analysis of uncertainties parameters through a binomial tree, a project value of €102
and other variables not modelled stochastically. million and three years to maturity deferment
option of €159 million (Model(1)). The model
Deferment Option Analysis considers a binomial tree that matches the returns
and assumes an equal initial investment value and
Given the project deferment option, the goal is risk-free rate. The changed project value differs
to determine when the investment becomes eco- by only about 5% (€97 million) from the previous
nomically viable. The present value is negative approach with an option value of €154 million
but the flexibility to defer the investment offers (Model(2)). The difference can be considered
management the possibility to benefit from ran- irrelevant, taking into account the initial invest-
dom movements in the project value. However, if ment value of €205 million.
market conditions change, it will affect the asym- The next section examines the robustness of the
metry in the obligation to invest (Trigeorgis, 1996). critical parameters such as project volatility, op-
The firm has a three-year investment license tion maturity, electricity demand and the price for
for starting the plant construction. If the option carbon dioxide emissions. Assuming all the other
expires before the end of the license, the invest- variables remain constant, sensitivity analyses
ment opportunity is lost. Although there are three show how each parameter individually moves and
market uncertainties (electricity price, natural gas compares them with the strategic net present value
price and the price for carbon dioxide emissions) that corresponds to baseline parameterisation.
only the first two are considered. Sensitivity
analysis will lead to the uncertainties which have Process Evaluation and
no market-equivalent. Sensitivity Analysis
This study uses a binomial tree that models
the stochastic process (GBM) over the time of the Several tests can be performed to assess the quality
project (Wang and Dyer, 2010). Because Panko of the process simulations. First, it is necessary
(1998) evidenced errors of between 20% and to corroborate if simulated project values follow
257
Power Systems Investments
a normal distribution. Second, it should be seen The relationship between electricity demand
whether the volatility over the expected project (δ) and price for carbon dioxide emissions (ρ)
returns remains constant from period to period. with the project value is reversed from those
The results of the x2 test with a p value of 5% shown by the two previous parameters (volatility
indicate that the null hypothesis of project returns and maturity). An increase in the value of these
following a lognormal distribution cannot be variables causes a reduction in the project value.
rejected. To examine volatility changes over the The variation in electricity demand leads to twice
several periods of the project, the paper calculates the adjustment in the project value of the same
the value of φB by applying expression (5) to all percentage change in the price for carbon dioxide
the periods. Because the difference between the emissions. Although there are significant changes
values of each period remain in the interval of in the project value, none of the sensitivity analy-
–0.2% and +0.8%, it can be considered that there sis procedures call into question the project with
are no significant variations in volatility from regard to its viability.
period to period. The following section describes some of the
For the sensitivity analysis, the study em- latent gaps in an ROA, which could be used as
ploys the expected threshold values of the cash indicators of future research lines. These perspec-
flow variation (πΔp) to capture the impact of the tives result from the connection between the DCF
variation in the uncertainties parameters. The and real options paradigms on risk assessment in
expression Δπ=((πΔp / πbp)-1)*100 translates the the decision-making process.
described method where πΔp is the strategic net
present value after revised parameterisation and
πbp corresponds to a baseline parameterisation. FUTURE RESEARCH DIRECTIONS
Table 2 considers deviations in the critical
parameters in a range between +10% and –75%. In a certain sense, electricity market is different
Each parameter change has a corresponding varia- from other commodities because it’s expensive to
tion in the strategic present value (πΔp). About the store electricity on a large scale. Consequently,
direction of variation it is possible to detect differ- electricity prices depend on the customers’ needs,
ent behaviours between the various parameters. resulting in large fluctuations in demand during
The values tested for the project volatility are the day and in balance problems between demand
the absolute values between 30% and 60%, with and production. For that reason, minimizing the
intervals of 10% that correspond to variations production costs and building a flexible adjustment
between –50% and –75%. between demand and supply are relevant factors to
A negative change in the project value’s vari- mitigate market risks and get a significant return
ance (σ2) reflects a change in the same direction on investments. Combined-cycle power plants
in the project’s value. Because the test values have low investment costs and short construction
of volatility deviate from the benchmark, this times when compared to coal-fired and nuclear
increases the magnitude of their effects on the plants (Fraser et al., 2000).
project value. The change in the option maturity This chapter analyze investments in gas-fired
(time deferral of the project) has the same effects power plants based on stochastic electricity and
on the project value as the volatility. However, the natural gas prices. Fleten and Nasakkala (2010)
same percentage change in the project volatility analyze the same variables and find that when the
causes an update in the project value twice as decision to build is considered, the abandonment
great as the change in the deferment period of option does not have significant value. This study
the project (𝜓). can also consider an abandonment option and ad-
258
Power Systems Investments
ditionally examine the effects of emission costs the index). The proper timing for its implemen-
on the value of installing CO2 capture technology. tation is not now. Therefore, the decision maker
Traditional DCF analysis in the energy sector should take the deferment option and delay its
can produce mistaken assessments by ignoring the implementation. Compared with other project
adaptability and possibility of cash flow change analyses (Frayer and Uludere, 2001), it seems
overtime. Some authors consider ROA imprecise that project volatility is slightly overvalued. This
and that might even complicate the procedure of feature seems to come from the high volatility of
identifying the best strategy. This article shows electricity price (Alvarado and Rajaraman, 2000)
that an ROA can supply added information on the and from a specific pattern of cash flows with
economic viability of energy projects, by valuat- high first investments and long periods of opera-
ing the deferment option. The energy market is tion. Furthermore, sensitivity analysis examines
uncertain and thereby an ROA is an essential tool changes in project volatility, period of production,
in strategic decision making. However, to use disruptions in electricity demand and changes in
consistently an ROA, there are some questions prices for carbon dioxide emissions and shows
to discuss. some robustness of the results.
For projects with equal returns, assuming Therefore, even with some conceptual con-
risk, neutrality and aversion, an ROA chooses traindications, ROA approach of Brandão et al.
the projects with a higher volatility. It contradicts (2005) can be considered fair. This study shows
some rules from the traditional literature, which that energy markets with long periods of opera-
encourage minimising the risk from an equal tion and a reasonable adhesion to the GBM, by
profitability. Also, an ROA creates some bias in the project returns, might call into question its
project selection. Increasing the approval rate of use in these markets. Although the values of op-
unprofitable projects in the short-term and decreas- tion evaluation have remained inside acceptable
ing the acceptance rate of projects that with new ranges, some shortfalls could be found.
information might offer new opportunities in the First, the study employs Monte Carlo simula-
long-term. Future research should try to overcome tions, which can be viewed as forward-looking
these problems in order to spread the use of ROA. processes, and option pricing problems need
In terms of technology, the ROA can be backward recursive solutions. So, these simula-
extended to choose between electricity generat- tions can be considered less appropriate to deal
ing technologies. The firm can decide when to with these kinds of option problems.
invest either in a Natural Gas Combined Cycle Second, Smith (2005) suggested some alter-
(NGCC) power plant or in an Integrated Gas- native approaches to estimate volatility, such as
ification Combined Cycle (IGCC) power plant. Longstaff and Schwartz (2001), Farias and Van
Furthermore, instead of assuming that fuel prices Roy (2003) and Ibanez and Zapatero (2004).
follow standard geometric Brownian motions can Since our study produces similar results, it is not
be assumed mean reversions, as in Abadie and obvious the extra complexity and computational
Chamorro (2008). needs needed by these approaches can produce
significantly superior results. Finally, it is broadly
accepted that simulation models have a limited
CONCLUSION ability to model complex options problems and
large possibilities of containing spreadsheet errors.
The article shows there is an investment project It is also consensual that lattices have relevant
to be carried out (profitability index of 1.50) with drawbacks on transparency and intuition.
a deferment option that contributes 51.62% (for
259
Power Systems Investments
Baltazar, D. (2009). Real options in engineer- Eurostat. (2009). Eurostat database. Retrieved
ing projects. Master Thesis (L53087). Lisbon, from http://epp.eurostat.ec.europe.eu/portal
Portugal: Technical University of Lisbon (IST). Eyedeland, A., & Wolyniec, K. (2003). Energy
Black, F., & Scholes, M. (1973). The pric- and power risk management: New developments
ing of options and corporate liabilities. The in modeling, pricing and hedging. New York, NY:
Journal of Political Economy, 81(3), 637–654. John Wiley and Sons.
doi:10.1086/260062 Farias, D., & Van Roy, B. (2003). The linear
Blyth, W. (2008). Use of real options as a policy programming approach to approximate dynamic
analysis tool, in analytical methods for energy programming. Operations Research, 51, 850–865.
diversity and security (pp. 69–83). London, UK: doi:10.1287/opre.51.6.850.24925
Elsevier. Fleten, S., & Nasakkala, E. (2010). Gas-fired
Borison, A. (2005). Real options analysis: Where power plants: Investment timing, operating flex-
are the emperor’s clothes? Journal of Applied Cor- ibility and CO2 capture. Energy Economics, 32(4),
porate Finance, 17(2), 17–31. doi:10.1111/j.1745- 805–816. doi:10.1016/j.eneco.2009.08.003
6622.2005.00029.x Fraser, N., Bernhardt, J., & Jewkes, E. (2000).
Brandão, L., Dyer, J., & Hahn, W. (2005). Using Engineering economics in Canada (2nd ed.).
binomial trees to solve real-option valuation prob- Toronto, Canada: Prentice Hall.
lems. Decision Analysis, 2(2), 69–88. doi:10.1287/ Frayer, J., & Uludere, N. (2001). What is it worth?
deca.1050.0040 Application of real options theory to the valuation
Brekke, K., & Schieldrop, B. (2000). Invest- of generation assets. The Electricity Journal, 14(8).
ment in flexible technologies under uncertainty. doi:10.1016/S1040-6190(01)00237-8
In Brennan, M., & Trigeorgis, L. (Eds.), Project Gardner, D., & Rogers, J. (1999). Planning
flexibility, agency and competition: New develop- electric power systems under demand. Manage-
ments in the theory of real options (pp. 34–49). ment Science, 45(10), 1289–1306. doi:10.1287/
Oxford University Press. mnsc.45.10.1289
260
Power Systems Investments
Gnansounou, E., Schmutz, A., & Dong, J. (2003). Reedman, L., Graham, P., & Coombes, P. (2006).
Investment decisions with flexibility in the liberal- Using a real-options approach to model technol-
ized electricity markets: An approach using real ogy adoption under carbon price uncertainty: An
options theory. SAFE Annual Conference. Zurich. application to the Australian electricity generation
sector. The Economic Record, 82(S1), S64–S73.
Ibanez, A., & Zapatero, F. (2004). Monte Carlo
doi:10.1111/j.1475-4932.2006.00333.x
valuation of American options through computa-
tion of the optimal exercise frontier. Journal of Rode, D., Fischbeck, P., & Dean, S. (2001). Monte
Financial and Quantitative Analysis, 39, 253–275. Carlo methods for appraisal and valuation: A case
doi:10.1017/S0022109000003069 study of a nuclear power plant. Journal of Struc-
tured and Project Finance, 7, 38–48. doi:10.3905/
Kehlhofer, R., Bachmann, R., Nielson, H., &
jsf.2001.320257
Warner, J. (1999). Combined-cycle gas & steam
turbine power plants (2nd ed.). Tulsa, OK: Pen- Smith, J. (2005). Alternative approaches for solv-
nwell. ing real options problems. Decision Analysis, 2(2),
89–102. doi:10.1287/deca.1050.0041
Lima, G., & Suslick, S. (2006). Estimating the
volatility of mining projects considering price and Smith, J., & Nau, R. (1995). Valuing risky projects:
operating cost uncertainties. Resources Policy, Option pricing theory and decision analysis. Man-
31, 86–94. doi:10.1016/j.resourpol.2006.07.002 agement Science, 14(5), 795–816. doi:10.1287/
mnsc.41.5.795
Longstaff, F., & Schwartz, E. (2001). Valuing
American options by simulation: A simple least- Spinney, P., & Watkins, G. (1996). Monte Carlo
squares approach. Review of Financial Studies, simulation techniques and electric utility re-
14, 113–147. doi:10.1093/rfs/14.1.113 source decisions. Energy Policy, 24, 155–163.
doi:10.1016/0301-4215(95)00094-1
Monaghan, B. (2009). How to trade natural gas.
Retrieved from http://www.paddypowertrader. Thompson, M., Davison, M., & Rasmussen,
com/index.php H. (2004). Valuation and optimal operation of
electric power plants in competitive markets.
MTSP. (2009). MTSPortugal database. Retrieved
Operations Research, 52, 546–562. doi:10.1287/
from http://www.mtsportugal.com/
opre.1040.0117
Nau, R., & McCardle, K. (1991). Arbitrage, ra-
Trigeorgis, L. (1993). Real options and interactions
tionality and equilibrium. Theory and Decision,
with financial flexibility. Financial Management,
33, 199–240. doi:10.1007/BF00132993
22(3), 202–223. doi:10.2307/3665939
OMEL. (2009). Operador del Mercado Ibérico
Trigeorgis, L. (1996). Real options: Managerial
de Energía. Retrieved from http://www.omel.es
flexibility and strategy in resource allocation.
Panko, R. (1998). What we know about spread- Cambridge, MA: MIT Press.
sheet errors. Journal of Organizational and End
Tseng, C., & Barz, G. (2002). Short-term gen-
User Computing, 10, 15–21.
eration asset valuation: A real options approach.
Operations Research, 50(2).
Wang, T., & Dyer, J. (2010). Valuing multifac-
tor real options using an implied binomial tree.
Decision Analysis, 7(2), 185–195. doi:10.1287/
deca.1100.0174
261
Power Systems Investments
Yang, M., & Blyth, W. (2007). Modelling invest- Cortazar, G. (2004). Simulation and numerical
ment risks and uncertainties with real options ap- methods in real options valuation, in Real Options
proach. In IEA (Ed.), Climate policy uncertainty and Investment under Uncertainty: Classical
and investment risk. Paris, France: IEA. Readings and Recent Contributions, Schwartz
and Trigeorgis (editors), USA: MIT Press.
Cox, J., Ross, S., & Rubinstein, M. (1979). Option
ADDITIONAL READING pricing: a simplified approach. Journal of Finan-
cial Economics, 7, 229–263. doi:10.1016/0304-
Amramand, M., & Kulatilaka, N. (1999). Real 405X(79)90015-1
options: Managing strategic investment in an
uncertain world. Cambridge, MA: Harvard Busi- Deb, R. (2000). Operating hydroelectric and
ness School Press. pumped storage units in a competitive environ-
ment, Electricity Journal.
Botterud, A., & Korpas, M. (2007). A stochastic
dynamic model for optimal timing of investment Deng, S., Johnson, B., & Sogomonian, A. (1999).
in new generation capacity in restructured power Exotic electricity options and the valuation of
systems. Electrical Power and Energy Systems, electricity generation and transmission assets.
29(2), 163–174. doi:10.1016/j.ijepes.2006.06.006 Proceedings of 32nd Hawaii International Confer-
ence on System Sciences.
Boyle, P. (1977). Options: A Monte Carlo Ap-
proach. Journal of Financial Economics, 4, Devin, K. (1995). Gas in electricity generation. En-
323–338. doi:10.1016/0304-405X(77)90005-8 ergy Exploration and Exploitation, 13, 149–157.
Brealey, R., & Myers, S. (2002). Capital invest- Dixit, A., & Pindyck, R. (1994). Investment under
ment and valuation (1st ed.). USA: McGraw-Hill. uncertainty. NJ: Princeton University Press.
Brennan, M., & Schwartz, E. (1985). Evaluat- Fuss, S., & Szolgayova, J. (2010). Fuel price and
ing natural resource investments. The Journal of technological uncertainty in a real options model
Business, 58(2), 135–157. doi:10.1086/296288 for electricity planning. Applied Energy, 87(9).
doi:10.1016/j.apenergy.2009.05.020
Castellacci, G. and Siclari, M. (2003). Asian bas-
ket spreads and other exotic averaging options. Gardner, D., & Rogers, J. (1999). Planning elec-
Energy Power Risk Management, March (2003). tric power systems under demand uncertainty
with different technology lead times. Manage-
Castellacci, G., Siclari, M., Ross, S. and Liao, W. ment Science, 45(10), 1289–1306. doi:10.1287/
(2004). Real option valuation of power generation mnsc.45.10.1289
assets and spread-switching options. Proceedings
of Forecasting Financial Markets. Geman, H. (2005). Energy commodity prices:
Is mean-reversion dead? Journal of Alterna-
Chen, H., & Baldick, R. (2007). Optimizing short- tive Investments, 8(1), 31–45. doi:10.3905/
term natural gas supply portfolio for electric utility jai.2005.591576
companies. IEEE Transactions on Power Systems,
22(1). doi:10.1109/TPWRS.2006.889144 Griffes, P., Hsu, M., & Kahn, E. (1999). Power
asset valuation: real options, ancillary services
Copeland, T., & Antikarov, V. (2001). Real op- and environmental risks. The New Power Markets.
tions: A practitioners guide. USA: Texere Pub- London, UK: Risk Books.
lishing.
262
Power Systems Investments
Hsu, M. (1998). Spark spread options are hot! Leppard, S. (2004). Valuation and risk manage-
The Electricity Journal, 11(2), 1–12. doi:10.1016/ ment of physical assets. In Kaminski, V. (Ed.),
S1040-6190(98)00004-9 Managing Energy Price Risk: The New Challenges
and Solutions (3rd ed.). Risk Books.
Hull, J. (2002). Options, futures and other deriva-
tives (5th ed.). USA: Prentice Hall. Leslie, K., & Michales, M. (1997). The real power
of real options. The McKinsey Quarterly, (3): 4–22.
Hull, J., & White, A. (1994a). Numerical proce-
dures for implementing term structure models I: Li, F., & Chiu, W. (2002). Valuing generation as-
single-factor models. Journal of Derivatives, 2(1), sets using real option competitive price analysis:
7–16. doi:10.3905/jod.1994.407902 A Step-by-Step Valuation Example for a Portfolio
of Generating Assets. In Ronn, E. (Ed.), Real
Hull, J., & White, A. (1994b). Numerical proce-
Options and Energy Management (pp. 597–634).
dures for implementing term structure models II:
Risk Waters Group Ltd.
two-factor models. Journal of Derivatives, 2(2),
37–48. doi:10.3905/jod.1994.407908 Luherman, T. (1998). Investment opportunities as
real options: Getting started in numbers. Harvard
Ingersoll, J., & Ross, S. (1992). Waiting to invest:
Business Review, July-Aug., 51-67.
investment under uncertainty. The Journal of Busi-
ness, 65(1), 1–29. doi:10.1086/296555 Madlener, R., Kumbaroglu, G., & Ediger, V.
(2004). Modeling technology adoption as an
Jacobs, J., & Schultz, G. (2002). Opportunities
irreversible investment under uncertainty: the
for stochastic and probabilistic modelling in the
case of the Turkish electricity supply industry.
deregulated electricity industry. In Greengard,
Energy Economics, 27(1), 139–163. doi:10.1016/j.
C., & Ruszczynski, A. (Eds.), Decision Making
eneco.2004.10.007
Under Uncertainty – Energy and Power (p. 128).
New York: Springer. Mcdonald, R., & Siegel, D. (1985). Investments
and the valuation of firms when there is an option
Kester, W. (1984). Today’s options for tomorrow’s
to shutdown. International Economic Review,
growth. Harvard Business Review, 62, 153–160.
26(2), 331–349. doi:10.2307/2526587
Kumbaroglu, G., Madlener, R., & Demirel, M.
McDonald, R., & Siegel, D. (1986). The value of
(2008). A real options evaluation model for the
waiting to invest. The Quarterly Journal of Eco-
diffusion prospects of new renewable power gen-
nomics, 101(4), 707–728. doi:10.2307/1884175
eration technologies. Energy Economics, 30(4),
1882–1908. doi:10.1016/j.eneco.2006.10.009 Merton, R. (1973). Theory of rational option
pricing. The Bell Journal of Economics and
Larsen, E., & Bunn, D. (1999). Deregulation in
Management Science, 4(Spring), 141–183.
electricity: Understanding strategic and regulatory
doi:10.2307/3003143
risks. The Journal of the Operational Research
Society, 50(4). Myers, S. (1977). Determinants of corporate bor-
rowing. Journal of Financial Economics, (5): 147.
Laurikka, H. (2006). Option value of gasification
doi:10.1016/0304-405X(77)90015-0
technology within an emissions trading scheme.
Energy Policy, 34(18), 3916–3928. doi:10.1016/j. Pindyck, R. (1999). The long-run evolution of
enpol.2005.09.002 energy prices. The Energy Journal (Cambridge,
Mass.), 20(2), 127. doi:10.5547/ISSN0195-6574-
EJ-Vol20-No2-1
263
Power Systems Investments
Pindyck, R. (2001). The dynamics of commod- Zambujal-Oliveira, J., & Duque, J. (2010). Opera-
ity spot and futures markets: A primer. The En- tional asset replacement strategy: A real options
ergy Journal (Cambridge, Mass.), 22(3), 1–29. approach. European Journal of Operational
doi:10.5547/ISSN0195-6574-EJ-Vol22-No3-1 Research. doi:.doi:10.1016/j.ejor.2010.09.011
Ronn, E. (Ed.). (2002). Real options and energy
management. USA: Risk Waters Group Ltd.
KEY TERMS AND DEFINITIONS
Roques, F., Nuttall, W., & Newbery, D. (2006).
Using probabilistic analysis to value power gen- Binomial options pricing model: Numerical
eration investments under uncertainty. Working method for the valuation of options proposed by
Paper, EPRG. Cox, Ross and Rubinstein (1979). The model uses
Rosenberg, M., Bryngelson, J., Sidorenko, N., & a lattice based model of the varying price over
Baron, M. (2002). Price spikes and real options: time of the underlying asset.
Transmission valuation. In Ronn, E. (Ed.), Real Combined Cycle: An electrical generating
Options and Energy Management (pp. 323–369). technology in which electricity is produced from
Risk Waters Group Ltd. otherwise lost waste heat, exiting from one or
more gas (combustion) turbines. The exiting heat
Schwartz, E., & Trigeorgis, L. (2004). Real options is routed to a conventional boiler or to a heat re-
and investment under uncertainty: classical read- covery steam generator for use by a steam turbine
ings and recent contributions. USA: MIT Press. to generate electricity.
Shahidehpour, M., Yami, H., & Zuvi, L. (2002). Historical Volatility: The annualized Stan-
Market operations in electric power systems: Fore- dard Deviation of percent changes in prices over
casting, scheduling and risk management. USA: a specific period. It is a hint of past volatility in
Wiley-IEEE Press. doi:10.1002/047122412X market prices.
Monte Carlo Simulation: Risk analysis
Shimko, D. (1994). Options on futures spread: techniques in which probable future events are
Hedging, speculation and valuation. Journal of Fu- simulated on a computer generating estimated
tures Markets, 14(2). doi:10.1002/fut.3990140206 rates of return and risk indexes.
Sun, N., Ellersdorfer, I., & Swider, D. (2008). Option: The right, but not the obligation, to
Model-based long-term electricity generation purchase, sale or contract an underlying asset at
system planning under uncertainty. Proc. IEEE In- a specified price during a defined time period.
ternational Conf. Electric Utility Deregulation and Real Options Analysis: Applies option valu-
Restructuring and Power Technologies, Nanjing. ation methods to investment decisions.
Risk-Adjusted Discount Rate: The discount
Weron, R. (2006). Modelling and forecasting rate that applies to a particular risky stream of
electricity load and prices: A statistical approach. income; the riskier the project’s income stream,
UK: John Wiley and Sons Ltd. the higher the discount rate.
Zambujal-Oliveira, J. (forthcoming). Validity of Spark Spread: The relationship between the
the homogeneity property in real assets replace- price of electricity and the price of natural gas or
ment procedures. International Review of Applied other fuel used to produce electricity. The spark
Financial Issues and Economics. spread reflects the costs, or expected costs, of
producing power.
