Sie sind auf Seite 1von 13

S. Durairaj and D.

Devaraj

REACTIVE POWER DISPATCH INCORPORATING


THYRISTOR CONTROLLED SERIES CAPACITORS
USING IMPROVED GENETIC ALGORITHM

S. Durairaj * and D. Devaraj †


Kalasalingam University, Tamilnadu, India

:‫اﻟﺨﻼﺻـﺔ‬
‫ﺗﻘﺪم هﺬﻩ اﻟﻮرﻗﺔ ﺧﻮارزﻣﻴﺎت ﺟﻴﻨﻴ ﺔ ﻣﻄ ﻮرة ﻟﺤ ﻞ ﻣ ﺸﻜﻠﺔ ﺗﻮزﻳ ﻊ اﻟﻘ ﺪرة اﻟﻐﻴ ﺮ ﻓﻌﺎﻟ ﺔ ﻓ ﻲ ﻧﻈ ﻢ اﻟﻘ ﻮى اﻟﺘ ﻲ ﺗﺘ ﻀﻤﻦ ﻧﺒ ﺎﺋﻂ اﻟﻤﻜﺜﻔ ﺎت اﻟﻤﺘ ﺼﻠﺔ ﺑ ﺎﻟﺘﻮاﻟﻲ و‬
‫ آﻤ ﺎ ﺗ ﻢ اﺳ ﺘﺨﺪام ﺧﻮارزﻣﻴ ﺔ‬.‫ و ﻟﻘﺪ ﺗﻢ اﻳﺠﺎد اﻟﻤﻮﻗﻊ اﻷﻣﺜﻞ ﻟﻬﺬﻩ اﻟﻨﺒﺎﺋﻂ ﺑﺎﺳﺘﺨﺪام ﺗﺤﺎﻟﻴﻞ اﻟﺤﺴﺎﺳﻴﺔ ﻟﻠﻨﻈ ﺎم ﺗﺤ ﺖ اﻟﺪراﺳ ﺔ‬.‫اﻟﻤﺘﺤﻜﻢ ﻓﻴﻬﺎ ﺑﺎﺳﺘﺨﺪام اﻟﺜﺎﻳﺮوﺳﺘﻮر‬
‫ و ﻟﻘﺪ أﺟﺮﻳﺖ ﺑﻌﺾ اﻟﺘﻌﺪﻳﻼت ﻓﻲ ﺗﻤﺜﻴﻞ ﻣﺘﻐﻴ ﺮات اﻟﺤ ﻞ‬.‫ﺟﻴﻨﻴﺔ ذات ﺗﺸﻔﻴﺮ ﺛﻨﺎﺋﻲ و اﻧﺘﻘﺎء ﺗﻨﺎﻓﺴﻲ ﻣﻊ ﻧﻘﻄﺘﻲ ﺗﻘﺎﻃﻊ ﻟﺤﻞ ﻣﺸﻜﻠﺔ اﻷﻣﺜﻠﺔ اﻟﻐﻴﺮ ﺧﻄﻴﺔ اﻟﻤﺮآﺒﺔ‬
IEEE 30-bus ‫ اﻷول ﻧﻈ ﺎم‬:‫ و ﻗ ﺪ ﺗ ﻢ اﺧﺘﺒ ﺎر اﻟﺨﻮارزﻣﻴ ﺔ اﻟﻤﻘﺘﺮﺣ ﺔ ﻋﻠ ﻰ ﻧﻈ ﺎﻣﻴﻦ‬.‫ﻟﻸﺧ ﺬ ﻓ ﻲ اﻻﻋﺘﺒ ﺎر اﻟﻄﺒﻴﻌ ﺔ اﻟﻤﺘﻘﻄﻌ ﺔ ﻟ ﻀﺒﻂ اﻟﻤﺤ ﻮﻻت و اﻟﻤﻜﺜﻔ ﺎت‬
.‫ وﻟﻘﺪ أﺛﺒﺘﺖ اﻟﺪراﺳﺔ ﻓﻌﺎﻟﻴﺔ أآﺜﺮ ﻟﻠﺨﻮارزﻣﻴﺔ اﻟﻤﻘﺘﺮﺣﺔ ﻣﻘﺎرﻧﺔ ﺑﺎﻟﻄﺮق اﻟﺘﻘﻠﻴﺪﻳﺔ‬.76-bus ‫واﻟﺜﺎﻧﻲ ﻧﻈﺎم اﻟﻘﻮى اﻟﻬﻨﺪي اﻟﻤﺨﻔﺾ‬

