Energy Conversion and Management 92 (2015) 82–91

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Combined emission economic dispatch of power system including solar

photo voltaic generation
Naveed Ahmed Khan a, Ahmed Bilal Awan c, Anzar Mahmood a,⇑, Member IEEE, Sohail Razzaq b,
Adnan Zafar a, Guftaar Ahmed Sardar Sidhu a
a
EE, COMSATS Institute of Information Technology Islamabad, Pakistan
b
EE, COMSATS Institute of Information Technology Abbottabad, Pakistan
c
EE, College of Engineering, Majmaah University, Al-Majmaah, Saudi Arabia

a r t i c l e i n f o a b s t r a c t

Article history: Reliable and inexpensive electricity provision is one of the significant research objectives since decades.
Received 28 August 2014 Various Economic Dispatch (ED) methods have been developed in order to address the challenge of con-
Accepted 10 December 2014 tinuous and sustainable electricity provision at optimized cost. Rapid escalation of fuel prices, depletion
of fossil fuel reserves and environmental concerns have compelled us to incorporate the Renewable
Energy (RE) resources in the energy mix. This paper presents Combined Emission Economic Dispatch
Keywords: (CEED) models developed for a system consisting of multiple Photo Voltaic (PV) plants and thermal units.
Economic Dispatch
Based on the nature of decision variables, our proposed model is essentially a Mixed Integer Optimization
Renewable energy
Particle Swarm Optimization
Problem (MIOP). Particle Swarm Optimization (PSO) is used to solve the optimization problem for a sce-
Solar PV generation nario involving six conventional and thirteen PV plants. Two test cases, Combined Static Emission Eco-
Combined Emission Economic Dispatch nomic Dispatch (SCEED) and Combined Dynamic Emission Economic Dispatch (DCEED), have been
Dynamic Emission Economic Dispatch considered. SCEED is performed for full solar radiation level as well as for reduced radiation level due
to clouds effect. Simulation results have proved the effectiveness of the proposed model.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction and related work
Significant research has been conducted throughout the world
for development of sustainable, renewable and efficient energy
systems in order to meet the requirements of increased population
and to reduce the extensive use of fossil fuels. Increasing energy
prices, environmental concerns and rapid depletion of the known
fuel reserves have significantly increased the scope of renewable
energy resources.
The power sector of Pakistan is designed as an interconnected
system and heavily relies on conventional sources of generation.
This system needs modifications and enhancements in order to
meet the twenty first century requirements. Pakistan’s energy
mix comprises of almost 67% thermal and 30% hydel resources.
According to Pakistan’s energy year book 2012 [1], total generated
electrical energy in Pakistan during 2010–2011 was 95,365 GW h
and share of different sources is: thermal power 64.3%; hydel
29.9% and nuclear and imported 5.8%. In thermal power, oil had
⇑ Corresponding author. Tel.: +92 03315079549.
E-mail addresses: anzarmahmood@comsats.edu.pk, anzarmahmood@gmail.com
(A. Mahmood).
URL: http://www.njavaid.com/anzar.aspx (A. Mahmood).
the largest share of 35.2% followed by natural gas 29.0% and coal
0.1%. On the other hand, the country has a large potential of solar
energy which has been estimated to be around 2900 GW in [2]. In
[3], the author discussed the energy scenario of Pakistan and
reviewed conventional and Renewable Energy (RE) resources of
the county in detail. The author has presented the supply, genera-
tion and exploitation of available resources in quantitative manner.
The paper is focused on RE development projects in the country,
recent progress, planning and public sector goals in this area. On
average, solar global insolation of 5–7 kW h/m2/day in almost
95% areas of Pakistan with persistence factor of over 85% has been
reported in [4,5].
Economic Dispatch (ED) is a vital and most frequent step in
power system operational planning [6]. ED is an optimization
problem that allocates power to each committed generating unit
so as to minimize the total operational cost, subject to constraints.
Various constraints include power balance, power limits of
generators, prohibited operating zones, ramp rate limits etc.
Several optimization techniques with equality and non-
equality constraints have been used for ED and reported in
literature [7].
History of ED dates back to 1920 [8]. Up till 1930 optimization
methods used were the base load method and best point loading.

