Particle Swarm Optimization Based Capacitor

Placement on Radial Distribution Systems

K. Prakash, Member, IEEE, and M. Sydulu

Abstract-- This paper presents a novel approach that determines

the optimal location and size of capacitors on radial distribution
systems to improve voltage profile and reduce the active power
loss. Capacitor placement & sizing are done by Loss Sensitivity
Factors and Particle Swarm Optimization respectively. The
concept of Loss sensitivity Factors and can be considered as the
new contribution in the area of distribution systems. Loss
Sensitivity Factors offer the important information about the
sequence of potential nodes for capacitor placement. These
factors are determined using single base case load flow study.
Particle Swarm Optimization is well applied and found to be very
effective in Radial Distribution Systems. The proposed method is
tested on 10, 15, 34, 69 and 85 bus distribution systems.
Index Terms--Capacitor Placement, Radial Distribution
Systems, Loss Sensitivity Factors and Particle Swarm

Optimization.
.
I. INTRODUCTION

S Distribution Systems are growing large and being

stretched too far, leading to higher system losses and
poor voltage regulation, the need for an efficient and effective
distribution system has therefore become more urgent and
important. In this regard, Capacitor banks are added on Radial
Distribution system for Power Factor Correction, Loss
Reduction and Voltage profile improvement.
With these various Objectives in mind, Optimal Capacitor
Placement aims to determine Capacitor location and its size.
Optimal Capacitor Placement has been investigated over
decades. Early approaches were based on heuristic
techniques. In the 80s, more rigorous approaches were
suggested as illustrated by Grainger [1],[2] and Baran Wu
[3],[4] formulated the Capacitor Placement as a mixed integer
non-linear program. In the 90s combinatorial algorithms were
introduced as a means of solving the Capacitor Placement
Problem and neural network technique based papers [5] and
[6] were investigated. Ng and Salama [7] have proposed a
solution approach to the capacitor placement problem based
on fuzzy sets theory. Using this approach, the authors
attempted to account for uncertainty in the parameters the

problem. They model these parameters by possibility

distribution functions. Chin [8] uses a fuzzy dynamic
programming model to express real power loss, voltage
deviation, and harmonic distortion in fuzzy set notation.
Sundharajan and Pahwa [9] used genetic algorithm for
capacitor placement.
Particle Swarm Optimization (PSO) was developed by James
Kennedy and Russell Eberhart [10]. It is based on metaphor
of social interaction, searches a space by adjusting the
trajectories of moving points in a multi dimensional space and
used for optimization of continues non linear problems. The
main advantages of the PSO are summarized as follows:
simple concept, easy implementation robustness to control
parameters and better computational efficiency when
compared with other heuristic algorithms.
Shi and Eberhart [11] uses an extra inertia weight term
which is used to scale down the velocity of each particle and
this term is typically decreased throughout a run. Shi and
Eberhart [12] also described an empirical study of PSO.
In this paper, Capacitor Placement and Sizing is done by Loss
Sensitivity Factors and Particle Swarm Optimization (PSO)
respectively. PSO is used for estimation of required level of
shunt capacitive compensation to improve the voltage profile
of the system. The proposed method is tested on 10, 15, and
34 bus radial distribution systems and results are very
promising. In this paper, Vector based Distribution Load Flow
method (VDLF) [13] with Sparsity Technique [14] is used.
With the support of sparsity technique the VDLF found to be
very effective.

II. SENSITIVITY ANALYSIS AND LOSS SENSITIVITY FACTORS

A new methodology is used to determine the candidate
nodes for the placement of capacitors using Loss Sensitivity
Factors. The estimation of these candidate nodes basically
helps in reduction of the search space for the optimization
procedure.
Consider a distribution line connected between p and q
buses.
p

K.Prakash, Research Scholar in National Institute of Technology, Deemed

University, Warangal (A.P), 506004 India (e-mail: prakashkam@yahoo.co.in).
M.Sydulu is with the Department of Electrical Engineering, National
Institute of Technology, Deemed University, Warangal (A.P), 506004 India (email: msydulu@nitw.ernet.in).