ABSTRACT
This paper presents an improved Genetic Algorithm (IGA) approach to solve the Reactive Power Dispatch (RPD)
in power system incorporating thyristor controlled series capacitor (TCSC) devices. In this paper, the optimal
location of the TCSC devices is found by sensitivity analysis for a given test system. A binary-coded GA with
tournament selection, two point crossover, and bit-wise mutation is used to solve this mixed-integer non-linear
optimization problem. Some modifications have been made in the representation of the solution variables in the
genetic population in order to take care of the discrete nature of transformer tap setting and capacitor bank.. The test
and validation of proposed algorithm are conducted on two test systems: one is the IEEE 30-bus system, and the
other is a reduced practical 76-bus Indian power system and is found to be more effective than the conventional
methods.
Key words: reactive power dispatch, optimization, genetic algorithm, line loss minimization, thyristor controlled
series capacitor, voltage profile

*Corresponding Author
Department of Electrical and Electronics Engineering
Kalasalingam University, Krishnankoil-626 190
Srivilliputtur (tk), Tamilnadu, India
E-mail: durairaj_eee@yahoo.com † E-mail: deva230@yahoo.com

Paper Received 28 April 2007; Accepted 9 February 2009

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 173
S. Durairaj and D. Devaraj

REACTIVE POWER DISPATCH INCORPORATING THYRISTOR CONTROLLED


SERIES CAPACITORS USING IMPROVED GENETIC ALGORITHM

1. INTRODUCTION
The purpose of the reactive power dispatch in power system is to identify the control variables which minimize
the given objective function while satisfying the unit and system constraints. This goal is achieved by proper
adjustment of reactive power variables like generator voltage magnitude (Vgi), reactive power generation of capacitor
bank (Qci), transformer tap setting (tk) and reactance of thyristor controlled series capacitor (Xtcsc). Scheduling of
reactive power in an optimum manner reduces circulating volt ampere reactive (VAR), thereby promoting a flatter
voltage profile which leads to appreciable power saving on account of reduced system losses. To solve the RPD
problem, a number of conventional optimization techniques have been proposed. They include Gradient method [1],
Quadratic programming (QP) [2], Interior point methods [3], and linear programming (LP) [4]. All these techniques
rely on convexity to find the global minimum. But due to the non-linear and non-convex nature of the RPD problem,
the methods based on these assumptions do not guarantee to find the global optimum. These traditional techniques
also suffer from bad starting points and frequently converge to a local minimum or even diverge. Recently,
evolutionary computation techniques like Genetic Algorithm (GA) [5], Evolutionary Programming (EP) [6] and
Evolutionary Strategy (ES) [7] have been applied to solve the optimal dispatch problems. Although, these methods
have solved the RPD problem successfully, these methods have considered all the control variables as continuous.
The power flow in a transmission line mainly depends on the difference of voltage angles of terminal buses. The
power flow can be controlled by adjusting any of the bus voltage angles. It is attractive for electrical utilities to have
a way of permitting more efficient use of the transmission lines by controlling the power flows. Until a few years
ago, the only means of carrying out this function were electromechnanical devices such as switched inductors or
capacitor banks and phase shifting transformers. However, specific problems related to these devices make them not
very efficient in some situations. The RPD problem which can be treated as a particular case of the Optimal Power
Flow (OPF) problem in which the control variables are the ones which directly related to the reactive power flow in
the system.
Several papers have reported the OPF incorporating flexible alternating current transmission system (FACTS)
devices. Noroozian and Andersson [8] have proposed a method for solving the power flow control problem
incorporating FACTS devices based on decomposition and locally measurable variables. Taranto et al. [9] have
proposed a decomposition method for representing FACTS devices in optimal power flow model. This methodology
is based on mathematical decomposition and network compensation techniques. This method dealt with the
representation of series compensators and phase shifters, but this method did not consider the specified line flow
constraints. Chung et.al, [10] described a method to incorporate the power flow control using FACTS for OPF
problem. Ambriz-Perez et al [11] solved the OPF incorporating FACTS devices using Newton’s method, leading to
highly robust iterative solutions.
In the present work, one of the FACTS devices, thyristor controlled series capacitor has been considered from a
static point of view to reduce the total system real power loss. It is important to ascertain the location for placement
of these devices because of their considerable cost. Hence, a method based on sensitivity approach, has been
suggested for optimal placement of the TCSC devices. In this work both the transformer setting and value of shunt
capacitance are treated as discrete variables through modifications in the representation of the solution variables. An
improved version of genetic algorithm is proposed to solve this mixed integer non-linear optimization problem. The
effectiveness of this algorithm is demonstrated through IEEE 30-bus and also in a practical Indian 76- bus systems.
LIST OF SYMBOLS
P loss Network real power loss
Pi, Qi Real and reactive powers injected into network at bus i
Gij, Bij Mutual conductance and susceptance between bus i and bus j
Gii,Bii Self- conductance and susceptance of bus i
Qgi Reactive power generation at bus i
QCi Reactive power generated by ith capacitor bank
tk Tap setting of transformer at branch k
Vi Voltage magnitude at bus i
θij Voltage angle difference between bus i and j