http://dx.doi.org/10.1016/j.enconman.2014.12.029
0196-8904/Ó 2014 Elsevier Ltd. All rights reserved.
 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91
In early 30s, equal incremental cost method was utilized to achieve
better results [8]. In those days analog computers were used for
computational effort. First computer for transmission loss penalty
factor was built up in 1954. By 1955 electronic differential analyzer
was developed. Digital computers were used for ED first time ever
in 1954 and are being used till date [9]. The authors in [10] have
reviewed the techniques of ED used during 1977–1988; optimal
power flow, dynamic dispatch, ED in relation to Automatic Gener-
ation Control (AGC) and ED with non-conventional sources have
been analyzed. Power system consisting of thermal generators
has been extensively used to evaluate ED problem. Input–output
cost curves of thermal generating units are required for ED. The
input–output cost curve of a thermal generating unit is obtained
by multiplying cost per unit heat and its input–output heat rate
curve [11]. In modern days multi-valve steam turbines and multi-
ple fuel turbines are often used in generating units.
The ED with piecewise quadratic cost function (EDPQ) and ED
with prohibited operating zones (EDPO) are the two non-convex
ED problems [12]. Valve point effects produce a ripple like non
convex input–output heat rate curve. Complex constrained ED is
addressed by intelligent methods including Genetic Algorithm
(GA), PSO [13,14], Neural Network (NN), Evolutionary Program-
ming (EP), Tabu search etc. [15–17]. Kennedy and Eberhart intro-
duced PSO in 1995 [18]. In this method, movement of particles is
dependent on local and social components of velocity. Moreover,
maximum value of velocity, Vmax, is also a significant parameter.
Its low value results in local exploitation while a higher value
results in global exploration. To obtain a better control over local
exploitation and global exploration, an inertia factor x is intro-
duced in [19].
ED with both cost and emission minimization becomes multi-
objective optimization problem and is named as Combined Emis-
sion Economic Dispatch (CEED). Using PSO, CEED has been solved
by Selvakumar et al. [20]. Zhao et al. [21] solved bid based ED using
Constriction Factor PSO (CFPSO) and inertia weight. In [22], a
hybrid PSO, a combination of PSO and Sequential Quadratic Pro-
gramming (SQP), is introduced in order to solve a non-convex con-
strained ED problem with valve point effects. In [23], CEED has
been solved using a novel PSO scheme taking into account the gen-
erator limits and power balance constraints. An improved PSO has
been proposed to solve ED problem of hydro-thermal co-ordina-
tion in [24]. Authors in [25] have proposed an Enhanced PSO
(EPSO) for hydro-thermal scheduling problem which takes into
account various constraints such as power balance, hydro and ther-
mal generation limit, reservoir storage volume, initial and terminal
storage limit, water balance equation and hydro discharge limit. In
[26], PSO has been used to evaluate CEED problem with equality
constraints handled by novel techniques and multi-objective opti-
mization problem converted into a single objective one.
A lot of research on Economic Dispatch (ED) problem has been
carried out during last five years. A few instances are as follows. In
[8], a non-convex ED problem has been addressed by various hybrid
optimization techniques. The problem has been addressed first by
developing an extensible and flexible computational framework
called ‘‘PED Frame’’, used as a platform for the computer implemen-
tation of different algorithms under consideration. This framework
has been used to implement Genetic Algorithm (GA) based models
and Hybrid models for ED. In [27], a PSO based technique with con-
striction factor (CFPSO) has been proposed for ED with valve point
effects; CFPSO technique proved to be fast converging. In [28], a
multi-objective CEED solution has been proposed by using Artificial
Bee Colony (ABC) algorithm. For the solution of the problem, multi-
objective CEED has been converted into single-objective CEED by
using penalty factor. In [29], iteration PSO with time varying accel-
eration coefficients (IPSO-TVAC) has been proposed for ED with
valve-point effects; Iteration term in velocity equation and time
83
varying acceleration coefficients improved the performance
(searching ability) of PSO technique. In [30], a novel optimization
methodology has been proposed to solve a large scale non-convex
ED problem. The proposed approach is based on a hybrid Shuffled
Differential Evolution (SDE) algorithm that combined the benefits
of shuffled frog leaping algorithm and differential evolution. The
proposed algorithm integrated a new differential mutation operator
in order to address the problem of ED.
In [31], Economic Environmental Dispatch (EED) has been car-
ried out using one Photo Voltaic (PV) plant and one wind turbine.
Authors used Strength Pareto Evolutionary Algorithm (SPEA) and
tests have been conducted for an IEEE bus system with 30 nodes,
8 machines and 41 lines. Dynamic Economic Emission Dispatch
(DEED) model with security constraints has been used for ED in
[32]. The authors have carried out their work on a system incorpo-
rating three thermal units, two solar PV plants and two wind tur-
bines. Authors in [33] have presented modified harmony search
algorithm to solve Combined Economic and Emission Load Dis-
patch (CEELD) problem. Practical constraints of real-world power
systems have been used and the experiments carried out on seven
systems in order to check the validity and behavior of the proposed
algorithm.
This paper presents a Combined Emission Economic Dispatch
(CEED) using 13 PV plants and 6 thermal units. Two test cases of
Static Combined Emission Economic Dispatch (SCEED) and
Dynamic Combined Emission Economic Dispatch (DCEED) have
been considered. SCEED is performed for full solar radiation level
as well as for reduced radiation level due to clouds effect whereas
DCEED for full radiation only. PSO is used for optimization of the
problem and simulation results have been computed in MATLAB.
The proposed model contains multiple solar plants unlike the work
discussed in [31,32]. Power demand data has been obtained from
Islamabad Electric Supply Company (IESCO) [34].
2. Problem formulation
This section is dedicated for problem formulation of CEED for a
power system having thermal and solar PV generations. As men-
tioned earlier, an ED problem can be formulated either statically
or dynamically. The mathematical formulation for both cases is
worked out in the following subsections.
2.1. Mathematical formulation of SCEED with solar power
CEED is a multi-objective optimization problem consisting of
both economic and environmental dispatch. The CEED problem
can be formulated as:
X
n
min G ¼ ðF i ðP i Þ þ Ei ðP i ÞÞ ð1Þ
i¼1
where G is objective function to be minimized, Fi(Pi) represents fuel
cost and Ei(Pi) denotes the emissions of ith generating unit. This
function is to be minimized subject to following constraints.
Equality constraint:
!
X
n
Pi  PL  Pd ¼ 0 ð2Þ
i¼1
where Pi is power generated by ith unit, PL represents power loss, Pd
is power demand and n is the total number of generating units.
Inequality constraint:
Pimin 6 Pi 6 Pimax ð3Þ
where Pimin and Pimax are the minimum and maximum power limits
of ith generating unit, respectively.
84 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91

Fi(Pi), Ei(Pi) in Eq. (1) and PL in Eq. (2) can be formulated as fol- Subject to
lows [35]. X
n X
m
Pd þ PL  Pi  Pgsj  Usj ¼ 0 ð13Þ
F i ðPi Þ ¼ ai P2i þ bi Pi þ ci þ jei  sinðf i  ðP imin  Pi ÞÞj$=h ð4Þ
i¼1 j¼1

where ai, bi, ci, ei and fi are fuel cost coefficients of ith generating
unit. Pimin 6 Pi 6 Pimax ð14Þ

Ei ðPi Þ ¼ /i P2i þ bi P i þ ci þ ei  expðdi  Pi Þ kg=h ð5Þ X

m
Pgsj  Usj 6 0:3  Pd ð15Þ
where /i, bi, ci, ei and di are emission coefficients of ith generating j¼1
unit.
Power losses can be calculated using the equation: 8Usj 2 f0; 1g
X
n X
n
PL ¼ Pi Bij P j ð6Þ where Ks is a constant used to make the last term of Eq. (12) com-
i¼1 j¼1 patible with the other terms. Moreover this allows us to control the
relative importance of the difference term compared to other terms.
where B is loss coefficient matrix.
By introducing a price penalty factor ‘h’, the multi-objective 2.2. Mathematical formulation of dynamic CEED with solar power
optimization function presented by Eq. (1) can be converted into
single objective optimization function. Therefore, by substituting DCEED is a more practical case in which it is aimed to allocate
Fi and Ei from Eqs. (4) and (5) respectively and introducing ‘h’ in powers to generating units for minimum cost of operation in a
Eq. (1), the CEED objective function can be defined as [28]: scheduling horizon over twenty-four hours a day. The ramp rate
n 
X limits are considered in this problem. In case of DCEED problem,
Min F C ¼ ai P2i þ bi P i þ ci þ jei  sinðf i  ðPimin  P i ÞÞj the mathematical formulation in Eq. (12) becomes:
i¼1
  $ X n 
N X
 2    
þhi ai P2i þ bi Pi þ ci þ ei  expðdi  Pi Þ ð7Þ Min F T ¼ ai Pti þ bi Pti þ ci þ ei  sin f i  Pimin  Pti 
h t¼1 i¼1
  2  
where hi is given as: þhi ai Pti þ bi Pti þ ci þ ei  exp di  Pti
ai P2imax þ bi Pimax þ ci þ jei  sinðf i  ðPimin  Pimax ÞÞj !
hi ¼ ð8Þ X
m X
m X
m

