The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No.

(2)

A Modified Particle Swarm Technique for

Distribution Systems Reconfiguration
A. Y. Abdelaziz S. F. Mekhamer F. M. Mohammed M. A. L. Badr
Department of Electrical Power and Machines, Faculty of Engineering, Ain Shams University, Cairo, Egypt

Abstract-This paper introduces the Particle Swarm blend of heuristics and optimization techniques. The blend
Optimization (PSO) algorithm to solve the optimal of the two types of technique permits the problem to
network reconfiguration problem for power loss retain a certain degree of accuracy, while assuring
reduction. The PSO is a relatively new and powerful convergence and an acceptable solution time.
intelligence evolution method for solving optimization In [1], a branch-exchange method that considered the on-
problems. It is a population-based approach. The PSO off conditions of the sectionalizing switches in discrete
was inspired from natural behavior of the bees on how numbers was developed [1]. Since the method is based on
they find the location of most flowers. The proposed PSO heuristics, it is not easy to take a systematic way to evaluate
algorithm is introduced with some modifications such as an optimal solution.
using an inertia weight that decreases linearly during the Two different methods with varying degree of accuracy to
simulation. This setting allows the PSO to explore a large approximate power flow in systems were proposed in [2]. The
area at the start of the simulation. A modification in the search method has an acceptable convergence characteristic.
number of iterations and the population size is also However, it can get stuck in local minimum. The method is
presented. To verify the effectiveness of the proposed PSO very time consuming due to the complicated combinations in
algorithm, comparative studies are conducted on two test large-scale systems.
distribution systems. The obtained results are compared An expert system for feeder reconfiguration, based upon
with those obtained using other approaches in the extensions of the rules of [1] was presented in [3], with the
previous work to examine the performance. potential of handling realistic operating constrains. The
approach taken is set up a decision tree to represent the
Keywords- Particle Swarm Optimization, Distribution various switching operations available. This strategy is
Systems Reconfiguration, Power Loss Reduction. efficient for trees that are not too large. However, as a search
tree becomes larger, a great amount of time can be spent
I. INTRODUCTION
searching for the optimal solution. To guarantee an optimal
The subject of minimizing distribution systems losses has solution an exhaustive tree search should be used.
gained a great deal of attention due to the high cost of A linear programming method using transportation
electrical energy and therefore, much of current research on techniques and a new heuristic search method for comparison
distribution automation has focused on the minimum-loss with previously developed heuristic techniques which are
configuration problem. There are many alternatives available based on an optimal load flow analysis were presented in [4].
for reducing losses at the distribution level: reconfiguration, This study indicates that linear programming, in the form of
capacitor installation, load balancing, and introduction of transportation algorithms, is not suitable for application to
higher voltage levels. This research focuses on the feeder reconfiguration since the power loss function is not
reconfiguration alternative. linear whilst heuristic approaches, although not optimal, can
Two types of switches are used in primary distribution provide substantial saving if properly formulated.
systems. There are normally closed switches (sectionalizing Based on partitioning the distribution network into groups
switches) and normally open switches (tie switches). Those of load buses, the line section losses between the groups of
two types of switches are designed for both protection and nodes are minimized [5]. By dividing the distribution network
configuration management. Network reconfiguration is the into groups of busses, the combinatorial nature of the
process of changing the topology of distribution systems by reconfiguration problem is overcome, while simultaneously
altering the open/closed status of switches. Because there are minimizing losses.
many candidate-switching combinations in the distribution In recent years, meta-heuristic methods have been studied
system, network reconfiguration is a complicated for solving combinatorial optimization problems to obtain an
combinatorial, non-differentiable constrained optimization optimal solution of global minimum. Typical meta-heuristic
problem. The change in network configuration is achieved by methods include Simulated Annealing (SA), Genetic
opening or closing of these two types of switches in such a Algorithm (GA), and Tabu Search (TS).
way that the ‘radiality’ of the network is maintained. A two-stage solution methodology based on a modified
The reconfiguration algorithms can be classified by the simulated annealing technique for solving the reconfiguration
solution methods that they employ: those based upon a blend problem of distribution systems was proposed in [6]. In [7], a
of heuristics and optimization methods, those making use of modified SA technique for network reconfiguration for loss
heuristics alone, and those using some from of artificial reduction in distribution systems was presented. An efficient
intelligence (AI). Numerous researchers advocate the use of a perturbation scheme and an initialization procedure