174 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

Sl Apparent power flow through the lth branch


gk Conductance of branch k
NB Total number of buses
NB-1 Total number of buses excluding slack bus
NPQ Number of PQ buses
Ng Number of generator buses
Nc Number of capacitor banks
NT Number of tap-setting transformer branches
Nl Number of branches in the system
X Reactance of the line where TCSC has been installed
Xtcsc Reactance of TCSC
Ntcsc Set of TCSC indices
2. PROBLEM FORMULATION
The purpose of the reactive power dispatch in power system is to identify the control variables which minimize
the system real power loss while satisfying the unit and system constraints. This goal is achieved by proper
adjustment of reactive power variables. This is mathematically stated as;

Minimize Ploss = ∑gk (Vi2 +Vj2 − 2ViVj cosθij ) (1)


k∈Nl
k =(i, j)

The real power loss given by (1) is a non-linear function of bus voltages and phase angles which are a function of
control variables. The minimization problem is subjected to the following equality and inequality constraints:
(i) Load flow constraints:
NB
Pi − Vi ∑V j (Gij cosθij + Bij sinθij ) = 0, i = 1,2,.......N B −1 (2)
j =1

NB
Qi − Vi ∑V j (Gij sinθij − Bij cosθij ) = 0, i = 1,2,......N PQ (3)
j =1

(ii) Voltage constraints:


Vimin < Vi < Vimax ; i ∈ NB (4)

(iii) Generator reactive power capability limit:


Qgimin < Qgi ≤ Qgimax ; i ∈ Ng (5)

(iv) Reactive power generation limit of capacitor banks


QCi min < QCi ≤ QCi max; i ∈ NC (6)

(v) Transformer tap setting limit:


tkmin < tk < tkmax ; k ∈ NT (7)

(vi) Transmission line flow limit


Sl < Slmax ; l ∈ Nl (8)

(vii) Reactance limit of TCSC


– 0.5X < XTCSC ≤ 0.5 X ; i ∈ N t csc (9)

The equality constraints given by Equations (2) and (3) are satisfied by running the Newton–Raphson power flow
algorithm. Generator bus terminal voltages (Vgi), transformer tap settings (tk), reactive power generation of capacitor
bank (Qci) and reactance limit of TCSC (Xtcsc) are the optimization variables and are self-restricted between the

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 175
S. Durairaj and D. Devaraj

minimum and maximum value by the optimization algorithm. The limits on active power generation at the slack bus
(Pgs), load bus voltages (Vload), reactive power generation (Qgi), and line flow (Sl) limits are state variables which are
satisfied by adding a penalty function and minimizing the combined function.
3. MODELING OF TCSC
TCSC is a series-controlled capacitive reactance that can provide continuous control of power on the ac line over
a wide range. From the system view point, the principle of variable-series compensation is simply to increase the
fundamental-frequency voltage across a fixed capacitor in a series-compensated line through appropriate variation of
firing angle. This enhanced voltage changes the effective value of the series capacitive reactance. Incorporating
TCSC affects the topology and hence the power flow distribution since the TCSC may have one of the two possible
characteristics: capacitive or inductive, respectively, to decrease or increase the reactance of the line XL. The
capacitance and the inductance are variable and their values are function of the reactance of the line in which the
device is located. TCSC is a series compensation component and is used to keep bus voltages and to ensure the
voltage stability of the system. By controlling the firing angle of the thyristors, one can change its apparent reactance
smoothly and rapidly. During steady state, the TCSC is represented as a variable series capacitance, which is able to
change the net susceptance of the connected line.
In this model, TCSC will be represented as a capacitor/inductor whose reactance can vary between –0.5 and + 0.5
times the nominal reactance of the branch. The effect of TCSC on the network can be seen as a controllable
reactance inserted in the related transmission line. In this work, the controllable reactance xtcsc, is directly used as the
control variable to be implemented in the power flow equation. The power flow equations of the line with a new line
reactance can be derived as follows:
2
( ( )
Pij = Vi Gij − Vi V j Gij cos θ ij + Bij sin θ ij ( )) (10)

2
( ( )
Qij = − Vi Bij − Vi V j Gij sin θ ij − Bij cos θ ij ( )) (11)

2
( ( )
Pji = V j Gij − Vi V j Gij cos θ ji − Bij sin θ ji ( )) (12)