264
Power Systems Investments
APPENDIX
Year 5 6 7 8 9 10 11 12 13 14 15 16 17
Net energy production (GWh) 1515 1510 1633 1544 1498 1478 1475 1539 1507 1477 1405 1335
Guaranteed production (k€) 7798 7798 7798 7798 7798 7798 7798 7798 7798 0 0 0
Electricity Price (€/MWh) 83.6 85.3 87 88.8 90.5 92.3 94.2 96.1 98 100 102 104
Year 18 19 20 21 22 23 24 25 26 27 28 29
Net energy production (GWh) 1194 1125 1210 1305 1392 1477 1567 1534 1501 1468 1468
Guaranteed production (k€) 0 0 0 0 0 0 0 0 0 0 0
Electricity Price (€/MWh) 108 110 113 115 117 119 122 124 127 129 132
265
266
Table 7. Revenues and costs of a combined-cycle plant investment (Baltazar, 2009).
Year 5 6 7 8 9 10 11 12 13 14 15 16 17
Natural 30 30 29 31 32 32 33 33 34 35 35 36 36
Gas Prices
(€/MWht)
PCI (%) 0.56 0.56 0.56 0.56 0.56 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55
CO2 Cost 28 28 29 29 29 31 32 34 35 37 39 40 42
(€/t)
CO2 Emis- 564,861 549,145 548,299 593,128 561,045 544,646 537,899 537,031 560,322 548,667 537,862 511,772 486,238
sion (t)
O&M 2,396 2,444 2,493 5,392 5,443 5,494 5,547 5,601 4,456 4,512 4,570 4,628 4,688
Fixed Cost
(€)
O&M Vari- 1,662 1,652 1,683 1,856 1,794 1,777 1,792 1,826 1,941 1,940 1,941 1,885 1,829
able Cost
(€)
Year 18 19 20 21 22 23 24 25 26 27 28 29 30
Natural 37 38 38 39 39 40 41 42 42 43 44 44
Gas Prices
(€/MWht)
PCI (%) 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.55
CO2 Cost 44 46 49 51 53 56 58 61 64 67 70 73
(€/t)
CO2 Emis- 460,064 435,682 410,289 441,541 476,612 508,928 540,050 573,170 561,554 549,909 538,391 538,391
sion (t)
O&M 4,748 5,602 5,665 5,729 5,795 5,862 5,931 14,172 7,621 7,693 7,768 6,880
Fixed Cost
(€)
O&M Vari- 1,768 1,708 1,642 1,800 1,983 2,160 2,336 2,529 2,529 2,528 2,526 1,888
able Cost
(€)
Power Systems Investments
Table 8. Cash flows of the combined cycle plant investment (Baltazar, 2009)
Year 1-4 5 6 7 8 9 10 11 12 13 14 15 16
Revenues 136,199 129,152 126,533 145,428 140,860 140,209 139,947 142,672 148,456 149,140 141,736 137,670
Costs 104,886 100,045 99,719 116,668 112,353 111,542 112,608 114,731 120,691 120,752 121,000 117,848
EBITDA 31,313 29,107 26,814 28,760 28,507 28,667 27,339 27,941 27,765 28,388 20,736 19,822
Deprecia- 9,785 9,792 9,792 9,792 9,722 9,511 9,511 9,511 9,511 9,511 9,511 9,511
tion
Power Systems Investments
EBIT 21,528 19,315 17,022 18,968 18,785 19,156 17,828 18,430 18,254 18,877 11,225 10,311
Taxes 5,705 5,119 4,511 5,027 4,978 5,076 4,724 4,884 4,837 5,002 2,974 2,732
Investment 205,427
Cash- 25,608 23,988 22,303 23,733 23,529 23,591 22,615 23,057 22,928 23,386 17,762 17,090
flows
Discount 17429 15117 13014 12823 11770 10927 9699 9156 8430 7962 5599 4988
Cash-
flows
Year 17 18 19 20 21 22 23 24 25 26 27 28 29
Revenues 133,463 128,772 124,916 120,215 129,324 139,395 148,419 158,032 168,072 165,920 163,847 161,476 160058
Costs 114,717 111,207 108,645 104,956 115,000 126,441 137,639 149,029 169,628 163,495 163,938 164,395 166732
EBITDA 18,746 17,565 16,271 15,259 14,324 12,954 10,780 9,003 -1,556 2,425 -91 -2,919 -6,674
Deprecia- 9,511 9,511 9,511 9,511 9,511 9,511 9,511 9,511 9,511 9,511 9,511 7,262 354
tion
EBIT 9,234 8,054 6,760 5,748 4,813 3,443 1,269 -508 -11,067 -7,086 -9,602 -10,181 -7,028
Taxes 2,447 2,134 1,791 1,523 1,275 912 336 0 0 0 0 0 0
Investment
Cash- 16,298 15,431 14,480 13,736 13,049 12,042 10,444 9,003 -1,556 2,425 -91 -2,919 -6,674
flows
Discount 4405 3864 3355 2947 2592 2215 1779 1420 -227 328 -11 -338 -716
Cash-
flows
Net Present -56904
Value
267
268
Chapter 9
Optimal Configuration and
Reconfiguration of Electric
Distribution Networks
Armin Ebrahimi Milani
Islamic Azad University, Iran
ABSTRACT
Power loss reduction can be considered as one of the main purposes for a distribution system’s design-
ers and operators, especially for recent non-governmental networks. Moreover, the nature of power
loss challenges different methods to solve this problem, while various studies indicate effectiveness
of reconfiguration and its high portion for this case. Thus, “reconfiguration” can be introduced as an
optimization procedure to obtain economical high quality operation by changing the status of sectional-
izing switches in these networks. Some major points such as using different switch types, considering
number of switching and time varying loads, which are almost neglected or not applied simultaneously
in most pervious essays, are the main motivation to propose this chapter. A heuristic practical scheme
is proposed to perform optimal reconfiguration, and all previous neglected topics are fully discussed.
Proposed method will apply to sample distribution networks, and the effectiveness of this method will
be discussed through several case studies and comparisons.
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Optimal Configuration and Reconfiguration of Electric Distribution Networks
procedure starts by closing all network switches have studied the hourly load patterns over an
which are then opened one after another so as to interval of a year in order to define the hourly
establish the optimum power flow in the network. load conditions for each season. They have used
This is while, many approximations of pervious Branch and Bound technique for obtaining mini-
method have been overcome in this algorithm and mum loss configuration considering hourly load
the computation takes less time. patterns over annual intervals. It was the year that
Civanlar et al. (1988) made use of method Kashem et al. (1999) investigated the load rear-
which is known as a “Branch Exchange” opera- rangement and proposed their famous method for
tion. This method suggested the opening of any load balancing in distribution systems.
switch was required to correspond to the closure Huang and Chin (2002) proposed an algorithm
of another switch, ensuring that the radial con- based on fuzzy operation in order to deal with
figuration of the distribution network would be feeder reconfiguration problem. They suggested
preserved. A year after, Baran and Wu (1989) this method to reduce power loss as well as to
succeeded to improve the method of Branch Ex- acquire the load balance. More Artificial intel-
change by offering two approximated power flow ligence based methods, such as (Ahuja, 2007;
Equations in 1989. Their power flow Equations Mendoza, 2009; Milani 2010) are other useful
introduced by recursive approximation of P, Q procedures proposed for reconfiguration of dis-
and V for each branch. tribution networks, recently.
Liu (1989) in the same year, Taylor, Chiang By review the full history of distribution
and Jeon (1990) in next years presented various network reconfiguration, can perceive a vast
heuristic method for distribution feeder recon- bibliography on this optimization problem. As it
figuration. Wagner (1993), compares different mentioned before, it includes general attempts to
major reconfiguration methods. This comparison find a global optimal solution, new approximate
depicts considerable economic advantages by us- methods (Civanlar, 1988), (Baran, 1989) and al-
ing presented heuristic methods especially for real ternative paradigms of approaching the problem,
time operations. He indicated that an important such as evolutionary computation (Nara, 1992)
loss reduction was obtained through simulations and other bio-inspired techniques (Ahuja, 2007).
in Canadian networks during a one year period. What is more considerable here is that most
Genetic Algorithm applied in distribution sys- of these approaches consider fixed demands at
tem reconfiguration for reduction of real power each load point (usually the maximum demands
loss by Nara et al. (1992) for the first time. They forecast for the planning period). However, loads
compared obtained results with the results of Simu- vary during any given planning period, with a dif-
lated Annealing (SA) method and concluded that ferent pattern for each bus, and these variations
although the genetic algorithm has less assurance must be considered in the search for a minimum
but it can be faster that SA. After introducing GA loss configuration. Also it seems considering
method and due to its effectiveness, many works different switch types will affect on optimal
done by refine or suggest heuristic techniques to configuration and reconfiguration of electric
this algorithm to compensate long solution time. distribution networks.
Sarfi et al. (1996) presented an algorithm
based on “Network Partitioning Method” after
their efforts to use refined genetic algorithm for MAIN FOCUS OF THE CHAPTER
reconfiguration of distribution networks. Before
that, Chen and Cho (1993) have performed an After a brief introduction to electric distribution
analysis of hourly reconfiguration scheme. They networks, their configuration and reconfiguration,
269
Optimal Configuration and Reconfiguration of Electric Distribution Networks
this chapter will focus on some issues involved definition radial network leaves the station and
with reconfiguration of electric distribution net- passes through the network area with no normal
works and provides practical answers to these connection to any other supply while an intercon-
major points. Through this investigation readers nected conFigure, such as mesh structure, will have
learn complete concepts of an optimal reconfigura- multiple connections to other points of supply.
tion procedure to reduce power loss and total costs, In general, the main advantages of radial con-
while they will get some basic information and figuration are its simplicity and its low cost (Ahuja,
knowledge about electric distribution networks’ Das & Pahwa, 2007). In radial configuration, the
configuration. Fully discussion about this subject number of disconnecting devices reduces and
in following sections supplies a perfect insight to design of a protection system is not complicated.
this problem and its issues. Moreover, this configuration is widely used for
effective coordination of its protective devices
(Kashem, Ganapathy & Jasmon, 1999). Figure 2
INTRODUCTION TO illustrates single-line diagram of a typical men-
DISTRIBUTION NETWORKS tioned distribution network.
AND THEIR CONFIGURATION The loop (or ring) system of distribution starts
at the substation and is connected to or encircles
Electricity distribution is the final stage in the an area serving one or more distribution transform-
delivery before retail of electricity to end users. ers or load centers. Unlike the radial network, the
A distribution system’s network carries electric- conductor of the ring system returns to the same
ity from the transmission system and delivers it substation. The loop system is more expensive to
to the consumers. In the other words, distribu- build than the radial type, but it is more reliable
tion network provides the final links between and may be justified in an area where continuity
the transmission system and demands. Figure 1 of service is of considerable importance. Figure
gives a basic insight from distribution position in
power systems and depicts the main configuration
of the network. Figure 1. Distribution position in power system
Typically, a distribution network would include and its main configuration
medium-voltage power lines, electrical substations
and transformers, low-voltage distribution wiring
and sometimes electricity meters. These networks
are generally divided into subsystems of radial or
loop feeders fitted with number of switches that
are normally closed or opened. Customers could
be supplied from different substations or by dis-
tributed power generations through different
routes. These paths are characterized by different
mixtures of commercial, industrial and residential
customers who might impose time-varying load
demands and different service reliability require-
ments (Yin & Lu, 2009).
Radial distribution networks are the most con-
ventional configurations of distribution systems.
In such networks, feeders are extended from
distribution substations to lateral feeders while all
service areas are supplied through feeders. By this
270
Optimal Configuration and Reconfiguration of Electric Distribution Networks
3 illustrates single-line diagram of a typical men- points are not interrupted (Rugthaicharoencheep
tioned distribution network. & Sirisumrannukul, 2009). The whole mentioned
operation for reaching a new optimized configura-
tion is defined as “Reconfiguration” of distribution
INTRODUCTION TO DISTRIBUTION networks.
NETWORK’S RECONFIGURATION Figure 4 illustrates a typical radial distribution
network before and after reconfiguration. For this
In original configuration of a distribution network sample network, the operator has followed loss
there are some sectionalizing switches (normally reduction purpose. In this case switches S14 &
closed3) as well as some tie switches (normally S15 are closed after reconfiguration. This is while;
opened4) for both protection and configuration S8 & S9 are opened to preserve the initial radial
management. By changing the status of these structure of network under study.
switches, the configuration of distribution sys- Generally, reconfiguration methods can be
tem is varied and loads are transferred among divided in to two groups: “General” and “Spe-
the feeders. This could be done while the radial cific”. In the specific methods, one initial answer
configuration format of electrical supply is still is obtained and is used in a specific algorithm to
maintained at the end of this process and all load reach further answers up to an improvement point
271
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Figure 4. Typical radial distribution network before (above) and after (below) reconfiguration
considering the constraints of the problem. In costs. Therefore, these benefits can achieve trough
general methods, an algorithm is used for solving two different cases as follow:
the problem and a large range of answer are ob-
tained, among which by performing an operation Configuration Mode
the most improved answer is selected as a final
one. Cost minimization can achieve by using differ-
The advantages obtained from feeder recon- ent switch types to conFigure a new distribution
figuration using any proposed methods are, for system. This procedure will follow by different
example, real power loss reduction, balancing switch type placement in a statistic manner i.e. an
system load, bus voltage profile improvement, automatic switch will be the best optimal chose
increasing system security and reliability, and for a place with high switching probability where
power quality improvement. manual ones can be selected for places with less
switch operating.
Different Switch Types and Number Note that in this case there are two different
of Switching Consideration costs which together determine the total cost;
first the cost should spend for change the state
To conFigure a distribution network can consider of switches and second the cost of automation.
two different switch types; manual and automatic Considering this differentiations in costs, com-
ones. Considering the combination of these two panies can achieve most benefits through their
different types of switches simultaneously will investigations by establish an optimal balance
result an optimal flexible planning and operating between above maintained costs.
of these networks with the most minimization in
272
Optimal Configuration and Reconfiguration of Electric Distribution Networks
CA: Cost of operating for Automatic type switches BASIC DETERMINISTIC LOAD FLOW
CM: Cost of operating for Manual type switches The Load Flow (LF) analysis has always been one
of the most important analyses in power systems
SSAx: State of automatic switch in the xth con- (Papaefthymiou, 2003). The exact formulation of
figuration LF problem concerns the determination of real
and reactive power flows in each line (branch)
SSMx: State of manual switch in the xth con- of the power system. The data used to solve this
figuration problem are the active and reactive loads at load
In this definition, SSAx and SSMx get only buses as well as the real power generation and
two values in each configuration; one or zero, voltage magnitudes at the generator buses. Since,
where represent state of close or open respectively. in this formulation, single values are used as
Moreover, in above objective function w1 and w2 inputs for the solution of the problem, the name
represent weighing factors refer to automatic and “Deterministic Load Flow (DLF)” has been used
273
Optimal Configuration and Reconfiguration of Electric Distribution Networks
in order to distinguish this methodology from Thus, the net current injected into the network
“Probabilistic Load Flow (PLF)” introduces and at bus i will be given by:
discussed in future sections.
Back to DLF problem, the general DLF for- N
I i = Yi 1V1 + Yi 2V2 + ... + YiNVN = ∑YinVn
mulation can be divided into two aspects: n =1
(4)
1. Determination of the State Vector of the
system: the magnitude and phase angle of Let Pi and Qi denote the net real and reactive
the unknown bus voltages in the system are power entering the network at bus i. Then,
computed.
2. Determination of the power flows on each N
∑YinVV i n |< (θin + δn − δi
line of the system: the state vector is used in N
Pi − jQi = Vi * ∑YinVn = or
n =1
N
this step for the determination of the active n =1
∑ |Vi | . |Vn | [cos(δn − δi ) + J sin(δn − δi )[Gin + jBin }
and reactive power flows in every branch. n =1
(5)
274
Optimal Configuration and Reconfiguration of Electric Distribution Networks
In these buses loads are connected, so Pi, and Qi A typical daily load profile and concept of its
are known from historical record, load forecast variations is shown in Figure 5. Considering
or measurements; δi and |Vi| are the unknown such these patterns for different loads of network,
quantities. a probabilistic density function (PDF) in each
load node can be obtained. It is obvious that by
Voltage Controlled Buses increasing the numbers of segments, results are
(Generator Buses, P-V Nodes) more accurate while more time consuming and
vice versa.
At each bus where a generator is connected, the Bell shaped curve shown in Figure 6 illustrates
active power generation can be controlled by a sample PDF plotted by Monte Carlo Method5
adjusting the prime mover, and the voltage mag- (Rubinstein & Kroese, 2008). MCM is a statisti-
nitude can be controlled by adjusting the generator cal simulation uses a random sequence of numbers
excitation. Therefore, Pi and |Vi| are known and to describe the statistical behavior of a variable.
δi and Qi are the unknown quantities. In the other words, for this purpose, MCM uti-
lizes repeated trials of the deterministic load flow
Slack Bus technique to determine the probability distribu-
tions of the nodal powers, line flows and losses.
The power injection at this bus is determined by The hypothesis of the Monte Carlo Method,
the power balance Equation of the system. In as a mathematical technique, infers the search of
the formulation of the power flow problem, the an efficient solution instead of an accurate solu-
power injection in this bus is not pre-specified or tion (Opazo Mora, Garcia-Santander & Lopez
scheduled. After the power flow problem has been Parra, 2008). Having “expected demand” and its
solved, the difference between the total power “standard deviation”, desired PDF can be explains
which is injected into the system at all other buses as an exponential Equation as follow:
and the total output plus the losses, are assigned
to the slack bus. The voltage angle of this bus (x −µ )2
−1/ 2
1 σ 2
serves as a reference for the angles of all other bus f (x ) = .e (8)
voltages. The other known quantity for this bus is σ. 2.π
the voltage magnitude |Vi|. Obviously, there is no
requirement to include the power flow Equations where,
for the slack bus in the power flow problem. x: Variable
With these definitions, the problem is trans- σ: Standard deviation of x
formed to a system of (2N-Ng-2) non-linear μ: Expected value of x
Equations with (2N-Ng-2) state variables to be In this Equation, to form a standard distribu-
calculated, where Ng is the number of voltage- tion, expected value of “x” can be consider as
controlled buses in the system. Due to the load value of the node with maximum occurrence
non-linearity of these Equations, power-flow probability through studied time interval.
calculations usually employ iterative techniques
such as “Gauss-Seidel” and “Newton-Raphson”
procedures (Papaefthymiou, 2003).
275
Optimal Configuration and Reconfiguration of Electric Distribution Networks
276
Optimal Configuration and Reconfiguration of Electric Distribution Networks
ing this series expansion, the mean of z may be To this system of Equations, it can be added
approximated as:
m
1
∞ ∑p k ,i =
n
, for each variable k (15)
ηz = E [z ] = E [h(x )] = ∫ ... ∫ h(x )fx (x ) i =1
−∞
∞ ∞ n
1 ∂ jh
... ∫ fx (x ) h(ηx ) + ∑ ∑ . j (x i − ηi )j .dx Therefore, there are 2m Equations with 2m
= ∫ j =1 i =1 j ! ∂x i unknowns for each variable k, forming a nonlin-
−∞
(10) ear system.
Solving these Equations, the moment of order
Since, j for the variable z, mz,j, can be then approximated
by:
∞
for every
PROPOSED PROBABILISTIC
q, …,j if ∃ km ≠ kn m, n ∈ {1, ..., l } . PROCEDURE
277
Optimal Configuration and Reconfiguration of Electric Distribution Networks
where,
278
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Where x are the standard measure variables easily applied to any kind of continuous or
and y are the real variables. The new Gaussian discrete problems.
function now becomes, • GA use probabilistic transition operates
rather than deterministic rules.
1
1 y −µpdf 2
.( ) • GA does not have any limitation relate to
2 σpdf
G (y ) = .e (23) continuity, differentiability and…, gener-
2π
ally exist in conventional methods.
The resulting probability density function, GA’s Advantages and its Limitations
f(y), is the actual distribution of power flow as:
Following list gives some advantages of genetic
∞
1 algorithm generally find in available researches:
f (y ) =
σpdf
∑ c .H (y ).G(y )
j =0
j j (24)
279
Optimal Configuration and Reconfiguration of Electric Distribution Networks
over rates, selection method and…which lation (or solution). Before can use these strings
strongly affect on optimal solution. there should be a way to encode and decode the
• Not good to identify the local optimum. initial information and final resulted solutions
• Not effective terminator respectively.
• Have some troubles to find the exact global Binary coding is one of these methods widely
optimum. It is very often when the popula- used to encode and decode each gene in a chro-
tions have a lot of subjects. mosome. While there are many other techniques
• Requires large number of fitness evalu- such as using real and decimal numbers, binary
ations which increases the computation coding is the most expanded presentation of a
time. chromosome, because it is simple and traceable.
• Like other artificial intelligence tech- Note that in all these methods searching process
niques, the genetic algorithm cannot assure will apply on coded strings. Figure 7 represents a
constant computation times. Even more, simple binary string including coded information.
the difference between the shortest and the In this string each gene defines the color of each
longest optimization computation time is point in Figure 8 where 0 and 1 represent white
much larger than with conventional gradi- and black colors respectively. Moreover, first gene
ent methods. in each section of this string determines quarter
number for each individual.
Basic GA Concepts
Selection
To start GA, many individual solutions are ran-
domly generated to form an initial population. Selection is a stage in genetic algorithm in which
The population size depends on the nature of the individuals are chosen from a population for later
problem, but typically contains several hundreds or reproduction. This process uses a fitness func-
even thousands of possible solutions. This initial tion to evaluate each individual corresponding
population and also future reprodused individuals to desired objective. Providing fitness values for
form crospanding chromosomes in the shape of each indivitual and sort these individuals by their
diffrent strings. coresponding fitness valuse in next step, selection
process ends by selecting induvituals with most
Chromosome fitness and forms a new population for reproduc-
tion process.
A chromosome or genotype of the genome as a While there are many different types of selec-
concept of GA consists a string of information tion, “Roulette Wheel” is the most commen and
which represent characteristic of each individual. simplest method to fulfil the selection process.
By this definition, each gene in a chromosome Figure 9 depics a sample roulette wheel. In this
introduces only one specific individual (or status) wheel, each individul get its own portion by related
where each chromosome represents specific popu- fitness persentage. Selection point will show the
280
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Crossover
For each new solution to be produced by crossover
operation, a pair or even a group of parents are
selected for breeding. While there are many dif-
ferent kinds of crossover, the most common type
is single point crossover.In single point crossover
one crossover position k is selected randomly from
[1, 2, …, N-1], where N is number of variables
in each individul. In this step, variables exchange
selected individul after each rotation. Following
between individuals about this point to produce
this process, the individual with the most fitness
new offsprings. This process ultimately results in
will have the most probability to select and trans-
the next generation population of chromosomes
fer to the next generation. This type of selection
that is different from privious generation. Figure 10
method can get the form of rout instead of rotation
illustrates this process, where k is asumed to be 5.
as shown in Figure 9.
Note that for this kind of crossover one can
Again, in this form selection can be assign by
also use more points. See Figure 11 for multi point
choosing a random number (for instance: a value
cross over at points k=5, 8 and 15.
between 0.0 and 1.0 for each iteration in above
example)
Mutation
Mutation is another genetic operator that alters one
Reproduction
ore more gene values in a chromosome from its
initial state. The purpose of mutation is to maintain
Reproduction is a process to generate latter gen-
diversity in the genetic algorith. This operator
eration of solutions from those picked previously
allows the algorithm to avoid local optimum by
by selection process (crossover) or by changing
preventing the indivituals to be all exactly the
some genes in a string (mutation). Although there
same. While Figure 12 illastrate a sample “Flip-
are some other GA operators apply to generate
Bit” mutation, Uniform, Non-Uniform, Boundary
new populations but “crossover” and “mutation”
and Gaussian are some other known methods for
are most commen sulotions as well as almost
mutation process.
insepraTable part of a genetic algorith. Type and
281
Optimal Configuration and Reconfiguration of Electric Distribution Networks
282
Optimal Configuration and Reconfiguration of Electric Distribution Networks
This is while, there are many other condicutions Considering these different triangle shapes for
suported by common softwares such as MATLAB various configurations under study, basic proposed
including “Function tolerance”, “Nonlinear con- fitness function can be form as follow:
straint tolerance”, “Stall time limit” and “Stall
generations”.
