93

CHAPTER 5

A NOVEL METHOD FOR LOAD SHEDDING TO IMPROVE THE

RELIABILITY OF THE TRANSMISSION LINE

5.1 INTRODUCTION

Transmission line challenges are many, but the ways and means to solve it
are specific. Load shedding can be considered as one potential solution for handling
the line issues. Planned load shedding or scheduled load shedding is not of research
interest. Unplanned load shedding due to eventuality is of research interest. In this
chapter, un-planned case of load shedding and their nuances are analyzed for effective
load shedding process. If not clearly done, then it may trigger a cascaded failure in the
power system as a whole. Once cascaded failure has occurred lot of technical,
economical, political crisis occurs for which solutions are not instant. So it is better to
prevent rather than to cure it.

When the power system is triggered by under- frequency or under- voltage

state, in order to avoid cascaded failure, load shedding is the quick as well as an
effective method adopted throughout the world.

In early stages, primarily the research orientation towards load shedding is

based on conventional methods and optimization techniques based on classical
formulas. Modal analysis approach was proven effective for under-voltage load
shedding problems.

But in recent years, the research focus is towards computational techniques

or intelligent techniques.
 94

5.2 REVIEW OF LITERATURES

In this section, brief overviews on review of literatures regarding load

shedding are presented.

Sunil Ku. Mahapatro (2014)102 in his research contributions has provided

strategies for load shedding using Fuzzy logic. His idea is to estimate in real time the
amount of load to be shed. Fuzzy algorithms are used to generate command vectors to
obtain the solution sets. Hoseinzadeh et al. (2014)44 in their research work have
analyzed stability in a decentralized under frequency load shedding situation. Their
paper presents the de-centralized coordination for under frequency load shedding and
under voltage load shedding. DigSilent Power Factory is used by them.

J. Kaewmanee et al. (2013)55 in their work have discussed on optimal

load shedding based on fuzzy decision theory. They have taken into considerations
many associated parameters like affected customers, voltage and frequency after the
load removal have been considered. Parvathi S and K Shanti Swarup (2013)80 in
their work have discussed optimal load or generation shedding for emergency control.
Tharangika Bambaravanage et al. (2013)104 in their contributions, have proposed a
new idea for under-frequency load shedding. They have created two cases. Case –I
and Case - II. In the first phase traditional and semi-adaptive methods were
implemented. In second phase fully adaptive methods were presented. They have
taken a practical power system in Srilanka.

Liu Zhou et al. (2012)60 have proposed a methodology for emergency

load shedding based on sensitivity analysis of relay margin. Their work is based upon
finding the sensitivity analysis between the relay operation margins and power system
state variables. Load dynamics is also considered. Test system is built in real time
digital simulator. Zuhaila Mat Yasin et al. (2013)118 in their research contributions
have worked on evolutionary programming – artificial neural network for gauging the
under voltage load shedding. The ANN parameters are selected using Quantum-
Inspired Evolutionary Programming (QIEP). The test system used by them is an
IEEE-69 bus system.
 95

S. Pahwa et al. (2013)78 in their outstanding work have provided with

effective strategies to avoid cascaded network failure in the power grid. In their
research they have used tree heuristic algorithms for effective load shedding. Tree
heuristic approach overcomes the drawback of an optimization algorithm. Also, it
provides an effective solution to avoid the cascading failures. Hasmaini Mohamad
et al. (2013)40 in their outstanding research contributions have proposed a new
scheme for islanding operations with adaptive load shedding. Their scheme is based
upon event response for two different situations. 1) Event based, is based on the
power imbalance 2) Response based, is based on the swing equation which relies on
the frequency and rate of change of frequency measurement.

Emmy Kuriakose and Filmy Francis (2013)26 in their research

contributions have proposed an optimal method for under frequency load shedding
using artificial neural networks. In their work a fast and an adaptitive load shedding
method is presented using ANN. The test system used by them is an IEEE -39 bus
system and the simulations are presented using MATLAB. Also, they have
highlighted ANFIS mechanism to predict the amount of load to be shed. Ying-Yi
Hong and Po-Hsuang Chen (2012)113 in their research contributions have applied
genetic-algorithm based approach for load shedding. The uncertainties in the bus
loads are modeled using fuzzy sets. A stand alone system using Diesel and wind
power generators are used for the proposed method.

A. Cheraghi Valujerdi and M. Mohammadian (2012)15 in their research

contributions have presented a novel load shedding scheme to improve transmission
line performance. The test system used by them is an IEEE – 14 bus and an IEEE -57
bus system. PSO technique is used to check the efficacy of the proposed method.
Mehdi K. Moghaddam and Parisa A Bahri (2012)64 in their research contributions
regarding load shedding have demonstrated a new idea for forecasting electricity
loads. In their research non-linear approaches of neural networks and decision trees
have been applied. Accordingly, a model has been fitted for load shedding. A
modified adaptation of mean absolute percentage error (MMAPE) method is also
applied for each model.
 96

Ilham Sentosa et al. (2012)46, in their research contributions have

presented SEM Modelling for banking services. Structural equation modeling analysis
helps to understand the direct impact of a particular parameter over the other
dependent parameter. H. Mokhlis et al. (2012)70 in their research contributions have
presented fuzzy based under-frequency load shedding for distribution network
connected with DG. Their proposed strategy takes into account of rate of change of
frequency and load prioritization.

Y. Wang et al. (2011)111 in their contributions have provided an effective

strategy to minimize load shedding. A multistage method is proposed by them to
solve the non-linear problem. The main idea behind this method is to convert non-
linear optimization problem to linear optimization problem. The test system used by
them is an IEEE-14 bus, IEEE- 118 bus and a real 2038-bus power system. Malek
AL-Majali (2011)61, in his research contributions has presented SEM modelling for
banking solutions. Structural equation modeling analysis helps to understand the
direct impact of a particular parameter over the other dependent parameter.

M. H. Moradi and M. Abedini (2010)71 in their research contribution

have presented optimal load shedding approach for distribution systems in a way to
improve system stability. Their objective is to reduce the amount of load to be shed.
Voltage deviation index is used to identify the sensitive buses. A. R. Malekpour
et al. (2008)62 in their research contributions have developed a methodology for
optimal load shedding particularly for distribution networks. Genetic algorithm based
approach is used by them to identify the optimal load shedding strategy. Their
proposed method is tested on a radial distribution system with 33 load points.

