This article was downloaded by: [National Institute of Technology - Warangal]

On: 25 June 2014, At: 08:24

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Electric Power Components and Systems
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uemp20
Optimal Reconfiguration in Radial Distribution System
Using Gravitational Search Algorithm
Y. Mohamed Shuaib
a
, M. Surya Kalavathi
b
& C. Christober Asir Rajan
c
a
Department of Electrical and Electronics Engineering, Jawaharlal Nehru Technological
University Hyderabad, B. S. Abdur Rahman University, Chennai, Tamilnadu, India
b
JNTUH College of Engineering, Hyderabad, India, Andhra Pradesh
c
Department of Electrical and Electronics Engineering, Pondicherry Engineering College,
Pondicherry, India
Published online: 24 Apr 2014.
To cite this article: Y. Mohamed Shuaib, M. Surya Kalavathi & C. Christober Asir Rajan (2014) Optimal Reconfiguration in
Radial Distribution System Using Gravitational Search Algorithm, Electric Power Components and Systems, 42:7, 703-715, DOI:
10.1080/15325008.2014.890971
To link to this article: http://dx.doi.org/10.1080/15325008.2014.890971
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Electric Power Components and Systems, 42(7):703715, 2014
Copyright C Taylor & Francis Group, LLC
ISSN: 1532-5008 print / 1532-5016 online
DOI: 10.1080/15325008.2014.890971
Optimal Reconguration in Radial Distribution
System Using Gravitational Search Algorithm
Y. Mohamed Shuaib,
1
M. Surya Kalavathi,
2
and C. Christober Asir Rajan
3
1
Department of Electrical and Electronics Engineering, Jawaharlal Nehru Technological University Hyderabad, B. S. Abdur
Rahman University, Chennai, Tamilnadu, India
2
JNTUH College of Engineering, Hyderabad, Andhra Pradesh, India
3
Department of Electrical and Electronics Engineering, Pondicherry Engineering College, Pondicherry, India
CONTENTS
1. Introduction
2. Load Flow
3. Mathematical Problem Statement
4. GSA
5. Test System
6. Result Analysis
7. Conclusion
References
Keywords: gravitational search algorithm, network reconguration, radial
distribution system, 33-bus radial distribution system, 69-bus radial
distribution system, tie Switches, I
2
R losses, distributed generation
Received 7 May 2013; accepted 19 January 2014
Address correspondence to C. Christober Asir Rajan, Department of
Electrical and Electronics Engineering, Pondicherry Engineering College,
Pondicherry, 605 014, India. E-mail: asir 70@pec.edu
Color versions of one or more of the gures in the article can be found online
at www.tandfonline.com/uemp.
AbstractThis article presents an innovative technique for solving
network reconguration problems with an objective of minimizing
network I
2
R losses for an explicit set of loads. Amid many per-
formance standards considered for optimal network reconguration,
voltage constraint is an important one. This problem calls for deter-
mining the best combination of feeders to be opened in the radial dis-
tribution system so it provides optimal performance in the preferred
settings. In solving this problem, the gravitational search algorithm
is used to recongure the radial distribution system; this algorithm
practices an optimal pattern for sustaining the radial nature of the
network at every stage of the solution, and it further allows procient
exploration of the solution space. The anticipated scheme minimizes
the objective function, which has been given in the problem formula-
tion to reduce I
2
R losses in addition to balancing loads in the feeders.
The solution technique involves determination of the best switching
combinations and calculation of power loss and voltage prole. The
practicality of the anticipated technique is validated in two distri-
bution networks, where attained results are compared by means of
available literature. Correspondingly, it is seen from the results that
network losses are reduced when voltage stability is enriched through
network reconguration.
1. INTRODUCTION
Distribution systems are customarily designed radially, and
there are two categories of switches usually found in the system
envisioned for both protection and conguration management.
These are called closed and open switches; closed switches are
termed sectionalizing switches, and open switches are termed
as tie switches. Aradial distribution system(RDS) engages dif-
ferent types of loads, such as industrial, commercial, domestic
etc. The demand prole of these loads may possibly vary from
time to time and will perhaps root imbalanced power ow in
the feeder, possibly leading to voltage collapse owing to low
voltages. When there is greater I
2
Rloss in the RDS, the voltage
in the buses may violate the voltage constraint. This can affect
the quality of the power supply and the stability of the system.
703
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

704 Electric Power Components and Systems, Vol. 42 (2014), No. 7
NOMENCLATURE
best(t) = best tness at generation t
F
d
i
(t ) = total force acting on ith agent
t
j
(t) = tness of jth agent at iteration t
M
ai
= active gravitational mass of ith agent
M
i i
= inertia mass of ith agent
M
pi
= passive gravitational mass of ith agent
NVB = number of buses that violate recommended volt-
age limits
P
acc
= real power accumulated
P
i
= real power load demand in bus i
P
i j (loss)
= real power loss in line connecting ith and jth bus
P
j
= real power injection at jth bus
Q
acc
= reactive power accumulated
Q
i
= reactive power load demand in bus i
Q
j
= reactive power injection at jth bus
R
i j
= resistance in line connecting ith and jth bus
V
i
= voltage magnitude of bus i
V
j
= voltage magnitude of jth bus
V
L
= upper limit of ith load bus
V
max
= maximum voltage limit
V
min
= minimum voltage limit
worst(t) = worst tness at generation t
x
d
i
= positions of ith agent in dth dimension

