38

CHAPTER 3

VOLTAGE STABILITY ENHANCEMENT

3.1 INTRODUCTION

Voltage stability analysis can be done in several ways. One

methodology is analyzing analytically on small networks by means of
mathematical bifurcations as stability criterion. Analysis of minimum eigen
value is an extraordinary case of this method. Eigenvectors of the system is
used occasionally in modal analysis. For larger networks modal analysis and
the smallest singular value can be used.

This chapter presents the identification of weak bus using modal

analysis technique. Index method is also used to find the weak buses such that
we can implement the FACTS devices for voltage stability enhancement. The
location for connecting the STATCOM is identified from the weak bus for
voltage stability enhancement.

Optimal location of TCSC and UPFC using GA, EGA and PSO is also
done for IEEE 14 bus system.

3.2 VOLTAGE STABILITY

Voltage stability is concerned with the ability of the power system to

maintain acceptable voltages at all buses in the system under normal
conditions and after being subjected to disturbance [2]. Power system will
enter instability state if a large disturbance or increase in load demand, or
change in system condition causes a progressive and unmanageable reduction
in voltage. The main cause causing instability is the inability of the power
system to congregate the requirement of reactive power.
 39

The following section deals with the various techniques available for
performing voltage stability analysis.

Various techniques exist in the literature for implementation of voltage

stability analysis. Some of the conventional methods widely used are
classified into the following types.

1. Q-V curve method.

2. P-V curve method.

3. Modal analysis

4. Continuation power flow method.

5. Index method.

6. Compensators

3.2.1 Q-V curve method

The Q-V curve technique is one among the popular methods to

examine voltage instability problems during the post transient period in power
systems. Two-bus equivalent representation is not required in Q-V curve
method. Reactive power at the bus is plotted against critical Voltage at that
bus. Generator bus is considered as test bus and power-flow program is run
for specified voltage ranges and the reactive power of a bus is found from the
power flow solutions and a curve is plotted against the specified voltage [2].

3.2.2 P-V curve method

P-V curve method is one of the extensively used methods for voltage
stability analysis. Active power margin exists before the point of voltage
instability can be calculated. Changes in real power consumption for change
in voltage are monitored in a radial system. In huge meshed networks, real
 40

power is the total active load in a area and V is the voltage of the critical bus.
Real power transferred throughout a transmission line is also studied by this
method [2].

3.2.3 Continuation Power flow

At the voltage collapse point, it is hard to obtain the power flow

solution because the Jacobian matrix becomes singular. To solve this problem
continuation power flow is used, where the power flow solution closer to
voltage collapse point is obtained. The varying parameter is the loading
factor; however, if system gets nearer to bifurcation the classical power flow
Jacobian fails. At the voltage collapse point the power flow Jacobian is made
nonsingular. Thus this method obviously goes around the collapse point, such
that the user traces the unstable side of the branch.

3.2.4 Modal Analysis

The modal analysis primarily depends on Jacobian matrix of the power

flow. Flowchart for the modal analysis method used is shown inFigure 3.1

At a given operating condition the voltage is stable for a system if for

all the buses in the system, the reactive power injection is increased at a bus
then voltage magnitude also increases in the same bus. A system is voltage
unstable; if the voltage magnitude decreases at a bus as the reactive power
injection is increased at the same bus. In other words, if Q-V sensitivity is
positive then the system is voltage stable for all bus and if Q-V sensitivity is
negative then it is unstable for at least one bus.
 41

Figure 3.1 Flowchart Representation of Modal Analysis

3.2.5 Index Method

The slow variation in reactive power loading towards its maximum

point causes the traditional load flow solution to reach its non- convergence
point. Beyond this point, the ordinary load flow solution does not converge,
which in turn forces the systems to reach the voltage stability limit prior to
bifurcation in the system. The margin measured from the base case solution to
the maximum convergence point in the load flow computation determines the
maximum loadability at a particular bus in the system. Some of the indices are
FVSI, Lmn and LQP

3.2.6 Compensators

Various compensators such as series and shunt compensators are used

for analysing the voltage stability. Nowadays FACTS devices are widely used
for compensation.
 42

3.3 MODAL ANALYSIS AND WEAK BUS IDENTIFICATION

3.3.1 Participation factors

The participation factor of the jth variable in the ith mode is defined as
the product of the jth´s components of the right and left eigenvectors
corresponding to the ith mode. The magnitude of participation factors is
dimensionless. They are independent state variables. In a mode the sum of the
participation factors of all variables and the sum of the participation of all
modes in a variable are equal to one [9].

The suitable description and purpose as to which load bus participates

in the selected modes become more important. This requires a tool, called the
participation factor, for classifying the weakest load buses that are making
major contribution to the selected modes.