Yuping Lu et al. (2006)116 in their contributions have proposed an

intelligent islanding techniques considering load balance. A new concept of cell is
proposed as well as heuristic logic is also proposed for an islanding model. Their
method is much suited for priority services. Shervin Shokooh et al. (2005)98 in their
contributions have briefed the need for an intelligent, fast and an optimal solution in
load shedding. They have highlighted an intelligent solution for load shedding. Also,
their methodology is based upon automated procedures.
 97

A.A. Mohd Zin et al. (2004)69 in their research work have highlighted
under-frequency load shedding and they have proven the results with a practical case
example. In their work, a real case example in Malaysia has been studied. UFLS has
been discussed by them. Yong Zheng and Chowdhury (2004)115 in their research
work have highlighted an expansion of transmission systems in deregulated scenario.
Their methodology is based on optimizing generation re-dispatch in case of
transmission line congestion. Economics behind transmission line expansion is also
discussed in their work.

G. Verbic and Gubina (2004)106 in their research work have presented a

new idea to prevent voltage collapse. Change in the power flow is considered to
calculate the voltage collapse criterion. Their algorithm can be easily executed in a
numerical relay. Their proposed methodology was implemented in an IEEE -118 bus
system. A. Carreras et al. (2002)11 in their research works have analyzed on
blackouts. Their model characterizes generators, loads, the transmission line network,
and the operating limits on these components. Also, they have analyzed the
interaction of two critical points and the complex nature of the solution space is also
discussed.

Ragu Balanathan et al. (1998)85 in their contributions have proposed a

strategy for under – voltage load shedding in power systems. The weakest node in the
system network is identified by VCPI – Voltage collapse Proximity indicator. Their
strategy is simple and also effective. Kaiumuzzaman Mollah and Nimal CK Nair 56
have provided a co-ordinated strategy for under voltage and under frequency load
shedding. Magnitude of the disturbance is taken into consideration for selection of the
parameters. Their proposed method is executed on an IEEE-39 bus test system.

Bottom line from the literature survey: From the above literature survey it is clear
that there are many workable methods proposed for load shedding. Many methods are
based on test cases and some methods are based on real time examples. Genetic
algorithm based approach is explored by many researchers. In this chapter, a novel
method for load shedding is proposed. This method is unique in its way. A path
analysis method is created for effective load shedding. Even though path model is
discussed by researchers it is not applied in Electrical Engineering. This research
 98

gap is well studied and path analysis model for effective load shedding is formed.
This is the unique feature to be underscored.

5.3 LOAD SHEDDING – TYPES

There are various types and forms of load shedding, some are discussed
below.

5.3.1 Conventional Shedding Types

Breaker Interlock Load Shedding: This is the simplest method of

carrying out load shedding. For this scheme, the circuit breaker interdependencies are
arranged to operate based on hardwired trip signals from an inter tie circuit breaker or
a generator trip. This method is often used when the speed of the load shedding is
critical. Even though, the execution of this scheme is fast, breaker interlock load
shedding possesses a number of inherent drawbacks.

Under-Frequency Relay Load Shedding

• In an electric power system, the power generated must be kept in

constant equilibrium with the power consumed; otherwise, a power
deficiency will occur.

• The system frequency decreases if load exceeds generation and

increases when power generation is greater than load demand.

Guidelines for setting up a frequency load shedding are common to both

large and small systems. The design methodology considers fixed load reduction at
fixed system frequency levels. Upon reaching the frequency set point and expiration
of pre-specified time delay, the frequency relay trips one or more load breakers. This
cycle is repeated until the system frequency is recovered, e.g., 10% load reduction for
every 0.5% frequency reduction. Since this method of load shedding can be totally
independent of the system dynamics, total loss of the system is an assumed
possibility.
 99

The significance of frequency control in load shedding

• In an electric power system, the power generated must be kept in

constant equilibrium with the power consumed; otherwise, a power
deficiency will occur.

• The system frequency decreases if load exceeds generation and

increases when power generation is greater than load demand.

• For frequency control in an interconnected power system, power

utilities employ a controlling mechanism to recover the frequency
during transient faults or severe load variations.

If this controlling mechanism is not properly done, it may result in

cascaded failure and the eventual blackout.

Programmable Logic Controller-Based Load Shedding: With a

Programmable Logic Controller (PLC) scheme, load shedding is initiated based on the
total load versus the number of generators online and/or detection of under-frequency
conditions. Each substation PLC is programmed to initiate a trip signal to the
appropriate feeder breakers to shed a preset sequence of loads. This static sequence is
continued until the frequency returns to a normal, stable level. A PLC-based load
shedding scheme offers many advantages such as the use of a distributed network via
the power management system, as well as an automated means of load relief.
However, in such applications monitoring of the power system is limited to a portion
of the network with the acquisition of scattered data. This drawback is further
compounded by the implementation of pre-defined load priority tables at the PLC
level that are executed sequentially to curtail blocks of load regardless of the dynamic
changes in the system loading, generation, or operating configuration. The system-
wide operating condition is often missing from the decision-making process resulting
in insufficient or excessive load shedding. In addition, response time (the time
between the detection of the need for load shedding and action by the circuit breakers)
during transient disturbances is often too long requiring even more load to be
dropped.
 100

PREVAILING LOAD SHEDDING SCHEMES

Under frequency load shedding (UFLS) techniques.

• According to the Institute of Electrical and Electronics Engineers

(IEEE) standards, ‘‘under frequency load shedding must be
performed quickly to arrest power system frequency decline by
decreasing the power system load to match available generating
capacity

• The European Network of Transmission System Operators for

Electricity (ENTSOE) has recommended the following steps for
under frequency load shedding.

• The first stage of automatic load shedding should be initiated at 49

Hz. At 49 Hz; at least 5% of total consumption should be shed.

 Load shedding can be done by under frequency method which we

generally mention it as UFLS strategy. Also, under voltage method generally
mentioned UVLS. But in coordinated strategy both UFLS and UVLS are taken
into effect which has improved results. There are also strong supports from
literature in determining the apt location for load shedding and the amount to
be shed for this non-linear optimization technique are used. Similarly, load
shedding is determined by identifying the weakest node in the network using a
voltage collapse proximity indicator (VCPI). The magnitude of the VCPI is
determined by estimating the ratio of the impedance of the load to the Thevenin
impedance looking into the network at a given voltage. There are also
approaches with specific interest on distribution side load shedding based on
distribution reliability index.
 101

Start

Measure Frequency / Voltage

No
Frequency < fmin /
Voltage < Vmin

Yes

Activate under Frequency

Relay / under Voltage Relay

Shed load

Frequency = 50 Hz No
Voltage within range

Yes

End

Figure 5.1 Flow chart explaining the under frequency / under voltage load
shedding.

Load Shedding
Techniques

Conventional Adaptive Load

Computational Intelligent Load
Load Shedding Shedding
Shedding Techniques
Techniques Techniques

Under Frequency Under Voltage Artificial Neural Fuzzy Adaptive Neuro Particle
Genetic
Load Shedding Load Shedding Networks Logic Fuzzy Inference Swarm
Algorithm
Techniques Techniques (ANN) Control System (ANFIS) Optimization

Figure 5.2 Different load shedding techniques.