1
Ffitness = (25)
Basic GA’s Flowchart 2
S + w . 1 +
trg 1
∑ (− 1)i
.a i
A flowchart describing main steps in basic genetic i =1
algorithm is illustrated in Figure 13 as follow:
283
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Figure 14. Comparison between power loss PDFs for two different configurations of network
284
Optimal Configuration and Reconfiguration of Electric Distribution Networks
In this Equation, Strg represents the space of These profiles are created randomly, with loads
a1 .a 2 ranging from 90% to 110% of the original loads.
approximated triangles (equal to ). Mean-
2 As the first scenario, only the lose reduction
while, W1 is considered as penalty factor for a1 is considered in the objective function. Table 1
and a2 where for the configuration with high prob- designates the numerical results of this experi-
ability and low power loss this factor makes a ment. With specified load pattern, resulted power
high fitness and vise versa. loss probability density function is depicted in
Several experiments and studies chose the Figure 16.
value of “10” for this factor which gives a normal- In the second scenario, two objectives of
ized fitness function with more accurate results. switching reduction and reduction in loss are ap-
By this fitness function, genetic algorithm will plied to optimization problem simultaneously.
search for optimum triangles which just represent Table 2 designates the numerical results of this
the least power lose values with the most total experiment. With specified load pattern, resulted
probability of occurrence simultaneously. This is power loss probability density function is de-
while, in some cases regarding to the condition of picted in Figure 17. Although the Table 2 ex-
the network and decision of the operator, in order presses less reduction in loss for this scenario than
to consider the cost of switching, the Equation (1) previous one, but implementing the second sce-
can be combined easily with above mentioned nario gives more benefits by reduce the number
Equation to form a single fitness function. of switching. This benefit is calculated as 405 $
includes income from total loss reduction and cost
of switching.
COMPUTATION RESULTS To calculate these values following assump-
AND DISCUSSION tions are used in this study:
To examine the efficiency and improvement, the • The cost of electricity is $6.5625 per kWh
suggested method is developed in MATLAB, on • To simplify the presentation, the network
a Pentium-4 PC (1.86GHz & 2GB RAM), and under study includes only one switch type
is performed on a sample three-feeder network. • Each switching cost is set to $203 which is
Original configuration of this network is illustrated one-twentieth of a new switch installation
in Figure 4, while the system parameters are acces- cost (Yin & Lu, 2009)
sible by Civanlar et al. (1988). Since it was pro-
posed for a study with fixed demands, to perform In order to compare the results of suggested
suggested method new different load profiles are method with other references, the proposed
added to each load sections which are remained method is applied to two other test systems. The
with no change all over the examination process. first one is a small system (Baran & Wu, 1989)
widely used as a first reference in case studies.
Minimum Power Maximum Power Mean Power Loss Total Number of Total Loss
Loss Loss (p.u) Switching Reduction
(p.u) (p.u)
Original 0.07 0.95 0.58 _ 0%
After Reconfiguration 0.01 0.73 0.35 6 35.7%
285
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Figure 16. Power loss probability density function for first scenario (before & after reconfiguration)
Minimum Power Maximum Power Mean Power Loss Total Number of Total Loss
Loss Loss (p.u) Switching Reduction
(p.u) (p.u)
Original 0.07 0.95 0.58 - 0%
After Reconfiguration 0.03 0.81 0.41 2 25.3%
Figure 17. Power loss probability density function for second scenario (before & after reconfiguration)
286
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Table 3 represents the loss reductions by proposed by proposed procedure in comparison to Lopez
method in comparison to other available method (2004) famous method with three different cases.
with several cases for this test network. This is
while, total number of switching and total income
are other results available in this Table to carry CONCLUSION
out a better comparison. In this examination the
income values are evaluated based on the energy In this chapter after review the background of
cost of $0.50 per kWh and the switching cost of “reconfiguration” thorough major presented meth-
$8.00 for each pair switching operation as are ods, electric distribution network, basic concepts
considered by Fang, Cai & Guo (2005). and its configuration are introduced. Furthermore
These results depict the relation between loss one becomes familiar with reconfiguration pro-
reduction and number of switching. It is clear that cedure and its main concepts. Two important
by increase the number of switching total power points proposed which are original and main
loss decrease and vise versa. Unlike the compared concerns in recent years. The proposed chapter
method, proposed process found optimal values fully discussed about these points and proposed
for number of switching and the power loss by new heuristic solutions to overcome mentioned
apply a balance between them. Again these con- issues in optimal reconfiguration process. This
siderations result in a huge amount of benefit is while, a precise tutorial followed to introduce
approximately 260,000 $ annually for the network basic Genetic Algorithm its parameters and main
under study. concepts including advantages and limitations.
Final comparisons relate to a real large sys- Finally, In order to examine the suggested
tem. For this network, Table 4 represents the loss solution several experiments have been done.
reductions and total number of switching resulted
Table 3. Comparison between the result of proposed method and other method
Table 4. Comparison between the result of proposed method and other method (Real network)
287
Optimal Configuration and Reconfiguration of Electric Distribution Networks
288
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Liu, C. C., Lee, S. J., & Venkata, S. S. (1989). Loss Papaefthymiou, G. (2003). Research on the
minimization for distribution feeders: Optimality uncertainty in electrical energy generation in
and algorithms. IEEE Transactions on Power distribution systems with high penetration from
Delivery, 4, 1281–1289. doi:10.1109/61.25615 distributed and renewable energy sources’genera-
tion. Netherlands: Delft University of Technology
Lopez, E., Opazo, H., Garcia, L., & Bastard, P.
and Eindhoven University of Technology.
(2004). Minimal loss reconfiguration based on
dynamic programming approach: Application Rubinstein, R. Y., & Kroese, D. P. (2008). Simu-
to real systems. IEEE Transactions on Power lation and the Monte Carlo method. Hoboken,
Systems, 19, 693–704. NJ: Wiley.
Mendoza, J. E., Villaleiva, L. A., Castro, N. A., & Rugthaicharoencheep, N., & Sirisumrannukul, S.
Lopez, E. A. (2009). Multi-objective evolutionary (2009). Optimal feeder reconfiguration with dis-
algorithms for decision making in reconfigura- tributed generators in distribution system by fuzzy
tion problems applied to the electric distribution multi objective and tabu search. International
networks. Studies in Informatics and Control, Conference on Sustainable Power Generation
18, 325–336. and Supply, 1, 1-7.
Merlin, A., & Back, H. (1975). Search for a Sarfi, R., Salama, M., & Chikhani, Y. (1996).
minimal-loss spanning tree configuration in Distribution system reconfiguration for loss reduc-
an urban power distribution system. 5th Power tion: an algorithm based on network partitioning
System Computation Conference (PSCC): Vol. theory. IEEE Transactions on Power Systems, 11,
1. (pp. 1-18). 504–510. doi:10.1109/59.486140
Milani, A. E., & Haghifam, M. R. (2010). A heuris- Shirmohammadi, D., & Hong, H. W. (1989).
tic approach for multi objective distribution feeder Reconfiguration for electric distribution net-
reconfiguration: Using fuzzy sets in normalization works for resistive line loss reduction. IEEE
of objective functions. [IJAEC]. International Transactions on Power Delivery, 4, 1492–1498.
Journal of Applied Evolutionary Computation, doi:10.1109/61.25637
1(1), 60–73. doi:10.4018/jaec.2010040103
Taylor, T., & Lubkeman, D. (1990). Imple-
Nara, K., Shiose, A., Kitagawa, M., & Ishihara, mentation of heuristic search strategies for
T. (1992). Implementation of a genetic algorithm distribution feeder reconfiguration. IEEE
for distribution systems loss minimum re-config- Transactions on Power Delivery, 5, 239–246.
uration. IEEE Transactions on Power Systems, 7, doi:10.1109/61.107279
1044–1051. doi:10.1109/59.207317
Usaola, J. (2008). Probabilistic load flow with
Opazo Mora, H., Garcia-Santander, L., & Lopez wind production uncertainty using cumulants
Parra, E. (2008). Minimal loss reconfiguration and cornish fisher expansion. Power Systems
considering random load: Applications to real Computation Conference (pp. 1-8).
networks. Ingeniare Journal of Engineering,
Wagner, T., Chikhani, A., & Hackman, R. (1991).
1(16), 264–272.
Feeder reconfiguration for loss reduction: An
application of distribution automation. IEEE
Transactions on Power Delivery, 6, 1922–1993.
doi:10.1109/61.97741
289
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Yin, S. A., & Lu, C. N. (2009). Distribution Golkar, M. A. (2003). A new probabilistic load-
feeder scheduling considering variable load flow method for radial distribution networks.
profile and outage costs. IEEE Transactions on European Transactions on Electrical Power, 3,
Power Systems, 24, 652–660. doi:10.1109/TP- 167–172. doi:10.1002/etep.4450130305
WRS.2009.2016300
Hadian, A., Haghifam, M. R., Zohrevand, J., &
Akhavan-Rezai, E. (2009). Probabilistic approach
for renewable DG placement in distribution sys-
ADDITIONAL READING tems with uncertain and time varying loads. IEEE
Power & Energy Society General Meeting (PES):
Ah King, R. T. F., Radha, B., & Rughooputh, H. Vol. 1. (pp. 1-8).
C. S. (2003). A real-parameter genetic algorithm
for optimal network reconfiguration. International Hines, W. W., Montgomery, D. C., Goldsman, D.
Conference on Industrial Technology: Vol. 2. (pp. M., & Borror, C. M. (2003). Probability and Statis-
237-288). tics in Engineering. New York: Wiley Publishing.
Allan, R. N., Leite da Silva, P. H., & Burchett, Huang, Y. C. (2002). Enhanced genetic algorithm-
R. C. (1981). Evaluation methods and accuracy based fuzzy multi-objective approach to distribu-
in probabilistic load flow solutions. IEEE Trans- tion network reconfiguration. IEEE Proceedings-
actions on Power Systems, 100, 2539–2546. Generation. Transmission and Distribution, 149,
doi:10.1109/TPAS.1981.316721 615–620. doi:10.1049/ip-gtd:20020512
Arrowsmith, D., Bernardo, M., & Sorrentino, F. Huo, L., Yin, J., Yu, Y., & Zhang, L. (2008). Dis-
(2005). Effects of variations of load distribution tribution network reconfiguration based on load
on network performance. IEEE International forecasting. International Conference on Intel-
Symposium on Circuits and Systems (ISCAS): ligent Computation Technology and Automation
Vol. 4. (pp. 3773-3776). (ICICTA): Vol. 1. (pp. 1039-1043).
Bueno, E. A., Lyra, C., & Cavellucci, C. (2004). Jianzhong, W., & Yixin, Y. (2003). Global optimi-
Distribution network reconfiguration for loss zation algorithm to time varying reconfiguration
reduction with variable demands. IEEE Transmis- for operation cost. Chinese Society of Electrical
sion and Distribution Conference and Exposition: Engineering, 11, 13–17.
Vol. 1. (pp. 384-389). Meliopoulos, A. P. S., Cokkinides, G. J., & Chao,
Chen, P., Chen, Z., & Bak-Jensen, B. (2008). X. Y. (1990). A new probabilistic power flow
Probabilistic load flow: A review. Third Interna- analysis method. IEEE Transactions on Power
tional Conference on Electric Utility Deregula- Systems, 5, 182–190. doi:10.1109/59.49104
tion and Restructuring and Power Technologies Morales, J. M., & Perez-Ruiz, J. (2007). Point
(DRPT): Vol. 1. (pp. 1586-1591). estimate schemes to solve the probabilistic power
Dickert, J., Hable, M., & Schegner, P. (2009). flow. IEEE Transactions on Power Systems, 22,
Energy loss estimation in distribution networks 1594–1601. doi:10.1109/TPWRS.2007.907515
for planning purposes. IEEE PowerTech, 1, 1–6.
doi:10.1109/PTC.2009.5281997
290
Optimal Configuration and Reconfiguration of Electric Distribution Networks
Neimane, V. (2001). Distribution Network Plan- Tanabe, T., Funabashi, T., Nara, K., Mishima, Y.,
ning Based on Statistical Load Modeling Applying & Yokoyama, R. (2008). A loss minimum recon-
Genetic Algorithms and Monte-Carlo Simula- figuration algorithm of distribution systems under
tions. IEEE Porto Power Tech Conference: Vol. three-phase unbalanced condition. IEEE Power
3. (pp. 5-9). and Energy Society General Meeting - Conver-
sion and Delivery of Electrical Energy in the 21st
Papaefthymiou, G., Schavemaker, P. H., van der
Century: Vol. 1. (pp. 1-4).
Sluis, L., Kling, W. L., Kurowicka, D., & Cooke,
R. M. (2005). Integration of stochastic generation Verbic, G., & Canizares, C. A. (2006). Probabilistic
in power systems. 15th Power Systems Computa- optimal power flow in electricity markets based on
tion Conference: Vol. 1. (pp. 773-781). a two-point estimate method. IEEE Transactions
on Power Systems, 21, 1883–1893. doi:10.1109/
Queiroz, L. M. O., & Lyra, C. (2009). Adaptive
TPWRS.2006.881146
hybrid genetic algorithm for technical loss re-
duction in distribution networks under variable Verbic, G., Schellenberg, A., Rosehart, W., &
demands. IEEE Transactions on Power Systems, Canizares, C. A. (2006). Probabilistic optimal
24, 445–453. doi:10.1109/TPWRS.2008.2009488 power flow applications to electricity markets.
International Conference on Probabilistic Meth-
Savier, J. S., & Das, D. (2007). Impact of Net-
ods Applied to Power Systems: Vol. 1. (pp. 1-6).
work Reconfiguration on Loss Allocation of
Radial Distribution Systems. IEEE Transactions Wanp, S., & Wang, C. (2001). Disbibution network
on Power Delivery, 22, 2473–2480. doi:10.1109/ reconfiguration based on interval algorithm to
TPWRD.2007.905370 maximize confidence of energy loss reduction.
Journal of Automation of Electric Power Systems,
Shouxiang, W., & Chengshan, W. (2007). Mod-
1(25), 27–31.
ern distribution system analysis. Beijing: China
Higher Education Press. Yu, H., Chung, C. Y., Wong, K. P., Lee, H. W.,
& Zhang, J. H. (2009). Probabilistic load flow
Su, C. L. (2005). Probabilistic load-flow com-
evaluation with hybrid latin hypercube sampling
putation using point estimate method. IEEE
and cholesky decomposition. IEEE Transactions
Transactions on Power Systems, 20, 1843–1851.
on Power Systems, 24, 661–667. doi:10.1109/
doi:10.1109/TPWRS.2005.857921
TPWRS.2009.2016589
Su, C. L. (2005). Cumulant-based probabilistic
Yu, Y., Qiu, W., & Liu, R. (2001). Distribution
optimal power flow (P-OPF) with Gaussian
system reconfiguration based on heuristic al-
and gamma distributions. IEEE Transactions
gorithm and genetic algorithm. Power System
on Power Systems, 20, 773–781. doi:10.1109/
Technology, 11, 19–22.
TPWRS.2005.846184
Zhu, J. (2009). Optimal reconfiguration of electri-
Su, C. L. (2005). Distribution probabilistic load
cal distribution network using the refined genetic
flow solution considering network reconfiguration
algorithm. In M.E. El-Hawary, Optimization of
and voltage control devices. 15th Power Systems
Power System Operation. Hoboken, New Jersey:
Computation Conference: Vol. 1. (pp. 773-781).
Wiley Publishing.
291
Optimal Configuration and Reconfiguration of Electric Distribution Networks
KEY TERMS AND DEFINITIONS calculate its parameters such as voltage angles
and magnitudes.
Electric Distribution Network: An electric Probability Density Function (PDF): A
distribution network is a system of cables and function that describes the relative chance for a
equipments which deliver electric power to the random variable to occur at a given point in the
end users. observation space.
Configuration: The process of design and
implement of an electric distribution system.
Reconfiguration: Switching and changing the ENDNOTES
status of network for better operation.
Optimization Problem: An optimization 1
1975 to present
problem is the problem of finding the best solu- 2
SSOM
tion from all feasible solutions. 3
N. C. switches
Switch Types: Maneuver switches in an elec- 4
N. O. switches
tric distribution network which can be manual 5
Monte Carlo Method (MCM), 1940s. Gen-
type or operated automatically. erally MCMs are a class of computational
Load Flow: An important tool involving algorithms that rely on repeated random
numerical analysis applied to power systems to sampling - wikipedia.org.
292
293
Chapter 10
A Descriptive Approach
for Power System Stability
and Security Assessment
A. G. Tikdari
University of Kurdistan, Iran
H. Bevrani
University of Kurdistan, Iran
G. Ledwich
Queensland University of Technology, Australia
ABSTRACT
Power system dynamic analysis and security assessment are becoming more significant today due to
increases in size and complexity from restructuring, emerging new uncertainties, integration of renewable
energy sources, distributed generation, and micro grids. Precise modeling of all contributed elements/
devices, understanding interactions in detail, and observing hidden dynamics using existing analysis
tools/theorems are difficult, and even impossible.
In this chapter, the power system is considered as a continuum and the propagated electromechanical
waves initiated by faults and other random events are studied to provide a new scheme for stability
investigation of a large dimensional system. For this purpose, the measured electrical indices (such as
rotor angle and bus voltage) following a fault in different points among the network are used, and the
behavior of the propagated waves through the lines, nodes, and buses is analyzed. The impact of weak
transmission links on a progressive electromechanical wave using energy function concept is addressed.
DOI: 10.4018/978-1-61350-138-2.ch010
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Descriptive Approach for Power System Stability and Security Assessment
Finally, a new power system emergency control framework based on descriptive study of electrical mea-
surements and electromechanical wave propagation in large electric power systems is introduced. Since,
fast and accurate detection of instability is essential in initiating certain emergency control measures,
the proposed methodology could be also useful to detect the contingency condition and performing the
well-known islanding and load shedding techniques. The chapter is supplemented by some illustrative
nonlinear simulations on large scale test systems.
294
A Descriptive Approach for Power System Stability and Security Assessment
295
A Descriptive Approach for Power System Stability and Security Assessment
distributed model contains a lot of elements that Rotor Angle Wave Propagation
each element includes one inductor and one ca-
pacitor. When a voltage source is applied at one As discussed, the angle wave propagation which
end of the line, the first capacitor is immediately is called electromechanical wave propagation
charged; but because of first inductor, the second starts when a generator rotor deviates from its
capacitor charges when the inductor is charged, base frame. Then, it propagates throughout the
and it makes a delay. This delay also exists for power system. The velocity of angle wave propa-
the next elements. gation is slower than voltage wave propagation
One of most important parameters in this issue (Thorp, 1998 and Phadke, 2008). As has argued
is the characteristic impedance which is calcu- in (Phadke, 2008), the nature of this phenomenon
lated by Equation 1: is not completely clear. It seems to be the result
of local inertias. As the waves may lead to loss of
L synchronism for under stressed power systems,
Z = (1) studying of them is one of the interested issues
C
in power systems stability analysis and security
assessment.
where, L and C are defined as line inductance and
Formulation of wave motion as nonlinear
capacitance values per length of line, respectively.
partial differential equations in a two-dimensional
When a voltage wave passes through a point in
surface is introduced in (Thorp, 1998). This for-
which the characteristic impedance is changed, the
mulation is in form of wave equations. The swing
magnitude of reflected wave and the magnitude
equation (Equation 3) of a generating unit can be
of the wave that lets trough are dependent on the
expressed as follows:
value of characteristic impedances of two lines
(for example when an overhead transmission line
2H .. .
is connected to an underground cable). δ + ωD δ = Pa = Pm − Pe (3)
The wave propagation velocity is another ω
important parameter that can be calculated with
Equation 2. where, H, ω, D, 𝛿, Pe, and Pm are inertia constant,
angular speed, damping factor, rotor angle, elec-
trical output power, and mechanical input power,
v= 1 (2)
LC respectively.
The above swing equation is introduced as a
second-order hyperbolic wave equation in (Thorp,
1998). Using of these equations lets us to intro-
duce a power system as a continuum system. The
296
A Descriptive Approach for Power System Stability and Security Assessment
methodology of extracting the continuum model energies of the lines belonging to the cutest; and
for a two-dimensional (2-D) power system is also, the kinetic energy is a function of voltage
well-established in (Parashar, 2004). For a mesh angle gradient of the cutest.
network power system, which the generators and When a power system is becoming unstable at
loads are located at distributed discrete points, the location of a cutest, it means that the angle of
introducing a 2-D continuum model needs to in- a portion of power system that connected to the
troduce the system parameters in form of smooth system (with the critical cutest) moves in opposite
functions. To distribute the parameters, a Gaussian direction of angle motion of the other parts of
filter which is the most common smoothing tool system. In this situation, the islanding control is
can be suggested (Parashar, 2004). the most effective control action. Therefore, the
unstable part of system is separated by tripping
Islanding Control the lines in the critical cutest. The detection of
critical cutest/weak link is the first step, and the
In fact, the angle instability is the loss of synchro- islanding control must be executed. Following the
nism of the system synchronous generators. Angle islanding process, we are in face of some islands
instability is usually started when the synchronism with excess load or generation. Load-generation
between two generators which are located at two imbalance in an isolated island leads the island
sides of a line is loosed. In other words, the angles into another form of instability, but with a slower
of two generators are separated, may because of dynamic than angle instability. Therefore, the
overloading a link which connects two generators, island formation may not be the final stabilizing
directly or indirectly. This link may be a weak link step. However, it can be considered as a way to
or a link which encountered a suddenly overload arrest the fast angle instability. Following the
due to some connections in the network. islanding, the other emergency control actions
There are many technical reports representing such as load shedding and generation tripping are
various methodologies for detection/prediction the usually needed (Tikdari, 2009 and Bevrani 2010).
angle instability situation. For example, using of Slow coherency is a suitable and usual meth-
energy function concept is suggested in (Padiyar, odology for islanding formation (You, 2004). Fol-
2006). Based on this concept, for the purpose lowing a fault, there are some groups of generators
of stability, a system should be able to convert based on their angle variations. The angle changes
whole of kinetic energy achieved throughout a in a group are similar, but the different groups of
disturbance into potential energy. On the other generators behave in opposite directions. This is
words, when in a stable swing, the potential energy the concept of slow coherency theory. One may
reaches its maximum value, the kinetic energy introduce every coherent group of generators as
should be zero. an island. The lines that connect the islands may
In complex power systems, there are usu- also provide the power system weak links (You,
ally more than one inter-islands connection 2004, Yang, 2006, Wang, 2004).
line. Therefore, to validate the system stability, Slow coherency is known as a good example
maintaining only one line is not adequate. Based for application of singular perturbation theory in
on the given idea in (Padiyar, 2006 and Wang, power systems (You, 2004, Yang, 2006, Wang,
2004), the change of power in all lines belong- 2004, Ourari, 2003 and Sowa, 2004). Using this
ing to a cutest called critical cutest determines method, the groups of generators with coherent
the stability or instability situation. As argued angle behaviors can be determined. In transient
in (Padiyar, 2006), the potential energy, under stability issues, the slow coherency can be also
certain assumption, can be considered as sum of used to find out the equivalent dynamic model of
297
A Descriptive Approach for Power System Stability and Security Assessment
a power system (Ourari, 2003 and Sowa, 2004). The methodology will be described in the next
In the slow coherency, it is assumed that state sections. As it can be seen, the main aim is looking
variables of an n-degree performance can be for an approach to help the system operators to
divided into two categories: r slowest states and immediate identify the proper emergency action.
n-r fast states. The r slowest states represent the Following dangerous events, the system operators
groups of generators with slowest coherency. In have some choices as emergency control actions.
slow coherency-based islanding, two assump- Sometimes a load shedding scheme or a genera-
tions can be considered to simplify the process tion rescheduling action can maintain the system
of finding coherent groups of generators: i) the stability. However, for very large disturbances, it
groups of coherent generators are independent to may be needed to separate the system into two or
size of disturbance; and ii) the groups of coherent more islands and control them separately.
generators are independent to the modeling ac-
curacy of the generator unit. These assumptions Solutions and Recommendations
are important to perform the continuum model
of power system, which is main interest in the Based on power-voltage curves, there is a
present chapter. maximum value for transmission line loadability
(Kundur, 1994 and Miller, 1982). A transmis-
sion line may encounter overloading following
PROBLEM ILLUSTRATION a contingency. But, it is important to know that
AND MODELING what happens when an overloading is appeared.