Reference Number: W09-0023 121

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

determining a better starting temperature for the simulated method, the genetic algorithm (GA), and the simulated
annealing approach were proposed. This method can get a annealing (SA). Numerical results in [14] show that ACSA
solution better than that obtained using the method presented method is better than the other two methods in terms of
in [5]. This solution algorithm gives a near optimal solution average power loss.
but this method does not work so well in the case of load Recently, power system applications have benefited from
variation. the powerful nature of Particle Swarm Optimization (PSO) as
A GA based method for feeder reconfiguration was a new optimization technique [15, 16]. The particle swarm
proposed in [8]. Strings which represent switch status, a algorithm is a method for optimizing hard numerical
fitness function consisting of total system losses, and penalty functions based on simulating the social behavior of bees and
values of voltage drop limit and current capacity limit were how can reach the location of most flower concentration.
formed. Sample results demonstrate that, although the In [17], a discrete particle swarm optimization (DPSO)
minimal loss solutions were obtained, solution time was algorithm was used as new algorithm for solving the
prohibitive. reconfiguration problems. The DPSO algorithm was applied
An artificial neural network based method for feeder to two test systems but, it was found that such method is often
reconfiguration was presented in [9]. However, such not efficient because the extremely large number of
technique can encounter difficulties, such as getting trapped unfeasible non-radial solutions appearing at each generation
in local minima, increased computational complexity, and not will lead to a long computing time before reaching an optimal
being applicable to certain objective functions. This led to the solution. DPSO algorithm uses the value 0 and 1 to denote
need of developing a new class of solution methods that can that the status of corresponding switch in the feeder is open or
overcome these shortcomings. closed, respectively.
A parallel tabu search (PTS) based method for feeder This paper proposes a modified PSO algorithm for
reconfiguration has been proposed in [10]. PTS introduces distribution system reconfiguration with a new variable
two parallel schemes. One is the decomposition of the expression design to overcome the drawbacks in the previous
neighborhood with parallel processors to reduce methods. To verify the effectiveness of the proposed method,
computational efforts. The other is the multiplicity of the tabu comparative studies are conducted on two test systems with
length to improve the solution accuracy. PTS algorithm gives encouraging results. The proposed method is applied to a 32-
results better than results obtained by SA, parallel Simulated node system and a 69-node system. The results obtained
Annealing (PSA), GA, and parallel Genetic Algorithm using the proposed PSO approach, are compared with results
(PGA). In [11], a TS algorithm for solving the problem of obtained using other modern techniques to examine the
network reconfiguration in distribution systems in order to performance.
reduce the resistive line losses under normal operating
II. PROBLEM FORMULATION
conditions was presented. A new method for checking system
radiality based on an upward-node expression, which has Generally, there are two types of switches in distribution
been developed in solving the problem of restorative planning systems: tie switch and sectionalizing switch. As shown in
of power system was proposed. In [12], an efficient hybrid Figure (1), switches in dotted branches connecting nodes (10-
algorithm of SA and TS method for feeder reconfiguration to 14), (5-11), and (7-16) are tie switches, and switches in other
improve the computation time and convergence property was continuous branches are sectionalizing switches. The tie
proposed. The authors of [13] have proposed a modified tabu switches are normally open and the sectionalizing switches
search (MTS) based method for reconfiguration of are normally closed. When the operating conditions have
distribution systems. The TS algorithm was introduced with been changed, feeder reconfiguration is performed by the
some modifications such as using a tabu-list with variable opening / closing of these two types of switches to reduce
size to prevent cycling and to escape from local minimum. resistive line losses.
Also, a constrained multiplicative move was used in the
search process to diversify the search process toward
unexplored regions.
An ant colony search algorithm (ACSA) to solve the optimal
network reconfiguration problem for power loss reduction has
been introduced in [14]. The ACSA is a relatively new and
powerful intelligence evolution method for solving
optimization problems. It is a population-based approach that
uses exploration of positive feedback as well as greedy
search. The ACSA was inspired from natural behavior of the
ant colonies on how they find the food source and bring them
back to their nest by building the unique trail formation. By
applying the ACSA, the near-optimal solution to the network
reconfiguration problem can be effectively achieved. The Figure (1): 16-node distribution system
network reconfiguration problems are solved using the ACSA