For the eigen value i of the matrix JR, if i and i correspond to the
right and left eigenvectors respectively, then the participation factor
measuring the participation of the kth bus in ith mode is defined as

Pki = ki ki (3.1)

V-Q sensitivity at ith mode is determined by the bus with highest Pki.
The area close to voltage instability is determined by bus participation factor
provided by the smallest eigen value of JR.

3.3.2 Location of STATCOM by Modal Analysis Technique

By Modal analysis technique the location of the STATCOM can be

identified using the subsequent procedure [8].

1. Using N-R method Voltage profile for base case is determined.

2. Modal analysis is done to predict the voltage collapse.

 43

3. Calculate Participation factor and identify weak bus.

4. Incorporating the STATCOM into the affected area.

3.3.3 Voltage Profile of IEEE 14 bus and 30 bus System

Power flow analysis is done for the base case for the test system. The
test system considered is IEEE 14 bus and IEEE 30 bus system. The Voltage
Profile graph which shows the voltages of all the buses in the system after the
load flow analysis is represented in the graph as shown in Figure 3.2. for 14
bus and Figure 3.3 for 30 bus system.

Figure 3.2 Voltage profile for 14 bus

Figure3.3 Voltage profile for 30 bus

 44

3.4.4 Eigen value and participation factor

Using MATLAB programming the eigen value and participation factor

for the respective buses are calculated. The Eigen value and the Participation
factor tabulation are shown in Table 3.1 and 3.2

Table 3.1: Eigen Value for 30 Bus

Load Load Load Eigen

Buses Eigen Value Buses Eigen Value Buses Value

3 110.2056 15 19.12576 23 1.023776

4 100.6465 16 19.78167 24 1.726683

6 65.95407 17 18.07852 25 8.785738

7 59.54311 18 16.37527 26 7.436012

9 37.81878 19 13.72794 27 3.580842

10 35.38626 20 13.63338 28 4.050713

12 23.42376 21 11.04466 29 6.020716

14 23.07394 22 0.50603 30 5.452661

Table 3.2(a): Participation Factor For 14 Bus

Load Bus Participation Load Bus Participation

Numbers Factor Numbers Factor
4 0.00907 11 0.110
5 0.00454 12 0.0225
7 0.0691 13 0.0351
9 0.191 14 0.325
10 0.232
 45

Figure 3.4 IEEE 30 bus system

Table 3.2(b): Participation Factor For 30 Bus

Load Bus Participation Load Bus Participation

Numbers Factor Numbers Factor
3 0.0004 19 0.0179
4 0.0005 20 0.0172
6 0.0005 21 0.0176
7 0.0002 22 0.0189
9 0.0037 23 0.0238
10 0.0121 24 0.0395
12 0.0037 25 0.1055
14 0.0081 26 0.1729
15 0.0111 27 0.1028
16 0.0079 28 0.0025
17 0.0115 29 0.1934
18 0.0165 30 0.2118
 46

The graphical representation of the Participation factor for the buses in

the test system is represented in Figure 3.5(a) and Figure 3.5(b)

Figure 3.5(a)Participation Factor for 14 Bus

Figure 3.5(b)Participation Factor for 30 Bus

 47

From the modal analysis calculation and the load flow analysis it is
found that the participation factor is more for the Bus 14, (Participation value
= 0.325) hence the bus is more sensitive to voltage collapse for 14 bus system.

For 30 bus system the participation factor is more for buses 26, 29 and
30 (Participation value = 0.1729, 0.1934 and 0.2118 respectively) compared
to other buses. Therefore the buses 26,29 and 30 is more sensitive to voltage
collapse in a 30 bus system.

3.4 LINE STABILITY INDICES

The main objective of this section is to develop indices based

algorithm for analysing voltage stability and find the maximum loadability,
critical line and the critical voltage for all the load buses. The most sensitive
bus is identified based on the maximum loadability. To develop an algorithm
for line outage contingency analysis based on indices for finding the most
critical line outage and maximum loadability of all the buses. Contingency
ranking is done based on the severity of all the line outages.

3.4.1 Fast Voltage Stability Index

The FVSI is derived from the voltage quadratic equation at the receiving
bus on a two- bus system [18].FVSI can be defined by

4Z 2Q j
FVSI ij 2 (3.2)
Vi X

where Z is the line impedance,

X is the line reactance,

Qjis the reactive power flow at the receiving end and

Vi is the sending end voltage.

 48

The line that gives index value closest to 1 will be the most critical line
of the bus and may lead to system wide instability scenario. This index can
also be used to determine the most sensitive bus on the system.