The relation between angle and power across
In this section, the problem that we are attempting a line is shown in Equation 4 (Kundur, 1994 and
to solve is described. Suppose a high stressed large Miller, 1982).
scale power system which its elements are work-
ing near their stability margins. When a system Es E r
is explored near the stability limits, disturbances P= Sin δ (4)
Z 0Sin θ
may easily force the system to a cascading failure
and even blackout. The goal is to widely monitor
system in order to effective protection against where, 𝜃 is the length of transmission line in
large disturbances. For this purpose, an online Radian; Es and Er are the voltage magnitudes
stability/security assessment program is needed. at two ends of the line; Z0 is the characteristic
On the other hand, following a contingency, impedance; P is the active power passes through
the voltage/angle deviations propagate through the the line; and 𝛿 is the angle difference between Es
power system. These traveling waves may pass a
nd Er. The 𝛿 is also known as transmission and
through the high stressed elements and trigger an load angle (Miller, 1982).
instability event. By using of wave propagation Based on Equation 4, if a line active power
phenomenon, a descriptive approach for power increases, the angle across that link also increased.
system stability assessment can be proposed. In In this situation, if two synchronous machines are
a power system, the operators and engineers will connected at two ends of the line, increasing of
be able to use this descriptive tool to track the 𝛿 which is the difference between the positions
system dynamic behaviors, to validate the system of rotors of the machine may lead to loss of their
performance against likelihood events, and to synchronism.
predict the next stable point. Therefore, there is also a maximum value for
load angle of a line. Equation 4 shows that this
298
A Descriptive Approach for Power System Stability and Security Assessment
Box 1.
.. .
δk + D δk = P mk −[2 − cos(δk − δk −1 ) − cos(δk − δk +1 )] − b[sin(δk − δk −1 ) + sin(δk − δk +1 )];
for k = 2, 3,..., N − 1
.. . (5)
δ1 + D δ1 = P m1−[2 − cos(δ1 − δN ) − cos(δ1 − δ2 )] − b[sin(δ1 − δN ) + sin(δ1 − δ2 )]
.. .
δN + D δN = P Nm −[2 − cos(δN − δN −1 ) − cos(δN − δ1 )] − b[sin(δN − δN −1 ) + sin(δN − δ1 )]
maximum value is 90°. However, based on experi- The value of Pkm can be achieved from steady
ences it is not safe to let 𝛿 to increase more than states values of angles (𝛿k = 𝛿k =0). The steady
30° if the transmission line is not compensated state values of angles for an N-bus ring system
(Miller, 1982). can be calculated as (Equation 6):
In transient stability studies, there are some
approaches for the system stability assessment, 2πk
δkeq = (6)
following a contingency. For example Equal area N
criterion is one of them which is only used for one
machine connected to an infinite bus or for two- Now, consider the 64-bus ring power system
machines system (Kundur, 1994). (Padiyar, 2006) given in (Thorp, 1998). A Gaussian disturbance
used energy function criterion for online stability around line 15-16 can be implemented as follows
assessment of a large power system. Major of (Equation 7).
proposed criteria say that a system can stay stable
if the generators can release the complete energy
1 2
value that they obtained following a contingency. δk = δkeq + e −0.1(k −15.5) (7)
2
Because of oscillations and uncertainty in a real
power system, the proposed algorithms for online
stability assessment that uses instance values The simulation results are illustrated in Figure
of power, angle and the other indices may need 2. This is a regeneration of the given example in
many considerations. While, in these situations it (Thorp, 1998). In Figure 2, the wave propagation
is needed to know the behavior of the indices that is illustrated versus time. However in Figure 3,
propagates through the power system like a wave. the angle of wave is plotted versus bus number for
different time slots. Figure 3b shows the normal-
System Modeling ized version of Figure 3a. As can be seen, when
a deviation appears in rotor angle of a generator,
A Ring System it propagates and in the traveling path, it may
encounter a weak link and may lead to a cascad-
If the number of elements in a power system is large ing failure.
with a set of distributed generators parameters, the
discrete model simulation results are also close 2-Dimensional System
to the continuum model (Thorp, 1998). Equation
5 and Equation 6 can be used for modeling of an Now, consider a meshed power system with a
N-bus ring power system: (see Box 1) configuration shown in Figure 4, in the Cartesian
299
A Descriptive Approach for Power System Stability and Security Assessment
Figure 3. Wave propagation; a) angle versus bus number for different time slots, and b) normalized plot
300
A Descriptive Approach for Power System Stability and Security Assessment
characteristic. Each point represents a bus and The system situation following the disturbance is
each connection represents a transmission line. shown in a few time slots.
For simplicity, assume that each bus consist of
a generator or a load, or nothing. To obtain the
necessary equations, assume that a generator with a PROPOSED METHODOLOGY
mechanical power of PAm is connected to the point
A (Figure 4). Considering the connections, PAe is This section describes the process of using island-
the sum of electrical powers transferred from bus ing formation, performing a continuum model for
A to its neighbor buses 1, 2, 3, and 4. Therefore, system stability assessment, and predicting suit-
PAe can be calculated as Equation 8: able emergency actions following a contingency.
The overall view of the algorithm is demonstrated
4
in Figure 6.
PeA = ∑ PeAk + PLA (8)
k =1
Angle instability is a fast instability phenom-
enon. Therefore, predicting its situation and
performing suitable actions are very important.
where (Equation 9),
Having a continuum model of a power system
can be helpful for predicting the trajectory of the
V AV k
PeAk = sin δ Ak , k = 1, 2, 3, 4 (9) disturbances, by using of disturbance conditions
Ak
x as initial states of the continuum model. Here, the
power system continuum model is used to provide
Now, the swing equations can be easily written descriptive tool for stability analysis in emer-
for each point (Bevrani and Tikdari, 2010). Us- gency conditions.
ing above descriptions, an example is presented As introduced in the previous section, the
in Figure 5 which illustrates the propagation of slow coherency theory can be used to identify
angle wave through a 2-D power system. At t = system islands. Determining the islands leads to
0, a disturbance occurs on the middle of the net- identify the weak links or critical cutset. Having
work then, it propagates throughout the system. knowledge on system weak links/islands helps one
301
A Descriptive Approach for Power System Stability and Security Assessment
302
A Descriptive Approach for Power System Stability and Security Assessment
islanding is not needed. However, for the higher time states of whole the network. In prediction
value of transmission angles, the other emergency mode, the online data are used as initial values
action will not be able to restore the system sta- of a continuum model, and the system states at a
bility. For more clarification, assume the angle certain time value later will be predicted. In the
across a link is increasing and excess 90°. After test mode, the operator or engineer can validate
that, decreasing of active power by, for example, a certain contingency based on real states of the
load shedding could not restore the system because system. In this mode the system real-data are
the angle may track the power at low side of P – 𝛿 used as initial values of a continuum model and
curve; so, 𝛿 increases, thus instability occurs. If the certain test is used as a deviation from initial
these circumstances to be predicted, the system state; then the post contingency condition will be
will undoubtedly need to be separated. The critical shown in HMI for a certain time interval.
angle values (thresholds) should be determined
based on desire level of security.
Operators and engineers can validate the power SIMULATION RESULTS
system stability at control rooms by observing the
wave propagation at human machine interface A Ring System and Islanding
(HMI). Three modes may be defined for this tool:
real-time, prediction, and test modes. To illustrate the concept of a coherent group of
In real-time mode, the real-time data which generators following a contingency, a 200-bus
are gathered from the network by the PMUs are ring system is simulated. For simplicity, it is as-
shown as a surface. The operator can see the real- sumed that there is only one generator or one load
303
A Descriptive Approach for Power System Stability and Security Assessment
at each bus. All generators are similar with equal picted in Figure 7 (config-2). Post-contingency
amount of power, and all loads are also equal. The wave propagation is illustrated in Figure 11. In
number of generators (NG) is equal to the number this case, the center of disturbance is bus 60;
of loads (NL), so NG = NL = 100. The system data however, it can be seen that the system is sepa-
is determined as follows (Equation 10): rated at line 75-76, which is a weak link. Actu-
ally, when a disturbance reaches a weak link
Pm = 0.3, PL = 0.3, H = 1, D = 0.2 (10) through its propagation trajectory, may lead to an
unstable operating point.
Figure 11b clearly shows the behavior of wave
The system is examined under two different propagation when it reaches a weak links. For
configurations. In the first configuration (con- plotting this figure, the angle variations versus
fig-1), all generators and loads are distributed bus number for each time slot is calculated. As it
throughout the power system by a uniform random can be seen, angle wave is reflected when it
function. In the second configuration (config-2), reaches to a transmission line without enough
the generators and loads are distributed in a three stability margin, and the system is separated ex-
areas ring system, as shown in Figure 7. In this actly at the weak point.
case, all line impedances are assumed to be fixed
at 0.1 p.u., except three lines that their impedances Application to 24-Bus Test System
are Z75-76 = 0.2. Z130-131 = 0.15, Z200-1 = 0.3. For both
configurations, a large disturbance (i.e. tripping Here, the 24-bus reliability test system (RTS)
line 200-1) is applied. The angle deviations are is used to investigate the effectiveness of the
illustrated for config-1 and config-2 in Figure 8 proposed strategy. Single line diagram of RTS is
(8.a and 8.b, respectively). The ring system is illustrated in Figure 12. The RTS with its full data
opened due to occurred disturbance. is introduced in (IEEE RTS Task Force 1979 and
For both cases, the angle deviation behavior 1999); and the generators data are selected the
across link 75-76 is illustrated in Figure 9. As same as given typical data in (Anderson, 2003).
shown, for the unstable case (config-2), the angle Here, the test system is divided into three areas.
across link 75-76 (i.e. 𝛿75-76 = 𝛿75 – 𝛿76), is con- While most of the generation is located in area
tinually growing and finally this situation leads 1, most of the load is located in area 3. In area 2,
toward separation and instability. However, for load and generation are approximately the same.
the stable case (config-1); although the angle Area 1 delivers its over generation into area 2
across link deviates, the system is remained in a and area 3 through three tie-lines: line 16-19, line
limited boundary and moves to a constant value. 16-14, and line 24-3.
The kinetic energy, potential energy, and total As a serious fault, the connections between
(kinetic plus potential) energy across link 75-76 area 1 and area 2 are loosed. Now, the angle in-
are also depicted in Figure 10. As already men- stability on link 24-3 can be considered as a good
tioned, the system to be stable if it can be able to example to examine the proposed methodology.
convert all amount of its kinetic energy achieved Assume lines 16-19 and 16-14 to be tripped at t
during a contingency into potential energy. This = 2 seconds. Following this large disturbance,
simulation (Figure 10) also shows that following line 3-24 will be encountered with a high over-
mentioned fault, config-2 is going to an unstable loading problem. This over-loading is larger than
condition, while config-1 holds its stability. its angle stability limit. Therefore, as shown in
As another example, assume a Gaussian dis- Figure 13, the angles of generators G1 and G15
tribution that affects the angles of a system de- located at two sides of the line, are separated; and
304
A Descriptive Approach for Power System Stability and Security Assessment
305
A Descriptive Approach for Power System Stability and Security Assessment
Figure 10. Energy across link 75: a) kinetic energy for config-1; b) kinetic energy for config-2; c) po-
tential and total energy for config-1; d) potential and total energy for config-2
the angle instability phenomena is immediately separated as well as separation of area 2 and area
occurred. To save system in this dangerous condi- 1. It demonstrates that the islanding between area
tion, one solution is the use of islanding control 1 and area 3 at link 3-24 is a good idea to protect
(You, 2003). the system stability. Figure 15 illustrates how an
The angle deviations reduced by PMUs from islanding improves the voltage behavior that has
the system buses are illustrated in Figure 14. As been suddenly depressed following the mentioned
it is shown, following the contingency the con- event.
nection between area 1 and area 3 (link 3-24) is
306
A Descriptive Approach for Power System Stability and Security Assessment
Figure 11. Wave propagation following a Gaussian disturbance which its center is bus 60 (the system
is separated at link 75-76): a) 2D plot, and b) 3D plot
307
A Descriptive Approach for Power System Stability and Security Assessment
308
A Descriptive Approach for Power System Stability and Security Assessment
Figure 15. Voltage response following disturbance (at 2 sec) and islanding plan (at 2.4 sec)
By using the error value, the continuum model can produce special patterns and may be used in
can be trained and improved. stability assessment problems, more effectively.
There are many researches that offer optimal is- Moreover, some important research needs in
landing. As another future work, these approaches future are the updating of existing emergency
may be included into the proposed methodology frequency control schemes for N-1 contingency,
to prepare a more effective approach. economic assessment/analysis the frequency regu-
In order to use wave propagation studies as lation prices (considering various control strate-
an automatic tool instead of descriptive tool by gies, penetration level, and installation location of
the power system operators, some additional renewable energy sources units), further study on
algorithms may needed. Pattern recognition frequency/voltage stability using dynamic demand
methods may be used in these situations. Some control and ratios of renewable energy sources
features can be extracted from the angle surface technologies, and quantification of reserve margin
variation, following different contingencies. They due to increasing renewable energy penetration.
309
A Descriptive Approach for Power System Stability and Security Assessment
CONCLUSION REFERENCES
In a power grid, when a propagation energy wave Anderson, P. M., & Fouad, A. A. (2003). Power
caused by a disturbance hits a weak link, a reflec- system control and stability. USA: IEEE Press.
tion is appeared and a part of energy is transferred Bank, J. N., Gardner, R. M., Tsai, S. J. S., Kook,
across the link. Based on this basic fact, the pres- K. S., & Liu, Y. (2007). Visualization of wide-area
ent chapter proposes an analytical descriptive frequency measurement information. In Proceed-
methodology to study the dynamical stability and ing of IEEE PES General Meeting, Tampa, FL.
security of a large scale power system, following
serious disturbances. The possibility of simulta- Bevrani, H. (2009). Power system control-an
neous remote measurement of bus voltage, rotor overview. In Bevrani, H. (Ed.), Robust power
angle and system frequency through synchronized system frequency control. New York, NY: Springer.
PMUs and intelligent electronic devices empha- doi:10.1007/978-0-387-84878-5_1
sizes the significance of the proposed descriptive
Bevrani, H., Ledwich, G., Dong, Z. Y., & Ford,
scheme for real-world complex power systems.
J. J. (2009). Regional frequency response analy-
In this chapter, first an overview on the electro-
sis under normal and emergency conditions.
mechanical waves in power systems, mathematical
Electric Power Systems Research, 79, 837–845.
formulation and their interesting properties with
doi:10.1016/j.epsr.2008.11.002
a brief literature review is presented. The ampli-
fication of a propagated wave due to reflections Bevrani, H., Ledwich, G., Ford, J. J., & Dong, Z.
or in combination with waves initiated from other Y. (2009). On feasibility of regional power system
disturbances is studied and it is shown how the emergency control plans. Energy Conversion and
occurrence of a large disturbance in a power system Management, 49(2), 193–204. doi:10.1016/j.
may lead to an uncontrolled tripping of generators enconman.2007.06.021
and cascading failures, and may finally result in a
Bevrani, H., & Tikdari, A. G. (2009). On the ne-
blackout; if proper actions are not taken.
cessity of considering both voltage and frequency
Finally, based on given descriptive study of
in effective load shedding scheme. IEEJ Technical
electrical measurements and electromechanical
Meeting, PSE-10-002, Fukui, Japan.
wave propagation in large electric power systems,
a power system emergency control scheme is in- Bevrani, H., & Tikdari, A. G. (2010). An ANN-
troduced to detect possible plans. The chapter is based power system emergency control scheme
supplemented by several simulations on standard in the presence of high wind power penetration.
24-bus, 200-bus, and 400-bus test systems. In L. F. Wang, et al., (Eds.), Wind power systems:
Applications of computational intelligence, (pp.
215-254). Heidelberg, Germany: Springer-Verlag:
ACKNOWLEDGMENT Series on Green Energy and Technology.
Bevrani, H., & Tikdari, A. G. (2010). Power sys-
This work is financially supported by Research
tem stability analysis based on descriptive study
Office at University of Kurdistan.
of electrical indices. In ASIJ 5th Conference, 6
March 2010, Tokyo.
310
A Descriptive Approach for Power System Stability and Security Assessment
Bevrani, H., Tikdari, A. G., & Hiyama, T. (2010). Phadke, A. G., & Thorp, J. S. (2008). Synchronized
An intelligent based power system load shedding Phasor measurements and their applications.
design using voltage and frequency information. New York, NY: Springer. doi:10.1007/978-0-
International Conference on Modelling, Identifi- 387-76537-2
cation and Control (ICMIC’10), (pp. 545-549).
Sluis, L. V. D. (2001). Travelling waves. In
Okayama, Japan.
Greenwood, A. (Ed.), Transients in power
FRCC. (2009). Automatic underfrequency load systems (pp. 31–56). John Wiley & Sons Ltd.
shedding program. (PRC-006-FRCC-01, 2009). doi:10.1002/0470846186.ch3
Retrieved from https://www.frcc.com/
Sowa, P., Azmy, A. M., & Erlich, I. (2004).
IEEE RTS Task Force of APM Subcommittee. Dynamic equivalents for calculation of power
(1979). IEEE reliability test system. IEEE Trans- system restoration. Retrieved from http://www.
actions on Power Apparatus and Systems, 98(6), math.wichita.edu
2047–2056. doi:10.1109/TPAS.1979.319398
Thorp, J. S., Seyler, C. E., & Phadke, A. G. (1998,
IEEE RTS Task Force of APM Subcommittee. June). Electromechanical wave propagation in
(1999). The IEEE reliability test system-1996. large electric power systems. IEEE Transactions
IEEE Transactions on Power Systems, 14(3), on CAS, 45(6), 614–622. doi:10.1109/81.678472
1010–1020. doi:10.1109/59.780914
Tikdari, A. G. (2009). Load shedding in the pres-
Kundur, P. (1994). Power system stability and ence of renewable energy sources. Dissertation,
control. New York, NY: McGraw-Hill. University of Kurdistan, Iran.
Miller, T. J. E. (1982). Reactive power control in Tsai, S. J. S., Zuo, J., Zhang, Y., & Liu, Y. (2005).
electric systems. New York, NY: Wiley. Frequency visualization in large electric power
systems. IEEE Power Engineering Society Gen-
Ourari, M. L., Dessaint, L. A., & Do, V. Q. (2003).
eral Meeting.
Coherency approach for dynamic equivalents of
large power systems. International Conference on Wagner, C. F., & McCann, G. D. (1950). Wave
Power Systems Transients – IPST 2003 in New propagation on transmission lines (Ed) electrical
Orleans, USA. transmission and distribution reference, central
station engineers of the Westinghouse electric &
Padiyar, K. R., & Krishna, S. (2006). Online detec-
manufacturing company. Lakeside Press.
tion of loss of synchronism using energy function
criterion. IEEE Transactions on Power Delivery, Wang, X., & Vittal, V. (2004). System islanding
21(1), 46–55. doi:10.1109/TPWRD.2005.848652 using minimal cutsets with minimum net flow.
Proceedings of the IEEE PES Power System
Parashar, M., & Thorp, J. S. (2004, June). Con-
Conference and Exposition, New York.
tinuum modeling of electromechanical dynamics
in large-scale power systems. IEEE Transactions Yang, B., Vittal, V., & Heydt, G. T. (2006).
on Circuit and Systems, 1851–1858. Slow-coherency-based controlled islanding—A
demonstration of the approach on the August
14, 2003 blackout scenario. IEEE Transac-
tions on Power Systems, 21(4). doi:10.1109/
TPWRS.2006.881126
311
A Descriptive Approach for Power System Stability and Security Assessment
You, H., Vittal, V., & Wang, X. (2004). Slow Diduch, J. Y., & Chang, C. P. L. (2008, April).
coherency based islanding. IEEE Transactions Islanding Detection Using Proportional Power
on Power Systems, 19(1), 483–491. doi:10.1109/ Spectral Density. IEEE Transactions on
TPWRS.2003.818729 Power Delivery, 23, 776–784. doi:10.1109/TP-
WRD.2007.915907
You, H., Vittal, V., & Yang, Z. (2003). Self-healing
in power systems: An approach using islanding Grobovoy, A., & Lizalek, N. (2002, April). As-
and rate of frequency decline based load shed- sessment of power system properties by wave ap-
ding. IEEE Transactions on Power Systems, 18, proach and structure analysis. Fifth International
174–181. doi:10.1109/TPWRS.2002.807111 Conference on Power System Management and
Control (pp. 365 – 370).
Kimbark, E. (1956). Power System Stability. New
ADDITIONAL READING York: Dover.
Abu-Elnaga, M. M., El-Kady, M. A., & Find- Kook, K. S., Liu, Y., & Bang, M. J. (2008, Septem-
lay, R. D. (1988, Nov). Stability assessment ber). Global behaviour of power system frequency
of highly stressed power systems using the in korean power system for the application of
sparse formulation of the direct method. IEEE frequency monitoring network. IET Generation.
Transactions on Power Systems, 3, 1655–1661. Transmission & Distribution., 2, 764–774.
doi:10.1109/59.192977 Makarov, Y. V. Reshetov, V.I. Stroev, A. Voropai,
Bin Qiu Ling Chen Centeno, V. Xuzhu Dong Yilu I. (2005, Nov.). Blackout Prevention in the United
Liu. (2001). Internet based frequency monitor- States, Europe, and Russia. Proceedings of the
ing network (FNET). IEEE Power Engineering IEEE. Vol. 93 (pp. 1942 – 1955).
Society Winter Meeting.vol.3 (pp. 1166 – 1171). Murphy, R. J. (1996, Jan). Disturbance record-
Cresap, R. L., & Hauer, J. F. (1981). Aril). Emer- ers trigger detection and protection. IEEE
gence of a New Swing Mode in the Western Power Computer Applications in Power, 9, 24–28.
System. IEEE Transactions on Power Apparatus doi:10.1109/67.475956
and Systems, PAS-100, 2037–2045. doi:10.1109/ Pariz, N. Shanechi, H.M. Vaahedi, E. (2003, May).
TPAS.1981.316481 Explaining and validating stressed power systems
De La Ree, J., Centeno, V., Thorp, J. S., & behavior using modal series. IEEE Transactions
Phadke, A. G. (2010, June). Synchronized Phasor on Power Systems, 18, 778–785. doi:10.1109/
Measurement Applications in Power Systems. TPWRS.2003.811307
IEEE Transactions on Smart Grid, 1, 20–27. Rajamani, K., & Hambarde, U. K. (1999, Jul).
doi:10.1109/TSG.2010.2044815 Islanding and load shedding schemes for cap-
Delin Wang Xiaoru Wang Yi Fang Wenbin Hao. tive power plants. IEEE Transactions on Power
(2006, Oct.). Study on Dynamic Characteristics of Delivery, 14, 805–809. doi:10.1109/61.772318
Electromechanical Wave in the Continuum Model
for Power System. International Conference on
Power System Technology.
312
A Descriptive Approach for Power System Stability and Security Assessment
Semlyen, A. (1974, March) Analysis of Distur- Vittal, V., Rajagopal, S., Fouad, A. A., El-Kady,
bance Propagation in Power Systems Based on M. A., Vaahedi, E., & Carvalho, V. F. (1988, Feb).
a Homogeneoue Dynamic Model. IEEE Trans- Transient stability analysis of stressed power
actions on Power Apparatus and Systems. Vol. systems using the energy function method. IEEE
PAS-93 (PP. 676 – 684). Transactions on Power Systems, 3, 239–244.
doi:10.1109/59.43206
Senroy, N., Heydt, G. T., & Vittal, V. (2006, Nov.).
Decision Tree Assisted Controlled Islanding. IEEE Wang Yuguo Zhao Wei Xie Yan. (2009, June).
Transactions on Power Systems, 21, 1790–1797. Dynamic feature extraction of power disturbance
doi:10.1109/TPWRS.2006.882470 signal based on time-frequency technology. Con-
trol and Decision Conference. CCDC ‘09. Chinese
Shu-Jen Tsai Li Zhang Phadke, A.G. Yilu Liu In-
(pp. 2300 – 2303).
gram, M.R. Bell, S.C. Grant, I.S. Bradshaw, D.T.