 102

Restore Mean

System Restore Normal

Frequency Limit
Deviation
Free Reserve
Activate Primary after outage
Control Free
Reserve

Activate if Secondary Correct

Take over if
Responsible Control
Responsible Free
Reserve

Tertiary
Take over Control

Activate on Long Time

Term Control

Figure 5.3 Different control mechanisms to maintain system frequency.

Under voltage load shedding (UVLS) techniques

• UVLS techniques are implemented to protect the power system

from voltage collapse.

• A look at major power blackouts that have occurred around the

world shows that most were caused by voltage instability problems.

• Voltage instability occurs due to either forced outage of the

generator or the line or overloading.
• When this happens, the reactive power demand in transmission
lines varies severely and may cause a blackout, if not recovered
quickly.

Limitation of conventional load shedding techniques

• Conventional load shedding techniques are limited by their

inability to provide optimum load shedding.

• They simply follow a preset rule in which a fixed amount of load is

shed when frequency deviates from the nominal value. The main
 103

disadvantage of this method is that it does not estimate the actual

amount of the power imbalance.

• The result is either over-shedding, which affects power quality, or

under-shedding, which leads to tripping of electricity service

Adaptive load shedding techniques

• Adaptive load shedding techniques employ a power swing equation

to shed the required amount of load.

• The power imbalance within the system can be obtained by using

this equation.

2H ∂f
∆P = × (5.1)
f ∂t

Where ∆P is the power imbalance, H is inertia constant of generator, f is nominal

frequency (Hz), and df /dt is the rate of change of frequency (Hz/s).

This equation can be applied to an isolated power system having only a

single generator as well as to an interconnected power system.

§ dv ·
¨ dq ¸
S = © ¹ pdiff
§ dv ·
¨ ¦ dq ¸
(5.2)
© ¹

Where, S - load to be shed.

dv - change in voltage w.r.t time.

dq - change in power w.r.t time.

Pdiff - power difference

 104

Start

Measure Frequency

No
Frequency < fmin

Yes

Measure power imbalance by

using power swing equation

Shed estimated load

No
Frequency = 50 Hz

Yes

End

Figure 5.4 Flow chart explaining adaptive load shedding.

Limitations of Adaptive load shedding

• Adaptive load shedding techniques enhance the reliability of

conventional load shedding.

• However, these techniques also suffer from un-optimum load

shedding due to variations in df/dt behavior. (Highly non – linear
variation).

• That is the reason it is unable to calculate the exact value of S.

 105

5.3.2 Computational intelligence techniques in load shedding.

• The term ‘computational intelligence techniques’ generally refers

to a set of techniques that are applied to mimic human intelligence.

• These techniques include artificial neural networks (ANN),

adaptive neuro-fuzzy inference system (ANFIS), fuzzy logic
control (FLC), genetic algorithms (GA), and particle swarm
optimizations (PSO).

• These techniques can easily solve those nonlinear, multi-objective

problems in power systems that cannot be solved by the
conventional methods with the desired speed and accuracy.

Many intelligent approaches are available and also many more new
approaches are in pipeline in this area of load shedding in power systems. This shows
the potential threat of load shedding (if not done properly, cascaded failure may
occur). Also, load shedding is having potential opportunity for research.

Out of the various computational intelligence techniques, fuzzy logic has

proven its simplicity and robustness in handling problems with non-linear behavior. In
load shedding the problem nature is non-linear. The rate of change of frequency df/dt
is a varying phenomenon. Moreover, the fundamental idea of choosing fuzzy logic is
that it is the mimic of human intelligence. Also, for scaling the varying non-linear
data to linear form fuzzy is good and simple. This is the strong reason for choosing
fuzzy logic for load shedding problems. Intelligent techniques are powerful than
conventional techniques, but among intelligent techniques there are advantages as
well as disadvantages among them.
 106

Table 5.1 Comparative analysis between different computational intelligence

techniques in load shedding.

Technique Advantages Drawbacks

Artificial neural networks Ensures optimum amount of Provide satisfactory results

(ANN) load shedding for known cases only

Genetic Algorithm (GA) Strong Global Optimization Takes long time to

Technique determine the amount of
load to be shed.

Particle Swarm Simple and Effective Interrupted By Partial

Optimization (PSO) Optimisation
(Source: [48], [57], [67], [83] and [105])

If proper rules are framed, fuzzy proves to be a powerful technique when

compared to other techniques.

Table 5.2 Comparative analysis between conventional and computational

intelligence techniques

No. Feature Conventional Computational

techniques techniques

1. Optimum load shedding Do not provide Have the ability to provide

optimum load shedding optimum load shedding

2. Complex power system Cannot deal efficiently Can deal efficiently with
with modern and modern and complex power
complex power systems systems
(Source: [48], [57], [67], [83] and [105])
 107

Table 5.3 Advantages and drawbacks of computational intelligence techniques

No. Technique Advantages Drawbacks

1. Artificial neural ANN can ensure an The ANN can provide satisfactory
networks (ANN) optimum amount of load results for known cases only and may
shedding fail to predict accurate results for
unknown or varying cases

2. Fuzzy logic control FLC can be used for load The membership parameters of FLC
(FLC) shedding application on a require prior system knowledge.
power system of any size Otherwise, it may fail to provide
optimum load shedding.

3. Adaptive Neuro- FLC parameters are It can only work with Sugeno-type
fuzzy inference optimized by using ANN, systems
systems (ANFIS) which may lead to
accurate load shedding

4. Genetic algorithms GA is a global GA takes a long time to determine the

(GA) optimization technique for load shedding amount. This relative
solving nonlinear, multi- slowness limits their usage for online
objective problems. application.

5. Particle swarm PSO computation is PSO is easily interrupted by partial

optimization (PSO) simple and can find the optimization.
optimum value.
(Source: [48], [57], [67], [83] and [105])

5.4 DESCRIPTION OF THE PROBLEM: CASE: I

In this section, a practical substation data are taken for the case. The data
from a Thiruvalam sub station is taken. These data are tabulated and computational –
fuzzy based load shedding is carried on it. The below is the methodology adopted for
load shedding.

1. Collection of data from the sub-station.

2. Tabulation of the values of Voltage, Reactive power and Real

power.

3. Find out the value of the maximum permissible value of voltage,

reactive power and real power.
 108

4. Calculate dV/dQ (voltage maximum -bus voltage/reactive power

maximum-bus reactive power).

5. Tabulate the different value of dV/dQ.

6. The summation of all value of the above-tabulated values will be

the denominator in the formula used.