3.4.2 Line Stability Index (LQP)

Line Stability Index, LQP is given as follows

X ij X ij 2
LQP 4 Pi Qj (3.3)
Vi 2 Vi 2

where

Xij is the line reactance,

Qj is the reactive power flow at the receiving bus,

Vi is the voltage on sending bus and

Pi is the active power flow at the sending bus.

Operating at secure and stable conditions requires the value of LQP

index to be maintained less than 1.

3.4.3 Line Stability Index (L mn)

This index was derived based on a two machine model of the power
system connected by a single transmission. This is represented
mathematically as

4 Q j X ij
L mn 2 (3.4)
Vi sin

Lines that represent Lmn close to 1 reaching their stability limits while
those near zero are stable.
 49

3.5 VOLTAGE STABILITY ANALYSIS (VSA)

The loading pattern is chosen so that each time the load is changed in
only one particular node, keeping the load at other nodes fixed at base case.
Several combinations of real and reactive load pattern are selected to achieve
this and they are listed as follows

Case 1: Change with reactive power loading

Case 2: Change with real power loading

Case 3: Change with both real and reactive power

The analysis is conducted on the IEEE 30 bus system. Load buses are
selected in order to investigate the effect of reactive power loading on FVSI
values which in turn identifies the most critical line with respect to a bus.
Reactive power at load buses are gradually increased from the base case until
their highest permissible load or maximum loadability limit which is the
maximum load that could be injected to a load bus previous to the power flow
solution diverges.

The impact of line outage can be done by contingency analysis in the

system. It is done by removing the lines in the system, in succession for all
pre-determined case. This is done in order to delay the divergence of the load
flow calculation; otherwise the load flow divergence will be faster and
produce an inaccurate outcome of contingency ranking. The process is similar
to voltage stability analysis. The only variation is that load flow calculation is
done with a line outage at a time and there is no need to raise the reactive
power load in the system. The buses are chosen in progression manner to
validate the severity of outages that could occur in the system. FVSI are
calculated on outage for every case. Output from all outage is sorted in
descending order. The outage that results the highest index is the most severe
contingency.
 50

Table 3.3 Voltage stability analysis based on FVSI for case 1

Critical Critical
Bus Bus
Bus Qmax Line Bus Line
Voltage FVSI Q (MVAR) Voltage FVSI
No. (MVAR) connecting No. max connecting
(p.u) (pu)
bus bus
3 286 0.776 1-3 0.998 19 92.5 0.6135 10-20 0.9995
4 470 0.7391 2-4 0.973 20 88.5 0.6903 10-20 0.9941
6 680 0.7864 6-8 0.983 21 159 0.6245 6-10 0.9915
7 280 0.7573 5-7 0.99 22 148 0.6181 6-10 0.9846
9 175 0.8288 9-11 0.981 23 88.2 0.6713 23-24 0.9794
10 173 0.7491 6-10 0.992 24 105 0.6571 22-24 0.9918
12 206 0.8187 12-13 0.988 25 60.8 0.672 24-25 0.9942
14 82.2 0.7788 14-15 0.998 26 28.5 0.7079 25-26 0.9936
15 149 0.6776 12-15 0.99 27 57.1 0.7486 28-27 0.9971
16 112 0.7219 12-16 0.986 28 292 0.7368 8-28 0.9923
17 154 0.588 6-10 0.979 29 34.5 0.6514 27-29 0.9916
18 86.5 0.6686 15-18 0.976 30 31.05 0.6606 27-30 0.9931

3.6 SIMULATION RESULTS

Voltage stability analysis is done on IEEE 30 bus system with line

stability index indicated by FVSI and the results are shown in Table 3.3. For
this purpose MATLAB software package has been used for coding. FVSI are
calculated for all line in the system for every load increase. The line which
has the biggest index with respect to increase in load is determined as the
most critical line. Further increase in the load will cause the line to contain an
index value larger than 1.00 resulting in whole system instable. The lines are
called critical lines that exhibit the maximum FVSI values for each test bus.
The greatest reactive load at FVSI value nearer to 1.00 is assigned as the
maximum permissible load. The maximum loadability of all the load buses is
referred in the Table 3.4. It is obvious that the line index increases as the
reactive power loading increased. Line 34 is the most critical line corresponds
 51

to any load change at bus 26. Bus 26 has the least maximum permissible load
of 0.028589p.u and it is ranked the highest in the system and also the buses 29
and 30 have the lowest maximum loadability, these buses also consider as
critical buses and marked as bold in Table.. Since the bus 3, 4 and 6 have a
maximum permissible load of 2.85768p.u, 3.00089p.u and 3.1546p.u
respectively. These buses are the most secure buses in the system according to
the large maximum loadability.