/i P2imax þ bi Pimax þ ci þ ei  expðdi  Pimax Þ þ P:UCost j  Pgsj  Ustj þ Ks Pgstj  Pgstj  Ustj
j¼1 j¼1 j¼1
The power generated by a solar plant can be represented as ð16Þ
[31]:
The ramp rate limits determine the range within which the gen-
    Si
Pgs ¼ P rated 1 þ T ref  T amb  /  ð9Þ eration of a thermal unit may increase or decrease. The power gen-
1000 eration of thermal units is constrained by the ramp rate limits as
where Prated is its rated power, Tref is the reference temperature, Tamb follows:
is the ambient temperature, / is temperature coefficient and Si is
the incident solar radiation. Pti  Pt1
i 6 URi ð17Þ
With m solar plants taking part in the dispatch, the solar share
(the scheduled solar power) is given as: Pit1  P ti 6 DRi ð18Þ
X
m where URi and DRi are the up rate and down rate of ith generating
Solar share ¼ Pgsj  Usj ð10Þ unit respectively.
j¼1
Due to ramp rate limits, the minimum and maximum generat-
where Pgsj is power available from jth solar plant and Usj denotes ing limits of thermal units are modified as follows:
status of jth solar plant which is either 1 (ON) or 0 (OFF).    
The cost of solar power is represented as follows. max Pimin ; URi  Pti 6 Pti 6 min Pimax ; Pit1  DRi ð19Þ

X
m
The power balance constraint for DCEED can be formulated as:
Solar cost ¼ PUCostj  Pgsj  Usj ð11Þ
j¼1 t
X
n X
m
Pd þ PtL  Pti  Pgstj  Ustj ¼ 0 ð20Þ
where PUCostj is per unit cost of jth solar plant. i¼1 j¼1
Along with cost minimization, another objective is to minimize
The share of solar power at any time, based on 30% upper limit
the difference between the total available solar power and the
[31], is constrained as:
solar share in order to achieve the maximum benefit of solar
availability. Therefore, with solar generation included in the X
m
t
dispatch, the objective function in Eq. (7) becomes: Pgstj  Ustj 6 0:3  Pd ð21Þ
j¼1
n 
X
Min F T ¼ ai P2i þ bi P i þ ci þ jei  sinðf i  ðPimin  P i ÞÞj
i¼1
8 Ustj 2 f0; 1g
 
þ hi ai P2i þ bi Pi þ ci þ ei  expðdi  Pi Þ
! 3. Optimization method
X
m X
m X
m
þ P:UCostj  Pgsj  Usj þ Ks Pgsj  Pgsj  Usj
j¼1 j¼1 j¼1
It is obvious from the above mentioned problem formulation
that CEED with solar generation is a Mixed Integer Optimization
ð12Þ Problem (MIOP). The decision variables for thermal machines are
 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91 85

continuous whereas the variables for solar plants are binary. In  After having value of updated velocity, position of each particle
order to solve this problem, PSO for MIOP is used in this work. is updated using following equation.
The PSO for MIOP is essentially a combination of classical PSO
and Binary PSO (BPSO). X kþ1
i ¼ X ki þ V kþ1
i ð28Þ
 Fitness is evaluated for updated position of each particle; pbest
3.1. Classical PSO and gbest are obtained for next iteration.
 The process is repeated until a convergence criterion is satisfied.
PSO is an optimization technique inspired by bird flocking. To
explain PSO we can imagine a swarm of birds searching for food. All the above mentioned steps for PSO procedure are depicted in
This swarm flocks to search the food randomly in a specific region. the flow chart given in Fig. 1.
All the birds are supposed to be searching for a single piece of food.
At any time during search, each bird has a position and velocity. 3.2. Binary PSO (BPSO)
Birds move with knowledge of distance to food but not its exact
location. Best Strategy planned by birds is to follow a bird nearest Binary version of PSO is used to optimize the problems having
to food. PSO makes use of above mentioned scenario to solve binary decision variables i.e. having values either 0 or 1. The steps
optimization problems. In PSO each bird is known as particle of BPSO procedure are same as that of real valued PSO except fol-
which is a possible solution in search space. Number of all particles lowing differences:
in a search space represents size of swarm (or population). Each
particle has a position in search space, velocity and fitness value.  As the variables in BPSO are binary, therefore particles are ini-
Fitness value for a particle is obtained by objective (fitness) tialized randomly for their binary positions.
function evaluation.
Following are the steps of PSO procedure. X i ¼ ½X i1 ; X i2 ; . . . . . . . . . ; X iN  ð29Þ

 Starts with decision of swarm/population size which is problem 8X i1 ; X i2 ; . . . . . . . . . ; X iN 2 f0; 1g

specific i.e. it depends on complexicity of problem.
 Particles are then initiallized randomly for their positions and Each dimension of a particle is assigned a binary value with a
valocities. In an N dimensional optimization problem, the posi- probability of 0.5 as following:

tion of an ith particle is an array of 1  N. i.e., 1; if rand > 0:5
X id ¼ f ðxÞ ¼ ð30Þ
0; otherwise
X i ¼ ½X i1 ; X i2 ; . . . . . . . . . ; X iN  ð22Þ
where d = 1, 2, . . .. . ., N
Similarly valocity is presented by,
Start
V i ¼ ½V i1 ; V i2 ; . . . . . . . . . ; V iN  ð23Þ
 Fitness for each particle is obtained by evaluating the objective
function given as: Inialize all parcles with random
posion and velocity vectors
F i ¼ f ðX i1 ; X i2 ; . . . . . . . . . ; X iN Þ ð24Þ
 Two best positions, known as pbest and gbest, are selected for
For each parcle posion “X”,
next iteration. pbest is personal best position obtained by a par- evaluate fitness “f(X)”
ticle so far and gbest is global best position among all pbest. In
case of first iteration, pbest is same as randomly initiallized
position of a particle while in case of next iterations, it is the Yes For each parcle, set pbest equal
position of a particle having best fitness value up to that specific First iteraon ? to its posion “X” i.e.
iteration. pbest=X
 Velocity of each particle is updated using following equation.
No
   
k k
V kþ1
i ¼x k
V ki þ C 1 r 1 pbest i  X ki þ C 2 r 2 gbest  X ki ð25Þ For each parcle, if f(X) is beer
than f(Pbest) then
Pbest=X
where V ki
is valocity of ith particle at iteration k. xk is a parameter
known as inertia weight at iteration k. C1 and C2 are acceleration
k
coefficients. r1 and r2 are random numbers between (0, 1). pbesti Set best of pbest as gbest
is best position of ith particle at iteration k. X ki is position of ith
particle at iteration k. gbestk is global best position at iteration k
and V kþ1
i is updated velocity at iteration k + 1.
V ki which is the velocity of ith particle at kth iteration should be Update velocity and posion of
each parcle
within range of its minimum and maximum values, i.e.,
V min 6 V ki 6 V max ð26Þ
Inertia weight, xk, at each iteration is modified using following No Convergence
criteria sasfied
equation.
?
xmax  xmin
xk ¼ xmax  k ð27Þ
iter max Yes

where xmax is maximum value of inertia weight. xmin is End

minimum value of inertia weight. itermax is maximum number of
iterations. Fig. 1. Flow chart of PSO algorithm.
86 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91