Lubkeman, D. Le Tang. (2007, Aug.). Frequency Xu, G. Vittal, V. Meklin, A. Thalman, J. E. (2010)
Sensitivity and Electromechanical Propagation Controlled Islanding Demonstrations on the
Simulation Study in Large Power Systems. WECC System. IEEE Transactions on Power
IEEE Transactions on Circuits and Systems. I, Systems. Vol. PP.
Regular Papers, 54, 1819–1828. doi:10.1109/
Yilu, L. (2006). A US-Wide Power Systems Fre-
TCSI.2007.902542
quency Monitoring Network. IEEE Power Systems
Taylor, C. W., & Erickson, D. C. (1997, Jan). Conference and Exposition. (pp. 159 – 166).
Recording and analyzing the July 2 cascading
Yingchen Zhang Markham, P. Tao Xia Lang
outage [Western USA power system]. IEEE
Chen Yanzhu Ye Zhongyu Wu Zhiyong Yuan Lei
Computer Applications in Power, 10, 26–30.
Wang Bank, J. Burgett, J. Conners, R.W. Yilu Liu.
doi:10.1109/67.560830
(2010, Sept.). Wide-Area Frequency Monitoring
Thomas, A. J., & Mahajan, S. M. (2009, Oct.). Network (FNET) Architecture and Applications.
Electromechanical Wave Analysis Through Tran- IEEE Transactions on Smart Grid. Vol. 1 (pp.
sient Magnetic Modeling. IEEE Transactions on 159 – 167).
Power Delivery, 24, 2336–2343. doi:10.1109/
Zhian Zhong Chunchun Xu Billian, B.J. Li Zhang
TPWRD.2009.2027504
Tsai, S.-J.S. Conners, R.W. Centeno, V.A. Phadke,
Thorp, J. S., Seyler, C. E., Parashar, M., & A.G. Yilu Liu. (2005, Nov.). Power system fre-
Phadke, A. G. (1998, Jan). The large scale elec- quency monitoring network (FNET) implementa-
tric power system as a distributed continuum. tion. IEEE Transactions on Power Systems, 20,
IEEE Power Engineering Review, 18, 49–50. 1914–1921. doi:10.1109/TPWRS.2005.857386
doi:10.1109/39.646962
Vanfretti, L. Aliyu, U. Chow, J.H. Momoh, J.A.
(2009, July). System frequency monitoring in the KEY TERMS AND DEFINITIONS
Nigerian power system. IEEE Power & Energy
Society General Meeting (pp. 1 – 6). Angle Instability: Angle instability is a very
fast instability that leads to loss of synchronism.
Emergency Control: Control of a power sys-
tem in dangerous conditions, where the system is
going to the instability or blackout.
313
A Descriptive Approach for Power System Stability and Security Assessment
Islanding Control: Separate of power system Slow Coherency: A method which is used in
into some isolated sub-systems following large islanding control problems. Using this method-
disturbances, in order to prevent a global black out. ology, the system islands and weak links can be
Load Shedding: An emergency control action identified.
to curtail a part of load, and is useful where the Wave Propagation: Propagation of a distur-
amount of load is larger than available generation. bance through the power system network like a
PMU: Power measurement unit is a device propagated wave.
which uses GPS to collect data from the various
points of network in a synchronous time.
314
315
Chapter 11
Analyses and Monitoring
of Power Grid
Rana A. Jabbar
Rachna College of Engineering and Technology, Pakistan
Muhammad Junaid
Rachna College of Engineering and Technology, Pakistan
M. A. Masood
Rachna College of Engineering and Technology, Pakistan
A. Bashir
Rachna College of Engineering and Technology, Pakistan
M. Mansoor
Rachna College of Engineering and Technology, Pakistan
ABSTRACT
Power system analyses and monitoring of power system engineering are as essential as oxygen for
human beings. This innovative approach deals with a 132 kV grid simulation in electrical transient
analyzer program (ETAP). The existing power distribution system in Pakistan consists of approximately
six thousand 11 kV feeders, which are mainly analyzed by software FDR-ANA (Feeder Analyses). This
software does not have capability to provide comprehensive analyses for integrated power system. The
case under study is 132 kV grid situated in Gujranwala electric power company (GEPCO), one of the
distribution companies (DISCO’s) of Pakistan electric power company (PEPCO) which has been selected
for comprehensive analyses using ETAP software. This software performs numerical calculations of large
integrated power system with fabulous speed, besides generating output reports. In a developing country
like Pakistan it is first time that analyses based Off-line monitoring has been made, which includes load
flow, harmonic, transient, short circuit and ground grid analyses. In load flow analysis, current flowing
in every branch, power factor, active and reactive power flow, line losses, voltage magnitude with angle
etc. have been calculated. During harmonic analysis, distorted current and voltage waveforms along
with their harmonic spectrum caused by non-linear loads have been recorded. Transient analysis has
been performed to record different waveforms like variation in bus frequency, bus real power loading,
bus voltage angle, and bus reactive power loading for short interval of time during transient conditions.
DOI: 10.4018/978-1-61350-138-2.ch011
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Analyses and Monitoring of Power Grid
In ground grid modeling, step, and touch potentials have been calculated in comparison with set stan-
dards. While performing short circuit analysis, all the possible short circuit faults like line to ground,
double line to ground, 3-phase faults etc. on ½ cycle, 1.5 to 4 cycle, and 30 cycle networks have been
performed to record the short circuit currents. These analyses have been executed using ETAP software,
based upon historical data obtained from original system that will be very helpful for system security
and reliability.
316
Analyses and Monitoring of Power Grid
simulation world widely, but ETAP has preferred connect, zoom in or out, display grid off or on,
here due to strong built-in properties regarding change element size, change element orientation,
power system studies (Sybille & Hoang, 2002; change symbols, hide or show protective devices,
Qinghuaz et al, 2009; Kjølle et al, 2002; Gatta et enter properties and set operating status, etc.
al, 2003; Gonen, 1968) ETAP’s electrical system diagram is a one-line
Earlier, considerable work has been done on representation of a three-phase system. The one-
transient stability studies, but power system of line diagram is the starting point for all studies. The
developing countries like Pakistan have never been electrical system can be constructed graphically by
simulated using software technique like ETAP on connecting the buses, branches, isolators, circuit
such a large scale. The major purpose for transient breakers and protective devices in any order from
stability studies on this power system is to find the one-line diagram edit toolbar.
dynamic performance which has great impact
in the design and operation of the system. After Single Line Diagram of the System
transient stability studies on 132 kV grid, machine
power angles and speed deviations, system electri- Figure 1 shows the single line diagram of the
cal frequency, real and reactive power flows of complete power system which is under study. It
the machines and buses, power flow of lines and is clear that there are two incoming lines of 132
transformers, as well as the voltage level of the
buses in the power system have been determined
thoroughly (Hongbin et al, 2002; IEEE 1995). Figure 1. Single line diagram of 132 kV grid
A ground grid study is first time addressed
in this research for a large power system. The
touch and step potential measurement will really
be helpful for safety of both grid staff and equip-
ment. It is worth mentioning that in this analysis
three-dimension view grid grounding scheme
has been used on such a large scale using ETAP.
In short circuit analyses on 132 kV grid mo-
mentary, steady-state, short circuit and other fault
interruption currents on ½ cycle network (sub
transient network), 1.5 to 4 cycle network (tran-
sient network), and 30 cycle network at different
buses have also been determined (Osahenvemwem
& Omorogiuwa, 2008).
BACKGROUND
317
Analyses and Monitoring of Power Grid
318
Analyses and Monitoring of Power Grid
319
Analyses and Monitoring of Power Grid
Figure 5. Load modeling for 200 kVA distribution Figure 6. Load modeling for50 kVA distribution
transformer transformer
320
Analyses and Monitoring of Power Grid
Over-Loaded Transformers
Voltage Drop
system. Similarly, Furnace-2 and Furnace-3 take
loads of 161.2 A and 136.5 A respectively (Jabbar
Table 7 shows the % voltage drop in different
et al, 2008a; Jabbar et al, 2008b).
distribution transformers (only few are given
here). Magnitude of voltage drop standard set by
utility is 5% which is significantly violated here.
SOLUTIONS AND
RECOMMENDATIONS
Load flow studies have been carried out on com- Monitoring Points kV MW MVAR % PF
plete power system using ETAP. However for Grid (Bus 3) 132 36.68 23.10 84.6
detailed monitoring purpose three important points A 11 5.520 3.793 83.1
are selected as shown in Figure 1. Point A is the B 11 0.379 0.253 83.0
secondary side of power transformer T1. Point B C 11 0.717 0.471 83.6
321
Analyses and Monitoring of Power Grid
Loading (output)
TF ID Capability (MVA) Loading (input) MVA Loading (input) % Loading (output) %
MVA
T114 0.200 1.251 625.3 0.952 476.2
T797 0.200 0.352 175.8 0.330 165.12
T122 0.200 0.452 225.9 0.416 208.231
T618 0.200 0.352 175.8 0.330 165.1
T803 0.200 0.452 225.9 0.416 208.3
T807 0.200 1.195 597.3 0.926 462.3
T809 0.200 0.352 175.8 0.330 165.3
322
Analyses and Monitoring of Power Grid
CKT/Branch
ID % Bus Voltage from % Bus Voltage to %Voltage drop (Mag.)
T114 73.7 96.8 23.08
T624 89.1 96.12 7.58
T625 90.4 96.2 6.27
T628 75 96.7 21.77
T797 90.8 96.7 5.89
T803 89.1 96.12 7.58
T819 75 96.7 21.77
T821 90.7 96.8 5.88
T831 75 96.7 21.77
T833 90.7 96.8 5.88
T839 89.1 96.12 7.58
C has exceeded the limit set by IEEE std. 519- Similarly, the current waveform and har-
1992 (i.e. 5% at the point of common coupling). monic spectrum captured at point B are shown in
The current waveform and its harmonic spec- Figure 12 and Figure 13 respectively.
trum captured at the monitoring point A are shown Voltage waveform with harmonic spectrum at
in Figure 8 and Figure 9 respectively. point B is given in Figure 14.
Voltage waveform and its harmonic spectrum
captured at monitoring point A are shown in
Table 9. % THDv & % THDi at monitoring points
Figure 10 and Figure 11. The below waveshape
is at the 11kV bus bar which shows the significant Monitoring Points % THDv %THDi
presence of 5th and 7th harmonic componets that A 2.13 3.41
also contribute very much in the disturbing the
B 3.0 2.25
power quality of the system.
C 13.1 15
323
Analyses and Monitoring of Power Grid
324
Analyses and Monitoring of Power Grid
325
Analyses and Monitoring of Power Grid
326
Analyses and Monitoring of Power Grid
327
Analyses and Monitoring of Power Grid
Due to presence of dc load voltage waveform Similarly, the voltage waveform and har-
is also distorted as a result of current waveform monic spectrum at point C are given in Figure 18
distortion. FFT spectrum is shown in Figure 15, and Figure 19 respectively.
at the monitoring point B, which shows that 5th, Harmonic analysis has been performed on
7th, 11th and 13th harmonic components contrib- complete system but due to space limitation it is
ute very much. not possible to discuss here any more.
Figure 16 and Figure 17 are the current wave-
form and its harmonic spectrum at point C which Transient Stability Analysis
is more distorted as compared to points A and B.
The reason behind this distorted pattern is sig- During transient stability analysis for the system
nificant presence of 5th, 7th, 11th and 13th etc. under consideration given in Figure 1, the stability
harmonic contents in the current waveform drawn limits of the power system can be inspected before,
by the furnace load. during and after system changes. Figure 20 is the
328
Analyses and Monitoring of Power Grid
section of power system modeled in ETAP which It is observed, during the transient stability
is selected for transient analysis, where a group analysis on the selected bus that real and reactive
of different motors is connected to a distribution power loading of the bus changes abruptly during
transformer. transient phase. Figure 22 and Figure 23 show
Figure 21 shows bus frequency variation at real and reactive power loading at specific time
specific time interval (0.5 – 0.7 sec) when group interval when group of different motors is con-
of different motors is connected to bus 232 during nected to the system.
transient stability analysis. It is clear from the During the switching of the motor load to the
Figure 3 that before 0.5 sec the system frequency power system it is observed that magnitude in bus
was 50 Hz. (i.e. 100%), after 0.5 sec – 0.7 sec voltage and angle is distorted very firmly which
during transient the system frequency decreases causes for the instability problems in the power
significanly and system becomes normal after system. Figure 24 and Figure 25 show bus voltage
that transient interval. magnitude and angle variations at perticular instant
329
Analyses and Monitoring of Power Grid
330
Analyses and Monitoring of Power Grid
331
Analyses and Monitoring of Power Grid
Table 10. Calculated step potential and touch potential on tolerable limits
332
Analyses and Monitoring of Power Grid
Short Circuit Fault Analysis Three different impedance networks are formed
for each faulted bus where calculations are made
Three-Phase Fault Calculations to calculate for sub transient, transient and steady
at Different Buses state short circuit currents. These networks are
½ cycle network (sub transient network), 1.5 - 4
For short circuit analysis some of the buses are cycle network (transient network) and 30 cycle
selected where an artificial short circuit fault network (steady state network).
is created. An equivalent voltage source at the
fault location is replaced by all external voltage
sources and machines internal voltage sources.
333
Analyses and Monitoring of Power Grid
Table 11. 3-phase short circuit current calculations in half cycle network
½ Cycle
Fault Bus ID ½ Cycle (kA Real) Imag/ Real kA symm Magnitude
(kA Imag)
232 5.661 -112.818 19.9 112.960
240 2.851 -12.084 4.2 12.416
693 0.881 -2.5588 2.9 2.705
695 0.993 -2.963 3.0 3.125
696 0.864 -2.503 2.9 2.648
700 2.045 -17.159 8.4 17.280
Table 12. 3-phase short circuit current calculations in 1.5 to 4 cycles network
Fault Bus ID 1.5-4 Cycle (kA Real) 1.5-4 Cycle (kA Imag) Imag/ Real kA symm Magnitude
240 2.632 -11.828 4.5 12.117
693 0.868 -2.548 2.9 2.692
695 0.977 -2.950 3.0 3.107
696 0.852 -2.494 2.9 2.635
700 1.747 -16.596 9.5 16.688
Fault Bus ID 30 Cycle (kA Real) 30 Cycle (kA Imag) X/R Ratio kA Symm Magnitude
232 3.712 -61.298 16.5 61.140
240 2.491 -11.509 4.6 11.775
693 0.858 -2.535 3.0 2.676
695 0.963 -2.932 3.0 3.086
696 0.842 -2.481 2.9 2.620
700 1.621 -15.964 9.8 16.046
This type of network is used to calculate sub Transient currents at selected buses in the power
transient currents at different buses in the sys- system are represented by using 1.5 to 4 cycle
tem. Detail of short circuit sub transient currents network in EATP. Transient calculations are given
calculations are given in the Table 11. in Table 12 at different buses.
It is important to note that bus 232 is rated at
0.415 kV and remaining buses are rated at 11 kV. Thirty-Cycle Network
334
Analyses and Monitoring of Power Grid
Table 14. 3-phase short circuit current & sequence voltages at faulted buses in sub transient networks
Fault Bus ID 3 Phase Fault %V kA Symm Line-Ground Fault (%V from Bus) Va Vb Vc
232 0.0 112.960 0.00 100.12 100.88
240 0.0 12.416 0.00 173.21 173.21
693 0.0 2.705 0.00 173.21 173.21
695 0.0 3.125 0.00 173.21 173.21
696 0.0 2.3648 0.00 173.21 173.21
700 0.0 17.280 0.00 173.21 173.21
Table 15. 3-phase symmetrical short circuit current calculations in 1/2 cycles network (sub transient
network)
Table 16. 3-phase short circuit current and sequence voltages at faulty buses in transient networks
Fault Bus ID 3 Phase Fault %V kA Symm Line-Ground Fault (%V from Bus) Va Vb Vc
232 0.0 112.960 0.00 100.75 100.57
240 0.0 12.416 0.00 173.21 173.21
693 0.0 2.705 0.00 173.21 173.21
695 0.0 3.125 0.00 173.21 173.21
696 0.0 2.648 0.00 173.21 173.21
700 0.0 17.280 0.00 173.21 173.21
13 shows steady state short circuit currents at Table 14 shows 3-phase fault currents and their
selected buses. respective sequence voltages at the faulty buses.
Three-phase symmetrical short circuit current
Three-Phase LG, LL, and LLG (1/2 at faulty buses are shown in Table 15 as:
Cycle Network) Fault Currents at
Selected Buses Three-Phase LG, LL, AND LLG (1.5 - 4
Cycle Network) Fault Currents at
The purpose of this analysis on the faulty buses is Different Buses
to calculate line-to-ground, line-to-line, double-
line-to-ground, three-phase short-circuit fault cur- Line-to-ground, line-to-line, double-line-to-
rents and sequence voltages at ½ cycles networks. ground, three-phase short-circuit fault currents
335
Analyses and Monitoring of Power Grid
Table 17. 3-phase symmetrical short circuit current calculations in 1.5-4 cycles network (transient network)
Table 18. 3-phase short circuit current & sequence voltages at faulty buses in transient networks
Fault Bus ID 3 Phase Fault %V kA Symm Line-Ground Fault (%V from Bus) Va Vb Vc
232 0.0 61.410 0.00 79.73 79.64
240 0.0 11.775 0.00 173.21 173.21
693 0.0 2.676 0.00 173.21 173.21
695 0.0 3.086 0.00 173.21 173.21
696 0.0 2.626 0.00 173.21 173.21
700 0.0 16.046 0.00 173.21 173.21
Table 19. 3-phase symmetrical short circuit current calculations in 30-cycles network (transient network)
and sequence voltage values in 1½ - 4 cycles Three-Phase LG, LL & LLG (30 Cycle
are represented through this study on the faulty Network) Fault Currents at Different
buses. Table 16 and Table 17 indicates 3-phase Buses
fault currents along with their respective sequence
voltages and 3-phase symmetrical short circuit Three-phase line-ground, line-line and double-
current calculations in 1.5 - 4 cycles network line-ground fault is calculated for 30-cycle net-
(transient network) at the selected faulted buses work are summarized in Table 18 and Table 19
respectively. respectively.
336
Analyses and Monitoring of Power Grid
Research conducted during this work is based Bashir, A., Jabbar Khan, R. A., Junaid, M., &
upon the historical data of a practical grid which Asghar, M. M. (2010). ETAP software based
has further been implemented in ETAP for off- transient, ground Grid and short circuit analyses
line monitoring. Keeping in view the latest trends, of 132 kV Grid. 3rd Global Conference on Power
on-line monitoring can also be performed on Control and optimization (PCO 2010). 2-4 Feb-
same network using same software tool along ruary, 2010, Gold Coast, Queensland, Australia.
with hardware setup. For accurate monitoring Brown, K., Shokooh, F., Abcede, H., & Don-
and analyses purpose, real time monitoring has ner, G. (1990). Interactive simulation of power
been recommended in future. Moreover, protec- system: ETAP techniques and applications (pp.
tive equipment installed at grid and sensitive 1930–1941). IEEE.
load equipment can also be modelled for relay
coordination analysis in ETAP. ETAP. (n.d.). Enterprise software solution for
power systems (Electrical Transient Analyzer
Program). Retrieved from http://etap.com/training
CONCLUSION /tutorials-training-videos.htm
Gatta, F. M., Iliceto, F., Lauria, S., & Masato, P.
ETAP Software based simulation of integrated
(2003). Modeling and computer simulation of
electric power system comprising of 132 kV grid
dispersed generation in distribution networks:
including load flow, harmonics, transient, ground
Measures to prevent disconnection during system
grid and short circuit analyses reveals innovative
disturbances. IEEE PowerTech Conference. June
contribution to the power industry stake holders
2003, Italy.
to access in-advance knowledge for acquiring
the security and quality of supply. Overloading Gonen, T. (1986). Electric power transmission
of power/distribution transformers, line conduc- system engineering (pp. 87–93). New York, NY:
tors/cables current carrying ability, power factor, Wiley.
supply demand gap, voltage drop at the tail end,
Hongbin, Z., Renmu, H., & Jian, Z. (2002). Ap-
technical losses, active and reactive power flow,
plication of different load models for the transient
voltage and current magnitudes, THD in voltage
stability calculation. Proc. 2002 IEEE Power
and current etc. can be analyzed and monitored
Engineering Society Transmission and Distribu-
at any desired location using this innovative ap-
tion Conf., (pp. 2014-2018).
proach. Moreover, the novel approach of ground
grid 3D analysis is really interesting and will be IEEE Task Force on Load Representation for
helpful for evaluating the exact value of step and Dynamic Performance. (1995). Bibliography on
touch potential for safety purpose. Simulation con- load models for power flow and dynamic perfor-
ducted in this research work can be implemented mance simulation. IEEE Transactions on Power
successfully at any desired section of the power Systems, 10.
system network for above mentioned parameters.
337
Analyses and Monitoring of Power Grid
338
Analyses and Monitoring of Power Grid
Grainger, J. J., & Stevenson, W. (1994). Power Tleis, N. (2007). Power Systems Modelling and
System Analysis. McGraw-Hill Science/ Engg/ Fault Analysis: Theory and Practice Book. ISBN:
Math. 0750680741
(2002). Natarajan (1st ed.). Computer Aided
Power System. CRC.
339
Analyses and Monitoring of Power Grid
Load Flow Analysis: Calculates the bus voltages, branch power factors, currents, and power flows
throughout the electrical system. Unlike traditional circuit analysis, a power flow study usually uses
simplified notation such as a one-line diagram and per-unit system, and focuses on various forms of AC
power rather than voltage and current. It analyzes the power systems in normal steady-state operation.
Load Flow Calculation Methods: ETAP provides three load flow calculation methods: Newton-
Raphson, Fast-Decoupled, and Accelerated Gauss-Seidel. Newton-Raphson method has been used
during the simulation.
∆P J J ∆δ
1 2
∆Q J J = ∆V (1)
3 4
Fast-Decoupled Method: Derived from the Newton-Raphson method. It takes the fact that a small
change in the magnitude of bus voltage does not vary the real power at the bus appreciably, and likewise,
for a small change in the phase angle of the bus voltage, the reactive power does not change apprecia-
bly. Thus the load flow equation from the Newton-Raphson method can be simplified into two separate
decoupled sets of load flow equations, which can be solved iteratively (Equation 2):
∆P = J ∆δ
1 (2)
∆Q = J ∆V
4
Gauss-Seidel Method: From the system nodal voltage equation (Equation 3):
l = Y V (3)
BUS
P + jQ = V T Y V (4)
BUS
Data Required for Load Flow Analysis: Following data is required during load flow analysis:
• Bus Data
• Branch Data
• Power Grid Data
• Synchronous Generator Data
• Inverter Data
• Synchronous Motor Data
• Induction Motor Data
• Static Load Data
• Capacitor Data
• Lumped Load Data
• Charger & UPS Data
340
Analyses and Monitoring of Power Grid
Harmonic Analysis: Due to the wide and ever increasing applications of power electronic devices,
such as variable speed drives, uninterruptible power supplies (UPS), static power converters, etc., power
system voltage and current quality has been severely affected in some areas. In these areas components
other than that of fundamental frequency can be found to exist in the distorted voltage and current
waveforms. These components usually are the integer multipliers of the fundamental frequency, called
harmonics. In addition to electronic devices, some other non-linear loads, or devices including saturated
transformers, arc furnaces, fluorescent lights, and cyclo-converters are also responsible for the deteriora-
tion in power system quality.
Harmonic Analysis Calculation Methods: ETAP harmonic analysis program fully complies with
the latest version of the following standards:
• IEEE Standards 519, IEEE Recommended Practices and Requirements for Harmonic Control in
Electrical Power Systems
• IEEE Standards 141, IEEE Recommended Practice for Electric Power Distribution for Power
Plants
• ANSI/IEEE Standard 399, IEEE Recommended Practice for Power System Analysis
∑F i
2
THD = 2
(5)
F1
Individual Harmonic Distortion (IHD): Simply calculates the ratio of the harmonic component to
the fundamental component. This value is sometimes used to track the effect of each individual harmonic
and examine its magnitude. IHD is determined by (Equation 6):
Fi
IHD = (6)
F1
Root Mean Square (RMS) Total: This is the square root of the sum of the squares of the magnitudes
of the fundamental plus all harmonics in the system. For a system with no harmonics at all, the total RMS
should be equal to the fundamental component RMS. The total RMS is determined by (Equation 7):
ϑ
RMS = ∑F i
2
(7)
1
Transient Stability Analysis: ETAP transient stability analysis program is designed to investigate
the stability limits of a power system before, during and after system changes or disturbances. The
program models dynamic characteristics of a power system, implements the user-defined events and
actions, solves the system network equation and machine differential equations interactively to find out
system and machine responses in time domain.