1-4244-1298-6/07/$25.00 2007 IEEE.

R+jX
kth line

q
Peff + jQeff

Active power loss in the kth line is given by [Ik2]*R[k],

distribution system.

which can be expressed as,

Plineloss[ q] =

(P

2
2
eff [ q] + Qeff [ q ] .R[ k ]
2

(V [q])

(1)

Similarly the reactive power loss in the kth line is given by

Qlineloss [q ] =

(P

2
eff [ q ] +

2
Qeff
[q ] . X [k ]

(V [q])2

(2)

Where, Peff[q] = Total effective active power

supplied beyond the node q.
Qeff [q] = Total effective reactive power
supplied beyond the node q.
Now, both the Loss Sensitivity Factors can be obtained as
shown below:
Plineloss (2 * Qeff [q ] * R[ k ])
=
Qeff
(V [q])2

(3)

Qlineloss (2 * Qeff [q ] * X [ k ])
=
Qeff
(V [q ])2

(4)

Candidate Node Selection using Loss Sensitivity Factors:

The Loss Sensitivity Factors ( Plineloss Qeff ) are calculated
from the base case load flows and the values are arranged in
descending order for all the lines of the given system. A
vector bus position bpos [i] is used to store the respective
end buses of the lines arranged in descending order of the
values ( Plineloss Qeff ). The descending order of
( Plineloss Qeff ) elements of bpos[i] vector will decide the
sequence in which the buses are to be considered for
compensation. This sequence is purely governed by the
( Plineloss Qeff ) and hence the proposed Loss Sensitive
Coefficient factors become very powerful and useful in
capacitor allocation or Placement. At these buses of bpos[i]
vector, normalized voltage magnitudes are calculated by
considering the base case voltage magnitudes given by
(norm[i]= V[i]/0.95). Now for the buses whose norm[i] value
is less than 1.01 are considered as the candidate buses
requiring the Capacitor Placement. These candidate buses are
stored in rank bus vector. It is worth note that the Loss
Sensitivity factors decide the sequence in which buses are to
be considered for compensation placement and the
norm[i]decides whether the buses needs Q-Compensation or
not. If the voltage at a bus in the sequence list is healthy (i.e.,
norm[i]>1.01) such bus needs no compensation and that bus
will not be listed in the rank bus vector. The rank bus
vector offers the information about the possible potential or
candidate buses for capacitor placement. Now sizing of
Capacitors at buses listed in the rank bus vector is done by
using Particle Swarm Optimization based algorithm. Table I
shows the active power Line Loss Sensitivity Coefficients
placed in descending order along with its bus identification
and normalized voltage magnitudes of 15 bus radial

TABLE I
LOSS SENSITIVITY COEFFICIENTS PLACED IN DESCENDING
ORDER OF A 15-BUS RADIAL DISTRIBUTION SYSTEM.
Bus
Norm[i]
Base voltage
Plineloss Qeff
No.
V[i]/0.95
(descending)
0.029661
0.016437
0.015485
0.008526
0.006182
0.005266
0.004134
0.003141
0.002966
0.002811
0.001678
0.001613
0.001342
0.001256

2
6
3
11
4
12
9
15
14
7
13
8
10
5

1.0224
1.0086
1.0069
0.9990
1.0009
0.9955
1.0188
0.99835
0.9985
1.0063
0.9942
1.007263
1.01768
0.99985

0.971283
0.958232
0.956669
0.949952
0.949952
0.945829
0.967971
0.94844
0.948608
0.956008
0.944517
0.956954
0.966897
0.949918