 Position is updated as following: Table 2

(   Emission coefficients of thermal generating units [36].
kþ1 1; if rand < sigmoid V kþ1
id
X id ¼ ð31Þ Machine no. F (kg/MW2 h) b (kg/MW h) c (kg/h)
0; otherwise
1 0.00419 0.32767 13.85932
The sigmoid function in above equation, used to scale velocities 2 0.00419 0.32767 13.85932
3 0.00683 0.54551 40.2669
between 0 and1, is calculated as:
4 0.00683 0.54551 40.2669
  1 5 0.00461 0.51116 42.89553
sigmoid V kþ1
id ¼ kþ1
ð32Þ 6 0.00461 0.51116 42.89553
1 þ eV id

3.3. PSO for MIOP applied to CEED with solar power

Table 3
Following are the steps to optimize the SCEED problem (Eqs. Power ratings and per unit rates of solar plants.
(12)–(15)) by means of PSO. Plant # Prated (MW) Unit rate ($/KW h) [37]
1 20 0.22
 Control parameters are selected. 2 25 0.23
 Initialization of position and velocity for each particle. Each par- 3 25 0.23
ticle contains continuous as well as binary variables. 4 30 0.24
5 30 0.24

 6 35 0.25
X i ¼ Pi1 ; Pi2 ; . . . . . . . . . ; Pin ; Usi1; Usi2 ; . . . . . . ; Usim ð33Þ 7 35 0.26
8 40 0.27
where Pid and UsiD are the power of dth thermal unit and Dth solar 9 40 0.27
plant in ith particle. 10 40 0.275
Each Pid is initialized randomly using following equation: 11 40 0.28
12 40 0.28
Pid ¼ LB þ rand  ðUB  LBÞ ð34Þ 13 40 0.28

LB and UB represent lower bound and upper bound of thermal

units respectively.
Each UsiD is initialized randomly using Eq. (30).
Where D = 1, 2, . . ., m
Table 4
Solar radiation, power demand and temperature for 17th day of July 2012.
 Velocity for each particle is initialized between 0 and 1. Fitness
Time Global solar Power demand Temperature (°C)
for each particle in Eq. (12) is evaluated and pbest and gbest are
radiation (W/m2) (MW) [34] [38]
selected.
1:00 0 965 30
 Velocity is updated using Eq. (25) while positions are updated
2:00 0 1142 29
using Eqs. (28) and (31) for thermal and solar generators 3:00 0 1177 28
respectively. 4:00 0 1198 28
 Fitness in Eq. (12) is evaluated for each updated position, pbest 5:00 5.4 1153 28
and gbest are obtained for next iteration. The process is repeated 6:00 101 1136 –
7:00 253.7 1138 29
until the convergence criterion is satisfied. 8:00 541.2 1060 31
9:00 530.4 1155 33
4. Test system 10:00 793.9 1244 34
11:00 1078 1088 35
12:00 1125.6 1240 36
In this section the proposed model has been implemented on 13:00 1013.5 1135 37
two test systems in order to investigate both SCEED and DCEED. 14:00 848.2 1318 37
15:00 726.7 1074 37
16:00 654 1190 38
4.1. Test system-I 17:00 392.9 1276 38
18:00 215.1 1154 37
The test system-I includes 6 thermal units and 13 solar plants 19:00 38.5 1333 35
and is supposed to be operated in Islamabad region of Pakistan. 20:00 0 1322 34
21:00 0 1269 34
The data for thermal units has been taken from [36] and is 22:00 0 1139 33
presented in Tables 1 and 2. 23:00 0 1202 32
Table 1 presents fuel cost coefficients as well as minimum and 0:00 0 1291 –
maximum power limits whereas Table 2 contains emission

Table 1
Fuel cost coefficients and generating capacities of thermal generating units [36].

Machine no. a ($/MW2 h) B ($/MW h) c ($/h) Pmin (MW) Pmax (MW)

1 0.15247 38.53973 756.79886 10 125
2 0.10587 46.15916 451.32513 10 150
3 0.02803 40.39655 1049.32513 40 250
4 0.03546 38.30553 1243.5311 35 210
5 0.02111 36.32782 1658.5696 130 325
6 0.01799 38.27041 1356.27041 125 315
 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91 87

Table 5
Data of thermal units.

Unit no. a ($/MW2 h) b ($/MW h) c ($/h) a (kg/MW2 h) b (kg/MW h) c (kg/h) Pmax (MW) Pmin (MW) UR (MW/h) DR (MW/h)
1 0.007 7 240 0.00419 0.32767 13.8593 500 100 80 120
2 0.0095 10 200 0.00419 0.32767 13.8593 200 50 50 90
3 0.009 8 220 0.00683 0.54551 40.2669 300 80 65 100
4 0.009 11 200 0.00683 0.54551 40.2669 150 50 50 90
5 0.008 10.5 220 0.00461 0.51116 42.8955 200 50 50 90
6 0.0075 12 190 0.00461 0.51116 42.8955 120 50 50 90

Table 6
Load demand for 24 h.

Hour 1 2 3 4 5 6 7 8 9 10 11 12
Demand (MW) 955 942 953 930 935 963 989 1023 1126 1150 1201 1235
Hour 13 14 15 16 17 18 19 20 21 22 23 24
Demand (MW) 1190 1251 1263 1250 1221 1202 1159 1092 1023 984 975 960

Table 7
Fitness value obtained in 10 runs for dispatch at t = 10:00 h.

xmin xmax
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2
0.5 277,640 274,330 278,700 271,740 – – – –
0.4 254,820 262,640 277,550 281,290 254,560 – – –
0.3 254,890 272,750 284,720 272,050 258,930 271,670 – –
0.2 281,310 286,340 271,130 257,140 262,870 258,120 26,0470 –
0.1 309,970 257,520 259,710 266,070 256,210 252,210 257,800 26,1030

Bold shows the minimum value of objective function for the selected parameters.

coefficients for the selected machines. The data for solar plants has by setting maximum and minimum values of x to 0.4 and 0.1
been presented in Tables 3 and 4. respectively, as evident from Table 7.
Table 3 presents power ratings and per unit costs of different The table presents the best values of objective function (Eq.
solar plants, approximated to be within the range provided in (12)) obtained with various settings of x.
[37]. Table 4 encompasses global solar radiation as well as temper- Solar plants are considered to be operating for 6 h a day, from
ature and load profiles of Islamabad for the 17th day of July 2012. 10:00 to 16:00 h, as In Pakistan, these hours provide maximum
In this paper, global solar radiation data has been generated using radiation and are free of shadow effects in almost all the seasons.
Geospatial Toolkit, data related to power demand of Islamabad Following are the results and discussions for both cases.
region has been taken from IESCO [34] and temperature profile
has been taken from [38]. The 17th day of July has been selected 5.1. Case I
arbitrarily from the only available demand data of July, 2012.
In this case, the simulations have been carried out for both full
4.2. Test system-II and reduced solar radiation; later is the case of cloudy weather.
Simulation results of static CEED are depicted in Figs. 2–4 as well
The test system-II is also comprised of 6 thermal units and 13 as in Tables 8–13. Graphs in Figs. 2–4 show simulation results in
solar plants. The data used for solar plants is the same as given terms of the fitness value (FT) versus iterations. As evident from
in test system-I whereas the data for thermal units and load Figs. 2–4, the algorithm converges within 1000 iterations which
demand has been taken from [39] and are presented in Tables 5
and 6 respectively.
5.9 10:00 hrs
10 11:00 hrs
5. Results 12:00 hrs
13:00 hrs
5.8 14:00 hrs
10 15:00 hrs
This section shows the results for proposed PSO based MIOP
FITNESS VALUE

model. The above mentioned method was implemented in 10

5.7

MATLAB R2013a. The proposed model has been implemented on

two cases as follows. 10
5.6

Case I: In case I, the proposed model has been implemented on

test system-I to investigate the problem of SCEED. 10
5.5

Case II: In case II, the proposed model has been implemented on
test system-II to investigate the problem of DCEED. 10
5.4