341
Analyses and Monitoring of Power Grid
Power System Stability: The property of a power system which insures that it remains in electro-
mechanical equilibrium throughout any normal and abnormal operating conditions.
Stability Limits: There are two types of stability limit for a power system, namely steady-state
stability limit and transient stability limit.
Steady-State Stability Limit: defined as the stability of a system under conditions of gradual or small
changes in the system. This stability can be either found by the load flow calculation for a steady-state
operation, or determined by a transient stability study if there are system changes or disturbances involved.
Transient Stability Limit: defined as the stability of a system during and after sudden changes or
disturbances in the system, such as short-circuits, loss of generators, sudden changes in load, line trip-
ping, or any other similar impact. The system is said to be transient stable if following a severe distur-
bance, all synchronous machines reach their steady-state operating condition without prolonged loss of
synchronism or going out of step with other machines.
Causes of Instability Problems: The major causes to industrial power system instability problems
include:
• Short-circuits
• Loss of a tie connection to a utility system
• Loss of a portion of in-plant co-generation (generator rejection)
• Starting a motor that is large relative to the system generating capacity
• Switching operations of lines, capacitors, etc.
• Impact loading (motors and static loads)
• A sudden large step change of load or generation
• Area-wide blackout
• Interruption of loads
• Low-voltage conditions
• Damage to equipment
• Relay and protective device malfunctions
Ground Grid Systems: Since the early days of the electric power industry, the safety of personnel
in and around electric power installations has been a primary concern. With ever increasing fault cur-
rent levels in today’s interconnected power systems, there is renewed emphasis on safety. The safety of
personnel is compromised by the rise in the ground potential of grounded structures during unbalanced
electric power faults. At such times, humans touching grounded structures can be subjected to voltages.
However, the magnitude and duration of the electric current conducted through the human body should
not be sufficient to cause ventricular fibrillation
Years of research on the effects of electric current on the human body have lead to the development
of standards of permissible values to avoid electrocution. The ground grid systems program utilizes the
following four methods of computation:
342
Analyses and Monitoring of Power Grid
• IEEE 80-2000
• EEE 665-1995
Calculations during Ground Grid Analysis: Following are the calculations performed during this
analysis:
Short Circuit Analysis: ETAP short-circuit analysis program analyzes the effect of three-phase,
line-to-ground, line-to-line, and line-to-line-to-ground faults on the electrical distribution systems. The
program calculates the total short-circuit currents as well as the contributions of individual motors,
generators, and utility ties in the system. Fault duties are in compliance with the latest editions of the
ANSI/IEEE standards (C37 series) and IEC standards (IEC 909 and others).
3-Phase Faults: Device Duty: This study calculates momentary symmetrical and asymmetrical RMS,
momentary asymmetrical crest, interrupting symmetrical RMS, and interrupting adjusted symmetrical
RMS short-circuit currents at faulted buses.
3-Phase Faults: 30 Cycle Network: This study calculates short-circuit currents in their RMS values
after 30 cycles at faulted buses.
LG, LL, LLG, & 3-Phase Faults: ½ Cycle: This study calculates short-circuit currents in their RMS
values at ½ cycle at faulted buses.
LG, LL, LLG, & 3-Phase Faults: 1.5 to 4 Cycle: This study calculates short-circuit currents in
their RMS values between 1.5 to 4 cycles at faulted buses.
LG, LL, LLG, & 3-Phase Faults: 30 Cycle: This study calculates short-circuit currents in their RMS
values at 30-cycles at faulted buses.
343
344
Chapter 12
Solving Fuzzy Optimization
Problems of Uncertain
Technological Coefficients
with Genetic Algorithms and
Hybrid Genetic Algorithms
Pattern Search Approaches
Pandian Vasant
University Technology Petronas, Malaysia
ABSTRACT
In this chapter a solution is proposed to a certain nonlinear programming difficulties related to the pres-
ence of uncertain technological coefficients represented by vague numbers. Only vague numbers with
modified s-curve membership functions are considered. The proposed methodology consists of novel
genetic algorithms and a hybrid genetic algorithm pattern search (Vasant, 2008) for nonlinear program-
ming for solving problems that arise in industrial production planning in uncertain environments. Real
life application examples in production planning and their numerical solutions are analyzed in detail.
The new method suggested has produced good results in finding globally near-optimal solutions for the
objective function under consideration.
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
modeling: uncertain systems in industry, nature 2. They deal with multiple solutions in one run.
and humanity; and facilitators for commonsense This is useful to achieve solutions rapidly in
reasoning in decision making in the absence of the presence of a large number of parameters;
complete and precise information. Their role is 3. They allow the exploration of multiple local
significant when applied to complex phenomena optima;
not easily described by traditional mathematical 4. GAs have successfully been applied to
methods, especially when the goal is to find a various optimization problems that involve a
good approximate solution (Bojadziev and Bo- large number of parameters, multiple criteria,
jadziev, 1998; Zamirian et. al, 2009). Modeling and complex criteria relationships.
using fuzzy sets has proven to be an effective
way for formulating decision problems where the It is commonly believed that the main driving
information available is subjective and imprecise principle behind the natural evolutionary process
(Zimmermann, 1992; Jimenez et. al, 2008). is the Darwin’s survival-of-the-fittest principle
A linguistic variable is a variable whose values (Eldredge, 1989). In most scenarios, nature ruth-
are words or sentences in a natural or artificial lessly follows two simple principles:
language (Zadeh, 1975). As an illustration, Age
is a linguistic variable if its values are assumed to 1. If by genetic processing an above-average
be the fuzzy variables labeled young, not young, offspring is created, it usually survives longer
very young, not very young, etc. rather than the than an average individual and thus has more
numbers 0, 1, 2, 3. (Bellman and Zadeh, 1977). opportunities to produce offspring having
The concept of a linguistic variable provides a some of its traits than an average individual.
means of approximate characterization of phe- 2. If, on the other hand, a below-average off-
nomena which are too complex or too ill-defined spring is created, it usually does not survive
to be amenable to description in conventional longer and thus gets eliminated quickly from
quantitative terms. The main applications of the the population.
linguistic approach lie in the realm of humanistic
systems—especially in the fields of artificial in- The two important main characteristics of
telligence, linguistics, human decision processes, genetic algorithms are provided:
pattern recognition, psychology, law, medical Exploration: The process of visiting entirely
diagnosis, information retrieval, economics and new regions of a search space, to see if anything
related areas (Zadeh, 1975). promising may be found there. Unlike exploita-
In this paper a novel, genetic algorithm (GA) tion, exploration involves leaps into the unknown.
approach is reported that was successfully used for Problems which have many local maxima can
solving problems originating from the uncertainty sometimes only be solved by this sort of random
of the technological coefficients in production search.
planning in industrial engineering. The main Exploitation: When traversing a search space,
reason for GA method is adopted in solving this exploitation is the process using information
problem is given below. gathered from previously visited points in the
Some of the advantages of GA over classical search space to determine which places might be
optimization methods include (Deb, 2001): profitable to visit next. An example is hill climbing,
which investigates adjacent points in the search
1. GA is less susceptible to the complexity of space, and moves in the direction giving the great-
the problem at hand than non-evolutionary est increase in fitness. Exploitation techniques
methods;
345
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
are good at finding local maxima (Maric, 2010; ibility, and because they are useful in promoting
Stanimrovic, 2010). representations and information processing in
Employing a GA is a global stochastic method a fuzzy environment. In this study an s-curve
based on the mechanism of natural selection membership function for the fuzzy numbers in
and evolutionary genetics and is used in various uncertain technological coefficients of the fuzzy
fields. For a specific problem such as nonlinear optimization is adopted (Vasant, 2003; Vasant &
programming, the combination of GA and other Barsoum, 2006).
methods such as the gradient based method can Decision-makers face up to the uncertainty
outperform the GA alone. This can be illustrated and vagueness with subjective perceptions and
by experimental results (Honggang and Jianchao, experiences in the decision-making process (Er-
1997; Jeya Mala et. al, 2010). Hybrid genetic al- tugrul and Karakasoglu, 2006). By using the fuzzy
gorithms have been adopted in this research work. optimization approach, uncertainty and vagueness
The GA approach that has been used in this from subjective perception and the experiences of
research work has some disadvantages regarding a decision maker can be effectively represented
global optimization. The major drawback of the and a more effective decision reached (Sanchez,
technique in this research work is that it is un- Jimenez and Vasant, 2007). A quadratic objective
able to locate the global optimum for a nonlinear function for a production planning problem was
problem with uncertain technological coefficients considered by Turabieh, Sheta and Vasant (2007).
for industrial production planning. Therefore we In order to describe the real world problem of
are motivated to use a more superior technique of production planning, a non-linear cubic function
hybrid evolutionary computation to solve nonlin- is considered in this research work.
ear industrial production planning problems with We formulate a non-linear cubic function for
uncertain technological coefficients. the fuzzy optimization problem of industrial pro-
The rest of this paper is structured as follows: duction planning. This problem cannot be solved
Section 2 describes the membership function for by the fuzzy linear programming approach. There-
the technological coefficients for the mathemati- fore in this research work global optimization
cal model of the problem. It’s followed by a brief techniques using genetic GAs have been adopted.
review of previous work on the GA approach for
production planning problems in Section 3. Sec-
tion 4 provides a description of the case study of GENETIC ALGORITHMS
chocolate manufacturing problems. Illustration on FOR PRODUCTION
the problem formulation and mathematical model PLANNING PROBLEMS
for the case study is given in Section 5. Section 6
explains the computational results and findings Since the late 1980s there has been a growing
on the solutions of the optimization problems of interest in Genetic Gas (stochastic optimization
production planning in an uncertain environment. algorithms based on the principles of natural
The paper ends with a conclusion in Section 7. [Darwinian] evolution). These have been used
widely for parameter optimization, classification
and learning. More recently, production plan-
FUZZY MEMBERSHIP FUNCTION ning has emerged as an area to which GAs can
be applied. A detailed introduction to GAs can
In applications it is often convenient to work be found in Goldberg (1989). One of the earliest
with s-curve membership functions for the fuzzy reported applications of GAs to production was
numbers because of their computational flex- reported by Davis (1985). It is a characteristic of
346
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
the robust optimization process that once fairly a ‘general purpose’ solution to the optimization
good solutions have been formed their features problem, with the peculiarities of any particular
will be carried forward into better solutions and example being accounted for in the fitness func-
lead ultimately to optimal solutions. It is in the tion without disturbing the logic of the standard
nature of GAs that new solutions are formed from optimization (GA) routine. This means that it is a
the features of known good solutions. Therefore, relatively straightforward and convenient to adapt
it follows that GAs are particularly attractive for the software implementation of the method to meet
production planning. Compared with other opti- the needs of particular applications.
mization methods, GAs are suitable for traversing
large search spaces since they can do this relatively Case Study Chocolate Manufacturing
rapidly and because the mutation operator diverts
the method away from local minima, which will Due to limitations in resources for manufacturing
tend to become more common as the search space a product and the need to satisfy certain conditions
increases in size. Suitability for large search spaces in manufacturing and demand, a problem of fuzzi-
is a useful advantage when dealing with produc- ness occurs in industrial systems. This problem
tion problems of increasing size, since the solution occurs, for example, in chocolate manufacturing
space grows very rapidly, especially when this is when deciding a mixed selection of raw materials
compounded by such features as alternative profit to produce varieties of products. This is referred
production optimization. It is important that these here to as the product-mix selection problem
large search spaces are traversed as rapidly as (Tabucanon, 1996).
possible to enable the practical and useful imple- There are a number of products to be manu-
mentation of automated production optimization. factured by mixing different raw materials and
If the optimization is done quickly, production using several varieties of processing. There are
managers can try out ‘what-if’ scenarios and limitations in raw material resources and facility
detailed optimization analyses besides being able usage for the different types of processing. The
to react to ‘crises’ as soon as possible. Traditional raw materials and facilities usage required for
approaches to production planning optimization manufacturing each product are expressed by
such as mathematical programming and ‘branch means of fuzzy coefficients. There are also some
and bound’ are computationally very slow in such constraints imposed by the marketing department
a massive search space. Researchers have found such as product-mix requirements, main product
that a key advantage of a GA is that it provides line requirements and lower and upper limits of
Table 1. Profit Coefficients ci, di and ei (Profit function in US $ per 103 units)
347
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
8
demand for each product. It is necessary to obtain Maximize 2 3
∑ (c x − di xi − ei xi )
i=1 i i
maximum profit with a certain degree of satisfac- Subject to:
tion for the decision-maker.
8 l
h
a − al 1
B
ij ij − 1 x − b ≤ 0,
∑ aij + ln i j j = 1, 2,..., 17
i=1 α C µ
Problem Formulation 8
6
∑ ri xi − 0.15 ∑ ri xi ≤ 0
i=7 i=1
Optimization techniques are primarily used in x − 0.6x ≤ 0
1 2
production planning problems in order to achieve x 3 − 0.6x 4 ≤ 0
x5 − 0.6x6 ≤ 0
optimal profit; a certain objective function is 0 ≤ xi ≤ ui , i = 1, 2,..., 8
maximized by satisfying a number of constraints. C = 0.001001001, B = 1,0.001£m £0.99, and 1£a £45.
The first step in an optimal production planning (1)
problem is to formulate the underlying nonlinear
programming (NLP) problem with uncertain tech- In the non-linear programming problem
nological coefficients by writing the mathematical (Equation 1), the variable vector x represents a
functions relating to the objectives and constraints. set of variables xi, i = 1, 2,…, 8. The optimization
Given a degree of possibility (μ), the fuzzy problem contains eight continuous variables and
constrained optimization problem can be formu- 21 inequality constraints. A test point xi satisfying
lated (Vasant, 2006) as a nonlinear constrained constraints is either feasible, or infeasible. The
optimization problem shown. Tables 1, 2, 3, and set satisfying the constraints is called the feasible
4 provide the input data for Equation 1 in the domain. The aim of the optimization is to maxi-
problem statement and the values for ci, di, ei, mize the total production profit for the industrial
aijh, aijl, ri, ui and bj. production planning problem. The practical mean-
The mathematical formulation (Equation 1), ing of the constraints is available in Tabucanon
of the problem is now described. The eight deci- (1996). The vagueness factor γ represents fuzzy
sion variables are: 250g of Milk Chocolate (x1), numbers in uncertain technological parameters.
100g of Milk Chocolate (x2), 250g of Crunchy The objective function appears in work by Lin
Chocolate (x3), 100g of Crunchy Chocolate (x4), (2007) and Chaudhuri (2007). The objective
250g of Chocolate with Nuts (x5), 100g of Choc- function represents the profit.
olate with Nuts (x6), Chocolate Candy (x7), and
Chocolate Wafer (x7). The objective is to find an
optimal profit value for the production planning COMPUTATIONAL RESULTS
problem with uncertain technological coefficients.
The production planning problem is subject to a Experimental studies are carried out for GAs and
number of constraints imposed by raw material hybrid GA pattern search (HGAPS) approaches.
availability, facility capacity, and the sales depart- In the following sections, the simulation results
ment. An upper and lower limit is imposed on for the GA approach is provided in detail and the
each of the decision variables (refer to Table 3). improved results of the HGAPS approach are
In Equation 1 there are 21 other inequality con- provided as well. In this research work MATLAB®
straints, as follows: (i) 17 constraints of facility toolboxes have been utilized for computation,
capacity (refer to Table 2) and resource variables simulation, 2D and 3D plots.
(refer to Table 4) (ii) 4 constraints by the sales
department (refer to Table 3).
The optimization formulation is,
348
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Table 2. Raw material and Facility usage required (per 103 units) (ãij = [alij, ahij]) and Availability (bj)
349
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Description of the Genetic The stopping criteria determines what causes the
Algorithm Parameter Setting algorithm to terminate. The following options
have been considered in the simulation of the GA.
Fitness function (Equation 2): Generations specifies the maximum number
of iterations the GA performs.
8 2 3 Time Limit specifies the maximum time in sec-
Maximize ∑ (ci xi − di xi − ei xi ) onds the genetic algorithm runs before stopping.
i=1
(2)
Table 4. Raw material availability (bj)
The fitness function is the objective function
that has to be maximized. Material or Facility Availability
Mutation functions make small random Cocoa (kg) 100000
changes in individuals in the population, which Milk (kg) 120000
provide genetic diversity and enable the GA to Nuts (kg) 60000
search a broader space. A step length is chosen Cons. sugar (kg) 200000
along each direction so that linear constraints and Flour (kg) 20000
bounds are satisfied. Alum. foil (ft2) 500000
Crossover combines two individuals, or par-
Paper (ft2) 500000
ents, to form a new individual, or child, for the
Plastic (ft )
2
500000
next generation. The following has been specified
Cooking (ton-hours) 1000
that performs the crossover in the crossover func-
Mixing (ton-hours) 200
tion field. Arithmetic crossover creates children
Forming (ton-hours) 1500
that are the weighted arithmetic mean of the two
Grinding (ton-hours) 200
parents. Children are feasible with respect to linear
Wafer making (ton-hours) 100
constraints and bounds.
Cutting (hours) 400
Packaging 1 (hours) 400
Packaging 2 (hours) 1200
Labour (hours) 1000
350
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
γ x1 x2 x3 x4 x5 x6 x7 x8 f time (s)
0.001 0.69 2.26 0.00 0.00 0.41 2.79 0.00 0.00 559.8 0.012
0.1 0.51 0.00 0.78 0.00 0.00 0.00 0.00 1.31 320.4 0.058
0.2 1.47 0.00 0.00 0.00 0.00 0.04 0.00 0.00 268.3 0.057
0.3 0.35 0.00 0.23 0.15 0.00 0.00 0.25 0.00 161.5 0.056
0.4 0.22 0.00 0.00 0.00 1.78 0.06 0.75 0.00 430.3 0.055
0.5 0.71 0.34 0.56 0.00 1.03 0.00 0.00 0.00 375.0 0.055
0.6 0.53 0.00 0.00 0.47 0.70 0.00 1.06 0.00 441.0 0.055
0.7 0.00 0.51 0.00 0.00 0.79 0.70 0.68 0.00 336.7 0.055
0.8 0.16 0.00 0.00 0.00 1.91 0.27 0.99 0.95 583.0 0.056
0.9 0.76 0.16 0.00 0.00 0.78 0.00 1.38 0.85 250.2 0.055
0.99 1.08 0.80 0.00 0.00 0.78 0.00 1.38 0.85 719.6 0.056
Fitness Limit: If the best value is less than or Function Tolerance: If the cumulative
equal to the value of Fitness Limit, the algorithm change in the fitness function value over Stall
stops. Generations is less than Function Tolerance, the
Stall Generations: If the weighted average algorithm stops.
change in the fitness function value over Stall Among the above options, the final results show
Generations is less than Function Tolerance, the that the Fitness Limit option has been utilized
algorithm stops. successfully in the simulation results.
Stall Time Limit: If there is no improvement Number of Runs: 20, 26, 31 and 95.
in the best fitness value for an interval of time Equation 1 was solved by the using the GA
in seconds specified by Stall Time Limit, the with the following parameters.
algorithms stops.
μ = Degree of possibility
γ x1 x2 x3 x4 x5 x6 x7 x8 f time (s)
0.001 0.18 0.00 0.00 0.00 1.46 1.80 0.00 0.39 379.2 0.420
0.1 0.00 0.11 0.81 0.00 0.00 0.28 0.29 0.00 213.5 0.212
0.2 0.00 0.90 0.00 0.00 0.72 0.30 1.55 0.00 510.6 0.201
0.3 0.00 0.00 0.00 0.93 0.00 0.00 0.00 0.78 131.8 0.200
0.4 0.60 0.94 0.00 0.00 0.08 0.00 0.00 0.00 197.3 0.205
0.5 1.35 2.14 0.17 0.00 0.28 0.39 0.00 0.18 523.7 0.202
0.6 0.00 0.00 0.82 0.00 0.00 0.00 0.00 0.69 182.5 0.233
0.7 0.00 0.32 0.51 0.00 0.10 0.00 0.69 1.88 416.5 0.210
0.8 0.32 0.00 0.03 0.00 0.00 0.48 0.00 0.39 128.9 0.203
0.9 0.05 0.00 0.24 0.72 0.15 0.13 0.00 1.31 236.2 0.207
0.99 1.20 0.00 0.00 0.39 0.00 0.00 2.64 0.00 791.4 0.310
351
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
352
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
353
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Figure 4 and Figure 5 show the simulation carried out by increasing the number of runs by
results for number of runs = 20, population size 31 with population size = 100 and number of
= 100 and generation = 10000. The computed generation = 10000.
results for best optimal value for fitness function, Table 7 reports the best optimal solution for
best feasible solutions for decision variables with fitness function f, and the best feasible decision
respect to γ is reported in Table 6. variables with respect to γ. The total CPU time =
Experimental simulation results for Figure 6, 3.155 seconds and average CPU time = 0.287
total CPU time = 2.603 seconds and average CPU seconds for running the simulation.
time = 0.237 seconds. Further experiments were
354
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
γ x1 x2 x3 x4 x5 x6 x7 x8 f time (s)
0.001 0.63 0.00 1.49 0.00 0.00 0.00 0.00 0.00 340.7 0.220
0.1 0.40 1.44 0.00 0.00 0.00 0.69 0.00 0.00 240.4 0.199
0.2 0.00 0.00 0.31 0.68 0.00 0.00 0.00 0.00 96.20 0.466
0.3 0.56 0.00 0.00 0.00 0.00 0.44 0.88 0.71 375.1 0.208
0.4 0.00 0.00 1.33 0.00 1.20 0.13 0.00 0.53 412.6 0.670
0.5 0.00 1.18 1.76 1.70 0.00 0.00 0.00 0.00 489.2 0.191
0.6 0.00 0.00 0.41 0.60 0.00 0.00 0.02 0.00 110.1 0.195
0.7 0.00 0.58 0.61 0.00 0.95 0.00 0.00 0.00 265.6 0.207
0.8 0.00 0.03 0.00 0.00 1.01 1.63 0.00 0.37 277.6 0.192
0.9 1.36 0.00 0.00 0.62 0.00 0.00 0.00 0.13 299.3 0.404
0.99 1.09 0.47 1.27 1.18 0.21 0.00 0.00 0.46 579.9 0.203
γ x1 x2 x3 x4 x5 x6 x7 x8 f time (s)
0.001 0.62 0.00 0.00 0.00 1.26 0.86 0.00 0.00 336.0 0.194
0.1 0.55 0.00 0.00 0.00 0.49 0.00 0.08 0.00 180.3 0.183
0.2 0.00 0.00 0.00 0.00 0.00 0.00 1.09 1.32 337.1 0.178
0.3 0.00 0.00 0.21 0.00 0.06 0.00 1.69 0.33 418.5 0.179
0.4 0.72 1.60 0.00 0.00 0.18 0.53 0.00 0.30 345.9 0.176
0.5 0.00 0.00 0.00 0.96 0.00 0.00 0.00 0.28 92.00 0.179
0.6 0.00 1.60 0.57 0.32 0.00 0.31 1.34 0.00 545.1 0.180
0.7 1.30 0.00 0.00 0.88 0.58 0.00 0.00 0.57 419.7 0.179
0.8 0.00 0.00 0.00 0.00 0.00 0.00 0.29 0.00 60.40 0.179
0.9 1.31 0.66 0.00 0.00 0.00 0.00 0.37 0.90 444.0 0.177
0.99 1.28 0.00 0.61 1.95 2.26 0.00 2.24 0.00 1222.6 0.177
The best individual solution for the decision A thorough investigation was carried out with
variables is shown in Figure 7 for the experiment following parameter settings in the genetic algo-
with increased population size = 100 and number rithms techniques:
of generation = 10000. Population size = 100; Generation = 10000;
Table 8 reports the best optimal value of fitness Elite Count = 20; Selection = Uniform, roulette
function at γ = 0.99 is 1222.6 and total CPU time and tournament; Fitness scaling = proportional,
= 1.981 seconds with average CPU time = 0.180 top and shift linear; Crossover functions = Inter-
seconds. There is a tremendous improvement in mediate, heuristic, arithmetic, single point, two
the fitness function value when the population point and Mutation function = Adaptive feasible.
size is increased to 100. Table 9 reports very important information on
the simulation results obtained for α = 1 to α =
25 by independent runs. The total CPU time (s)
355
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
and an average CPU time (s) for each α from 1 Figure10 shows the solution for a run number
to 41 with respect to the range γ = 0.001 to γ = of 26. In this case, the number of generations used
0.99 provided in the last two rows of the Table 9. is 1000000 and the population size is 100.