2
( ( )
Q ji = − V j Bij + Vi V j Gij sin θ ji + Bij cos θ ji ( )) (13)

where,
Rij X new
Gij = , Bij = (14)
Rij2 + X new
2
Rij2 + X new
2

4. OVERVIEW OF GENETIC ALGORITHM


A genetic algorithm is a general purpose optimization algorithm based on the mechanics of natural selection and
genetics. Genetic algorithms are favored for search problems which require the identification of a globally optimal
solution. They are based on the natural evolution process and operate by combining the best solutions at each stage
into the next generations of solutions. This process is iterative until the criteria for an optimal solution are met. The
popularity of the genetic algorithm approach can be attributed to the fact that optimal solutions can be found without
the search becoming trapped in local minima or maxima. Additionally, genetic algorithms can be applied to
problems with a large number of variables to be considered. A fitness value, derived from the problem’s objective
function is assigned to each member of the population. Individuals that represent better solutions are awarded higher
fitness values, thus enabling them to survive more generations. Starting with an initial random population, successive
generations of populations are created by the genetic operators: reproduction, crossover, and mutation to yield better
solutions which approach the optimal solution to the problem. Figure 1 shows the computation diagram of the
proposed GA-based algorithm for solving an optimization problem.
4.1. Reproduction
Selection is an essential component of GAs, playing an important role especially in solving hard optimization
problems. Selection plays a major role in GAs since it determines the direction of search where other genetic
operators propose new search points in an undirected way. The selection methods include fitness proportionate
selection, tournament selection, rank selection, etc. All these methods use genetic operations in one form or another
to create new search points. Tournament selection is used in this work. In tournament selection, ‘n’ individuals are
selected at random from the population and the best of the n is inserted into the new population for further genetic
processing. This procedure is repeated until the mating pool is filled. Tournaments are often held between pairs of
individuals (n = 2), although larger tournaments can be held.

176 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

Figure 1. Flow chart of GA-based algorithm


4.2. Crossover Operation
Crossover is a recombination operator. In the crossover operator, new strings are created by exchanging
information among strings of the mating pool. In crossover operators, two strings are picked from the mating pool at
random and some portions of the strings are exchanged between the strings. The crossover operator is mainly
responsible for the global search property of the GA. Crossover basically combines substructures of two parent
chromosomes to produce new features, with the selected probability typically in the range of 0.6–1.0. If the good
strings are created by crossover, there will be more copies of them in the next mating pool generated by the
reproduction operator. But if the good strings are not created by crossover, they will not exist in the mating pool too
long because reproduction will select against those strings in subsequent generations.
4.3. Mutation
Mutation is a background operator which produces spontaneous random changes in various chromosomes. After
performing the crossover, mutation will take place. Mutation is intended to prevent falling of all solutions in the
population into a local optimum of the solved problem. Mutation operation randomly changes the offspring resulted
from crossover. In case of binary encoding, it is possible to switch a few randomly chosen bits from 1 to 0 or from 0
to 1. “Uniform mutation” operator is applied to the real variables. It randomly selects one of the variables ui, from a
parent and sets it equal to a uniform random number between the variables lower (umin) and upper (umax) bounds.
5. IMPROVED GENETIC ALGORITHM (IGA)
After the application of the basic genetic operators (parent selection, crossover, and mutation), the advanced and
problem-specific operators are applied [12] to produce the new generation. All chromosomes in the initial population
are created at random. Population statistics are then used to adaptively change the crossover and mutation
probabilities. If premature convergence is detected, the mutation probability will be increased and the crossover
probability will be decreased. The contrary happens in the case of high population diversity. One of the most
important issues in the genetic evolution is the effective rearrangement of the genotype information. In a Simple
Genetic Algorithm (SGA), crossover is the main genetic operator for the exploitation of information while mutation
brings new missing bit structures. This SGA 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. Also, the
possibility for the SGA to get trapped in local optima increases and the convergence speed of the SGA decreases. To

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 177
S. Durairaj and D. Devaraj

overcome these difficulties a suitable combination of the basic, advanced and problem specific genetic operators
must be introduced in order to enhance the performance of the GA. The details of problem specific operators used in
this work are presented in the following subsections.
5.1. Gene Cross-Swap Operator (GCSO)
This operator 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. Figure
2 illustrate the crossover operation by GCSO.
1st variable Nth variable
1 0 1 1 0 . . . . . 1 1 1 0 0

1 0 0 1 1 . . . . . 1 0 0 0 1
Figure 2. Gene Cross Swap Operator
5.2. 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 to operate at the same output level.
5.3. 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 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. Figure 3
illustrate the operation by GIO.
… Nth variable …
. . . . . 1 1 0 0 1 . . . . .