Control settings used for PSO were: C1, C2 = 2; r1, r2 = random

0 500 1000 1500
numbers between 0 and 1; Maximum number of iterations = 1500; ITERATION
swarm size = 10. Maximum and minimum values of velocity are
0.5 ⁄ Pmax and 0.5 ⁄ Pmin respectively. Best results were obtained Fig. 2. Simulation results for full solar radiation.
88 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91

10:00 hrs Table 9

11:00 hrs
10
5.9
12:00 hrs
Results of CEED with solar power for 1088 MW demand at 11:00 h.
13:00 hrs
5.8
14:00 hrs Thermal P1 (MW) 10.1062
10
FITNESS VALUE

15:00 hrs
generation P2 (MW) 10
5.7 P3 (MW) 99.1
10
P4 (MW) 168.682
10
5.6 P5 (MW) 235.8781
P6 (MW) 246.7809
5.5
10
Solar generation Us1, Us2, . . ., Us13 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0,
5.4 0
10
Solar power share (MW) 317.471
5.3
10 Cost Fuel cost ($/h) 1.0e + 04 ⁄ 3.9426
0 500 1000 1500 Emission cost ($/h) 1.0e + 04 ⁄ 2.7712
ITERATION Solar cost ($/h) 1.0e + 04 ⁄ 8.2286
Total cost ($/h) 1.0e + 05 ⁄ 1.4942
Fig. 3. Simulation results for 15% reduced solar radiation. Others Emissions (kg/h) 1.0e + 03 ⁄ 0.6073
Demand-generation gap 0.0181
(MW)

10:00 hrs
10 5.7 11:00 hrs
12:00 hrs
13:00 hrs
14:00 hrs
15:00 hrs
Table 10
FITNESS VALUE

10 5.6
Results of CEED with solar power for 1240 MW demand at 12:00 h.

10 5.5 Thermal P1 (MW) 10

generation P2 (MW) 10.2191
P3 (MW) 194.9316
10 5.4 P4 (MW) 177.4014
P5 (MW) 224.8683
10 5.3 P6 (MW) 303.5647
Solar generation Us1, Us2, . . ., Us13 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 0,
0 500 1000 1500
1
ITERATION Solar power share (MW) 319.1076

Fig. 4. Simulation results for 30% reduced solar radiation. Cost Fuel cost ($/h) 1.0e + 04 ⁄ 4.6762
Emission cost ($/h) 1.0e + 04 ⁄ 3.8326
Solar cost ($/h) 1.0e + 04 ⁄ 8.2436
Total cost ($/h) 1.0e + 05 ⁄ 1.6752
Others Emissions (kg/h) 1.0e + 03 ⁄ 0.8607
corresponds to a maximum of 3.56 s using 1.8 GHz core i5 proces-
Demand-generation gap 0.0927
sor. Generation of thermal units in MW is given in Tables 8–13 for (MW)
the timings 10:00, 11:00, . . .. . ., 15:00 respectively. Us1,. . .., Us13
correspond to status of solar plants which is either ON (repre-
sented by 1) or OFF (represented by 0). Power balance constraint
violation is represented by demand-generation gap. Positive value
of demand-generation gap means that generation is greater than Table 11
demand while the negative value corresponds to generation not Results of CEED with solar power for 1135 MW demand at 13:00 h.
coping up with the demand. Thermal P1 (MW) 10.8593
It can be seen from all tables that the proposed algorithm is well generation P2 (MW) 118.1312
behaved. For instance, in Table 8 the thermal generation values for P3 (MW) 147.9272
P4 (MW) 186.3632
P5 (MW) 150.7713
P6 (MW) 221.0182
Solar generation Us1, Us2, . . ., Us13 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1,
1
Table 8
Solar power share (MW) 300.0974
Results of CEED with solar power for 1244 MW demand at 10:00 h.
Cost Fuel cost ($/h) 1.0e + 04 ⁄ 4.4136
Thermal P1 (MW) 120.4479 Emission cost ($/h) 1.0e + 04 ⁄ 3.0728
generation P2 (MW) 92.2947 Solar cost ($/h) 1.0e + 04 ⁄ 7.8311
P3 (MW) 155.8062 Total cost ($/h) 1.0e + 05 ⁄ 1.5318
P4 (MW) 76.4153
P5 (MW) 257.9089 Others Emissions (kg/h) 1.0e + 03 ⁄ 0.6395
P6 (MW) 302.2846 Demand-generation gap 0.1678
(MW)
Solar generation Us1, Us2, . . ., Us13 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1,
1
Solar power share (MW) 238.825
Cost Fuel cost ($/h) 1.0e + 04 ⁄ 5.2626
Emission cost ($/h) 1.0e + 04 ⁄ 4.2322 units P1–P6 are well within constraint limits. It can be noted from
Solar cost ($/h) 1.0e + 04 ⁄ 6.2322 Tables 8–13 that the solar power share is well within the upper
Total cost ($/h) 1.0e + 05 ⁄ 1.5727 bound. The optimized cost values are consistent with the respec-
Others Emissions (kg/h) 1.0e + 03 ⁄ 0.8808 tive shares of thermal and solar power generation. The algorithm
Demand-generation gap 0.0173 increases or decreases the solar share based on available solar radi-
(MW)
ation and temperature at any time as evident from Fig. 5.
 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91 89

Table 12 350
Results of CEED with solar power for 1318 MW demand at 14:00 h.

Thermal P1 (MW) 65.2834 300

generation P2 (MW) 97.2893
P3 (MW) 250
250
P4 (MW) 107.6407
P5 (MW) 252.7949

Solar share (MW)

P6 (MW) 297.7576 200
Solar generation Us1, Us2, . . ., Us13 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0,
1
150
Solar power share (MW) 247.1655
Cost Fuel cost ($/h) 1.0e + 04 ⁄ 5.5082
100
Emission cost ($/h) 1.0e + 04 ⁄ 4.7011
Solar cost ($/h) 1.0e + 04 ⁄ 6.4662
Total cost ($/h) 1.0e + 05 ⁄ 1.6675 50
Others Emissions (kg/h) 1.0e + 03 ⁄ 1.0376
Demand-generation gap 0.0686
0
(MW)
10:00 11:00 12:00 13:00 14:00 15:00
hours hours hours hours hours hours
Full Radiaon 15% Reduced 30% Reduced

Fig. 5. Solar shares at different solar radiation levels.