7. Each value in the above table will be the numerator in the formula.

8. Find out Pdiff (Shaft power-Electrical power) using the formula

mentioned in equation 5.1.

9. Find out the value of load to be shed using the classical formula
mentioned in equation 5.2.

Table 5.4 Thiruvalam substation data and load to be shed values

Voltage Q Mvar Pdiff in MW Load to be

KV Shed MW
101 49 9 3.18
101 49 18 18.59
103 54 3 6.13
103 29 33 14.09
104 45 27 17.19
104 50 24 17.19
105 39 24 10.32
105 39 17 10.59
106 45 30 11.43
106 30 39 28.19
106 59 35 15
107 40 39 21.2
107 44 18 8.29
107 54 3 2.64
108 45 27 5.71
109 34 27 0.448
 109

From the inputs received from the substation, the above table 5.4 is
obtained. Here, a prominent point to be observed is that in the above table the load to
be shed is calculated from classical formulae for load shedding.

The above methodology can be expressed in a flow as follows.

Start

Measure the voltage, reactive power,

real power, frequency of the bus

No
Check if frequency < 49.5 Hz ‘or’ Load shedding not
if voltage < 0.97 pu required
Yes

Begin load shedding procedure

Calculate the dV/dQ and P(diff)

Calculate the load to be shed

using the formula

Stop

Figure 5.5 Flow chart explaining basic load shedding procedure

 110

5.4.1 Thiruvalam
hiruvalam Substation One Line Diagram

Figure 5.6 Thiruvalam sub-station (Courtesy: TNEB)

5.4.2 Results and Interpretations

I

The main bottom line in this research, case is that the application of fuzzy
as an intelligent system for load shedding problems. Here,
Here the problem taken for the
study is already explained in the above Table 5.4 and the results obtained through
conventional methodology is also presented in the same table. The main idea in this
case is that, whether fuzzy helps in arriving an optimality
op and an improved decision
process in addressing the non-linearity.
non

The membership functions

function are assigned as follows.

Input voltage (L1)

190-195-200 — LOW (MF1)

200-205-210 — MEDIUM (MF2)

210-215-220 — HIGH (MF3)

 111

Reactive Power (L2)

25-31-38 — LOW (MF1)

38-45-52 — MEDIUM (MF2)

52-59-66 — HIGH (MF3)

Pdiff (L3)

0-7-14 — LOW (MF1)

14-21-28 — MEDIUM (MF2)

28-35-42 — HIGH (MF3)

Output

LOAD TO BE SHED (MW)

0- 4.5-9 — LOW (MF1)

9-13.5-18 — MEDIUM (MF2)

18-22.5-27 — HIGH (MF3)

The membership functions are formed as above. Then a fuzzy linguistic

control table is formed as below.

Table 5.5 Fuzzy Linguistic Control for 230 KV Bus System

Input 1 Input 2 Input 3

Output
Voltage MVAR MW
High High High High
High High Medium Medium
Medium Medium High Medium
 112

Table 5.6 Fuzzy Linguistic Control for 110 KV Bus System

Input 1 Input 2 Input 3

Output
Voltage MVAR MW
Medium Low Low Low
Low Low Medium Medium

Fuzzy linguistic rules are formed, for the input parameters

parameters and for the
different conditions the fuzzy outputs are obtained and are as shown below. These
fuzzy outputs are collectively
ctively tab
tabulated below which are presented in the Table 5.7
and Table 5.8.. Table 5.7 shows the results
sults for 110 KV bus and Table 5.8 shows the
results for 230 KV bus.

Figure 5.7 Condition 1: If the voltage is high, Reactive Power is high and Real
Power is high then output is high.

Voltage (KV) =219 Q (MVAR) = 58.5 Pdiff=34

Pdiff=34.1
.1 Load to be shed= 25
 113

Figure 5.8 Condition 2: If the voltage is high, Reactive power is high and Real
power is medium then output is medium. .

Voltage (KV) =216 Q (MVAR) =62.1 Pdiff= 23 Load to be Shed=17

 114

Figure 5.9 Condition 3: If the voltage is medium, Reactive Power is Low, Real
Power is Low then Output is Low.

Voltage=210 Q (MVAR) =37 Pdiff=13 Load to be shed=8

 115

Figure 5.10 Condition 4: If the voltage is Low, Reactive Power is low and Real
Power is medium then Output is medium.

Voltage =195 Q (MVAR

MVAR) =27 Pdiff= 24 Load to be Shed=18
 116

Figure 5.11 Condition 5: If Voltage is Medium, Reactive Power is medium

and Real Power is high then output is medium.

Voltage (KV) =210 Q (MVAR) =39 Pdiff=35

Pdiff=35 Load to be shed= 16
 117

So far the data of 110 KV and 230 KV bus systems are scaled using fuzzy
expert system. From the data obtained from the substation, this clearly indicates that
there is a significant benefit over the conventional methods. Also, taking three critical
parameters say voltage, reactive power, shaft power or real power and applying these
data in fuzzy expert system gives the output of the value of load to be shed, which is
much less than the system without fuzzy. This is applied to both 110KV and 230 KV
systems and the results reinforce the same idea. Here Matlab fuzzy tool box is applied
for data scaling and to minimize the load shedding. With the above procedure an
effective value of the amount of load to be shed is arrived. The results are tabulated
below.

Table 5.7 Comparative results between Conventional Method and fuzzy

system approach (for 110 KV bus)

Load to be Shed Load to be shed

Sl. No. Voltage (KV)
(conventional) in MW (Fuzzified) in MW
1. 101 13.83 3.18
2. 103 7.44 6.13
3. 104 20.12 17.19
4. 105 24.12 10.59
5. 106 22.1 15
6. 107 24.18 21.20
7. 108 26.17 5.71
8. 109 15.32 4.48
 118

Table 5.8 Comparative results between conventional method and fuzzy

system approach (for 230 KV bus)

Load to be Shed Load to be shed

Sl. No. Voltage (KV)
(conventional) in MW (Fuzzified) in MW
1. 199 14.72 0.705
2. 203 13.83 15.8
3. 205 20.12 16.6
4. 209 20.88 14
5. 211 24.18 23.31
6. 213 13.32 11.3
7. 215 26.37 18
8. 217 16.2 15.8

The above results are shown graphically below where the yellow line
indicates the system without fuzzy and green line indicates the system with fuzzy.
Two separate graphical displays are shown each for 110 KV bus and 230 KV.

Conventional
Method

Fuzzy expert
system

Figure 5.12 Graph That Represents The Comparison Between The

Conventional Method And Fuzzy Expert Systems, Output Value
Of Amount Of Load To Be Shed For 110KV Bus System.
 119

Conventional
Method

Fuzzy expert
system

Figure 5.13 Graph that represents the comparison between the co

conventional
method and fuzzy expert systems, the output value of the amount
of load to be shed for 230
230KV bus system.