Reference Number: W09-0023 122

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

That is, a tie switch may be closed for the purpose of However, many advances in PSO development elevated its
transferring loads to different feeders, and, at the same time, a capabilities to handle a wide class of complex optimization
sectionalizing switch should be opened to maintain the radial problems involved in engineering and science.
structure of the distribution network. For example, in Figure PSO is a population-based evolutionary technique that has
(1), when the loads of feeder 2 become heavy under normal many key advantages over other optimization techniques like
operating conditions, the tie switch connecting nodes (5-11) [15, 16]:
may be closed to transfer the load at bus 11 from feeder 2 to  It is a derivative-free algorithm unlike many conventional
feeder 1 and at the same time the sectionalizing switch techniques.
connecting nodes (9-10) must be opened to maintain the  It has the flexibility to be integrated with other optimization
radial structure of the network. techniques to form a hybrid tool.
The objective of the reconfiguration is to minimize the  It is less sensitive to the nature of the objective function,
distribution losses with turning on / off sectionalizing i.e., convexity or continuity.
switches. The reconfiguration problem has the following
constrains:
 It has less parameter to adjust unlike many other competing
evolutionary techniques.
1. Power flow equations.
2. Upper and lower bounds of nodal voltages.  It has the ability to escape from local minima.
3. Upper and lower bounds of line currents.  It is easy to implement and program with basic
4. Feasible conditions in terms of network topology. mathematical and logic operations.
Mathematically, the problem can be formulated as follows:  It can handle objective functions with stochastic nature.
Cost function:  It does not require a good initial solution to start its
L ( Pi2  Qi2 ) iteration process.
Min Z =  rI (1) Different variants of the PSO algorithm were proposed but
i 1 Vi 2 the most standard one is introduced by Shi and Eberhart in
Subject to: [19]. Key attractive feature of PSO is its simplicity as it
g(x) = 0 (2) involves only two model equations. In PSO, the coordinates
min
Vi < Vi <
max
Vi (3) of each particle represent a possible solution associated with
min max
two vectors, the position ( s i ) and velocity ( vi ) vectors. The
Ii <I i < Ii (4)
det(A) = 1 or -1 radial system (5)
size of vectors si and vi is equal to the problem space
det(A) = 0 not radial (6) dimension. A swarm consists of number of particles “or
where, possible solutions” that proceed (fly) through the feasible
Z: Cost function solution space to explore optimal solutions. Each particle
L: No. of transmission lines updates its position based on its own best exploration; best
Pi: Active power loss at bus i swarm overall experience, and its previous velocity vector
Qi: Reactive power at bus i according to the following model:
Vi: Voltage at bus i ( pbest i  s ik ) ( gbest i  s ik ) (7)
v k  1  wv k  c * rand () *
i i 1 2  c * rand () *
t t
Ii: Line current at line i
g(x):Power flow equations
1 1
min s ik  s ik  v i k * t (8)
Vi : Lower voltage limit
max
where,
Vi : Upper voltage limit
vik is the ith velocity component at iteration k
min
Ii : Lower current limit rand() is random number between 0 and 1
max
Ii : Upper current limit sik is the current position in the ith dimension
A: Bus incidence matrix c1 , c 2 are the acceleration coefficients, that are usually
ri : Section resistance set to 2.0
pbest i is the personal best position in the ith dimension
III. PARTICLE SWARM
gbesti is the global best position in the ith dimension
Kennedy and Eberhart first introduced particle swarm
optimization (PSO) in 1995 as a new heuristic method. The t is the time step
original objective of their research was to graphically w is the inertia weight
simulate the social behavior of bird flocks and fish schools. The modifications to the proposed PSO algorithm are
As their research progressed, they discovered that with some explained in details in the following sections:
modifications, their social behavior model can serve as a
powerful optimizer. The first version of PSO was intended to
handle only nonlinear continuous optimization problems.