Rank bus vector of 15-bus Radial Distribution System

contains set of sequence of buses given as
{6,3,11,4,12,15,14,7,13,8,5}

III. PARTICLE SWARM OPTIMIZATION

Particle Swarm Optimization (PSO) is a meta heuristic
parallel search technique used for optimization of continues
non linear problems. The method was discovered through
simulation of a simplified social model. PSO has roots in two
main component methodologies perhaps more obvious are ties
to artificial life in general, and to bird flocking, fish schooling
and swarming theory in particular.
It is also related, how ever to evolutionary computation and
has ties to both genetic algorithms and evolutionary
programming. It requires only primitive mathematical
operators, and is computationally inexpensive in terms of both
memory requirements and speed.
It conducts searches using a population of particles,
corresponding to individuals. Each particle represents a
Candidate solution to the capacitor sizing problem.
In a PSO system, particles change their positions by flying
around a multi dimensional search space until a relatively
unchanged position has been encountered, or until
computational limits are exceeded. In social science context, a
PSO system combines a social and cognition models.
The general elements of the PSO are briefly explained as
follows:
Particle X(t): It is a k-dimensional real valued vector which
represents the candidate solution. For an ith particle at a time t,
the particle is described as Xi(t)={Xi,1(t), Xi,2(t), Xi,k(t)}.
Population: It is a set of n number of particles at a time t
described as {X1(t), X2(t) Xn(t)}.
Swarm: It is an apparently disorganized population of moving
particles that tend to cluster together while each particle
seems to be moving in random direction.

Particle Velocity V(t): It is the velocity of the moving

particle represented by a k-dimensional real valued vector
Vi(t)= {vi,1(t), vi,2(t) vi,k(t)}.
Inertia weight W(t): It is a control parameter that is used to
control the impact of the previous velocity on the current
velocity.
Particle Best (pbest): Conceptually pbest resembles
autobiographical memory, as each particle remembers its own
experience. When a particle moves through the search space,
it compares its fitness value at the current position to the best
value it has ever attained at any time up to the current time.
The best position that is associated with the best fitness
arrived so far is termed as individual best or Particle best. For
each Particle in the swarm its pbest can be determined and
updated during the search.
Global Best (gbest): It is the best position among all the
individual pbest of the particles achieved so far.
Velocity Updation: Using the global best and individual best,
the ith particle velocity in kth dimension is updated according
to the following equation.
V[i][j]=K*(w*v[i][j]+c1*rand1*(pbestX[i][j]X[i][j])+c2*rand2*(gbestX[j]-X[i][j])).
where,
K constriction factor
c1, c2 weight factors
w Inertia weight parameter
i particle number
j
control variable
rand1, rand2 random numbers between 0 and 1
Stopping criteria: This is the condition to terminate the
search process. It can be achieved either of the two following
methods:
i.
The number of the iterations since the
last change of the best solution is
greater than a pre-specified number.
ii.
The number of iterations reaches a prespecified maximum value.
IV. ALGORITHM FOR CAPACITOR PLACEMENT AND
SIZING USING LOSS SENSITIVITY FACTORS AND
PARTICLE SWARM OPTIMIZATION
Step1: Run the base case Distribution load flow and
determine the active power loss.
Step2: Identify the Candidate buses for placement of
capacitors using Loss Sensitivity Factors.
Step3: Generate randomly n number of particles, where
each particle is represented as particle[i]={Qc 1,Qc 2,.,Qc j}
Where j represents number of candidate buses.
Step4: Generate the particle velocities (v[i]) between vmax
and vmax.
Where, vmax = (capmax-capmin)/N
Capmax= maximum capacitor rating in kvar
Capmin= minimum capacitor rating in kvar
N= number of steps to move the particle from
one position to the other.
Step5: Set the Iteration count, iter=1.