From the Tables

Table 5.7 and 5.8 it is clear that fuzzy scaling of critical
parameters and writing the fuzzy linguistic rules properly,
properly, has clearly given good
results. The amount of load to be shed is minimized.
minimized. This method or this work ha
has
been developed from the data obtained from the substation and the discussions with
the field engineers. It is a logical thought by some
some ways and means the load shedding
has to be minimized. Even though load shedding is the
the last resort and practically
followed method to relieve congestion, there is a constant
c nstant urge both from field
engineers and academic circle to analyze and find some
s ways to minimize
imize the load
shedding. In this regard, this work can be respected. By building robust electronics in
such a way that the fuzzy outputs are tuned to a fuzzy
fu controller,
roller, which is again
governed by another control device with the feedback system and data acquisition
system to monitor the bus parameters, when these forms a robust loop, this idea can
be tested and if the results are satisfactory,
satisfactory this can be implemented in distribution
substations.
 120

5.5 DESCRIPTION OF THE PROBLEM: CASE: II

The same fuzzy logic based load shedding is applied over the IEEE-26
BUS system. In this case the test system selected is a classical test system.

5.5.1 IEEE -26 Bus System

The test system is shown below

Figure 5.14 IEEE -26 BUS system (Source : Hadi Saadat, 201039)

The following are the tables that represent the load data, dv/dt, dq/dt,df/dt
data for the above test case.
 121

Table 5.9 Load data for IEEE 26 Bus System

S. No. Voltage in p.u Reactive power in Frequency

p.u
1. 1.025 0.41 48.7
2. 1.020 0.15 48.5
3. 1.035 0.5 49.5
4. 1.050 0.1 49.2
5. 1.045 0.3 49.7
6. 0.999 0.29 48.5
7. 0.994 0 50
8. 0.997 0 50
9. 1.009 0.5 49.5
10. 0.989 0 50
11. 0.997 0.15 48.5
12. 0.993 0.48 48.7
13. 1.014 0.15 48.5
14. 1.000 0.12 48.5
15. 0.991 0.31 48.6
16. 0.983 0.27 48.5
17. 0.987 0.38 48.7
18. 1.007 0.67 48.7
19. 1.004 0.15 48.5
20. 0.980 0.27 48.7
21. 0.977 0.23 48.7
22. 0.978 0.22 48.7
23. 0.976 0.12 49.3
24. 0.968 0.27 48.7
25. 0.974 0.13 49.4
26. 1.015 0.2 49.8
 122

Table 5.10 Change in voltage, reactive power and frequency for IEEE 26 bus
system

S. No. dv/dt dq/dt df/dt

1. 0.025 0.59 1.3
2. 0.02 0.85 1.5
3. 0.035 0.5 0.5
4. 0.05 0.9 0.8
5. 0.045 0.7 0.3
6. 0.001 0.71 1.5
7. 0.006 1 0
8. 0.009 0.5 0.5
9. 0.011 1 0
10. 0.003 0.85 1.5
11. 0.007 0.52 1.3
12. 0.014 0.85 1.5
13. 0 0.88 1.5
14. 0.009 0.69 1.4
15. 0.017 0.73 1.5
16. 0.013 0.62 1.3
17. 0.007 0.33 1.3
18. 0.004 0.85 1.5
19. 0.020 0.73 1.3
20. 0.023 0.77 1.3
21. 0.022 0.78 1.3
22. 0.024 0.88 0.7
23. 0.032 0.73 1.3
24. 0.026 0.87 0.6
dV dQ
¦ dt
= 0.4 ¦ dt
= 19.63
 123

Table 5.11 Pdiff and the load to be shed for IEEE -26 Bus System

S. No. Pdiff Load to be shed

1. 0.26 0.49
2. 0.31 0.32
3. 0.1 0.31
4. 0.16 0.39
5. 0.06 0.17
6. 0.31 0.02
7. 0 0
8. 0 0
9. 0.1 0.08
10 0 0
11. 0.31 0.04
12. 0.26 0.15
13. 0.31 0.23
14. 0.31 0
15. 0.28 0.16
16. 0.31 0.32
17. 0.26 0.24
18. 0.26 0.24
19. 0.31 0.06
20. 0.35 0.42
21. 0.33 0.44
22. 0.37 0.46
23. 0.35 0.42
24. 0.35 0.68
25. 0.33 0.44
26. 0.04 0.03
 124

Pdiff and the load to be shed are also found by the conventional rule set as
already discussed in the equation (5.2).

5.5.2 Results and Interpretations

The following is the methodology adopted in this case; fuzzy rules are
formed based on the methodology.

It is briefly given as follows.

• After calculations, the amount of load to be shed is known.

• Now a new column in which load to be shed and in another column

the Pdiff values are entered.

• From the table, the minimum and maximum ranges of values of

each variable are calculated to create the membership function.

• Membership functions are created for voltage, frequency, Pdiff,

reactive power by dividing the ranges into 3 categories, for simple
calculations.

• Membership function for the above consideration is LOW,

MEDIUM and HIGH.

• Fix the corresponding values for low, medium and high.

• Create rules based on membership function. (If-then rules)

• Create a fuzzy linguistic control table.

• Apply the values in fuzzy MATLAB software and create the rules
from the table.

• It must be noted that the rules which are created must satisfy

• all the values within the given range

• For the given value the amount of load shed to be shed is

displayed.

• The final results obtained from the above fuzzy process are better
than the conventional methods.
 125

The membership function is formed as follows.

Input Voltage (L1)

0.968-1.05  Range

0.968-0.981-0.99 - LOW (MF1)

0.995-1.008-1.022 - MEDIUM (MF2)
1.022-1.036-1.05  HIGH (MF3)

Reactive Power (L2)

0-0.5  Range

0-0.083-0.166  LOW (MF1)

0.166-0.249-0.33  MEDIUM (MF2)

0.33-0.413-0.5  HIGH (MF3)

Frequency (L3)

48.7-50  Range

48.7-48.915-49.13  LOW (MF1)

49.13-49.34-49.56  MEDIUM (MF2)

49.56-49.78-50  HIGH (MF3)

Power Difference (L4)

0-0.37  Range

0-0.06-0.12  LOW (MF1)

0.12-0.18-0.24  MEDIUM (MF2)

0.24-0.3-0.37  HIGH (MF3)

 126

Load to be shed: (Output)

0-0.68  Range

0-0.12-0.23  LOW (MF1)

0.23-0.35-0.46  MEDIUM (MF2)

0.46-0.57-0.68  HIGH (MF3)

Fuzzy linguistic control table is formed as below

Table 5.12 Fuzzy linguistic control table

Input 1 Input 2 Input 3 Input 4 Output

Voltage Reactive power Frequency Pdiff Load Shed
Medium Medium High Low Low
Low Medium High Low Low
High High Medium Low Medium
High Low Medium Medium Medium
Low Medium Low High High

The membership function for the load data is shown below. The
procedures to create membership function for the load data is same as followed above.