Reference Number: W09-0023 123

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

A. Inertia Weight The chosen position for each particle represents the tie line
The original PSO velocity update equation can be obtained by number, in the configuration tie , to be closed, when this tie
setting w  1 . The inertia weight governs how much of the is closed this will create a loop in the network. To restore the
previous velocity should be retained from the previous time system back to radial structure, one of the sectionalizing
step. The best performance, however, was again obtained by switches in the loop must be opened. In this work, the
using an inertia weight that decreases from 0.9 to 0.4 during sectionalizing switch with minimum power loss has been
the first 1500 iterations [20]. In this paper, further empirical chosen to be opened, and it is found that this assumption
experiments have been performed with an inertia weight set leads to good results.
to decrease linearly from 0.9 to 0.2 during the course of a In this paper, the velocity of the particle must be bounded by
simulation. This setting allows the PSO to explore a large this interval [-mv mv]. It is not necessary for the velocity to
area at the start of the simulation, when the inertia weight is be integer as it is used only to update the particle's position.
large, and to refine the search later by using a smaller inertia
C. Variable Expression Design
weight. In addition, damping the oscillations of the particles
around gbest is another advantage gained by using a In applying modern heuristic methods, such as SA and GA to
decreasing inertia weight. These oscillations are recorded solve the problem of distribution network reconfiguration, it
when a large constant inertial weight is used. Accordingly, is very important to choose a good variable expression. This
damping such oscillations assists the particles of the swarm to is also true for the PSO-based distribution network
converge to the global optimal solution. In brief, the inertia reconfiguration problem. An initial attempt is to choose all
weight can be likened to the temperature parameter feeders with switches as a set of variables to represent the
encountered in SA [21]. For the value of inertia weight (w), it solution of the problem. With such a variable expression,
is assumed to decrease linearly during the course of the each element of the solution vector represents one feeder with
simulation from 0.9 to 0.2 according to: a switch. The value 0 or 1 of one element in the solution
w  wmax  vector denotes that the status of corresponding switch in the
w( i )   min ( i  1 )  wmax
 (9) feeder is open or closed, respectively [17]. It was found that
 itermax  1  such a variable expression is often not efficient because the
where, extremely large number of unfeasible non-radial solutions
w(i): The inertia weight at iteration i. appearing at each generation will lead to a long computing
wmin : The minimum inertia weight (final), default = 0.2. time before reaching an optimal solution. A good variable
expression design, which can restrict each trial solution to be
wmax : The maximum inertia weight (initial), default = 0.9. radial networks in distribution network reconfiguration, is
itermax : Iteration by which inertial weight should be at final very important to improve the efficiency of search process as
indicated in the proposed algorithm for checking radiality of
value, default = 1500.
the system in the following section.
B. Repair Algorithm
D. Algorithm for Checking the Radial Topology of the
After position updating of each particle, a position check is System
carried out to make sure that none of the particles have flied
out of the search space bounds or violated the constraints, i.e. In this section a new algorithm based on the bus incidence
all the generated solutions are feasible. If a violation is matrix  is proposed for checking the radiality of trial
detected, a repair algorithm is used to force the violating solutions. The flow chart of the algorithm is shown in Figure
particle to return to the feasible region as follows: (2). A graph may be described in terms of a connection or
 If position of particle j (pos (j)) is greater than the incidence matrix. Of particular interest is the branch-to-node
maximum position (mp), then pos (j) = mp. Else if pos (j) is incidence matrix Â, which has one row for each branch and
less than one, then pos (j) = 1. Else, end. one column for each node with an entry a ij in row i and
 If velocity of particle j (Vel (j)) is greater than the
column j according to the following rules:
maximum velocity (mv), then Vel (j) = mv.
 Else if Vel (j) is less than the minimum velocity (-mv), then a ij = 0 if branch i is not connected to node j (10)
Vel (j) = -mv. Else, end. a ij = 1 if branch i is directed away from node j (11)
In this paper, the position of the particle represents the a ij = -1 if branch i is directed toward node j (12)
number of the tie line to be closed. Therefore, the values of
the positions of the particles must be integers, positive and These rules formalize the procedure used to set up the
bounded by this interval [1 mp] where, mp is the total number coefficient of Â. In network calculation, a reference node
of the tie lines. So, pos (j) must be approximated to the must be chosen. The column corresponding to the reference
nearest integer number. node is omitted from  and the resultant matrix is denoted by
A. If the number of branches is equal to the number of nodes
then, by applying the previous rules a square branch-to-node