Unfortunately, there is no solution for the fitness Figure 11 shows the optimal results for α
function for α = 45 at γ = 0.99. This is the draw values from 39 to 57. The optimal value occurs
back of this GA algorithm for the higher value of at α = 41 and γ = 0.5 with fitness value 1302.59.
α in an independent run. The CPU time for this solution is 0.056 seconds.
A 3D plot for the fitness value, vagueness The simulation was carried out for 95 runs and
factor and level of satisfaction for α = 1 to α = 41 the following result obtained by using a popula-
is shown in Figure 8. The result obtained for α = tion size of 200 and the number of generations
45 at γ = 0.99 motivates to further investigate this equal to 1,000,000.
simulation with a single run for α = 1 to α = 41. The number of runs carried out in this simula-
Figure 9 shows a 3D solution obtained with a tion results for the optimal fitness function value
single run of the GA for α = 1 to α = 41. The are 20, 26, 31 and 95. This is the largest number
optimal value for the fitness function is 1086 at of runs that have been considered in these simu-
α = 41 and γ = 0.99. The solution for the problem lation results.
is further investigated for α values between 39 The optimal value for the fitness function is
and 59. Figure 9 was obtained for the solution of 1302.59, obtained at α = 41 and γ = 0.5 and the
fitness function f. computational time is 0.055 seconds.
Population size of 2000 and number of gen- The simulation was run for α = 1 to α = 41 and
eration of 100000 with arithmetic cross over and the results for the fitness function is 1302.59 at
adaptive feasible mutation have been used in the α = 21 and γ = 0.9. The computational time for
above result. The optimal solution for the fitness running this solution is 0.056. Figure 12, Figure
function is 925.53 at α = 41 and γ = 0.2. Compu- 13, Figure 14, and Figure 15 give complete views
tational time for this solution is 0.148 seconds. of the various dimensions of the solutions.
356
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
The contour in Figure 16 clearly explains the fitness limit reached or an optimal solution
situation where by the optimal solution for the is located during the genetic search pro-
fitness function is stuck at 1221 for α = 39 and α cess, then stop; otherwise, go to Step 2.
= 59. In order to to escape from this solution, • Step 5: Continue with general pattern
another hybrid method is needed to further improve search techniques (Vasant & Barsoum,
the final optimal solution. 2009).
• Step 6: End
Results and Discussion for Hybrid
Genetic Algorithm Pattern Search Figure 17 provides the best optimal solution
Approach for the fitness function with respect to a level of
satisfaction of α = 13.813.
The hybrid GA and general pattern search is a Table 10 provides the solution for the best
combination of a GA and a general pattern search. fitness function and the decision variables at α =
Below is the algorithm for this approach. 13.813. The CPU time to run the simulation for
α = 13.813 is 16 minutes. The best optimal solu-
• Step 1: Initial population: Generate initial tion for the fitness function is 180776.1 at γ = 0.1.
population randomly. Figure 18. explains the detailed solution for
• Step 2: Genetic operators: Selection: sto- the decision variables versus level of satisfaction.
chastic uniform, Crossover: arithmetic The CPU time for α = 13.813 at γ = 0.99 is 78.375
crossover operator Mutation, Adaptive fea- seconds.
sible mutation operator The solution for the decision variable x8 is
• Step 3: Evaluation: Best offspring and fit- unrealistic since for the majority values of γ, x8 is
ness function. zero. Therefore, this solution needs some improve-
• Step 4: Stop condition: If a pre-defined ment. For this purpose, the following hybrid
maximum generation number, time limit, methods is thoroughly investigated.
357
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
358
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Fitness values for various values of α from 1 to The main contribution of the hybrid HGAPS
41 are given in Figure 19. The optimal value for approach is that it is able to locate the best near
the fitness function occurs at α = 9 and γ = 0.99 is global optimal solution for the fitness function
197980.93. The total CPU time for the simulation which is far superior to the GA alone. On the
of Figure 19 is 2 hours 18 minutes. other hand, it is unable to provide a reasonable
Figure 20 provides a holistic view of the 3D computational CPU time for the fitness function
mesh plot solution for various values of α, and γ values.
with respect to the level of satisfaction. The CPU
time for the 3D mesh plot is 2 hours 52 minutes.
359
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
360
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
γ x1 x2 x3 x4 x5 x6 x7 x8 f
0.001 0.62 0.80 0.94 0.00 0.21 0.24 0.00 0.00 366.4
0.1 346.7 580.5 276.2 460.4 148.7 247.9 189.2 4.54 180776.1
0.2 0.43 0.90 0.73 0.58 0.04 0.68 0.57 0.00 475.5
0.3 314.8 577.5 291.0 485.7 104.2 173.7 175.9 5.34 173329.7
0.4 0.00 0.00 0.00 0.61 0.51 1.69 0.59 0.00 351.2
0.5 278.2 466.2 282.0 521.1 156.5 260.8 164.2 0.00 170805.1
0.6 280.6 497.6 269.3 493.6 146.5 244.2 161.4 0.00 169598.1
0.7 258.1 602.6 260.3 472.7 128.5 214.2 153.6 0.00 166739.9
0.8 305.2 512.4 238.8 398.0 136.8 228.1 159.4 0.00 166566.4
0.9 257.7 444.3 269.2 450.5 146.1 243.6 148.2 0.00 163208.3
0.99 283.8 566.7 153.1 255.2 106.4 177.3 123.5 0.00 144803.8
361
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
To overcome these major drawbacks, a hybrid hybridize technique HGAPS in order to find a
approach has been adopted in this study for the global optimal solution for the objective func-
global optimal solutions. The major drawback of tion. The final optimal solution for the objective
the GA algorithms is that is not able to provide a function is very promising and is superior to the
global optimal solution for the fitness function. GA approach alone. The major goal of this paper
This is possibly due to premature convergence. is finally achieved successfully.
Further investigation has been carried out by the
362
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
363
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Deb, K. (2001). Multi-objective optimization using Lin, F. T. (March 2007). Personal communication.
evolutionary algorithms. New York, NY: Wiley. Maric, M. (2010). An efficient genetic algorithms
Eldredge, N. (1989). Macro-evolutionary dynam- for solving the multi level incapacitated facility
ics: Species, niches, and adaptive peaks. New location problem. Computing and Informatics,
York, NY: McGraw-Hill. 29, 183–201.
Ertuğrul, I., & Karakaşoğlu, N. (2006). Fuzzy MATLAB. (2007). User’s guide. The Math Works.
TOPSIS method for academic member selection Parmee, I. C. (2003). Poor-definition, uncertainty
in engineering faculty. International Joint Con- and human factors—A case for interactive evolu-
ferences on Computer, Information, and Systems tionary problem reformulation. Proceedings of the
Sciences, and Engineering (CIS2E 06). 3rd IEC Workshop of the Genetic and Evolutionary
Goldberg, D. E. (1989). Genetic algorithms in Computation Conference (GECCO).
search optimization and machine learning. To-
ronto, Canada: Addison Wesley.
364
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Peidro, D., Mula, J., & Poler, R. (2010). Fuzzy Turabieh, H., Sheta, A., & Vasant, P. (2007).
linear programming for supply chain planning Hybrid optimization genetic algorithm (HOGA)
under uncertainty. International Journal of In- with interactive evolution to solve constraint
formation Technology & Decision Making, 9, optimization problems for production systems.
373–392. doi:10.1142/S0219622010003865 International Journal of Computational Science,
1(4), 395–406.
Rennard, J. P. (2000). Introduction to genetic algo-
rithms. Retrieved May 12, 2008, from http://www. Vasant, P. (2003). Application of fuzzy linear
rennard.org/alife /english /gavintrgb.html#Evol programming in production planning. Fuzzy
Optimization and Decision Making, 3, 229–241.
Sadeh-Koniecpol, N., Hildum, D., Laliberty,
doi:10.1023/A:1025094504415
T. J., Smith, S., McA’Nulty, J., & Kjenstad, D.
(1996). An integrated process-planning/produc- Vasant, P. (2006). Fuzzy production planning and
tion-scheduling shell for agile manufacturing. its application to decision making. Journal of Intel-
(Technical Report CMU-RI-TR-96-10), Robotics ligent Manufacturing, 15(1), 5–12. doi:10.1007/
Institute, Carnegie Mellon University. s10845-005-5509-x
Sanchez, G., Jimenez, F., & Vasant, P. (2007). Vasant, P. (2011). Hybrid optimization for deci-
Fuzzy optimization with multi-objective evolu- sion making in an uncertain environment. LAP
tionary algorithms: A case study. Proceedings of LAMBERT Academic Publishing, 244 pages,
the 2007 IEEE Symposium on Computational Germany.
Intelligence in Multi-criteria Decision Making
Vasant, P., & Barsoum, N. N. (2006). Fuzzy
(pp. 58-64). Honolulu, Hawaii.
optimization of units products in mix-products
Stanimrovic, Z. (2010). A genetic algorithm ap- selection problem using FLP approach. Soft
proach for the capacitated single allocation P-Hub Computing Journal, 10(2), 144–151. doi:10.1007/
median problem. Computing and Informatics, s00500-004-0437-9
29, 117–132.
Vasant, P., & Barsoum, N. N. (2009). Hybrid
Swaminathan, J. M., Smith, S., & Sadeh-Koniecpol, general pattern search and simulated annealing
N. (1998). Modeling the dynamics of supply for industrial production planning problems. 3rd
chains: A multi-agent approach. Decision Scienc- Global Conference on Power Control and Optimi-
es, 29, 607–632. doi:10.1111/j.1540-5915.1998. zation, 2-4th February 2010, Gold Coast, Australia.
tb01356.x
Zadeh, L. A. (1965). Fuzzy sets. Information
Tabucanon, M. T. (1996). Multi objective pro- and Control, 8, 338–353. doi:10.1016/S0019-
gramming for industrial engineers. Mathemati- 9958(65)90241-X
cal programming for industrial engineers (pp.
Zadeh, L. A. (1975). The concept of a linguis-
487–542). New York, NY: Marcel Dekker, Inc.
tic variable and its application to approximate
Takagi, H. (2001). Interactive evolutionary com- reasoning. Information Sciences, 8, 199–249.
putation: Fusion of the capabilities of EC compu- doi:10.1016/0020-0255(75)90036-5
tation and human evaluations. Proceedings of the
IEEE, 89(9), 1275–1296. doi:10.1109/5.949485
365
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Zamirian, M., Kamyad, A. V., & Farahi, M. H. Buckley, J. J., & Feuring, T. (2000). Evolutionary
(2009). A novel algorithm for solving optimal algorithm solution to fuzzy problems: Fuzzy linear
path planning problems based on parametrization programming. Fuzzy Sets and Systems, 109, 35–53.
method and fuzzy aggregation. Physics Letters. doi:10.1016/S0165-0114(98)00022-0
[Part A], 373, 3439–3449. doi:10.1016/j.phys-
Elamvazuthi, I., Vasant, P., & Ganesan, T. (2010).
leta.2009.07.018
Fuzzy linear programming using modified logistic
Zimmermann, H. J. (1992). Fuzzy set theory and membership function. International Review of
its applications. Boston, MA: Kluwer Academic Automatic Control, 3(4), 370–377.
Publishers.
Jen, M., Ida, K., Lee, J., & Kim, J. (1997). Fuzzy
nonlinear goal programming using genetic algo-
rithm. Computers & Industrial Engineering, 33,
ADDITIONAL READING 39–42. doi:10.1016/S0360-8352(97)00036-3
Abraham, A., Vasant, P., & Bhattacharya, A. Jimenez, F., Cadenas, J. M., Verdegay, J. L., &
(2008). Neuro-Fuzzy Approximation of Multi- Sanchez, G. (2003). Solving fuzzy optimization
Criteria Decision-Making QFD Methodology. problems by evolutionary algorithms. Information
In Kahraman, C. (Ed.), Fuzzy Multi-Criteria Sciences, 152, 303–311. doi:10.1016/S0020-
Decision Making Theory and Applications with 0255(03)00074-4
Recent Developments (pp. 301–324). Springer. Jimenez, F., Sanchez, G., & Vasant, P. (2008).
doi:10.1007/978-0-387-76813-7_12 Fuzzy Optimization via Multi-Objective Evolu-
Aliev, R. A., Fazlollahi, B., Guirimov, B. G., & tionary Computation for Chocolate Manufactur-
Aliev, R. R. (2007). Fuzzy-genetic approach to ing. In Kahraman, C. (Ed.), Fuzzy Multi-Criteria
aggregate production-distribution planning in Decision Making Theory and Applications with
supply chain management. Information Sciences, Recent Developments (pp. 523–538). Springer.
177, 4241–4255. doi:10.1016/j.ins.2007.04.012 doi:10.1007/978-0-387-76813-7_20
Bhattacharya, A., Abraham, A., & Vasant, P. Liang, T. F. (2008). Interactive multi-objective
(2008). FMS Selection Under Disparate Level- transportation planning decisions using fuzzy
of-Satisfaction of Decision Making Using an linear programming. Asia Pacific Journal of
Intelligent Fuzzy-MCDM- Model. In Kahraman, Operational Research, 25, 11–31. doi:10.1142/
C. (Ed.), Fuzzy Multi-Criteria Decision Making S0217595908001602
Theory and Applications with Recent Develop- Madronero, M. D., Peidro, D., & Vasant, P.Vendor
ments (pp. 263–280). Springer. doi:10.1007/978- selection problem by using an interactive fuzzy
0-387-76813-7_10 multi-objective approach with modified s-curve
Bhattacharya, A., Vasant, P., & Susanto, S. (2007). membership functions. Computers & Mathemat-
Simulating theory of constraint problem with a ics with Applications (Oxford, England), 60,
novel fuzzy compromise linear programming 1038–1048. doi:10.1016/j.camwa.2010.03.060
model. In A. Elsheikh, A. T. Al Ajeeli, & E. M. Modal, S., & Maiti, M. (2002). Multi-item fuzzy
Abu-Taieh. Simulation and Modeling: Current EOQ models using genetic algorithm. Comput-
Technologies and Applications (pp. 307-336). ers & Industrial Engineering, 44, 105–117.
IGI Publisher. doi:10.1016/S0360-8352(02)00187-0
366
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
Peidro, D., Mula, J., & Poler, R. (2010). Fuzzy Vasant, P., Barsoum, N., Kahraman, C., & Di-
linear programming for supply chain planning mirovski, G. (2007). Application of fuzzy optimi-
under uncertainty. International Journal of In- zation in forecasting and planning of construction
formation Technology & Decision Making, 9, industry. In Vrakas, D., & Vlahavas, I. (Eds.),
373–392. doi:10.1142/S0219622010003865 Artificial Intelligent for Advanced Problem Solv-
ing Technique. IGI publisher.
Sasaki, M., & Gen, M. (2003). Fuzzy multiple
objective optimal system design by hybrid ge- Vasant, P., Bhattacharya, A., & Abraham, A.
netic algorithm. Applied Soft Computing, 2(3F), (2008). Measurement of Level-of-Satisfaction
189–196. doi:10.1016/S1568-4946(02)00068-6 of Decision Maker in Intelligent Fuzzy-MCDM
Theory: A Generalized Approach. In Kahraman,
Varela, L. R., & Ribeiro, R. A. (2003). Evaluation
C. (Ed.), Fuzzy Multi-Criteria Decision Making
of simulated annealing to solve fuzzy optimiza-
Theory and Applications with Recent Develop-
tion problems. Journal of Intelligent & Fuzzy
ments (pp. 235–262). Springer. doi:10.1007/978-
Systems, 14, 59–71.
0-387-76813-7_9
Vasant, P. (2010). Hybrid simulated annealing
Vasant, P., Elamvazuthi, I., & Webb, J. F. (2010).
and genetic algorithms for industrial production
Fuzzy technique for optimization of objec-
management problems. International Journal
tive function with uncertain resource variables
of Computational Methods, 7(2), 279–297.
and technological coefficients. International
doi:10.1142/S0219876210002209
Journal of Modeling, Simulation, and Scien-
Vasant, P. (2010). Innovative hybrid genetic tific Computing, 1(3), 349–367. doi:10.1142/
algorithms and line search method for indus- S1793962310000225
trial production management. In Chis, M. (Ed.),
Vasant, P., & Kale, H. (2007). Introduction to
Computation and Optimization Algorithms in
fuzzy logic and fuzzy linear programming. In A.
Software Engineering: Application and Tech-
Frederick & P. Humphreys (Eds.), Encyclopedia
niques (pp. 142–160). Hershey, PA: IGI Global.
of Decision Making and Decision Support Tech-
doi:10.4018/978-1-61520-809-8.ch008
nologies. IGI publisher.
Vasant, P. (2010). Hybrid optimization techniques
Yang, L., Jones, B. F., & Yang, S. H. (2007).
for industrial production planning. Journal of
A fuzzy multi-objective programming for op-
Computer Science & Technology, 10(3), 150–151.
timization of the fire station locations through
Vasant, P., & Barsoum, N. (2009). Hybrid genetic genetic algorithms. European Journal of Op-
algorithms and line search method for indus- erational Research, 181, 903–915. doi:10.1016/j.
trial production planning with non-linear fitness ejor.2006.07.003
function. Engineering Applications of Artificial
Zamirian, M., Kamyad, A. V., & Farahi, M. H.
Intelligence, 22, 767–777. doi:10.1016/j.engap-
(2009). A novel algorithm for solving optimal
pai.2009.03.010
path planning problems based on parametrization
Vasant, P., & Barsoum, N. (2010). Hybrid pattern method and fuzzy aggregation. Physics Letters.
search and simulated annealing for fuzzy produc- [Part A], 373, 3439–3449. doi:10.1016/j.phys-
tion planning problem. Computers & Mathemat- leta.2009.07.018
ics with Applications (Oxford, England), 60,
1058–1067. doi:10.1016/j.camwa.2010.03.063
367
Solving Fuzzy Optimization Problems of Uncertain Technological Coefficients with Genetic Algorithms
368
369
Pandian Vasant was born in Sungai Petani, Malaysia in 1961. Currently, he is a Senior Lecturer of
Engineering Mathematics for Electrical & Electronics Engineering Program and Fundamental & Applied
Sciences Department at University Technology Petronas in Tronoh, Perak, Mlaysia. He has graduated
in 1986 from University of Malaya (MY) in Kuala Lumpur, obtaining his BSc Degree with Honors (II
Class Upper) in Mathematics, and in 1988 also obtained a Diploma in English for Business from Cam-
bridge Tutorial College, Cambridge, England. In the year 2002 he has obtained his MSc (By Research)
in Engineering Mathematics from the School of Engineering & Information Technology of University
of Malaysia Sabah, Malaysia, and has a Doctoral Degree (2008) from University Putra Malaysia in
Malaysia. After graduation, during 1987-88 he was Tutor in operational research at University Science
Malaysia in Alor Setar, Kedah and during 1989-95 he was Tutor of Engineering Mathematics at the same
university but with Engineering Campus at Tronoh, Perak. There after during 1996-2003 he became a
lecturer in Advanced Calculus and Engineering Mathematics at Mara University of Technology, in Kota
Kinabalu. He became Senior Lecturer of Engineering Mathematics in American Degree Program at
Nilai International College (Malaysia), during 2003-2004 before taking his present position at Univer-
sity Teknologi Petronas in Tronoh. His main research interests are in the areas of Optimization Methods
and Applications to Decision Making and Industrial Engineering, Fuzzy Optimization, Computational
Intelligence, and Hybrid Soft Computing. Vasant has co-authored 200 research papers and articles in
national journals, international journals, conference proceedings, conference paper presentation, and
special issue guest editor, lead guest editor for book chapters’ project, conference abstract, edited book
and book chapters. In the year 2009, Vasant was awarded top reviewer for the journal Applied Soft
Computing (Elsevier). He has been Co-editor for AIP Conference Proceedings of PCO (Power Control
and Optimization) conferences since 2008 and editorial board member of international journals in the
area of Soft Computing, Optimization and Computer Applications. Currently he’s a lead managing
editor for GJTO (Global Journal Technology & Optimization) and organizing committee member (PCO
Global) for PCO Global conferences.
Nader Barsoum obtained his PhD from University of Newcastle upon Tyne, UK, 1989 in Eclectic
Machines and Drives, and his first three degrees from Alexandria University, Egypt in Power Engineer-
ing 1976, Applied Mathematics 1979 and Engineering Mathematics 1983. He has academically jointed
University Sains Malaysia, University Malaysia Sabah, University of Western Sydney, and Curtin Uni-
versity of Technology in the past 20 years. He obtained School Research and Developments Excellence
Award 2010, Highest Research Performance Incentive Award 2010, Best Researcher Award 2009, ANAK
Sarawak Award 2009, and University Encouragement Award 1992. He completed several funded projects
About the Contributors
such as: Implementation of an Optimal Energy Saving System with solar tracker 2010, Development
of Smart Drive Build in AC Motors for Optimal Operations 2010, Investigation of Electrical Properties
of Palm Oil Using Kerr-effect Technique 2009, Hardware design of body sensor device 2008, Low cost
hydrogen fuel cell for hybrid renewable energy system 2006, and Simulation of Micro-Controller for
Induction Motor Drive 1995. Besides supervising many Master and PhD degrees, he selected several
times as an external examiner and published over 86 papers in both international scientific journals
and conferences. He is the President of Malaysian Solar Energy Society since 2010, Editor In Chief of
Global Journal Of Technology and Optimization since 2010, General Chairman of an Annual Global
Conference on Power Control and Optimization since 2008, and Guest Editor of International Journal
Of Computers & Mathematics With Applications.
Jeffrey Webb has a Ph.D. degree in photothermal physics from Strathclyde University, Scotland,
U.K. He is now an Associate Professor in the School of Engineering, Computing and Science, Swin-
burne University of Technology, Sarawak (Malaysia). His research experience is in the areas of optics,
electromagnetic waves, ferroelectric materials, and nanoelectronics.
***
Muhammad Mansoor Asghar did his BSc Electrical Engineering from University of Engineering
& Technology, Lahore, Pakistan in 2009. He is also student in MSc Electrical Engineering. Currently
he is serving as Lecturer/Lab Engineer in Rachna College of Engineering & Technology, Gujranwala,
Pakistan. He has published more than five international research papers. His research interests include
power system analysis, renewable technologies, power electronics, et cetera.
Adnan Bashir did his BSc and MSc Electrical Engineering from University of Engineering & Tech-
nology, Lahore, Pakistan in 2006 and 2009, respectively. Currently he is serving as Lecturer in Rachna
College of Engineering & Technology, Gujranwala, Pakistan. He has published more than three inter-
national research papers. His research interests include power system analysis, renewable technologies,
power electronics, et cetera.
Roman V. Bershanskiy (M’ 2002) graduated from Moscow Power Institute as an engineer-electrician
in 2002 and since that time has been working with Joint Stock Company “System Operator of Unified
Energy System”, JSC, SO UES, Moscow, Russia. Now he is a postgraduate student at Energy Systems
Institute, Russian Academy of Sciences, Irkutsk. His scientific interests are connected with real-time
control problems, using FACTS and WAMS for control of power systems, detection of weak nodes,
and cut sets.