Input Voltage (L1)

1.015-1.05 - Range
1.015-1.021-1.02 - Low (MF1)
1.027-1.033-1.039 - Medium (MF2)
1.039-1.045-1.05 - High (MF3)

Reactive Power (L2)

0.33-2.24 - Range
0.33-0.65-0.97 - Low (MF1)
0.97-1.29-1.61 - Medium (MF2)
1.61-1.93-2.24 - High (MF3)
 127

Frequency (L3)

48.5-49.7 - Range
48.5-48.7-48.9 - Low (MF1)
48.9-49.1-49.3 - Medium (MF2)
49.3-49.5-49.7 - High (MF3)

Power Difference (L4)

0.06-0.31 - Range
0.06-0.1-0.14 - Low (MF1)
0.14-0.18-0.22 - Medium (MF2)
0.22-0.26-0.31 - High (MF3)

Load to be shed: (Output)

0.04-0.89 - Range
0.04-0.18-0.32 - Low (MF1)
0.32-0.46-0.6 - Medium (MF2)
0.6-0.74-0.89 `- High (MF3)

The control table adopted is given below.

Table 5.13 Fuzzy linguistic control table

Input 1 Input 2 Input 3 Output

Voltage Frequency Pdiff Load to be Shed
Low Low Low Low
Low Low High Medium
Medium Medium Medium Low
High High Medium Low
High Low High High
Low High Low Low
 128

The generator data, transformer data and dv/ dt values for an IEEE 26 bus
system is provided below.

Table 5.14 Generator data

S. No. Voltage Reactive Power Frequency

1. 1.025 2.24 48.7
2. 1.020 1.25 48.5
3. 1.035 0.63 49.3
4. 1.050 0.49 49.5
5. 1.045 1.24 48.5
6. 1.015 0.33 49.7

Table 5.15 Transformer data

Transformer Description Tap Setting per Unit

2–3 0.960
2 – 13 0.960
3 – 13 1.017
4–8 1.050
4 – 12 1.050
6 – 19 0.950
7–9 0.950

Table 5.16 Transformer data

S. No. dv/dt dq/dt Pdiff Load to be shed

1. 0.025 1.24 0.27 0.08
2. 0.020 0.25 0.31 0.38
3. 0.035 0.37 0.14 0.20
4. 0.050 0.5 0.10 0.15
5. 0.045 0.24 0.31 0.89
6. 0.015 0.33 0.06 0.04
 129

A snapshot of the fuzzy output is shown below as Figure 5.15 and it is

summarized in the Table 5.17
5.1 shown below.

Figure 5.15 Fuzzy Outputs

Table 5.17 Comparative results between conventional method and fuzzy

system approach
pproach

S. No. Voltage Frequency Load to be shed Load to be shed

(conventional) (fuzzified)
1. 1.025 48.7 0.49 0.4
2. 0.983 48.5 0.32 0.2
3. 0.978 48.7 0.46 0.3
4. 1.014 48.5 0.23 0.02
5. 0.968 48.7 0.68 0.4
 130

The amount of load to be shed is minimized. This method has been

developed from the data obtained through an IEEE 26 bus system. It is a logical
thought by some ways and means the load shedding has to be minimized. Even
though load shedding is the last resort and practically followed method to relieve
congestion, there is a constant urge both from field engineers and academic circle to
analyze and find some ways to minimize the load shedding. In this regard, this work
can be respected.

From the above two cases, it is clear that fuzzy has proven its efficacy in
load shedding. Based on the above, a new model for load shedding is formed – Path
model for effective load shedding. This model not only validates the above two cases
but also proposes a new idea for load shedding problems. This can be taken as a
novelty in this research for the reason that, even though path model is applied in other
areas like life sciences, etc. It is not applied earlier in electrical engineering side and
for load shedding problems. In this way it is a novelty. It is explained as follows.

5.6 A NOVEL METHOD FOR LOAD SHEDDING –PATH MODEL

Power system networks are operating very close to their limits. In order to
prevent cascaded failures and eventually blackouts, load shedding is used as a tool.
There are many methodologies which discusses the effective ways for load shedding.
In this work, a new perspective is presented – path analysis. Path analysis is a
technique under structured equation modelling which explains about dependent
variable and independent variables. Here the output results are modelled and
presented in path analysis, through which the significant variable which is crucial in
determining the output can be identified. The correlations between crucial parameters
which are having an influence in output are also analyzed here.

Path analysis was originally developed by a geneticist Sewall Wright in

1920s to examine the effects of hypothesized models in phylogenetic studies. Here,
systems of equations were written by him based on correlations among variables. It is
a statistical tool used primarily to examine the comparative strength of direct and
indirect relationships among the variables. These variables are critical in influencing
the output. In this research, three cases have been done with fuzzy logic as the
fundamental technique. In the first work a real case example is analyzed.
 131

5.6.1 Formation of Model Fit Summary for Case –I

The data from Thiruvalam substation were taken and the results were
already discussed in section 5.4 and the results are enumerated in the Tables 5.7 and
5.8.

These results are applied over path analysis and the following model was
obtained. The basic theoretical model is given below. Where H1, H2, H3, H4, H5
represents regression coefficients. Also, from the below model it is clear that the
output (load to be shed) is dependent on voltage values, reactive power and also Pdiff.
Where Pdiff is the difference in shaft power. Here, Pdiff is used as the moderating
variable. Through this section, it is been analyzed how a path analysis can be
effectively used to deeply understand how much critical is voltage, reactive power
and Pdiff in deciding the values of load to be shed. From this model, it is observed
that the input variables such as voltage, reactive power do not have a significant
impact on the output (load to be shed). When the Pdiff is used as a moderator variable
the impact on output is significant. The value of H5 > 0.5 it means its impact is more
compared to other variables. In our case, the value is 0.82. Here e1 and e2 are
assigned correction factors in order to address the non-linear behavior of power
systems, because from classical equations it is clear that Pdiff is dependent on df/dt,
which is highly non- linear. By assigning these correction factors e1 and e2 the
accuracy of results are improved.

Figure 5.16 Theoretical model of path analysis

 132

Figure 5.17 Practical model of path analysis for Case I

The overall model fit summary is given below.