Step6: Run the load flows by placing a particle i at the

candidate bus for reactive power compensation and
store the active power loss (pl).
Step7: Evaluate the fitness value (base power loss-pl) of the
particle i and compare with previous particle best
(pbest) value. If the current fitness value is greater than its
pbest value, then assign the pbest value to the current value.
Step8: Determine the current global best(gbest) maximum
value among the particles individual best (pbest)
values.
Step9: Compare the global position with the previous
global position. If the current global position is greater than
the previous, then set the global position to the current global
position.
Step10: Update the velocities by using
v[i][j]=K*(w*v[i][j]+c1*rand1*(pbestparticle[i][j]particle[i][j])+c2*rand2*(gbestparticle[j]-particle[i][j]).
Where,
particle[i] position of individual i
pbestparticle[i] best position of individual i
gbestparticle best position among the swarm
v[i] velocity if individual i
Step11: If the velocity v[i][j] violates its limits (-vmax,
vmax), set it at its proper limits
Step12: Update the position of the particle by adding the
velocity (v[i][j]) to it.
Step13: Now run the load flow and determine the active
power loss (pl) with the updated particle.
Step14: Repeat step 7 to step 9
Step15: Repeat the same procedure for each particle from
steps from 6 to 13.
Step16: Repeat steps from 6 to 13 until the termination
criteria are achieved.

V. TEST RESULTS
The proposed method for loss reduction by capacitor
placement is tested on 10bus [15], 15bus [13], 34bus [16], 69
bus [3] and 85bus radial distribution systems. The various
constants used in the proposed algorithm are
capmin=200kvar, capmax=1200kvar, K=0.7259, c1=c2=2.05
and w=1.2.The test results are shown below in various tables.
The 10 bus test system with the proposed method is compared
with the paper [15] in which the total kvar placed is 5500 kvar
with a loss reduction of 10.06% where as the proposed
method for the identified locations the total kvar placed only
3186 kvar that too with a loss reduction of 11.17% as shown
in Table II. When the proposed method is tested on 15 bus
system and compared with the paper [18], nearly similar
results are obtained as shown in Table III. The proposed
method is also tested on 34 bus test system and compared
with the heuristic method [16] and Fuzzy Expert Systems
(FES) approach [20], results obtained are more promising. as
shown in Table IV. In practice the capacitor size should be in
discrete in value. With this in mind, for a 34 bus system, when
buses 19, 22 and 20 are compensated by 800kvar, 800kvar
and 450kvar instead of 781kvar, 803kvar and 479kvar
respectively as shown in Table IV, the load flow results

indicate the active power loss of 168.87kw instead of

168.89kw can be achieved. Similarly the other systems can
also be compensated by round off discrete capacitor values.
Table V shows the test results of the proposed method on 69
bus system with a loss reduction of 32.23% and Table VI
shows the test results of the 85bus radial distribution system
with a loss reduction of 48.27%.This proposed PSO based
capacitor placement with Loss sensitivity Factors method
saves not only initial investment on the capacitors but also
running cost as it places the capacitors in a minimum number
of locations.