Reference Number: W09-0023 124

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

matrix is obtained. The non reference nodes of a network are 5. Decrease the inertia weight (w) linearly as explained in
often called independent nodes or buses, and when it is said section (III.A).
that the network has N buses, this means that there are N 6. Repeat steps 2–5 until a termination criterion is satisfied.
independent nodes not including the reference node. The A The flow chart of the particle swarm algorithm is shown in
matrix has the row-column dimension B  N for any network Figure (3).
with B branches and N nodes excluding the reference node. A computer program is developed to implement the
By assuming, that there is a branch between this reference proposed PSO algorithm using MATLAB 6.5, and executed
node and the root of the network; this will lead to a square on a Pentium III 700-MHz PC with 128-MB RAM.
matrix if the initial structure of the network is radial. The new Regarding the population size (number of particles) used
proposed method is based on the value of the determinant of during the simulation; most of previously published PSO
A. It is found that, if the determinant of A is equal to 1 or -1, literature recommended the use of a population of about 10 to
then the system is radial. Else if the determinant of A is equal 20 agents. It has been found during simulation that using only
to zero, this means that either the system is not radial or 5 agents are enough to reach a satisfactory result. For the
group of loads are disconnected from service. maximum number of iterations to be done, it was found that a
The advantage of using the proposed method range of 50 to 100 iterations is suitable for most cases.
 It reduces the computation time. Although in some other cases especially for relatively small
 It restricts each trial solution to be radial network in dimensional problems, a fewer number of iterations, may be
distribution network reconfiguration. from 20 to 40 iterations is enough to reach a satisfactory
 It can be used to determine the branches of the loop formed result.
by closing a tie line.

Figure (2): Checking system radiality algorithm

IV. THE APPLICATION OF PSO TO SOLVE THE
RECONFIGURATION PROBLEM
The PSO algorithm is described in general as follows:
1. Input data and initialize parameters. For each particle, the
position and velocity vectors will be randomly initialized
with the same size as the problem dimension.
2. Measure the fitness (power loss) of each particle (pbest)
and store the particle with the best fitness (gbest) value
by running the load flow program. Figure (3): A flow chart of the proposed PSO algorithm
3. Update velocity and position vectors according to
equation (7) and (8) for each particle. The PSO parameters used during the simulation study for the
4. Perform violation check. If violation detected then, apply reconfiguration of the distribution network are summarized in
''Repair Algorithm'' explained in section (3.2), else end. Table 1.