Table 3.4 Bus Ranking based on Maximum loadability (VSA)

FVSI FVSI
Rank Rank
Case 1 Case 2 Case 3 Case 1 Case 2 Case 3
1 26 26 30 13 22 15 10
2 30 30 29 14 15 22 24
3 29 29 26 15 17 17 22
4 27 25 18 16 21 21 17
5 25 27 14 17 10 12 21
6 14 18 27 18 9 10 4
7 18 19 25 19 12 9 9
8 23 23 20 20 7 3 12
9 20 14 19 21 3 28 3
10 19 20 23 22 28 7 7
11 24 24 15 23 4 4 28
12 16 16 16 24 6 6 6

3.7 VOLTAGE STABILITY ENHANCEMENT IN DEREGULATED

ENVIRONMENT

The objective of this section is to apply an algorithm and simulate to

find the power allocated for each of the generators and find the optimal and
best location for the FACTS controllers such that overall system cost are
minimized. The algorithm used is enhanced Genetic Algorithm to find the
 52

type of the controller to be connected and conventional Newton-Raphson

power flow analysis to find the optimal location of the devices and its rating.

3.8 COST FUNCTION

The objective of this work is to find simultaneously the optimal

generation, choice and location of FACTS controllers so as to minimize the
overall cost function, which comprises of generation cost and investment cost
of FACTS controllers.

3.8.1 Generation Cost Function

The generation cost function is represented by a quadratic polynomial

as follows [70].

2
C2 (PG) = 0 1PG 2PG (3.5)

Whereas PG is the output of the generator (MW), and 0, 1 and 2 are

cost coefficients.

3.8.2 FACTS Controller Cost Function

Based the Siemens AG Database the cost function for the controller
that has been selected to use are as follows

The cost function for UPFC is:

C1UPFC = 0.0003s2 – 0.2691s + 188.22 (US$ / kvar) (3.6)

The cost function for TCSC is:

C1TCSC = 0.0015s2 – 0.7130s + 153.75 (US$ / kVar) (3.7)

 53

The rating of the device is given by

RTCSC = rf * 0.45 – 0.25 (Mvar) (3.8)

RUPFC = rf * 180 (MVar) (3.9)

Where

C1UPFC and C1TCSC are in US$ / kVar, rf– rating factor and s is the
operating range of the FACTS controller in MVar.

3.9OPTIMAL POWER FLOW(OPF) WITH FACTS CONTROLLERS

The formulation of the optimal allocation of FACTS controllers can be

expressed as follows [70]

Minimize CTotal = C1 (f) + C2 (PG) (3.10)

Subjected to E(f,g) = 0 (3.11)

B1 (f) < b1, B2 (g) < b2 (3.12)

Where

CTotal: the overall cost objective function which includes the average
investment costs of FACTS devices C1 (f) and the generation cost C2(PG).

E (f.g): the conventional power flow equations.

B1(f) and B2(g): are the inequality constraints for FACTS controllers
and the conventional power flow respectively.

f and PG: are vectors that represent the variables of FACTS controllers
and the active power outputs of the generators.

g: represents the operating state of the power system.

 54

The unit for generation cost is US$/Hour and for the investment cost of
FACTS controllers are US$. They must be unified into US$/Hour. Normally
the FACTS controllers will be in service for many years. However only a part
of its life time is employed to regulate the power flow. In this work three
years is employed to evaluate the cost function.

The average value of the investment costs are calculated as follows [70]

C1(f) =C(f)/{8760 x 3 } (3.13)

As mentioned above, power system parameters can be changed using

FACTS controllers. These different parameters derive different results on the
objective function in Equation 3.10. Also, the variation of FACTS locations
and FACTS types has also influences on the objective function. Therefore,
using the conventional optimization methods is not easy to find the optimal
location of FACTS devices, types and control parameters simultaneously.

To solve this problem, optimization technique such as Genetic

Algorithm, Enhanced Genetic Algorithm and Particle Swarm Algorithm are
employed in conjunction with conventional NR power flow method.

3.10 GENETIC ALGORITHM

Genetic Algorithms are general purpose optimization algorithms

[34]based on the mechanics of natural selection and genetics. They operate on
string structures (chromosomes), typically a concatenated list of binary digits
representing a coding of the control parameters (phenotype) of a given
problem. Chromosomes themselves are composed of genes. The real value of
a control parameter, encoded in a gene, is called an allele.

GA’s are an attractive alternative to other optimization methods

because of their robustness. There are three major differences between
 55

Genetic Algorithm and conventional optimization algorithms. First, Genetic

Algorithm operates on the encoded string of the problem parameters rather
than the actual parameters of the problem. Each string can be thought of as a
chromosome that completely describes one candidate solution to the problem.
Second, Genetic Algorithm uses a population of points rather than a single
point in their search. This allows the Genetic Algorithm to explore several
areas of the search space simultaneously, reducing the probability of finding
local optima. Third, Genetic Algorithm do not require any prior knowledge,
space limitations, or special properties of the function to be optimized, such as
smoothness, convexity, unimodality, or existence of derivatives. They only
require the evaluation of the so-called fitness function (FF) to assign a quality
value to every solution produced.