. . . . . 0 0 1 1 0 . . . . .
Figure 3. Gene Inverse Operator
6. GENETIC ALGORITHM IMPLEMENTATION FOR RPD PROBLEM
While applying GA to solve a particular optimization problem, two main issues need to be addressed:
• Representation of the solution variables and
• The formation of the fitness function.
These issues are explained in this section.
6.1. Representation
Each individual in the genetic population represents a candidate solution. In the binary-coded GA, the solution
variables are represented by a string of binary alphabets. The size of the string depends on the precision of the
solution required. For problems with more than one decision variables, each variable is represented by a sub-string
and all the sub-strings are concatenated together to form a larger string. In the RPD problem under consideration the
solution variables consist of generator bus voltage (Vgi), the transformer tap setting (tk), and the reactive power
generation of capacitor (Qci) and reactance of TCSC (Xtcsc). With binary representation, an individual in the GA
population will look like the following:
100001 101111… 111111 001 110 … 100
Vg1 Vg2 ... Vgn t1 t2 ... tn
101 100 … 111 0.34 -0.23 … 0.123
Qc1 Qc2 ... Qcn Xtcsc1 Xtcsc2 … Xtcscn
While evaluating the fitness value of each individual, the binary strings have to be decoded into its actual value.
The continuous variables are decoded using the expression:

178 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

k
ui = ui min + (ui min − ui max ). (15)
2 N ui − 1
The conventional method in solving a RPD problem is to approximate the set of discrete control variables u by
continuous variables, and then the problem becomes an ordinary non-linear programming problem. After the
continuous problem is solved, its solution is approximated to the closest discrete value. In practical applications, this
approach usually conduces to sub optimal solutions that may be far from the global optimum. However, three
problems arise from this approach [13]. First, there is no guarantee that the rounded solution is the optimal solution.
Second, the solution may become infeasible after rounding, i.e., some constraints may be violated. Finally, these
methodologies will not work properly for discrete variables that have very large step sizes such as switched capacitor
banks. These complications in the VAR dispatch problem necessitate a discrete solution to determine the exact
optimal solution which is physically realizable. Therefore, the discrete variables controls like tap changing
transformer and susceptance of capacitor taking ‘M’ values ui1, ui2 ,...uiM are decoded using the expression

⎡ M ⎤
ui = uim with, m = int ⎢ N ⋅ k + 0.5⎥
⎣2 ui

and
log 2 M ≤ Nui ≤ log 2 M + 1

where ‘k’ is the decimal number to which the binary number in a gene is decoded and Nui, is the gene length (number
of bits) used for encoding control variable ui.
6.2. Fitness function
GA searches for the optimal solution by maximizing a given fitness function and therefore an evaluation function
which provides a measure of the quality of the problem solution must be provided. Initially, the individuals in the
genetic population are generated randomly between the lower and upper limits of the variables. The performance of
each individual in the population is evaluated according to its ‘fitness’, which is defined as the non-negative figure of
merit to be maximized. It is associated directly with the objective function value. In the TCSC placement problem
under consideration the objective is to minimize the total power loss while satisfying the constraints (2) to (9). For
each individual, the equality constraints (2) and (3) are satisfied by running Newton-Raphson algorithm. The
constraints on the control variables are taken into account through the proper representation and the constraints on
the state variables are taken into consideration by adding a quadratic penalty function to the objective function. With
the inclusion of penalty function, the new objective function then becomes,
N PQ Ng N
l
Min.FT = Ploss + K v ∑ ( V i − Vilim ) 2 + K q ∑ ( Q gi − Q gi
lim ) 2 + K
f ∑
( Sl − Sl
lim 2
) (16)
i =1 i =1 l =1

The first term on the right hand side of Equation (16) corresponds to the objective function value and the
remaining terms are penalty functions for the different state variables. For the candidate solutions with no limit
violation, the penalty function values become zero and the value of loss function is equal to the objective function
value. On the other hand, if there is any violation in the lower or upper limit of the state variables, a corresponding
penalty value based on the amount of violation is added to the objective function. The success of the penalty function
approach lies in the proper choice of the penalty factors. Using the above penalty function approach, one has to
experiment to find a correct combination of penalty factors Kv, Kq, and Kf, where Kv, Kq, and Kf are the penalty factors
for the bus voltage limit violation, generator reactive power limit violation, and line flow violation respectively. As
pointed out by researchers in the past, there is no widely accepted method for the systematic selection of the
penalties, and experimentation is used by most researchers. In general, the construction of a penalty function must
follow the so-called minimum penalty rule. In this work, the penalty co-efficient was selected based on
experimentation. However, in order to reduce the number of penalty parameters, the constraints are normalized and
only one penalty factor is used. In the above objective expression, xi lim of a quantity (xi) is defined by the following
expression:
xilim = ximax , if xi > ximax

= ximin , if xi < ximin

During the GA-run, GA searches for a solution with maximum fitness function value. Hence, the minimization
objective function given by Eqn. (16) is transformed to a fitness function (f) to be maximized as,
f = K / FT (17)