Reference Number: W09-0023 125

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

Table 1: PSO Parameters used During the Simulation

Parameter Value
Number of particles 5
Initial velocity of the agent 0.0
Maximum inertia weight 0.9
Minimum inertia weight 0.2
c1 & c2 2.0
A. Application to the IEEE 32-Node Distribution System Figure (5): Optimal configuration of the 32-node system
The 32-node test distribution system is a 12.66 kV with one using the PSO algorithm
main feeder, 3 laterals, 32 buses, and 5 tie lines. The system Figure (6) shows the voltage profile improvement achieved
data is given in [2]. The schematic diagram of the system is by the proposed PSO algorithm. As shown, most of the bus
shown in Figure (4). voltages have been improved after feeder reconfiguration.
The initial system real power loss was 202.6 kW. By applying The minimum bus voltage was equal to 0.9131 p. u. before
the proposed PSO algorithm, the final power loss is 139.5 reconfiguration and the minimum bus voltage is raised to
kW. 0.9378 p. u. after reconfiguration.
It is shown from the simulation results listed in Table 2 that
the power loss after reconfiguration is reduced by 31.1% of
its initial value. These results are identical to those obtained
by the hybrid SA + TS method proposed in [12] and the MTS
method proposed in [13]. The optimal configuration using the
PSO algorithm is shown in Figure (5).
Table 2: 32-node system results using the PSO algorithm
Voltage
Power
System profile Tie Lines
Loss
(p. u.)
Before 202.6 kW V=1 7-20 Figure (6): Voltage profile for the 32- bus system after using
reconfiguration V = 0.9131 8-14 the PSO algorithm
11-21
17-32 The total number of iterations of the program is 40 iterations
24-28 and eleven of them lead to power loss reduction as shown in
After 139.5 kW V=1 6-7
Figure (7). Also, ( itermax ), iteration by which inertial weight
reconfiguration V = 0.9378 13-14
using [SA+TS] 8-9 (w) should be at final value is iteration no. 25 and after that
[12] 31-32 number the inertial weight will be constant and equal to 0.2.
24-28
After 139.5 kW V=1 6-7
The obtained results using the proposed PSO algorithm have
reconfiguration V = 0.9378 13-14 been reached after 50 trails. Table 3 shows the statistics of the
using MTS 8-9 cost function in the simulation results.
algorithm [13] 31-32
24-28
After 139.5 kW V=1 6-7
reconfiguration V = 0.9378 13-14
using proposed 8-9
PSO algorithm 31-32
24-28

Figure (7): The best moves recorded during the search

process for the 32-node system using the PSO algorithm
B. Application to the IEEE 69-Node Distribution System

Figure (4): Initial configuration of the 32-node distribution The 69-node distribution test system is a 12.66 kV system
system with 69 buses and 7 laterals. The system data is given in [6].
The schematic diagram of the test system is shown in Figure
(8).

Reference Number: W09-0023 126

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

Table 3: Cost function statistics for the 32-node system using reconfiguration; it is raised to 0.982 p.u. The total number of
the PSO algorithm iterations of the program is 70 iterations and 10 of them lead
to power loss reduction as shown in Figure (11). Also,
Best cost Average cost Worst cost
function function function ( itermax ), iteration by which inertial weight (w) should be at
139.5 kW 141 kW 143 kW final value is iteration no. 50 and after that number the inertial
weight will be constant and equal to 0.2. The obtained results
using the proposed PSO algorithm have been reached after 50
trails.
Table 5 shows the statistics of the cost function in the
simulation results.

Figure (8): Initial configuration of the 69-node distribution

system
The initial system real power loss was 20.88 kW. By applying
the proposed PSO algorithm, the final power loss is 9.4 kW.
It is found from the simulation results listed in Table 4 that
the power loss after reconfiguration is reduced by 55% of its
initial value. These results are identical to the results obtained
by in [6, 13 and 17]. The optimal configuration using the PSO Figure (9): Optimal configuration of the 69-node distribution
algorithm is shown in Figure (9). system using the PSO algorithm (Normal load)
Table 4: 69-node system results using the PSO algorithm Table 5: Cost function statistics for the 69-node system using
the PSO algorithm
Voltage
Power Tie
System profile Best cost Average cost Worst cost
Loss Lines
(p. u.) function function function
Before 20.89 kW V=1 7-20 9.4 kW 10 kW 13 kW
reconfiguration V = 0.9131 8-14
11-21
17-32
24-28
After 9.4 kW V=1 6-7
Reconfiguration V = 0.9378 13-14
using SA [6] 8-9
31-32
24-28
After 9.4 kW V=1 6-7
reconfiguration V = 0.9378 13-14
using MTS 8-9
algorithm [13] 31-32 Figure (10): Voltage profile for the 69- node distribution
24-28 system using the PSO algorithm
After 9.4 kW V=1 6-7
Reconfiguration V = 0.9378 13-14
Using DPSO 8-9
[17] 31-32
24-28
After 9.4 kW V=1 6-7
reconfiguration V = 0.9378 13-14
using proposed 8-9
PSO algorithm 31-32
24-28
Figure (10) shows the voltage profile improvement achieved
by the proposed PSO algorithm. As shown, most of the bus Figure (11): The best moves recorded during the search
voltages have been improved after feeder reconfiguration. process for the 69-node system using the PSO algorithm
The minimum bus voltage was equal to 0.9724 p.u. and after