Assuming an initial random population produced and evaluated,

genetic evolution takes place by means of three basic genetic operators:

1) Parent selection

2) Crossover

3) Mutation

Parent selection is a simple procedure whereby two chromosomes are

selected from the parent population based on their fitness value. Solutions
with high fitness values have a high probability of contributing new offspring
to the next generation. The selection rule used in this approach is a simple
roulette-wheel selection.

Crossover is an extremely important operator for the Genetic

algorithm. It is responsible for the structure recombination (information
exchange between mating chromosomes) and the convergence speed of the
Genetic algorithm and is usually applied with high probability (0.6–0.9). The
chromosomes of the two parents selected are combined to form new
 56

chromosomes that inherit segments of information stored in parent

chromosomes. Until now, many crossover schemes, such as single point,
multipoint, or uniform crossover have been proposed in the literature.
Uniform crossover has been used in our implementation.

While crossover is the main genetic operator exploiting the

information included in the current generation, it does not produce new
information.

Mutation is the operator responsible for the injection of new

information. With a small probability, random bits of the offspring
chromosomes flip from 0 to 1 and vice versa and give new characteristics that
do not exist in the parent population. In this approach, the mutation operator
is applied with a relatively small probability (0.0001-0.001) to every bit of the
chromosome.

Figure 3.6 Simple genetic algorithm (SGA)

 57

The FF evaluation and genetic evolution take part in an iterative

procedure, which ends when a maximum number of generations is reached, as
shown in Figure 3.6.

When applying Genetic algorithms to solve a particular optimization

problem (OPF in this case), two main issues must be addressed

1) The encoding, i.e., how the problem physical decision variables are
translated to a Genetic Algorithm chromosome and its inverse
operator, decoding;

2) The definition of the FF to be maximized by the Genetic Algorithm is

formed by an appropriate transformation of the initial problem
objective function augmented by penalty terms that penalize the
violation of the problem constraints.

3.10.1 Encoding

In the SGA, shown in Fig. 3.7, after the application of the basic genetic
operators (parent selection, crossover, and mutation) the advanced and
problem-specific operators are applied to produce the new generation. All
chromosomes in the initial population are created at random (every bit in the
chromosome has equal probability of being switched ON or OFF).

Due to the decoding process selection, the corresponding control

variables of the initial population satisfy their upper–lower bound or discrete
value constraints. Population statistics are then used to adaptively change the
crossover and mutation probabilities. If premature convergence is detected the
mutation probability is increased and the crossover probability is decreased.
The contrary happens in the case of high population diversity.
 58

3.10.2 Fitness Function

GAs is usually designed so as to maximize the FF, which is a measure

of the quality of each candidate solution. The objective of the OPF problem is
to minimize the total operating cost.

Therefore, a transformation is needed to convert the cost objective of

the OPF problem to an appropriate FF to be maximized by the GA. The OPF
functional operating constraints are included in the GA solution by
augmenting the GA FF by appropriate penalty terms for each violated
functional constraint. Constraints on the control variables are automatically
satisfied by the selected GA encoding/decoding scheme.

Therefore, the GA FF is formed as follows [71]

= (3.14)
( )

= ( , ). ( , )) (3.15)

Where

FF-fitness function;

A- constant;

Fi(PGi) - fuel cost of unit i

H(.) -Heaviside (step) function;

NG -number of units;

Nc -number of functional operating constraints.

Penj – penalty function for functional operating constraint j

wj– weighting factor for functional operating constraint j

 59

3.11 ENHANCED GENETIC ALGORITHM (EGA)

In the EGA, [37] after the application of the basic genetic operators
(parent selection, crossover, and mutation) the advanced and problem-specific
operators are applied to produce the new generation.

All chromosomes in the initial population are created at random (every

bit in the chromosome has equal probability of being switched ON or OFF).
Due to the decoding process selected, the corresponding control variables of
the initial population satisfy their upper–lower bound or discrete value
constraints. However, the initial population candidate solutions may not
satisfy the functional operating constraints or even the load flow constraints
since the random, within limits, selection of the control variables may lead to
load flow divergence.

Population statistics computed for the new generation include

maximum, minimum, and average fitness values and the 90% percentile.