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 179
S. Durairaj and D. Devaraj

where K is a large constant. This is used to amplify (1/ FT), the value of which is usually small, so that the fitness
value of the chromosome will be in a wider range.
7. CASE STUDY
To demonstrate the effectiveness of the proposed GA based reactive power optimization approach, the IEEE 30-
bus system and a practical Indian 76-bus system are used as the test systems. The GA-based algorithm was
implemented in MATLAB 6.5 programming language and numerical tests are carried on a Pentium IV Computer.
Two different cases were considered to show the effectiveness of the proposed method. In Case 1, the proposed
algorithm was applied to solve the RPD problem in IEEE 30-bus system and in a practical 76 bus system with out
considering TCSC devices. In Case 2, GA based approach was applied to solve the RPD problem in the above test
systems with the inclusion of TCSC devices. For placement of the TCSC devices using the sensitivity approach, top
priority lines were selected. TCSC device has been installed on different branches one by one based on the real
power transmission loss and further the model has been applied to multiple TCSC devices. The optimal location,
which is the location at which loss is minimum can be identified as optimal locations. After deciding the optimal
placement of the FACTS devices, the conventional RPD (with out TCSC devices) and RPD considering TCSC
devices were run to determine the optimal settings of control variables for loss minimization.
7.1. IEEE 30-bus system
Figure 4 shows the IEEE 30-bus system which consists of 6 generator buses, 24 load buses and 41 transmission
lines of which 4 branches (6–9), (6–10), (4–12), and (28–27) are with the tap setting transformer. The transmission
line parameters of this system and the base loads are taken from [1]. For the RPD problem, the candidate buses for
reactive power compensation are 10, 12, 15, 17, 20, 21, 23, 24, and 29. The lower voltage magnitude limits at all
buses are 0.95 p.u. and the upper limits are 1.1 for all the PV buses and 1.05 p.u. for all the PQ buses. The lower and
upper limits of the transformer in tappings are 0.9 and 1.1 p.u. respectively.

Figure 4. IEEE 30- bus system


The best result of the GA was obtained with the following control parameters:
No. of generations : 100
Population size : 40
Crossover probability : 0.8
Mutation probability : 0.01
Problem specific operator : 0.2

180 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

To identify the optimal locations for TCSC devices, reactance of single line were simulated on the system one by
one and the value of the real power transmission loss was evaluated for each case. From this analysis, it is found that
the branches (4–2), (6–3), (15–8), and (19–9) are found to be most sensitive branches selected for the placement of
TCSC devices. The values of transmission loss for all the branches are given in Table 1. To illustrate the
convergence of the algorithm, the relationship between the best fitness value of the results and the average fitness are
plotted against the generation number in Figure 5. From this figure, it can be observed that the proposed algorithm
converges towards the optimal solution quickly. The minimum transmission loss corresponding to the optimal
solution is 4.5 MW. This result is less than the value reported in [14, 18] for the same real power settings. The
optimal control variable setting is given in Table 2. Corresponding to these control variable settings, it is found that
there are no limit violations in any of the state variables. For comparison, the reactive power dispatch problem is also
solved without considering TCSC devices and the loss obtained in this case is 4.6 MW. This shows the effectiveness
of the TCSC devices in reducing the transmission loss in the system.
Table 1. Transmission Losses with Placement of TCSC in Different Lines
TCSC in line Optimal value of Xi limit of TCSC (–0.5 to + 0.5 ) Ploss (MW)
1 0.5 4.7030
2 –0.3571 4.6664
3 0.2143 4.6646
4 –0.2143 4.5940
5 0.3571 4.6254
6 –0.2143 4.5161
7 –0.5000 4.6676
8 0.3571 4.8515
9 0.5000 4.8673
10 0.3571 4.8116
11 0.2143 4.7357
12 –0.2143 4.6895
13 0.5000 4.8588
14 0.5000 5.0063
15 0.3571 4.3983
16 –0.3571 4.5953
17 –0.2143 4.8129
18 0.3571 4.6745
19 –0.2143 4.5864
20 0.3571 4.6644
21 –0.5000 5.0860
22 –0.5000 4.7353
23 0.3571 4.6643
24 –0.2143 4.7849
25 –0.0714 7.0185
26 0.3571 4.6637
27 –0.2143 4.8123
28 0.3571 4.9770
29 0.3571 4.6654
30 0.3571 5.0363
31 –0.0714 6.9873
32 0.0714 4.7945
33 –0.2143 4.7203
34 0.3571 4.6653
35 –0.5000 4.7070
36 –0.5000 4.6072
37 0.3571 4.8624
38 –0.0714 4.6703
39 0.3571 4.8550
40 0.0714 4.7048
41 –0.2143 4.8273

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 181
S. Durairaj and D. Devaraj