Reference Number: W09-0023 127

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

V. CONCLUSION [5] R. J. Sarfi, M. M. A. Salama and A. Y. Chikhani,

“Distribution System Reconfiguration for Loss
This paper has proposed the PSO algorithm, as a new
Reduction: An Algorithm Based on Partitioning
evolutionary technique, for reconfiguration of distribution
Theory,” IEEE Transactions on Power Systems, Vol. 11,
systems. The main advantage of solving such problems using
No. 1, February 1996, pp 504-510.
PSO over the conventional mathematical methods is its
[6] H. D. Chiang and R. Jean-Jumeau, “Optimal Network
simplicity. The results obtained during simulation showed
Reconfigurations in Distribution Systems: Part II,” IEEE
that the proposed PSO algorithm is capable of finding an
Transactions on Power Delivery, Vol. 5, No. 3, July
optimal or a near-optimal solution to the two cases studied in
1990, pp. 1568-1574.
this paper. Moreover, the algorithm is capable of finding a
[7] Hong-Chan Chang and Cheng-Chien Kuo, “Network
feasible solution to the 32-bus system and the 69-bus system.
Reconfiguration in Distribution Systems using Simulated
The main advantage of the proposed algorithm is that no
Annealing,” Electric Power Systems Research, Vol. 29,
penalty function is used to handle the violating (infeasible)
No. 3, May 1994, pp. 227-238.
solutions making it a problem independent, i.e. it is applicable
[8] K. Nara, A. Shiose, M. Kitagawa and T. Ishihara,
to any other optimization problem regardless of the shape of
“Implementation of Genetic Algorithm for Distribution
the feasible region or the constraints (linear or non-linear),
Systems Loss Minimum Reconfiguration,” IEEE
although linear constraints are the only constraints considered
Transactions on Power Systems, Vol. 7, No. 3, August
in this paper. A good variable expression design, which can
1992, pp. 1044-1051.
restrict each trial solution to be radial networks in distribution
[9] H. Kim, Y. Ko and K. Jung, “Artificial Neural Network
network reconfiguration, has been presented to improve the
Based Feeder Reconfiguration for Loss Reduction in
efficiency of search process. PSO algorithm is introduced
Distribution Systems,” IEEE Transactions on Power
with some modifications such as using an inertia weight that
Delivery, Vol. 8, No. 3, July 1993, pp. 1356-1366.
decreases linearly from 0.9 to 0.2 during the course of a
[10] H. Mori and Y. Ogita, “A Parallel Tabu Search Method
simulation. This setting allows the PSO to explore a large
for Reconfiguration of Distribution Systems,”
area at the start of the simulation. A modification in the
Proceedings of the 2000 IEEE Power Engineering
number of iterations and the population size is also,
Society Summer Meeting, Vol. 1, pp. 73-78.
presented. The modified PSO algorithm proposed in this
[11] K. K. Li, T. S. Chung, G. J. Chen and G. Q. Tang, “A
paper requires only 5 agents; however, it can converge to a
Tabu Search Approach to Distribution Network
satisfactory result. Consequently, the proposed PSO
Reconfiguration for Loss Reduction,” Electric Power
algorithm outperforms the standard PSO in terms of
Components and Systems, Vol. 32, No. 6, June 2004, pp.
computational speed and memory requirements, as it requires
571-585.
only 25% of the agents recommended by most of the
[12] Young-Jae Jeon and Jae-Chul Kim, “Application of
previously published PSO. The validity and effectiveness of
Simulated Annealing and Tabu Search for Loss
the proposed PSO algorithm is well proved by two sample
Minimization in Distribution Systems,” Electric Power
test systems. For the 32-node system in comparison with [12,
and Energy Systems, Vol. 26, No. 1, January 2004, pp.
13] same results are obtained. Also, for the 69-node system in
9-18.
comparison with [6, 13, 17] same results are obtained. The
[13] S. F. Mekhamer, A. Y. Abdelaziz, F. M. Mohammed and
execution time of the developed software program is
M. A. L. Badr, “A New Intelligent Optimization
acceptable for practical applications.
Technique for Distribution Systems Reconfiguration”,
REFERENCES Proceedings of the Twelfth International Middle-East
Power Systems Conference MEPCON’2008, South
[1] S. Civanlar, J. J. Grainger and S. H. Lee, “Distribution
Valley University, Egypt, March 2008, pp. 397-401.
Feeder Reconfiguration for Loss Reduction,” IEEE
[14] Ching-Tzong Su, Chung-Fu Chang and Ji-Pyng Chiou,
Transactions on Power Delivery, Vol. 3, No. 3, July “Distribution Network Reconfiguration for Loss
1988, pp. 1217-1223. Reduction by Ant Colony Search Algorithm,” Electric
[2] M. E. Baran and F. F. Wu, “Network Reconfiguration in
Power Systems Research, Vol. 75, No. 2-3, pp. 190–199,
Distribution Systems for Loss Reduction and Load
August 2005.
Balancing,” IEEE Transactions on Power Delivery, Vol.
[15] M. R. Alrashidi and M. El-Hawary, "A Survey of
4, No. 2, April 1989, pp. 1401-1407.
Particle Swarm Optimization Applications in Power
[3] T. Taylor and D. Lubkeman, “Implementation of System Operations," Electric Power Components and
Heuristic Search Strategies for Distribution Feeder Systems, Vol. 34, No. 12, pp. 1349-1357, December
Reconfiguration,” IEEE Transactions on Power
2006.
Delivery, Vol. 5, No. 1, January 1990, pp. 239-246.
[16] Y. del Valle, G. K. Venayagamoorthy, S. Mohagheghi, J.
[4] T. P. Wagner, A. Y. Chikhani and R. Hackam, “Feeder
C. Hernandez and R. G. Harley, "Particle Swarm
Reconfiguration for Loss Reduction: An Application of
Optimization: Basic Concepts, Variants and
Distribution Automation,” IEEE Transactions on Power Applications in Power Systems", IEEE Transactions on
Delivery, Vol. 6, No. 4, October 1991, pp. 1922-1931.