Population statistics are then used to adaptively change the crossover

and mutation probabilities. If premature convergence is detected the mutation
probability is increased and the crossover probability is decreased. The
contrary happens in the case of high population diversity.
 60

Figure 3.7 Enhanced Genetic Algorithm Flow Sequence

3.11.1 Advanced and Problem-Specific Genetic Operators

One of the most important issues in the genetic evolution is the

effective rearrangement of the genotype information. In the Simple Genetic
Algorithm, crossover is the main genetic operator responsible or the
exploitation of information while mutation brings new non-existent bit
structures. It is widely recognized that the Simple Genetic Algorithm scheme
is capable of locating the neighborhood of the optimal or near-optimal
solutions, but in general, requires a large number of generations to converge.
This problem becomes more intense for large-scale optimization problems
with difficult search spaces and lengthy chromosomes, where the possibility
for the Simple Genetic Algorithm to get trapped in local optimal increases and
the convergence speed of the Simple Genetic Algorithm decreases.
 61

At this point, a suitable combination of the basic, advanced, and

problem-specific genetic operators must be introduced in order to enhance the
performance of the Genetic Algorithm. Advanced and problem-specific
genetic operators usually combine local search techniques and expertise
derived from the nature of the problem.

A set of advanced and problem-specific genetic operators has been

added to the Simple Genetic Algorithm in order to increase its convergence
speed and improve the quality of solutions. Our interest was focused on
constructing simple yet powerful enhanced genetic operators that effectively
explore the problem search space. The advanced features included in this GA
implementation are as follows.

1) Fitness Scaling: In order to avoid early domination of extraordinary

strings and to encourage a healthy competition among equals, a scaling
of the fitness of the population is necessary. In this approach, the
fitness is scaled by a linear transformation.

2) Elitism: Elitism ensures that the best solution found thus far is never
lost when moving from one generation to another. The best solution of
each generation replaces a randomly selected chromosome in the new
generation.

3) Hill Climbing: In order to increase the GA search speed at smooth

areas of the search space a hill-climbing operator is introduced, which
perturbs a randomly selected control variable. The modified
chromosome is accepted if there is an increase in FF value; otherwise,
the old chromosome remains unchanged. This operator is applied only
to the best chromosome (elite) of every generation.

In addition to the above advanced features, which are called

“advanced” despite their wide use in most recent GA implementations to
 62

distinguish between the Simple genetic algorithm and our EGA, operators
specific to the OPF problem have been added.

All problem-specific operators introduce random modification to all

chromosomes of a new generation. If the modified chromosome proves to
have better fitness, it replaces the original one in the new population.
Otherwise, the original chromosome is retained in the new population. All
problem-specific operators are applied with a probability of 0.2. The
following problem-specific operators have been used.

Gene Swap Operator (GSO) This operator randomly selects two

genes in a chromosome and swaps their values. This operator swaps the active
power output of two units, the voltage magnitude of two-generation buses,
etc. Swapping among different types of control variables is not allowed.

Gene Cross-Swap Operator (GCSO) The GCSO is a variant of the

GSO. It randomly selects two different chromosomes from the population and
two genes, one from every selected chromosome, and swaps their values.
While crossover exchanges information between high-fit chromosomes, the
GCSO searches for alternative alleles, exploiting information stored even in
low-fit strings.

Gene Copy Operator (GCO) This operator randomly selects one

gene in a chromosome and with equal probability copies its value to the
predecessor or the successor gene of the same control type. This operator has
been introduced in order to force consecutive controls (e.g., identical units on
the same bus) to operate at the same output level.

Gene Inverse Operator (GIO) This operator acts like a sophisticated

mutation operator. It randomly selects one gene in a chromosome and
inverses its bit-values from one to zero and vice versa. The GIO searches for
 63

bit structures of improved performance, exploits new areas of the search space
far away from the current solution, and retains the diversity of the population.

Gene Max-Min Operator (GMMO) The GMMO tries to identify

binding control variable upper/lower limit constraints. It selects a random
gene in a chromosome and, with the same probability (0.5), fills its area with
1s or 0s.

3.12 PARTICLE SWARM OPTIMIZATION (PSO)

PSO is a population-based stochastic optimization technique

developed by Dr. Eberhart and Dr. Kennedy in 1995, inspired by the social
behavior of bird flocking or fish schooling. These phenomena can also be
observed on insect colonies, e.g. bees. It is applicable to solving a number of
problems where local methods fail or their usage is ineffective as in this case.
One of the most important features of PSO is the ability of optimizing large
complex multi-criterial combinatorial problems where the problem with the
design of criterial function occurs, for example, it is hard to derive or is not
continuous.