Table 13
Results of CEED with solar power for 1074 MW demand at 15:00 h.

Thermal P1 (MW) 82.7064

generation P2 (MW) 60.696
by comparing the results in Tables 9 and 13 where the load
P3 (MW) 249.2579
P4 (MW) 96.2554 demands are 1088 MW and 1074 MW respectively.
P5 (MW) 182.7257 The value of Ks has been experimentally set to 1.0e + 03 which
P6 (MW) 190.6486 results in maximum solar share of 319.1076 MW which is 25.73%
Solar generation Us1, Us2, . . ., Us13 1, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 0, of load demand, a value near to the solar share upper bound. By
1 reducing the value of Ks, the importance of difference in available
Solar power share (MW) 211.7604 solar power and the solar share in Eq. (12) gets reduced and vice
Cost Fuel cost ($/h) 1.0e + 04 ⁄ 4.5057 versa; the resulting solar share varies accordingly.
Emission cost ($/h) 1.0e + 04 ⁄ 3.2438 Fig. 5 shows the solar share for various levels of solar radiation.
Solar cost ($/h) 1.0e + 04 ⁄ 5.5399
Total cost ($/h) 1.0e + 05 ⁄ 1.3289
The solar radiation often gets reduced due to clouds, depending on
various parameters like thickness, height, amount, etc. of clouds.
Others Emissions (kg/h) 1.0e + 03 ⁄ 0.7149
Demand-generation gap 0.0504
As we have taken into account the global solar radiation which is
(MW) less affected by clouds as compared to beam radiation, therefore
tests have been carried out for assumed reductions of 15% and
30% in solar radiation. It is evident from Fig. 5 that the solar share
gets reduced for reduced solar radiation, in an expected manner.

The solar share is increased or decreased by turning ON or OFF

the appropriate solar units. For instance, the number of solar units
that are OFF in both Tables 8 and 10 is 4. Turning OFF the solar unit
number 1, 3, 8 and 11 at time 10:00 h, as clear from Table 8, results
in the solar share of 238.825 MW. On the other hand, Table 10
describes that Turning OFF the solar unit number 3, 7, 8 and 12 at
time 12:00 h accounts for the solar share of 319.1076 MW. As the
solar share is increased, thermal share gets reduced for a given load
demand. Therefore by increasing the solar share, the solar cost is
increased whereas fuel cost, emission cost and emissions get
reduced; which is evident from Tables 8 and 10 where the load
demands are approximately equal, i.e. 1244 MW and 1240 MW
respectively. In Table 8, the solar cost is 1.0e + 04 ⁄ 6.2322 $/h and
the fuel cost, emission cost and emissions are 1.0e + 04 ⁄
5.2626 $/h, 1.0e + 04 ⁄ 4.2322 $/h and 1.0e + 03 ⁄ 0.8808 kg/h
respectively, with solar share of 238.825 MW whereas in Table 10,
the solar cost is increased to 1.0e + 04 ⁄ 8.2436 $/h while the
fuel cost, emission cost and emissions are reduced to
1.0e + 04 ⁄ 4.6762 $/h, 1.0e + 04 ⁄ 3.8326 $/h and 1.0e + 03 ⁄
0.8607 kg/h respectively, for increased solar share of
319.1076 MW. The influence of solar share on total cost is much
greater as compared to thermal share because of higher per unit
costs of solar generating units. Therefore, the total cost is higher
in Table 10 as compared to that in Table 8. This effect can also be
seen in Fig. 2, where the fitness value increases from time 10 to
12 as the solar share is increased. Similar relations can be found
5.2. Case II
Table 14 presents the results of DCEED problem which is similar
to SCEED except an additional constraint of ramp rate limits and
the problem has been expanded over the time horizon of 1 day.
The results in Table 14 satisfy all the constraints discussed in Sec-
tion 2.2. The system operates from hour 1 to hour 10 and from
hour 15 to hour 24 with only thermal generation and therefore is
dealt with as an ordinary DCEED problem. The solar power contrib-
utes from hour 10 to hour 15 as in case of SCEED. The solar share
varies in a manner similar to that of SCEED. The results at hour
12 can be compared with that of hour 14 due to comparable load
demands of 1235 MW and 1251 MW respectively. The larger solar
share of 335.063 MW results in less fuel cost, emission cost and
emissions while higher solar cost and total cost at hour 12 as com-
pared to the respective quantities at hour 14, where solar share is
239.192 MW, for the same reasons discussed earlier in the case of
SCEED.
When solar generation is included to or removed from the sys-
tem, the load ramp seen by the other plants gets increased. The
greater the amount of solar share included or removed, the larger
the load ramp seen by thermal which may cause failure of opera-
tion due to ramp rate limits of thermal units. Fig. 6 shows the load
ramps seen by thermal units with and without solar power
included in the system.
 90
Table 14
Results for DCEED.

Hour P1 (MW) P2 (MW) P3 (MW) P4 (MW) P5 (MW) P6 (MW) Us1, Us2, . . .., Us13 Solar share Fuel cost ($/h) Emission cost ($/ Solar cost ($/h) Total cost ($/h) Emissions ($/h)
(MW) h)
1 298.7907 139.766 170.8483 50 200 95.5949 1.0e + 04 ⁄ 1.1237 1.0e + 04 ⁄ 0.802 1.0e + 04⁄1.9257 1.0e + 03 ⁄ 0.9651