Reference Number: W09-0023 128

The Online Journal on Electronics and Electrical Engineering (OJEEE) Vol. (1) – No. (2)

Evolutionary Computation, Vol. 12, No. 2, April 2008,

pp. 171-195.
[17] S. Sivanagaraju, J. V. Rao and P. S. Raju, “Discrete
Particle Swarm Optimization to Network
Reconfiguration for Loss Reduction and Load
Balancing,” Electric Power Components and Systems,
Vol. 36, No. 5, May 2008, pp. 513–524.
[18] J. Kennedy and R. Eberhart, "Particle Swarm
Optimization," Proceedings of the 1995 IEEE Neural
Networks Conference, IEEE Service Center, Piscataway,
NJ, pp. 1942-1948.
[19] Y. Shi and R. Eberhart, "A Modified Particle Swarm
Optimizer," Proceedings of the 1998 IEEE Evolutionary
Computation Conference, pp. 69-73.
[20] Y. Shi and R. C. Eberhart, "Empirical Study of Particle
Swarm Optimization," Proceedings of the 1999
Evolutionary Computation Congress, IEEE Service
Center, Piscataway, NJ, pp. 1945-1949.
[21] F. Van den Bergh, "An Analysis of Particle Swarm
Optimizer," Ph. D. Dissertation, Faculty of Natural and
Agricultural Science, University of Pretoria, South
Africa, 2001.

Reference Number: W09-0023 129