PSO however does not need thisas it only requires the evaluation of
each solution by the fitness function depending on the set of optimized
parameters. This function is also used by GA and so is the idea of the
initialization of parameter setup as a random generation. The main advantage
of PSO compared to GA is the simpler method of providing new solutions
based only on two variables - velocity and position related by two linear
equations. Each possible solution is represented by a particle, which flies
through the searched space, which is limited by restrictive maximum and
minimum values, toward the current optimal position. The particle has its
direction and speed of movement (velocity) but it can also randomly decide to
move to the best position of all positions or to its own best position. Each
 64

particle holds information about its own position (which represents one
potential solution), the velocity and the position with the best fitness function
it ever has flown through.

A. Implementation

The program was implemented in the MATLAB environment. The

position here represents one potential solution, the velocity shows the trend of
this particle, and both parameters are represented by a vector in the program
implementation. The particles were coded by natural numbers. The position of
each element in the vector space represents the number of the node in which a
shunt capacitor should be placed whose value designates the capacitor type.
The whole set of particles at a time is called the population. The subset made
of newly born particles is called the generation.

The first generation of particles is produced with random position and

velocity. Particle velocity is checked whether it is within the limits. The top
speed can be different for each unit of velocity vector. If the velocity
component exceeds the maximum allowed value, then it is set to the top
value. After this correction, the solution is evaluated by the fitness function.
The fitness function plays a key role in the program; therefore it is necessary
to describe it in more details.

B. Fitness and penalization functions

The fitness function evaluates the quality of solutions and it

incorporates numerous parameters, such as the capital cost of capacitors,
expenses covering the power losses in the network per year, and function .
The power losses are calculated by steady state analysis of the network. The
output of the fitness function is total yearly operational costs of the network.
The lower the fitness function value, the better the solution. The fitness
function is calculated by the following equation [71]
 65

= (3.16)
( )

Where

FF- fitness function;

A- constant;

Fi(PGi)- fuel cost of unit i

H(.)-Heaviside (step) function;

NG- number of units;

Nc- number of functional operating constraints (2).

= ( , ). ( , )) (3.17)

C. Next population

After evaluation, the solutions can be sorted with respect to their

fitness functions and it is possible to develop a new generation.

The first cycle ends after the creation of a new velocity vector and the
calculation of the new position of the particular particle.

The new vector of velocity is calculated by the formula:

= + + (3.18)
 66

Figure 3.8. The principle of PSO

Where denotes the new vector of velocity, is the original vector of

velocity, c1 and c2 are the constants which are set to the weight of differences
of positions, n1 and n2 are the random variables, is the best position of
particle, is the current position of particle, and best is the best
position of all particles.

The new position is determined by the formula:

= + (3.19)

D. Border

Each particle should be kept in a confined space corresponding to the

parameter limitations. This problem is solved in this program by one of four
methods. In the first case, the particle arriving in the forbidden area returns to
its previous position. In the second case, the particle is held on the border. In
the third case, the particle is bounced back to the allowed space. Bouncing
back to the allowed space can be perfect or imperfect. Regarding the
imperfect bounce, it is possible for the particle to end up in a random position.

In the fourth case, the particle can fly through the forbidden area back
to the allowed space, but on the other side of the allowed space. This
approach can be used in the case of a very specific limited space.
 67

E. Parallel operation

One of the advantages of PSO utilization is the possibility to introduce

parallel operations with mutual coupling, which enables searching in a larger
area of feasible solutions and thereby finding the optimum solution more
quickly. Parallel operation means that instead of a single branch of evolution
several branches are created. These branches influence one another during the
evolution only after a given number of generations when the temporary best
solution of all the branches is transferred to the other branches. Thus the
evolution of the branches is independent, but they can also use the results of
the other branches. This modification limits a potential deadlock of the
algorithm in the local minimum.

A MATLAB coding is developed for each algorithm inter linked with

the conventional Newton Raphson’s method for load flow study. IEEE 14-bus
system is taken to verify the effective operation of the algorithm. The Figure
3.9 shows the line diagram of IEEE 14-bus system.

Figure.3.9 Line diagram of IEEE 14-bus system

 68

The total population size in each algorithm is selected as 30, the

mutation probability as 0.01 and crossover probability as 1.0.

Table 3.5 Result obtained for Genetic Algorithm

Controller Bus
S. No.
Type Rating (p.u) nl nr
1 UPFC -0.681 9 14

2 TCSC -0.977 4 9

3 UPFC -0.353 4 7

4 UPFC 0.6501 3 4
5 TCSC 0.7654 4 9

From the coding results obtained with genetic algorithm it is found that
each time the coding is executed a new result is obtained as tabulated in
Table3.5. From the observation made from the Table 3.5 we could say that
with genetic algorithm analysis the use of TCSC controller with rating of
1.0p.u to -1.0p.u at the line connecting the bus 4 to bus 9.