N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91

2 321.0318 113.6014 162.5339 94.8329 200 50 1.0e + 04 ⁄ 1.1028 1.0e + 04 ⁄ 0.7931 1.0e + 04 ⁄ 1.8959 1.0e + 03 ⁄ 0.992
3 201.0318 124.6422 223.4963 105.3914 200 98.4383 1.0e + 04 ⁄ 1.1355 1.0e + 04 ⁄ 0.8084 1.0e + 04 ⁄ 1.9439 1.0e + 03 ⁄ 0.8493
4 255.8099 130.6614 164.1621 87.6478 200 91.7187 1.0e + 04 ⁄ 1.099 1.0e + 04 ⁄ 0.7511 1.0e + 04 ⁄ 1.8502 1.0e + 03 ⁄ 0.8396
5 335.8099 103.7868 172.4276 50 200 72.9757 1.0e + 04 ⁄ 1.0935 1.0e + 04 ⁄ 0.781 1.0e + 04 ⁄ 1.8746 1.0e + 03 ⁄ 1.0239
6 338.0287 151.024 80 97.2356 202.2665 94.4452 1.0e + 04 ⁄ 1.1497 1.0e + 04 ⁄ 0.8407 1.0e + 04 ⁄ 1.9903 1.0e + 03 ⁄ 1.0183
7 316.362 174.9805 144.0999 50 201.2571 102.3006 1.0e + 04 ⁄ 1.1711 1.0e + 04 ⁄ 0.8672 1.0e + 04 ⁄ 2.0383 1.0e + 03 ⁄ 1.0355
8 344.7516 84.9805 198.2418 100 200 95.0262 1.0e + 04 ⁄ 1.2021 1.0e + 04 ⁄ 0.8725 1.0e + 04 ⁄ 2.0746 1.0e + 03 ⁄ 1.1124
9 395.6592 78.4906 228.5335 83.7331 246.861 92.7227 1.0e + 04 ⁄ 1.3348 1.0e + 04 ⁄ 1.0559 1.0e + 04 ⁄ 2.3907 1.0e + 03 ⁄ 1.4124
10 297.8072 50.2762 141.5124 78.3317 296.861 50 0, 1, 1, 0, 1, 1, 1, 0, 1, 235.034 1.0e + 04 ⁄ 1.1002 1.0e + 03 ⁄ 8.58 1.0e + 04 ⁄ 6.1374 1.0e + 05 ⁄ 0.8096 1.0e + 03 ⁄ 0.9895
0, 1, 1, 1
11 236.8295 50.7576 203.368 95.6474 237.9027 69.1967 1, 1, 0, 1, 1, 0, 1, 0, 1, 307.23 1.0e + 04 ⁄ 1.0633 1.0e + 03 ⁄ 7.5297 1.0e + 04 ⁄ 7.9931 1.0e + 05 ⁄ 0.9809 1.0e + 03 ⁄ 0.8419
0, 1, 1, 1
12 308.5653 81.1024 106.8937 91.1342 262.2717 50 1, 1, 0, 0, 0, 1, 1, 1, 1, 335.063 1.0e + 04 ⁄ 1.0758 1.0e + 03 ⁄ 8.0197 1.0e + 04 ⁄ 8.8286 1.0e + 05 ⁄ 1.0706 1.0e + 03 ⁄ 0.944
1, 0, 1, 1
13 379.7585 50.1201 87.6007 74.3231 226.1978 71.9097 1, 0, 0, 1, 1, 1, 0, 0, 1, 300.097 1.0e + 04 ⁄ 1.0616 1.0e + 03 ⁄ 7.7716 1.0e + 04 ⁄ 7.9026 1.0e + 05 ⁄ 0.9741 1.0e + 03 ⁄ 1.0588
1, 1, 1, 1
14 418.2248 91.3466 120.0901 105.821 216.0514 60.2609 1, 1, 0, 1, 1, 1, 0, 1, 0, 239.192 1.0e + 04 ⁄ 1.1992 1.0e + 03 ⁄ 9.1687 1.0e + 04 ⁄ 6.1791 1.0e + 05 ⁄ 0.8295 1.0e + 03 ⁄ 1.2713
1, 1, 1, 0
15 444.9107 118.6177 149.631 51.6389 204.8054 98.7275 1, 1, 1, 1, 1, 1, 0, 1, 0, 194.683 1.0e + 04 ⁄ 1.2654 1.0e + 03 ⁄ 9.845 1.0e + 04 ⁄ 4.9354 1.0e + 05 ⁄ 0.7185 1.0e + 03 ⁄ 1.4115
0, 1, 0, 1
16 500 119.6441 214.631 95.7249 200 120 1.0e + 04 ⁄ 1.4918 1.0e + 04 ⁄ 1.2477 1.0e + 04 ⁄ 2.7395 1.0e + 03 ⁄ 1.7997
17 433.6712 118.8918 228.3949 145.7249 200 94.3172 1.0e + 04 ⁄ 1.4485 1.0e + 04 ⁄ 1.2011 1.0e + 04 ⁄ 2.6496 1.0e + 03 ⁄ 1.5945
18 366.9561 168.8241 179.9709 137.3386 249.1787 99.7314 1.0e + 04 ⁄ 1.4367 1.0e + 04 ⁄ 1.1722 1.0e + 04 ⁄ 2.6089 1.0e + 03 ⁄ 1.3839
19 306.7417 200 244.9709 112.5704 200 94.7169 1.0e + 04 ⁄ 1.3762 1.0e + 04 ⁄ 1.1047 1.0e + 04 ⁄ 2.4808 1.0e + 03 ⁄ 1.2984
20 386.7417 110 214.7914 86.5183 201.3953 92.5533 1.0e + 04 ⁄ 1.2836 1.0e + 04 ⁄ 0.9736 1.0e + 04 ⁄ 2.2572 1.0e + 03 ⁄ 1.3123
21 352.6291 50 227.9327 50 251.3953 91.043 1.0e + 04 ⁄ 1.2126 1.0e + 04 ⁄ 0.9471 1.0e + 04 ⁄ 2.1597 1.0e + 03 ⁄ 1.2323
22 311.8292 100 185.0233 50 248.9536 88.1939 1.0e + 04 ⁄ 1.1646 1.0e + 04 ⁄ 0.8635 1.0e + 04 ⁄ 2.0281 1.0e + 03 ⁄ 1.0502
23 237.8993 108.2409 250.0233 58.8365 200 120 1.0e + 04 ⁄ 1.1564 1.0e + 04 ⁄ 0.8613 1.0e + 04 ⁄ 22.0178 1.0e + 03 ⁄ 0.963
24 265.8164 134.7064 200.7121 108.765 200 50 1.0e + 04 ⁄ 1.1285 1.0e + 04 ⁄ 0.8274 1.0e + 04 ⁄ 21.9559 1.0e + 03 ⁄ 0.9526
 N.A. Khan et al. / Energy Conversion and Management 92 (2015) 82–91
250
200
150
100
50
Load ramp
0
9_10 10_11 11_12 12_13 13_14 14_15 15_16
-50
-100
-150
-200
-250 Hour
Without solar With solar
Fig. 6. Load ramps seen by thermal units at different hours.
6. Conclusions
We have presented a new dispatch model to solve CEED prob-
lem for a system containing conventional thermal and solar PV
plants. Two case studies with six thermal units and thirteen solar
plants, employing PSO as an optimization tool, have been investi-
gated. SCEED problem has been investigated for full and reduced
solar radiation and DCEED problem is solved for full radiation only
with constraints of thermal generator limits, power balance and
renewable energy limits. However, the ramp rate limits have been
treated as an additional constraint in the case of DCEED. The larg-
est solar shares of 319.1076 MW (25.73% of 1240 MW) and
335.063 MW (27.13% of 1235 MW) have been recorded at
12:00 h in case of SCEED and DCEED, respectively. It confirms that
higher solar radiations contribute larger solar shares in both the
cases. Larger solar share for a given load demand results in higher
solar cost and total cost as well as lower fuel cost, emission cost
and emissions. The dispatch during 12:00 h resulted in highest
operation cost of 167,520 $/h because of highest share of solar
power. In case of DCEED, although the load ramp seen by thermal
units were increased at the points of addition and removal of solar
generation, the algorithm converged successfully and solved
DCEED without violating any constraint. The simulation results
demonstrate satisfactory operation of the proposed model. For
the sake of simplicity, in this work, power losses have been ignored
as well as thermal units with simple convex characteristics have
been selected. The future work is aimed at investigating the CEED
problem involving thermal units with non-convex characteristics,
taking losses into account and implementation of proposed model
on large power systems. New constraints identification and inte-
gration in the dispatch problem is also under consideration.
References
[1] Pakistan, Hydro carbon development institute of, Pakistan energy yearbook
2012. Islamabad: Ministry of petroleum and natural resources Pakistan; 2012.
[2] Mahmood A, Javaid N, Zafar A, Riaz RA, Ahmed S, Razzaq S. Pakistan’s overall
energy potential assessment, comparison of TAPI, IPI and LNG gas projects.