TABLE II
COMPARISION OF PREVIOUS METHOD [15] AND PROPOSED
METHOD FOR OF 10BUS RADIAL DISTRIBUTION SYSTEM.
BASE CASE ACTIVE POWER LOSS=783.77 KW
Fuzzy Reasoning based[15]
Proposed PSO based
Bus No
Size (kvar)
Bus No
Size (kvar)
4
1050
6
1174
5
1050
5
1182
6
1950
9
264
10
900
10
566
Total kvar placed
4950
Total kvar placed
3186
Active power loss(kw)
704.88
Active power loss(kw)
696.21
TABLE III
COMPARISION OF PREVIOUS METHOD [18] AND PROPOSED
METHOD FOR OF 15BUS RADIAL DISTRIBUTION SYSTEM.
BASE CASE ACTIVE POWER LOSS=61.79 KW
Method proposed in[18]
Proposed PSO based
Bus No
Size (kvar)
Bus No
Size (kvar)
3
805
3
871
6
388
6
321
Total kvar placed
1193
Total kvar placed
1192
Active power loss(kw)
32.6
Active power loss(kw)
32.7
TABLE IV
COMPARISION OF PREVIOUS METHODS AND PROPOSED METHOD
FOR OF 34BUS RADIAL DISTRIBUTION SYSTEM.
BASE CASE ACTIVE POWER LOSS=221.723 KW
Heuristic based [16]
FES based[20]
Proposed PSO based
Bus No
kvar
Bus No
kvar
Bus No
kvar
26
1400
24
1500
19
781
11
750
17
750
22
803
17
300
7
450
20
479
4
250
------------------Total kvar
2700
Total kvar
2700
Total kvar
2063
Power loss(kw) 168.47
Power loss(kw) 168.98
Power loss(kw) 168.8
TABLE V
PROPOSED PSO METHOD TESTED ON 69 BUS RADIAL DISTRIBUTION
SYSTEM.
BASE CASE ACTIVE POWER LOSS=225 KW
Bus No
Kvar
46
241
47
365
50
1015
Total Kvar
1621
Active Power Loss (kw)
152.48
TABLE VI
PROPOSED PSO METHOD TESTED ON 85 BUS RADIAL DISTRIBUTION
SYSTEM.
BASE CASE ACTIVE POWER LOSS=315.714 KW
Bus No
Kvar

8
58
7
27
Total Kvar
Active Power Loss (kw)

796
453
314
901
2464
163.32

VI. CONCLUSION
In this paper, an algorithm that employs Particle Swarm
Optimization, a meta heuristic parallel search technique for
estimation of required level of shunt capacitive compensation
to improve the voltage profile of the system and reduce active
power loss. Loss Sensitivity Factors are used to determine the
optimum locations required for compensation. The main
advantage of this proposed method is that it systematically
decides the locations and size of capacitors to realize the
optimum sizable reduction in active power loss and significant
improvement in voltage profile. Test results on 10, 15, 34, 69
and 85 bus systems are presented. The method places
capacitors at less number of locations with optimum sizes and
offers much saving in initial investment and regular
maintenance.
VII. ACKNOWLEDGMENT
The authors gratefully acknowledge the support and facilities
extended by the Department of Electrical Engineering,
National Institute of Technology, Warangal and Vaagdevi
College of Engineering, Warangal (A.P) India.
VIII. REFERENCES
[1]

J.J. Grainger and S. Civanlar, Volt/var control on Distribution systems

with lateral branches using shunt capacitors as Voltage regulators-part I, II
and III, IEEE Trans. Power Apparatus and systems, vol. PAS-104, no.
11, pp. 3278-3297, Nov. 1985.
[2] J.J .Grainger and S.H Lee, Capacitor release by shunt capacitor
placement on Distribution Feeders: A new Voltage Dependent Model,
IEEE Trans .PAS, pp 1236-1243 May 1982.
[3] M. E Baran and F. F. Wu, Optimal Sizing of Capacitors Placed on a
Radial Distribution System, IEEE Trans. Power Delivery, vol. no.1, pp.
1105-1117, Jan. 1989.
[4] M. E. Baran and F. F. Wu, Optimal Capacitor Placement on radial
distribution system, IEEE Trans. Power Delivery, vol. 4, no.1, pp. 725734, Jan. 1989.
[5] N. I. Santoso, O. T. Tan, Neural- Net Based Real- Time Control of
Capacitors Installed on Distribution Systems, IEEE Trans. Power
Delivery, vol. PAS-5, no.1, pp. 266-272, Jan. 1990.
[6] P. K. Dash, S. Saha, and P. K. Nanda, Artificial Neural Net Approach for
International forum on Applications of Neural Networks to Power
Systems, pp. 247-250, 1991.
[7] H.N. Ng, M.M.A. Salama, Fuzzy Optimal Capacitor Sizing and
Placement, Canadian Conference on Electrical and Computer
Engineering, pp. 680-683, 1995.
[8] H.C.Chin, Optimal Shunt Capacitor Allocation by Fuzzy Dynamic
Programming, Electric Power Systems Research, pp.133-139, Nov.
1995.
[9] Sundharajan and A. Pahwa, Optimal selection of capacitors for radial
distribution systems using genetic algorithm, IEEE Trans. Power
Systems, vol. 9 no.3, pp.1499-1507, Aug. 1994.
[10] J. Kennedy and R. Eberhart, Particle Swarm Optimization, proceedings
of IEEE International Conference Neural Networks, vol.,IV, pp.19421948,1995.
[11] Y.Shi and Eberhart, A modified particle swarm optimization, pro. IEEE,
Int. Conf. Evol. Comput. pp. 69-73, May 1998.