Table 3.6 Result obtained for Enhanced Genetic Algorithm

Controller bus
S. No.
Type Rating (p.u) nl nr
1 TCSC -0.277 2 3
2 TCSC -1 4 7
3 UPFC 0.9989 1 5
4 UPFC -0.251 1 5
5 TCSC 0.5972 5 6
 69

From the result obtained shown in Table 3.6, it is found that the use of
UPFC controller will be more efficient with a rating of 1.0p.u to -0.25p.u at
the line connecting the bus 1 to bus 5. Though the TCSC controller is
repeated more number of times than UPFC controller the bus location for
TCSC controller is different in each time when TCSC controller is selected.

Table 3.7 Result obtained for Particle Swarm Optimization

Controller bus
SI Rating
Type nl nr
(p.u)
1 UPFC -0.0496 5 6
2 UPFC -0.2579 4 9
3 UPFC -0.1543 5 6
4 TCSC -0.1129 5 6
5 UPFC -0.9757 4 9

Table 3.7 shows the coding result obtained for particle swarm
optimization. The controller selected by this algorithm is UPFC with a rating
ranging from 0.5p.u to -0.5p.u at the line connecting bus 5 to bus 6.

3.13 VOLTAGE STABILITY ENHANCEMENT

To improve the voltage stability of the system, the Voltage Stability

Index (FVSI) which estimates the stability margin of the system is included
into the objective function. The value of FVSI varies between 0 and 1
corresponds to no load and voltage collapse. Reduction in FVSI value
indicates voltage stability improvement.
 70

FVSI has to lie between 0 and 1 to maintain stability in power system.

It has to be minimized to imply voltage stability improvement. Lower the
index higher is the stability of the system.

From the Bus ranking from Voltage stability analysis and Line outage
contingency analysis, it is evident that the optimum locations of the
STATCOM are bus 26, 29 and bus 30. Buses 26, 29 and 30 has been
identified as weak bus through Maximum loadability as well as modal
analysis, hence we can conclude that the FACTS device STATCOM can be
placed at buses 26, 29 and 30 and to improve the voltage stability.

Table 3.8 STATCOM at Bus 30 (Q=31.09 MVAr)

Voltage FVSI
Bus Line
No. With Without No. With Without
FACTS FACTS FACTS FACTS
30 0.9836 0.7896 38 0.8436 0.9972
29 0.8941 0.6993 36 0.4351 0.6104
26 0.9925 0.8671 39 0.4315 0.5824

From table 3.8, it is clear that after connecting a STATCOM at bus 30

with reactive power of 31.09MVAR, the voltages at buses 26, 29 and 30 is
increased. The voltage is increased from 0.8671p.u to 0.9925p.u at bus 26.
after connecting STATCOM at bus 26. The FVSI values at the corresponding
buses are reduced from 0.5824 to 0.4315 indicating that the loadability of
buses 26,29 and 30 is increased.
 71

Table 3.9 STATCOM at Bus 26 (Q=28.58 MVAr)

Voltage FVSI
Bus Line
No. With Without No. With Without
FACTS FACTS FACTS FACTS
30 0.9732 0.9021 34 0.9759 0.9982
29 0.9625 0.8732 36 0.2468 0.4025
26 0.9869 0.7156 33 0.2302 0.3967

From table 3.9, after connecting a STATCOM at bus 26 with reactive

power of 28.58MVAR, the voltages at buses 26, 29 and 30 is increased.
Further it is concluded after connecting STATCOM at bus 26 the FVSI values
at bus reduced from 0.3967 to 0.2302 and voltage increased from 0.7156p.u to
0.9869p.u indicating that the loadability of buses 26 and other weak bus is
increased.

3.14 SUMMARY

Voltage stability analysis has been carried out by modal analysis and
indices method. The test system considered for performing the studies are
IEEE 14 bus and IEEE 30 bus system.

In modal analysis technique, the eigen values are estimated from the
Jacobian matrix formed by N-R power flow method. The participation factor
is estimated by using the eigen value. The bus corresponding to the highest
participation factor is considered as weak bus. For IEEE 14 bus system 14th
bus is identified as weak bus and for IEEE 30 bus system buses 26, 29 and 30
are the weak buses.

In the indices method the relation between Q and V has been used for
formulation of different indices. Based on the indices value the critical bus
 72

and critical lines have been estimated for IEEE 30 bus system in the critical
buses are 26,29 and 30.

To improve the stability in a deregulated environment TCSC and

UPFC devices have been implemented in IEEE 14 bus system. The rating,
location and the cost function of FACTS devices have been developed. The
cost minimization of TCSC and UPFC has been optimized using GA, EGA,
and PSO.

By connecting STATCOM at bus 30 and 26 the voltage profile as well

as loadability of the weak bus is enhanced.