Table 5.18 Model fit summary for Case I (Various models)

Model NPAR CMIN DF P CMIN/DF RMR GFI AGFI PGFI

Default model 9 0.615 1 0.433 0.615 2.889 0.980 0.803 0.098
Saturated model 10 0.000 0 0.000 1.000
Independence 4 20.660 6 0.002 3.443 22.908 0.679 0.466 0.408
model

Table 5.19 Base line comparisons

Model NFI ∆1 RFI ρ1 IFI ∆2 TLI ρ2 CFI

Default model 0.970 0.821 1.020 1.158 1.000
Saturated model 1.000 1.000 0 1.000
Independence model 0.000 0.000 0.000 0.000 0.000
 133

Table 5.20 Parsimony-adjusted measures

Model P Ratio PNFI PCFI NCP LO 90 HI 90

Default model 0.167 0.162 0.167 0.000 0.000 5.869
Saturated model 0.000 0.000 0.000 0.000 0.000 0.000
Independence model 1.000 0.000 0.000 14.660 4.440 32.449

Table 5.21 Parsimony-adjusted measures – FMIN

Model FMIN F0 LO90 HI90

Default model 0.041 0.000 0.000 0.391
Saturated model 0.000 0.000 0.000 0.000
Independence model 1.377 0.977 0.296 2.163

Table 5.22 Parsimony-adjusted measures – RMSEA

Model RMSEA LO 90 HI 90 PCLOSE

Default model 0.000 0.000 0.625 0.442
Independence model 0.404 0.222 0.600 0.003

Table 5.23 Parsimony-adjusted measures – AIC

Model AIC BCC BIC CAIC

Default model 18.615 27.615 25.568 34.568
Saturated model 20.000 30.000 27.726 37.726
Independence model 28.660 32.660 31.750 35.750
 134

Table 5.24 Parsimony-adjusted measures – ECVI

Model ECVI LO90 HI90 MECVI

Default model 1.241 1.267 1.658 1.841
Saturated model 1.333 1.333 1.333 2.000
Independence model 1.911 1.229 3.097 2.177

Table 5.25 Parsimony-adjusted measures – HOELTER

Model HOELTER 05 HOELTER 01

Default model 94 162
Independence model 10 13

Table 5.26 Consolidated results for Case: I

Regression Weights: (Group number 1 - Default model)

Estimate S.E. C.R. P

pdiff ← voltage 1.205 1.035 1.164 .244
pdiff ← Reactive power -.613 .282 -2.173 .030
output ← Pdiff .561 .129 4.347 ***
output ← voltage -1.462 .541 -2.705 .007
output ← Reactive power .072 .162 .442 .658

From the model fit summary the CMIN/DF is 0.615 with the P value of
0.433 > 0.05 which indicates this model is free from the bad fit index. Similarly, the
table Root Mean Square Error Approximation (RMSEA) shows the value 0.000
indicates the model is free from the bad fit index. As far as good fitness of the index is
concerned, the value belongs to 0.980 which indicates the model has the goodness of
fit index.
 135

From the estimate we can definitely conclude that reactive power has a
significant impact on Pdiff since the value is less than 0.05. In this case it is 0.03 
0.05 (at the 5 % level of significance). It can be concluded that voltage has an impact
on Pdiff. Similarly, the impact on output with respect to Pdiff is also significant since
the P value is *** (from final estimate table). *** indicates the results are highly
authentic.

5.6.2 Formation of Model Fit Summary for Case –II

In the second case, the load shedding technique using fuzzy logic is
applied over classical IEEE 26 bus system and the crucial results are given below.

Figure 5.18 IEEE 26 bus system

In the above classical test system, based on the fuzzy control rules load
shedding procedure is done and the results are summarized as follows.
 136

Table 5.27 Comparative results between conventional and fuzzy approach

(for case 2)

Frequency Load to be Shed Load to be Shed

S. No. Voltage P.U
Hz (Conventional) (Fuzzified)
1. 1.025 48.7 0.49 0.4
2. 0.983 48.5 0.32 0.2
3. 0.978 48.7 0.46 0.3
4. 1.014 48.5 0.23 0.02
5. 0.968 48.7 0.68 0.4

These results are applied over path analysis and the following model was obtained.

-0.49

0.33

Figure 5.19 Practical Model of Path Analysis for Case II

 137

Table 5.28 Model fit summary for Case II - CMIN

Model NPAR CMIN DF P CMIN/DF

Default model 9 0.736 1 0.391 0.736
Saturated model 10 0.000 0
Independence model 4 17.813 6 0.007 2.969

Table 5.29 Model fit Summary for Case II – RMR, GFI

Model RMR GFI AGFI PGFI

Default model 0.001 0.986 0.857 0.099
Saturated model 0.000 1.000
Independence model 0.005 0.756 0.594 0.454

Table 5.30 Base line comparisons

Model NFI ∆1 RFI ρ1 IFI ∆2 TLI ρ2 CFI

Default model 0.959 0.752 1.016 1.134 1.000
Saturated model 1.000 1.000 0 1.000
Independence model 0.000 0.000 0.000 0.000 0.000

Table 5.31 Parsimony - Adjusted Measures

Model P Ratio PNFI PCFI NCP LO 90 HI 90

Default model 0.167 0.160 0.167 0.000 0.000 6.24
Saturated model 0.000 0.000 0.000 0.000 0.000 0.000
Independence model 1.000 0.000 0.000 11.813 2.756 28.467
 138

Table 5.32 FMIN

Model FMIN F0 LO90 HI90

Default model 0.029 0.000 0.000 0.250
Saturated model 0.000 0.000 0.000 0.000
Independence model 0.713 0.473 0.110 1.139

Table 5.33 RMSEA

Model RMSEA LO90 HI90 PCLOSE

Default model 0.000 0.000 0.500 0.406
Independence model 0.281 0.136 0.436 0.010

Table 5.34 AIC

Model AIC BCC BIC CAIC

Default model 18.736 23.236 30.059 39.059
Saturated model 20.000 25.000 32.581 42.581
Independence model 25.813 27.813 30.845 34.845

Table 5.35 ECVI

Model ECVI LO90 HI90 MECVI

Default model 0.749 0.760 1.010 0.929
Saturated model 0.800 0.800 0.800 1.000
Independence model 1.033 0.670 1.699 1.113
 139

Table 5.36 HOELTER

Model HOELTER 05 HOELTER 01

Default model 131 226
Independence model 18 24

Table 5.37 Consolidated results for Case II

(Regression weights (Group 1: Default model)

Estimate S.E. C.R. P Label

← Voltage -2.838 9.78 -2.902 0.004 H1
Pdiff
← Reactive power 0.182 0.128 1.419 0.156 H2

← PDiff 0.820 0.296 2.767 0.006 H5

Output ← Voltage 0.148 1.676 0.088 0.930 H3

← Reactive power 0.133 0.198 0.674 0.500 H4

From the model fit summary, the CMIN/DF is 0.736 with the P value of
0.391 which is more than 0.05 (0.391>0.05) indicates this model is free from the bad
fit index. Similarly the table RMSEA (Root mean square error approximation) show
the value is 0.000 indicates the model is free from the bad fit index. Table RMR, GFI
indicates good fitness of the index. As far as good fitness of the index is concerned
the value belongs to 0.986 indicates the model has the goodness of fit index. From the
overall table estimate we can definitely conclude that voltage has significant impact
on Pdiff [since P value is less than 0.05, which indicates 5% level of coefficient].
From this it is concluded that voltage has an impact on Pdiff. Similarly, the impact on
output with respect to Pdiff is also significant since the P value is 0.006 which is less
than 0.05 (5% level of significance).