Figure 5. Convergence characteristics of GA based RPD


Table 2. Optimal Control Variables for the IEEE 30- Bus System
Variables Proposed Genetic Algorithm
With TCSC Without TCSC
V1 1.0524 1.0429
V2 1.0429 1.0405
V5 1.0238 1.0214
V8 1.0262 1.0262
V11 1.0167 1.0119
V13 0.9833 0.9976
T11 1.0500 1.0500
T12 0.9750 1.0750
T15 1.0750 1.0750
T36 1.0000 0.9500
Qc10 4 0
Qc12 1 0
Qc15 3 0
Qc17 4 3
Qc20 3 7
Qc21 4 5
Qc23 1 0
Qc24 0 0
Qc29 2 9
X(4-2) –0.5 -
X (6-3) 0.3571 -
X(15-8) –0.3571 -
X(19-9) –0.5 -
Ploss 4.5 MW 4.6MW
Moreover, voltage collapse problems are a serious concern for power system engineers and planners. Voltage
collapse [15] is mathematically indicated when the system jacobian becomes singular. The collapse points are
indicative of the maximum loadability of the transmission lines or the available transfer capability (ATC). The
TCSCs can significantly enhance loadability of transmission network, thus obviating voltage collapse at existing
power-transfer levels. While the TCSC reduces the effective line reactance, thereby increasing the power flow, it

182 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

generates reactive power with increasing through-current, thus exercising a beneficial influence on the neighboring
bus voltage. The bus voltage magnitudes obtained after the placement of TCSC are shown in Figure 6 and it is
observed that the voltages of the buses have been improved after reactive power dispatch. The loss obtained with
different methods and also with the proposed algorithm is given in Table 3. These results show that the optimal
dispatch solutions determined by the proposed algorithm lead to lower active power loss than that found by other
methods, which confirms that the proposed algorithm is capable of determining the global or near-global optimum
dispatch solution.

Figure 6. Voltage profile diagram for IEEE 30- bus system before and after the placement of TCSC devices
Table 3. Comparison of Minimum Transmission Loss Obtained with Other Methods
Techniques Ploss (MW)
Broyden method [16] 5.3
Canonical Genetic algorithm (CGA) [18] 4.98
Adaptive Genetic algorithm (AGA) [18] 4.92
Evolutionary programming [18] 5.01
Evolutionary strategy [18] 4.97
Particle swarm optimization [16] 5.0
Improved genetic algorithm [16] 4.95
Evolutionary algorithm [17] 5.09
Linear programming [18] 7.56
Non-Linear programming [18] 5.73
Proposed GA based method
(without TCSC) : 4.6 MW
(with TCSC) : 4.5 MW

7.2. Practical Indian 76-bus system


In this case, the proposed approach was applied to solve the RPD problem in a practical Indian power system.
The system under consideration is a regional grid of Indian power system, consisting of 13 Generator buses, 63 load
buses, 116 transmission lines of which 7 are with tap changing Transformers. Switchable VAR compensators are
located at 12 places. The total load on the system is 3668 MW and 2591 MVAR. To obtain the optimal values of the
control variables the GA-based algorithm was run with different control settings (Table 4). The algorithm reaches a
minimum loss of 46.3 MW. RPD is also solved but with out considering TCSC devices in the system and the loss
obtained in this case are 48 MW. Corresponding to the control variables, it was found that there was no limit
violation in any of the state variables. The best and worst reactive power dispatch solutions found from ten GA run
are tabulated in Table 5. From the comparison, it is found that the proposed method is more effective in solving the
RPD problem than the conventional methods.

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 183
S. Durairaj and D. Devaraj

Table 4. Optimal Control Variables for the Practical 76-Bus System


Variables Proposed improved Genetic
algorithm
with TCSC without TCSC
devices devices
1.0017 1.0025
1.0474 1.0033
Vi 1.0720 1.0786
1.0748 1.0786
1.0897 1.0890
1.0597 1.0272
1.0468 1.0500
1.0175 1.0235
1.0236 1.0191
1.0236 1.0988
1.0885 1.0857
1.0226 1.0998
1.1000 1.0953
0.9500 0.9750
1.000 1.0000
Ti 0.9750 0.9750
0.9750 1.0000
0.9250 0.9000
1.0250 0.9750
0.9750 1.0750
9 20
20 2
18 2
0 2
19 20
Ci (Mvar)
20 20
3 19
3 19
8 19
19 20
20 19
20 20
X (75-140) –0.5 -
X (75-59) –0.2143 -
X (32-67) 0.2143 -
X (58-76) 0.5 -
X (25-69)
0.2143 -
X (45-9)
–0.5 -
Ploss 46.3 MW 48 MW
Table 5. Comparison of Best and Worst Results in IEEE-30 and Practical 76-Bus Systems
Worst
System Best solution Average solution
solution
IEEE 30-bus system (Loss in MW) 4.5 6.2178 5.261
Practical Indian 76-bus system (Loss in MW) 46.3 56.7838 51.45