Hassan Bevrani received the M.Eng. (Hons.) degree from K. N. Toosi University of Technology,
Tehran, Iran, in 1997, and the Ph.D. degree from Osaka University, Osaka, Japan, in 2004, both in electri-
cal engineering. From 2004 to 2006, he was a Postdoctoral Fellow at Kumamoto University, Kumamoto,
Japan. From 2007 to 2008, he was a Senior Research Fellow at Queensland University of Technology,
Brisbane, Qld., Australia. From 2000, he has been an academic member of University of Kurdistan. Cur-
rently, he is a Professor in Kumamoto University (Japan). His special fields of interest include intelligent
370
About the Contributors
and robust control applications in power system and power electronic industry. Prof. Bevrani is a senior
member of Institute of Electrical and Electronics Engineers (IEEE), and a member of the Institute of
Electrical Engineers of Japan (IEEJ) and the Institution of Engineering and Technology (IET).
Selcuk Cebi received his MSc degree on mechanical engineering from Karadeniz Technical University
in 2004 and he received his PhD degree on industrial engineering from Istanbul Technical University
in 2010. His academic career started in Karadeniz Technical University in 2002 as a research assistant
until 2004. Then, he worked at Istanbul Technical University as a research assistant between 2005 and
2010. Now, he is working at Karadeniz Technical University as an Assistant Professor. Dr. Cebi has
interests in decision support systems, multiple criteria decision making, and design topics, and he has
over 20 publications in SCI indexed journals during his academic career.
Muhammad Ali Masood Cheema was born in Gujranwala, Pakistan, in 1984. He did his BSc and
MSc Electrical (Power) Engineering with Honours from University of Engineering & Technology,
Lahore, Pakistan in 2008 and 2010, respectively. He was awarded two Gold Medals during his BSc in
Electrical Engineering. He served in Pakistan’s largest transformer manufacturing company, PEL, for
one year as Design Engineer. Currently he is serving as Lecturer in Rachna College of Engineering &
Technology, Gujranwala, Pakistan. He has published more than twelve international research papers
including IEEE PES, Springer Journal, and Elsevier Journal. His research interests include power qual-
ity, electric machine design, control systems, power system analysis, distributed generation and load
modeling, et cetera.
Vo Ngoc Dieu received his B.Eng. and M.Eng. degrees in electrical engineering from Ho Chi Minh
City University of Technology, Ho Chi Minh city, Vietnam, in 1995 and 1999, respectively, and his
D.Eng. degree in energy from Asian Institute of Technology (AIT), Pathumthani, Thailand in 2007. He
is Research Associate at AIT and lecturer at Department of Power Systems, Faculty of Electrical and
Electronic Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh City, Vietnam. His
interest is applications of AI in power system optimization.
Dmitry N. Efimov is a senior researcher in the Energy Systems Institute of the Russian Academy
of Science (ESI SB RAS), Irkutsk, Russia. He was born in 1961 in Irkutsk. He graduated from Irkutsk
State Technical University in 1987. Since then he has been with ESI SB RAS. He received his degree
of Candidate of Technical Sciences from the ESI SB RAS in 1998. From 1998 till 2000, D.N.Efimov
was an Executive Director of the International Research Energy Laboratory, Budapest, Hungary. His
research interests include simulation of transients in electric power systems, development, operation,
and dynamic properties of the large power interconnections.
Alexander Z. Gamm graduated from the Electrotechnical Institute, Novosibirsk, Russia, where he
received the Ph.D and D.Sc. degrees. During 1961-1962, he was a scientific worker in the Transport
Power Institute, Novosibirsk. Since then, he has been with the Energy System Institute (ESI), Russian
Academy of Sciences, Irkutsk. At present, he is a Chief Researcher. His special fields of interest include
state estimation, optimization, and real-time control problems. Dr. Gamm is an Academician of the Rus-
sian Academy of Electrical Sciences and member of International Energy Academy.
371
About the Contributors
Anna M. Glazunova graduated from the Irkutsk Polytechnical Institute in 1982. Since 1986 she has
been working as a researcher at Energy Systems Institute. She received her PhD in 2002. Her research
interest is methods of state estimation of electricity use.
Irina I. Golub (M’1995, SM’1997) graduated from Moscow Power Institute as an engineer-elec-
trician. She has been working at Siberian Energy Institute, Irkutsk since 1972. Her scientific interests
are connected with real-time control problems, especially in the field of observability of electric power
systems and allocation of measurements of operating parameters of electric networks. She is a leading
researcher, University Professor, and Doctor of technical sciences.
U. Häger received his diploma degree in Electrical Engineering in 2006 from the Univ. of Dortmund,
Germany. He is working at Technische Universität Dortmund, Institute of Power Systems and Power
Economics, on his doctoral degree in the field of wide area congestion management by use of fast power
flow controllers.
Mahmoud-Reza Haghifam received his B.Sc., M.Sc., and Ph.D. degrees in 1988, 1990, and 1995,
respectively, all in electric power engineering. Currently, he is the Full Professor of electrical engineering
at Tarbiat Modares University, Tehran, Iran and worked on Electric Distribution Systems’ Reconfiguration
during several essays and projects. He is a senior member of IEEE and research fellow of Alexander Van
Humboldt, Germany. His research interest includes electric distribution analysis and planning, electric
distribution system operation and automation, power system analysis and planning, development of soft
computing and bio-inspired computing techniques for power system planning, power system reliability
and reactive power control.
Muhammad Junaid was born in Gujranwala, Pakistan, in 1981. He did his BSc, MSc Electrical
(Power) Engineering with Honours from University of Engineering & Technology, Lahore, Pakistan
in 2008 and 2010 respectively. Currently he is serving as Lecturer in Rachna College of Engineering
& Technology (RCET), Gujranwala. He has published more than fifteen international research papers
including IEEE PES, Springer Journal, and Elsevier Journal. He is also author of International Book
Chapter titled “Harmonics Modelling & Simulation.” During 4th Global Conference on Power Control
& Optimization held in Malaysia, he was declared Best Presenter. Keeping in view his valuable research
contributions, RCET has given him Best Research award. His research interests include power quality,
power system analysis, power system operation & control, power electronics, et cetera.
Cengiz Kahraman received his MSc and PhD degrees in industrial engineering from Istanbul Technical
University. He is in the editorial board of some journals such as Journal of Human and Ecological Risk
Assessment, International Journal of Computational Intelligence Systems, New Mathematics and Natural
Computation, Journal of Enterprise Information Management, Technological and Economic Development
of Economy, Baltic Journal on Sustainability, International Journal of Strategic Property Management,
Journal of Multiple Valued Logic and Soft Computing, and he has made the guest editorships of special
issues of some international journals such as Information Sciences, Journal of Enterprise Information
Management, and International Journal Of Approximate Reasoning. He is the editor of such Springer
books as “Fuzzy Applications in Industrial Engineering,” “Fuzzy Multicriteria Decision Making: Theory
and Applications with Recent Developments,” and “Fuzzy Engineering Economics with Applications.”
372
About the Contributors
Fouad Kamel is a Senior Lecturer at the University of Southern Queensland in Toowoomba, Faculty
of Engineering and Surveying, Department of Electrical and Computer Engineering since February 2008.
He graduated Diploma Engineer and PhD in photovoltaic systems from Hanover University in Germany
1984, Dr. Fouad worked as a Lecturer and Associate Professor at the Suez Canal University in Egypt
during 1985-1999. In 1999, he moved to New Zealand and worked there between 2000 and 2007 for
tertiary education and research at Christchurch Polytechnic Institute of Technology and the Southern
Institute of Technology. Dr. Fouad has a history of publications in different renewable energy areas.
Rana Abdul Jabbar Khan graduated from University of Engineering and Technology, Lahore,
Pakistan in 1991. He completed ME from University of Wollongong, Australia in 1995 and PhD from
RMIT University Melbourne, Australia in 2003. He has been serving as SDO and XEN in Water and
Power Development Authority (WAPDA), largest organization of Asia, responsible for generation,
transmission, and distribution of electricity country wide. Currently he is on deputation and serving
Rachna College of Engineering and Technology (RCET) as Principal since 2005. He is principal author
of an open access book chapter, titled “Harmonics Modelling and Simulation”. He is a senior member
of IEEE, IEAust, IEE, and PEC. He has published several papers in IEEE journals. Honoring his valu-
able research contribution he has been nominated for Presidential Award 2009 by the government of
Pakistan. His research interest includes power quality, power system simulation, renewable technologies
and deregulation of electricity, et cetera.
İhsan Kaya is currently an Assistant Professor in Yıldız Technical University Department of Industrial
Engineering. He received his MSc and PhD degrees on industrial engineering from Selçuk University
and Istanbul Technical University, respectively. His main research areas are fuzzy set theory, process
capability analyses, quality control, decision making, and fuzzy statistics.
Irina N. Kolosok graduated from St. Petersburg Technical University. Since 1972 she has been with
Energy Systems Institute (ESI), Russian Academy of Sciences, currently as a leading researcher. She
received the PhD (1986) and DSc (2004) degrees. Her scientific interests are: real-time control prob-
lems, especially in the field of state estimation of electric power systems (EPS), SCADA systems, and
application of AI-methods for on-line EPS control.
Victor G. Kurbatsky (M’08) was born on May, 27th, 1949 in Russia. He is PhD, Professor, and
Doctor of Science. He is Leading Researcher at the Energy Systems Institute of the Russian Academy
of Sciences, Irkutsk, Russia. Victor Kurbatsky received his degree of Candidate of Technical Sciences
at SibNIIE (Novosibirsk) in 1984 and Doctor of Technical Sciences at the Energy Systems Institute (Ir-
kutsk) in 1997. His research interests include: electromagnetic compatibility and power quality in electric
networks and application of artificial intelligence techniques in power systems. Professor Kurbatsky is
the author of several monographs and manuals and more than 280 scientific papers.
373
About the Contributors
Gerard Ledwich received the Ph.D. degree in Electrical Engineering from the University of Newcastle,
Newcastle, Australia, in 1976. He has been the Chair Professor in Power Engineering at Queensland
University of Technology, Brisbane, Australia, since 2006. Previously, he was the Chair in Electrical
Asset Management from 1998 to 2005 at the same university. He was the Head of Electrical Engineering
at the University of Newcastle from 1997 to 1998. Previously, he was associated with the University of
Queensland from 1976 to 1994. His interests are in the areas of power systems, power electronics, and
controls. Prof. Ledwich is a Fellow of the Institution of Engineers Australia (I.E.Aust.).
Igor Litvinchev received his M.Sc. degree from Moscow Institute of Physics and Technology (Fizteh),
Moscow, Russia, and Ph.D. and Dr.Sci. degrees in systems modelling and optimisation from Computing
Centre, Russian Academy of Sciences, Moscow. He has held visiting positions at universities in Brazil,
Mexico, as well as positions at various universities and research centres in Russia. His research focuses
on large-scale system modelling, optimisation, and control. Dr. Litvinchev is a member of Russian
Academy of Natural Sciences and Mexican Academy of Sciences.
Marwan Marwan received the B.Eng. degree from Hasanuddin University, Makassar, Indonesia,
and the M.Eng. degree from the Queensland University of Technology, Brisbane, Australia, in 2000
and 2006, respectively, all in electrical engineering. He is working at The State Polytechnic of Ujung
Pandang Makassar Indonesia as a Lecturer in Energy Conversion Department from 2001 to the present.
Presently, he is preparing Ph.D. at the University of Southern Queensland in Toowoomba, Australia.
Armin Ebrahimi Milani received his B.Sc. degree (Electric Power Engineering) from Islamic Azad
University, South Branch, Tehran, Iran, in 2006 and M.Sc. degree in Electric Power Engineering from
Islamic Azad University, Science and Research Branch, Tehran, Iran, in 2010. He worked his thesis on
Electric Distribution Systems’ Reconfiguration. Currently, he is with department of Electrical Engineering,
Azad University, Iran. He is a member of IEEE and senior member of Young Researchers Club, Tehran,
Iran. His research interest includes power system optimization and control, electric distribution systems,
power system restructuring, development of soft computing and bio-inspired computing techniques for
power system planning, operation, and control applications.
J. Zambujal-Oliveira received M.S. degrees in Economics at New University of Lisbon (FE) and in
Management Science at Technical University of Lisbon (ISEG). In 2007, he received a Ph.D. degree in
Management Science from the Technical University of Lisbon (ISEG). Before coming to the Department
of Engineering and Management at Technical University of Lisbon (IST), Prof. J. Zambujal-Oliveira
was employed as a Systems Engineer at banks, insurance companies and consulting firms, and as an
Assistant Professor of management at the University of Madeira in Funchal. Prof. J. Zambujal-Oliveira’s
research interests include real options analysis and international taxation. In particular, he enjoys work-
ing with graduate students and colleagues in developing, analyzing, and testing numerical methods for
solving stochastic differential equations, developing and analyzing models in transfer prices systems
and developing, analyzing, and testing computational methods in mathematical finance.
Abdeen Mustafa Omer (BSc, MSc, PhD) is a qualified Mechanical Engineer with a proven track
record within the water industry and renewable energy technologies. He has been graduated from
374
About the Contributors
University of El Menoufia, Egypt, BSc in Mechanical Engineering. His previous experience involved
being a member of the research team at the National Council for Research/Energy Research Institute in
Sudan and working director of research and development for National Water Equipment Manufactur-
ing Co. Ltd., Sudan. He has been listed in the WHO’S WHO in the World 2005, 2006, and 2007. He
has published over 200 papers in peer-reviewed journals, 50 review articles, and 40 chapters in books.
Weerakorn Ongsakul received his B.Eng. degree in electrical engineering from Chulalongkorn
University, Bangkok, Thailand, in 1988, and his M.S. and Ph.D. degrees in electrical engineering from
Texas A&M University, College Station, in 1991 and 1994, respectively. Currently, he is an Associate
Professor of Energy Field of Study, and Dean of School of Environment, Resources and Development
at Asian Institute of Technology (AIT), Pathumthani, Thailand. His interests are in computer applica-
tions to power systems, parallel processing applications, AI applications in power systems, and power
system deregulation.
Alexei S. Paltsev graduated from Irkutsk State Technical University specializing in computer science
in 2005. Since 2000 he has been with the Energy Systems Institute (ESI), Russian Academy of Sciences,
currently as a leading Engineer. He finished his postgraduate courses at ESI in 2008. His scientific in-
terests are multi-agent systems and EPS state estimation.
Daniil A. Panasetsky (SM’07) is a postgraduate student at the Energy Systems Institute m of the
Russian Academy of Sciences, Irkutsk, Russia. He graduated with honors from Irkutsk State Technical
University specializing in electrical engineering in 2006. His research interests include stability analy-
sis of power systems, emergency control, FACTS devices, and application of artificial intelligence to
power systems.
Dmitri B. Popov is a leading engineer of Energy Systems Institute (Siberian Energy Institute until
1997) of the Russian Academy of Sciences, Irkutsk, Russia. He was born in 1967 in Irkutsk. In 1984
he was granted the certificate in mathematical programming at the Institute of System Dynamics and
Control Theory of Russian Academy of Sciences. In 1995 he graduated from Irkutsk State Technical
University specializing in electrical engineering. D.B. Popov joined the Siberian Energy Institute in 1989.
His research interests are transient stability analysis, database application development, users interface
design, program interface development, asynchronous conditions, and prevention by customization of
automatic equipment.
Socorro Rangel received her M.Sc. degree from Campinas State University (UNICAMP), Campinas,
Brazil, and Ph.D. degree from Brunel University, Uxbridge, England. She is currently an Associate Pro-
fessor at the State University of São Paulo (UNESP), São José do Rio Preto, Brazil. Her main research
interests lies on large-scale system modelling and optimisation.
Christian Rehtanz received his diploma degree in Electrical Engineering in 1994 and his Ph.D. in
1997 at Technische Universität Dortmund, Germany. From 2000 he was with ABB Corporate Research,
Switzerland and from 2003, Head of Technology for the global ABB business area Power Systems.
From 2005 he was Director of ABB Corporate Research in China. From 2007 he is Professor and Chair
375
About the Contributors
for power systems and power economics at Technische Universität Dortmund. His research activities
include technologies for network enhancement and congestion relief like stability assessment, wide-area
monitoring, protection, and coordinated FACTS- and HVDC-control.
Mohammad Saleh received B.Sc. and MSc degrees in electrical engineering from Tarbiat Moallem
University (Tabriz), and University of Kurdistan (Sanandaj), Iran, in 2007 and 2011, respectively. His
current research interests include Grid integration of wind energy converters.
Ghasem Tikdari was born on 1985 in Tikdar, Kerman, Iran. He received the B.S. degree from Bahonar
University, Kerman, Iran in 2007 in Electronic Engineering and received the M.S. degree from University
of Kurdistan, Sanandaj, Iran in 2009 in Electrical Power System Engineering. His research interests are
on power system emergency control, renewable energy sources, artificial intelligence techniques, and
electronic and power electronic device designing. He also experienced in some industrial works which
use Distributed Control System (DCS), Supervisory Control and Data Accusation (SCADA), and field-
bus communication protocols.
Nikita V. Tomin, PhD, (M’08) was born on December 18th in 1982, in Russia. Dr. Tomin is a Senior
Researcher at the Energy Systems Institute of the Russian Academy of Science, Irkutsk, Russia. In 2007
he defended his PhD thesis at the Energy Systems Institute SB RAS (Irkutsk). Dr. Tomin specializes in
the field of artificial intelligence technologies in electric power systems. He is the author and co-author
of more than 60 scientific papers.
Nikolai I. Voropai (M’1996, SM’1998; F’2009) is Director of the Energy Systems Institute (Siberian
Energy Institute until 1997) of the Russian Academy of Sciences, Irkutsk, Russia. He is also Head of
Department at Irkutsk Technical University. He was born in Belarus in 1943. He graduated from Len-
ingrad (St. Petersburg) Polytechnic Institute in 1966 and has been with the Siberian Energy Institute
since. N.I. Voropai received his degree of Candidate of Technical Sciences at Leningrad Polytechnic
Institute in 1974, and Doctor of Technical Sciences at the Siberian Energy Institute in 1990. His research
interests include: modeling of power systems; operation and dynamic performance of large intercon-
nections; reliability, security and restoration of power systems; development of national, international
and intercontinental electric power grids. N.I. Voropai is a member of CIGRE, a Fellow of IEEE, and a
member of PES. He is the IEEE PES Region 8 Zone East Representative.
376
377
Index
B D
degree of satisfaction 348, 368
bad data 7, 9-10, 12, 22-23
device duty 343
binomial options pricing model 264
Differential Evolution (DE) 51, 59, 90-91, 169,
biomass energy 127-128, 135
188, 206, 220, 243, 261, 312
boundary branch 16-18
Direct Load Control (DLC) 141
boundary nodes 7-13
Discount Cash Flow (DCF) 248-249, 252, 255-259
box agents 38
Distribution Companies (DISCO’s) 315-316
branch data 340
Divandarreh 204-210
bus data 340
Double-Line Ground (LLG) 335-336, 343
Doubly-Fed Induction Generator (DFIG) 199-201,
C 203, 209, 215
calculated scheme 7-9, 12-13, 19, 50 drive train model 202
capacitor data 340 Dual-Input Power System Stabilizer Model (PSS2A)
209
Index
E forecasting model 22
Fuel Constrained Economic Load Dispatch (FELD)
Earth energy 117 57, 72-73, 88, 94
economic load dispatch 57-58, 88, 91, 93-94 functional decomposition 1, 7, 9, 13, 49
electrical circuit 10, 163 fuzzy logic 55, 59, 91, 166, 169, 188, 196, 344,
electrical component 163 364, 367
electrical energy 135, 137-138, 148-150, 156-158, fuzzy optimization 344, 346, 364-368
160-161, 187, 201, 210, 289, 291 fuzzy sets 54, 188-189, 192, 196-197, 289, 344-
electrical generation 163 345, 364-366
Electrical Transient Analyzer Program (ETAP) 315- fuzzy set theory 344, 366
322, 328-329, 331, 337-338, 340-341, 343 fuzzy TOPSIS 171, 174, 188, 364
electrical vehicle 162-163
electric distribution network 287, 292 G
electricity network 145, 163, 200
Electricity Price (EP) 59, 90-91, 96, 140, 147, 149, Gaussian Filter 297
152-154, 159, 161, 163, 169, 206, 208, 249- Gauss-Seidel Method 340
257, 259 generalized assignment problem 220, 228, 243-246
electric motors 163 generator agent 35-38
Electric Power Research Institute (EPRI) 140 generator node 26-27, 30
Elimination and Choice Expressing Reality (ELEC- Genetic Algorithms (GA) 21-22, 52, 59, 91, 128,
TRE) 171, 176, 184, 187-188, 194 169-170, 176, 186, 189-190, 193, 269, 279-
emergency control 1-6, 20, 32, 34, 43, 46, 49, 51, 281, 283, 288, 291, 344-351, 355-357, 359,
54-56, 214, 294, 297-298, 302, 310, 313-314 361-362, 364-365, 367-368
Emergency Demand Response Program (EDRP) geothermal energy 117, 130, 133, 135
141-142, 158 geothermal power 127, 129, 133, 135
Emerging Energy Technology (EET) 167, 169, 176, global warming 95-98, 100, 103-104, 113, 127,
185-186, 197 129, 133, 164-165
Enercon E-53 206 Graphical User Interface (GUI) 317
energy cost analysis 207 greenhouse effect 95, 97, 103, 113, 136
Energy Efficiency (EE) 55, 96-97, 110, 112, 118- greenhouse gases 96, 108, 118, 127, 136, 165
120, 122, 124, 126-127, 129-130, 133, 145, ground grid analysis 316, 331, 343
159, 162, 165, 181, 183, 192 ground grid systems 342
energy function 44, 54, 57, 60-62, 65-67, 69-70, 74, Gujranwala Electric Power Company (GEPCO)
81, 88, 90, 94, 293-294, 297, 299, 311, 313 315-316
energy management 137, 154, 162-163, 263-264
Environmental Performance (EP) 59, 90-91, 96, H
110, 159, 169, 206, 208, 253, 256 harmonic analysis calculation methods 341
equal area criterion 51, 299 Harmonic Analysis (HA) 315-316, 322, 328, 341
ETAP software 315-317, 337-338 Heating, Ventilating and Air-Conditioning (HVAC)
Evolutionary Programming (EP) 59, 90-91, 96, 159, 96
169, 206, 208, 256 historical volatility 254-255, 257, 264
exergy 122-124, 188 hopfield lagrange network 57, 61-62, 64, 94
Hopfield Neural Network (HNN) 57, 59-61, 69, 80,
F 84, 90-94
fast-decoupled method 340 Human Machine Interface (HMI) 303
Feeder Analyses (FDR-ANA) 315 Hybrid GA Pattern Search (HGAPS) 348, 359,
Fixed-Speed Induction Generator (FSIG) 199, 209 361-362
Fixed Speed Wind Turbine (FSWT) 201, 217, 219 hybrid genetic algorithms 344, 346, 367
Florida Reliability Coordinating Council (FRCC) hydropower 135
307, 311 hydrothermal economic load dispatch 88, 94
378
Index
379
Index
380
Index
transient stability 44-46, 209, 211, 214-216, 218, VlseKriterijumska Optimizacija I Kompromisno
295, 297, 299, 313, 317, 328-329, 337-338, Resenje (VIKOR) 171, 191, 194
341-342 voltage magnitudes 7, 10, 13, 16, 273, 298
transient stability analysis 215, 313, 328-329, 341
transient stability limit 342 W
transmission system 32, 35, 37, 60, 90, 150, 163,
270, 337 WAMS (Wide-Area Measurement System) 7-8, 21
Two-Dimensional (2-D) 296-297, 301 water privatisation 107
Weibull Distribution 198, 219
U Western European Union for the Coordination of
Transmission of Electricity (UCTE) 6, 43, 45
uncertain coefficients 368 wholesale markets 5
Under Frequency Load Shedding (UFLS) 307 wind energy 130, 135, 189, 198-202, 204, 210,
Uninterruptible Power Supplies (UPS) 43, 51, 340- 213-219
341 wind farms 198-200, 209-211, 214-218
Wind Power Systems (WPSs) 199-200, 207-208,
V 210, 212-214, 219, 310
wind rose diagram 204, 219
vagueness factor 348, 352, 356, 359-360, 368 World Energy Council (WEC) 96-98, 130, 132
Variable Speed Wind Turbine (VSWT) 201-203,
218-219
381