Renew Sustain Energy Rev 2014;2014(31):182–93.
[3] Sheikh M. Energy and renewable energy scenario of Pakistan. Renew Sustain
Energy Rev 2010;14(1):354–63.
[4] Shamshad K. Solar insolation over Pakistan. J SES (Taiyo Enerugi)
1998;24(6):30.
[5] Sheikh M. Renewable energy resource potential in Pakistan. Renew Sustain
Energy Rev 2009;13(9):2696–702.
[6] Morshed MJ, Asgharpour A. Hybrid imperialist competitive-sequential
quadratic programming (HIC-SQP) algorithm for solving economic load
dispatch with incorporating stochastic wind power: a comparative study on
91
heuristic optimization techniques, techniques. Energy Conver Manage
2014;84:30–40. ISSN 0196-8904, vol. 84, No. August 2014, p. 30–40.
[7] Xiong G, Li Y, Chen J, Shi D, Duan X. Polyphyletic migration operator and
orthogonal learning aided biogeography-based optimization for dynamic
economic dispatch with valve-point effects. Energy Convers Manage
2014;80(April):457–68.
[8] Malik TN. Economic dispatch using hybrid approaches. Taxila: University of
Engineering & Technology; 2009.
[9] Happ H. Optimal power dispatch – A comprehensive survey. Power Appar Syst
IEEE Trans 1977;96(3):841–54.
[10] Chowdhury BH, Rahman S. IEEE Trans 1990;5(4):1248–59.
[11] Wood AJ, Wollenberg BF. Power generation, operation, and control. John Wiley
& Sons; 2012.
[12] Lin W-M, Cheng F-S, Tsay M-T. Nonconvex economic dispatch by integrated
artificial intelligence. Power Syst IEEE Trans 2001;16(2):307–11.
[13] Jeyakumar D, Jayabarathi T, Raghunathan T. Particle swarm optimization for
various types of economic dispatch problems. Int J Electr Power Energy Syst
2006;28(1):36–42.
[14] Selvakumar AI, Thanushkodi K. A new particle swarm optimization solution to
nonconvex economic dispatch problems. Power Syst IEEE Trans
2007;22(1):42–51.
[15] Lee KY, Arthit S-Y, June HP. Adaptive Hopfield neural network for economic.
IEEE Trans Power Syst 1998:519–26.
[16] Park Y-M, Jong-Ryul W, Jong-Bae P. A new approach to economic load dispatch
based on improved evolutionary programming. Eng Intell Syst Elect Eng
Commun 6.2 1998:103–10.
[17] Mantawy AH, Soliman SA, El-Hawary ME. A new Tabu search algorithm for the
long term hydro scheduling problem. In: Proceedings of the large engineering
systems conference on power engineering; 2002.
[18] Kennedy J. Particle swarm optimization. Encyclopedia Mach Learn
2010:760–6.
[19] Shi Y, Eberhart R. A modified particle swarm optimizer. In: Evolutionary
computation proceedings, 1998. IEEE world congress on computational
intelligence. The 1998 IEEE international conference on, 1998.
[20] Kumar AIS, Dhanushkodi K, JayaKumar J, Paul CKC. Particle swarm
optimization solution to emission and economic dispatch problem. In:
TENCON 2003. Conference on convergent technologies for the Asia-Pacific
region, 2003.
[21] Bo Z, Yi-jia C. Multiple objective particle swarm optimization technique for
economic load dispatch. J Zhejiang Univ Sci A 2005;6(5):420–7.
[22] Victoire TAA, Jeyakumar AE. Reserve constrained dynamic dispatch of units
with valve-point effects. Power Syst IEEE Trans 2005;20(3):1273–82.
[23] Abido M. Multiobjective particle swarm optimization for environmental/
economic dispatch problem. Electric Power Syst Res 2009;79(7):1105–13.
[24] Titus S, Jeyakumar AE. Hydrothermal scheduling using an improved particle
swarm optimization technique considering prohibited operating zones. Int J
Soft Comput 2007;2(2):313–9.
[25] Yuan X, Wang L, Yuan Y. Application of enhanced PSO approach to optimal
scheduling of hydro system. Energy Convers Manage 2008;49(11):2966–72.
[26] Alrashidi M, El-Hawary M. Impact of loading conditions on the emission-
economic dispatch. In: Proceedings of world academy of science: engineering
& technology, vol. 41; 2008.
[27] Lim SY, Montakhab M, Nouri H. Economic dispatch of power system using
particle swarm optimization with constriction factor. Int J Innovat Energy Syst
Power 2009;4(2):29–34.
[28] Sonmez Y. Multi-objective environmental/economic dispatch solution with
penalty factor using Artificial Bee Colony algorithm. Sci Res Essays
2011;6(13):2824–31.
[29] Mohammadi-Ivatloo B, Rabiee A, Soroudi A, Ehsan M. Iteration PSO with time
varying acceleration coefficients for solving non-convex economic dispatch
problems. Int J Electrical Power Energy Syst 2012;42(1):508–16.
[30] Srinivasa Reddy AAVK. Shuffled differential evolution for large scale economic
dispatch. Electric Power Syst Res 2013;96:237–45.
[31] Brini S, Abdullah HH, Ouali A. Economic dispatch for power system included
wind and solar thermal energy. Leonardo J Sci 2009;14:204–20.
[32] ElDesouky A. Security and stochastic economic dispatch of power system
including wind and solar resources with environmental consideration. Int J
Renew Energy Res 2013;3(4):951–8.
[33] Jeddi B, Vahidinasab V. A modified harmony search method for environmental/
economic load dispatch of real-world power systems. Energy Convers Manage
2014;78:661–75.
[34] IESCO, Load management curve July, 2012. Islamabad: Islamabad Electric
Supply Company; 2012.
[35] Emmanuel DM, Nicodemus AO. Combined economic and emission dispatch
solution using ABC_PSO hybrid algorithm with valve point loading effect. Int J
Sci Res Publ 2012;2(12):1–9.
[36] Manteaw ED, Odero NA. Multi-objective environmental/economic dispatch
solution using ABC_PSO hybrid algorithm. Int J Sci Res Publ 2012;2:12.
[37] REN. Renewables, global status report, 2012. Renewable energy policy
network for the 21st century. REN21 Secretariat, Paris, France; 2012.
[38] <http://www.wunderground.com/history/airport/OPRN/2012/7/17/DailyHistory.
html> [accessed April 2014].
[39] Elaiw AM, Xia X, Shehata AM. Application of model predictive control to
optimal dynamic dispatch of generation with emission limitations. Electric
Power Syst Res 2012:31–44.