184 The Arabian Journal for Science and Engineering, Volume 34, Number 1B April 2009
S. Durairaj and D. Devaraj

8. CONCLUSION
This paper has presented the application of genetic algorithm to reactive power dispatch including TCSC in
power systems. Sensitivity of the real power loss with respect to line reactance of TCSC has been used for the
optimal placement of FACTS devices. A set of advanced and problem-specific genetic operators has been added to
the simple genetic algorithm to increase its convergence speed and improve the quality of the solutions. The
performance of the proposed algorithm has been demonstrated through its application on the IEEE 30-bus power
system and also in a practical Indian 76 bus system which shows that the proposed Genetic algorithm is able to
undertake global search with a fast convergence rate. Simulation results demonstrate that the proposed approach is
effective than other conventional methods and efficient for solving the reactive power dispatch problem in large
scale power systems also.
REFERENCES
[1] H.W. Dommel and W.F. Tinny, “Optimal Power Flow Solutions”, IEEE Trans. on Power App. & Sys, PAS- 87,
(1968), pp. 1866–1876.
[2] V.H. Quintana and M. Santos, “Reactive Power Dispatch by Successive Quadratic Programming”, IEEE
Transactions on Energy Conversion, 4(3)(1989), pp. 425–434.
[3] S. Granville, “Optimal Reactive Power Dispatch through Interior Point Methods”, IEEE Trans. on Power Systems,
9(1)(1994), pp. 136–146.
[4] G.R.M. Da Costa, “Optimal Reactive Dispatch through Primal-Dual Method”, IEEE Trans. on Power Systems,
12(2)(1997), pp. 669–674.
[5] Y. Lin., L. Ma, and J. Zhang, “Reactive Power Optimization by GA / SA / TS Combined Algorithms”, Electric
Power & Energy Systems, 24(2002), pp. 765–769.
[6] Q.H. Wu and Ma J.T., “Power System Optimal Reactive Power Dispatch Using Evolutionary Programming”, IEEE
Trans., Power Systems, 10(3)(1995), pp. 1243–1248.
[7] K.Y. Lee and F.F. Yang, “Optimal Reactive Planning Using Evolutionary Algorithms: a Comparative Study for
Evolutionary Programming, Evolutionary strategy, Genetic Algorithm and linear programming”, IEEE Trans. on
Power Syst., 13(1)(1998), pp. 101–108.
[8] M. Noroozian, L. Anguist, M. Ghandhari, and G. Andersson, “Improving Power System Dynamics by Series
Connected FACTS Devices”, IEEE Trans. on Power Delivery, 12(4)(1997), pp. 1635–1641.
[9] G. N. Taranto, L.M.V.G. Pinto, and MVF. Pereira, “Representation of FACTS Devices in Power System
Economics Dispatch”, IEEE Trans. on Power Syst., 7(2)(1992), pp. 572–576.
[10] S.Y. Ge and T.S. Chung, “Optimal Active Power Flow Incorporating FACTS Devices with Power Flow Control
Constraints”, Electrical Power and Energy Systems, 20(5)(1998), pp. 321–326.
[11] H. Ambriz-Perez, E. Acha, C.R. Fuerte-Esquivel, “Advanced SVC Model for Newton–Raphson Load Flow and
Newton Optimal Power Flow Studies”, IEEE Trans. on Power Syst., 15(1)(2000), pp. 129–136.
[12] A. G. Bakirtzis, P.N. Biskas, C.E. Zoumas, and V. Petridis, “Optimal Power Flow by Enhanced Genetic
Algorithm”, IEEE Trans. on Power Syst., 17(2)(2002), pp. 229–236.
[13] D. Devaraj, B. Yegnanarayana, “A Combined Genetic Algorithm Approach for Optimal Power Flow”, Proc. of
11th National Conference on Power Systems, IISc, Bangalore, December 2000, pp. 524–528.
[14] D.B. Das, and C. Patvardhan, “Reactive Power Dispatch with a Hybrid Stochastic Search Technique”, Int. J.Elect.
Power Energy Syst., 24(2002), pp. 731–736.
[15] R.C. Bansal, T.S. Bhatti, and D.P. Kothari, “Artificial Intelligence Techniques for Reactive Power Voltage Control
in Power Systems- A Review”, International Journal of Power and Energy Systems, 23(2)(2003), pp. 81–88.
[16] L.L. Lai, T.Y. Nieh, D. Vujatovic, Y.N. Ma, Y.P. Lu, Y.W. Yang, and H. Braun, “Swarm Intelligence for Optimal
Reactive Power Dispatch”, IEEE/PES Transmission and Distribution Conference & Exhibition: Asia and Pacific,
Dalian, China, 2005, pp. 1–5.
[17] M.A. Abido and J.M. Bakhashwain, “Optimal VAR Dispatch Using Multi-Objective Evolutionary Algorithm”,
Electrical Power and Energy Systems, 27(2005), pp. 13–20.
[18] D.B. Das and C. Patvardhan, “A New Hybrid Evolutionary Strategy for Reactive Power Dispatch”, Electric Power
System Research, 65(2003), pp. 83–90.

April 2009 The Arabian Journal for Science and Engineering, Volume 34, Number 1B 185

Das könnte Ihnen auch gefallen