[12] Y.Shi and Eberhart, Empirical study of particle swarm optimization,

pro. IEEE, Int. Conf. Evol. Comput., NJ, pp. 1945-1950, 1999
[13] D.Das, D. P. Kothari, and A. Kalam, Simple and efficient method for
load flow solution of radial distribution networks, Electrical Power &
Energy Systems, vol. 17. N0.5,pp 335-346, Elsevier Science Ltd 1995.
[14] M. Sydulu, An Effective Algorithm for Power System Topological
Observability proceedings of IEEE TENCON 98, International
Conference on Global Connectivity in Energy, Computer Communication
and Control, 17-19 Dec. 1998.
[15] Ching-Tzong Su and Chih-Cheng Tsai, A New Fuzzy- Reasoning
Approach to Optimum Capacitor Allocation for Primary distribution
Systems, proceedings of IEEE International Conference on Industrial
Technology, 1996.
[16] M.Chis, M. M. A. Salama and S. Jayaram, Capacitor Placement in
distribution system using heuristic search strategies, IEE Proc-Gener,
Transm, Distrib, vol, 144, No.3, pp. 225-230, May 1997.
[17] M. M. A. Salama and A.Y. Chikhani, An expert system for reactive
power control of a distribution system, part1: System configuration. IEEE
Trans. Power Delivery, vol. 7, no.2, pp. 940-945, Apr. 1992.
[18] M. H. Haque, Capacitor placement in radial distribution systems for loss
reduction, IEE Proc-Gener, Transm, Distrib, vol, 146, No.5, Sep. 1999.
[19] J.R.Laframboise, G. Ferland, A.Y. Chikhani and M. M. A. Salama, An
expert system for reactive power control of a distribution system, part2:
System implementation, IEEE Trans. Power Systems, vol. 1, no.3, pp.
1433-1441, Aug. 1995.
[20] H.N.Ng, M.M.A. Salama and A.Y.Chikhani, Capacitor Allocation by
Approximate Reasoning: Fuzzy Capacitor Placement, IEEE Trans.
Power Delivery, vol. 15, no.1, pp. 393-398, Jan. 2000.

IX. BIOGRAPHIES
K.Prakash received his B.E (Electrical and Electronics Engineering, 1999)
from University of Madras, M.Tech (Power Systems, 2003) from National
Institute of Technology, Warangal. He is currently pursuing his PhD (Power
Systems Engineering) in National Institute of Technology, Warangal, Andhra
Pradesh, INDIA. His areas of interest include Distribution system studies, Meta
Heuristic Techniques in Power Systems and Economic operation of Power
Systems.

Maheswarapu Sydulu received his B.Tech (Electrical Engineering,1978),

M.Tech (Power Systems,1980), PhD (Electrical Engineering Power
Systems,1993), all degrees from Regional Engineering College, Warangal,
Andhra Pradesh, INDIA. His areas of interest include Real Time power system
operation and control, ANN, fuzzy logic and Genetic Algorithm applications in
Power Systems, Distribution system studies, Economic operation, Reactive
power planning and management. Presently he is working as Professor and Head
of Electrical Engineering Department, National Institute of Technology,
Warangal (formerly RECW).