5.6.3 Formation of Model Fit Summary for Case –III

In the third case, the rate of change of voltage (dv/dt) and the rate of
change of reactive power (dq/dt) and its significance in the output (load to be shed) is
 140

analyzed. This analysis is a crucial analysis in understanding the behavior of load

shedding. The third example is tested in IEEE -26 bus system as presented in the
second case. The data are from the second case example.

Table 5.38 Rate of change of voltage, Rate of change of reactive power and
Rate of change of frequency.

Sl.No dv/ dt dq/ dt df/ dt

1. 0.025 0.59 1.3
2. 0.02 0.85 1.5
3. 0.035 0.5 0.5
4. 0.05 0.9 0.8
5. 0.045 0.7 0.3
6. 0.001 0.71 1.5
7. 0.006 1 0
8. 0.003 1 0
9. 0.009 0.5 0.5
10. 0.011 1 0
11. 0.003 0.85 1.5
12. 0.007 0.52 1.3
13. 0.014 0.85 1.5
14. 0 0.88 1.5
15. 0.009 0.69 1.4
16. 0.017 0.73 1.5
17. 0.013 0.62 1.3
18. 0.007 0.33 1.3
19. 0.004 0.85 1.5
20. 0.020 0.73 1.3
21. 0.023 0.77 1.3
22. 0.022 0.78 1.3
23. 0.024 0.88 0.7
24. 0.032 0.73 1.3
25. 0.026 0.87 0.6
26. 0.015 0.8 0.2

This is applied over path analysis and the model fit is as presented below.
 141

dv

dq

Figure 5.20 Practical Model of Path Analysis for Case III

From this model it is observed that the input variables

variables such as change of
voltage, change of reactive power have an impact on output (load to be shed) but
relatively
ly less when compared with the P
Pdiff which is used as a moderator variable
then the impact on output is much significant. Here e1 and e2 are assigned correction
factors in order to address the non-
non linear behavior of power systems, because from
classical equations it is clear that Pdiff is dependent
dependent on df/dt, which is highly non
non-
linear. By assigning these correction factors e1 and
and e2 the accuracy of results are
improved. The overall model fit summary is given below.

Table 5.39 Model fit summary for Case III

I (Various models)

Model NPAR CMIN DF P CMIN/DF RMR GFI AGFI PGFI

Default model 9 0.258 1 0.612 0.258 0.000 0.995 0.949 0.099
Saturated model 10 0.000 0 0 0.000 1.000
Independence 4 14.411 6 0.025 2.402 0.005 0.798 0.664 0.479
model
 142

Table 5.40 Base line comparisons

Model NFI ∆1 RFI ρ1 IFI ∆2 TLI ρ2 CFI

Default model 0.982 0.893 1.055 1.530 1.000
Saturated model 1.000 0.000 1.000 0.000 1.000
Independence model 0.000 0.000 0.000 0.000 0.000

Table 5.41 Parsimony-adjusted measures

Model P Ratio PNFI PCFI NCP LO 90 HI 90

Default model 0.167 0.164 0.167 0.000 0.000 4.457
Saturated model 0.000 0.000 0.000 0.000 0.000 0.000
Independence model 1.000 0.000 0.000 8.411 0.902 23.565

Table 5.42 FMIN

Model FMIN F0 LO90 HI90

Default model 0.010 0.000 0.000 0.178
Saturated model 0.000 0.000 0.000 0.000
Independence model 0.576 0.336 0.036 0.943

Table 5.43 Parsimony-adjusted measures – RMSEA

Model RMSEA LO 90 HI 90 PCLOSE

Default model 0.000 0.000 0.422 0.623
Independence model 0.237 0.078 0.396 0.035
 143

Table 5.44 Parsimony-adjusted measures – AIC

Model AIC BCC BIC CAIC

Default model 18.258 22.758 29.581 38.581
Saturated model 20.000 25.000 32.581 42.581
Independence model 22.411 24.411 27.444 31.444

Table 5.45 Parsimony-adjusted measures – ECVI

Model ECVI LO90 HI90 MECVI

Default model 0.730 0.760 0.938 0.910
Saturated model 0.800 0.800 0.800 1.000
Independence model 0.896 0.596 1.503 0.976

Table 5.46 Parsimony-adjusted measures – HOELTER

Model HOELTER HOELTER

0. 05 0.01
Default model 373 644
Independence model 22 30

Table 5.47 Consolidated results for Case: III

Regression Weights: (Group number 1 - Default model)

Estimate S.E. C.R. P

pdiff ← dv 0.184 1.202 0.153 0.878
pdiff ← dq -0.116 0.148 -0.787 0.431
output ← dv 3.038 1.426 2.131 0.033
output ← dq -0.102 0.177 -0.576 0.565
output ← Pdiff 0.792 0.237 3.341 ***
 144

From the model fit summary the CMIN/DF is 0.258 with the P value of
0.612 which is more than 0.05 (0.612>0.05) indicates this model is free from the bad
fit index. Similarly the table RMSEA (Root mean square error approximation) show
the value is 0.000 indicates the model is free from the bad fit index. Table RMR, GFI
indicates good fitness of the index. As far as good fitness of the index is concerned
the value belongs to 0.995 indicates the model has the goodness of fit index. From the
overall table estimate we can definitely conclude that change of voltage has
significant impact on output [since P value is less than 0.05, which indicates 5% level
of coefficient]. In this case, it is 0.03  0.05 . Also, it can be concluded that Pdiff has
a very strong impact on output – load to be shed. Since the value is *** It proves the
authenticity of the results.

5.7 SUMMARY

Path analysis is a very useful as well as a powerful tool to understand the

behavior of any system. In engineering, particularly in power system engineering, the
tool is seldom applied. In this work the tool is applied for fuzzy based load shedding
problems and the authenticity of the results are verified. Moreover, three cases have
been taken and a path fit summary (path analysis) was done. The results of all the
three cases exhibit goodness of fit (GFI). It can be reiterated that Pdiff (difference in
power) is very crucial in deciding the amount of load to be shed. Further, the
application of this tool can be critically explored in power system problems.