j
= phase angle at jth bus
One way to sustain the safety and reliability of the system is to
recongure the RDS. This alteration of the network topology
is done by means of altering the status of the open and closed
switches. Certain switches can be operated remotely despite
the fact that the others are opened or closed by a lineman.
When scheduling network reconguration, the demand prole
of different consumers are considered to perceive if a particu-
lar conguration is safe, is reliable, and has adequate capacity
to supply all the customers. Once the feeder reconguration
is done, overburdening of the feeder is reduced, I
2
R loss is
minimized, and the voltage prole of the system is improved,
thereby leading to voltage stability enhancement.
Venkatesh and Ranjan [1] presented a solution technique
that used a fuzzy adaptation of evolutionary programming
(FEP). This method is suitable when considering optimization
of multiple objectives. Sahoo and Prasad [2] advanced the al-
teration of network topology using a fuzzy genetic approach
(FGA) to maximize the voltage stability of the network. The
FGApractices an apt coding and decoding scheme for sustain-
ing the radial nature of the network at every stage of genetic
evolution in addition to using a fuzzy-rule-centered mutation
controller for efcient search of the solution space. AbulWafa
[3] described a load ow based on graph theory where the de-
veloped load owalgorithmwas integrated into a newheuristic
search methodology for nding the minimum loss congured
network. Kashemet al. [4] proposed a branch loss-change tech-
nique where they derived a loss-change formula to determine
the change of losses in the system when a branch exchange
is performed. The best branch exchange to be implemented
is chosen at each successive operation that gives a maximum
loss reduction without any constraints being violated.
There are many ways of handling a heuristic algorithm
based on the modication or adaptation of the algorithm. The
authors in [5, 6] used heuristic algorithms for the recongu-
ration problem to achieve the lowest I
2
R loss in the network,
but the algorithm is called heuristic until the best solution is
proven to be the best. Rao et al. [7] presented a meta-heuristic
harmony search algorithm (HSA) that was used concurrently
to recongure and to detect optimal locations for installa-
tions of distributed generation (DG) units. Nasiraghdam and
Jadid [8] introduced a fresh multi-objective articial bee
colony (MOABC) algorithm to explain RDS reconguration
and hybrid (photovoltaic/wind turbine/fuel cell) energy system
sizing. MOABC outlines a potential solution for the optimiza-
tion problem as a food source and the tness value of the solu-
tion as the nectar amount of the allied food source. Swarnkar
et al. [9] investigated a method centered on adapted ant colony
optimization (AACO) for the reconguration of RDS. AACO
overcomes the drawbacks of the conventional ant colony opti-
mization technique by encoding the discrete ant by means of
the graph theory.
Savier and Das [10] discussed the impact of network recon-
guration on loss allocation based on a fuzzy multi-objective
approach, where loss allocation and network reconguration
where considered together, in which the loss allocation is by
the quadratic loss allocation pattern and the network recon-
guration is framed using a fuzzy multi-objective problem.
Kashem et al. [11] stated a geometrical approach for loss min-
imization in which each loop in a network is represented as a
circle, which is again derived from the relationship between
the change of loss due to the branch exchange and the power
ows in the branches. Abdelaziz et al. [12] projected real ant-
behavior-inspired ant colony optimization implemented in the
hypercube framework and a musician-behavior-inspired HSA
to address the objective function. Wu et al. [13] minimized
power loss in an RDS by network reconguration in the pres-
ence of distributed generators, using an ant colony algorithm
(ACA) to do so. Kumar and Jayabarathi [14] addressed the
feeder loss problembased on a bacterial-foraging optimization
algorithm (BFOA). The optimization problem was considered
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 705
as a non-linear problem and used to nd the optimal solution.
The BFOA was inspired by the social foraging behavior of Es-
cherichia coli. The authors utilized this BFOA to congure an
RDS to keep the load balancing so that I
2
R loss is minimized.
Li et al. [15] recommended a tabu search (TS) approach to
obtain near-optimal solutions of combinatorial optimization
problems, which makes it appropriate to resolve the problem
of RDS reconguration.
Gomes et al. [16] offered an approach for RDS recong-
uration based on optimum power ow in which the branch
statuses are represented by continuous functions. Their arti-
cle described a methodology that uses an optimal power ow
(OPF) program based on a heuristic algorithm to solve the
objective function. Rao et al. [17] prescribed a newly devel-
oped technique to nd the optimal switching status of the
recongured RDS, which was conceptualized using the musi-
cal process of searching for an impeccable state of harmony;
the algorithm is called the HSA. The HSA practices stochas-
tic random search as an alternative to gradient search, which
removes the necessity of derivative information. Gupta et al.
[18] reduced I
2
R loss by creating an initial population for par-
ticle swarm optimization (PSO) using a heuristic method in
which the particles are adapted with the help of graph the-
ory to generate feasible individuals. Installation costs were not
considered in the approach. A step-by-step heuristic algorithm
based on sensitivity indexes was carried out by Rosseti et al.
[19] to allocate DGwith reconguration in electric distribution
system to minimize energy losses.
In the light of the above progress, this article presents an
optimization technique called the gravitational search algo-
rithm (GSA), which has been developed as a solution tech-
nique for feeder reconguration. The application of the GSA
for minimization of I
2
R losses due to network reconguration
is formulated as a multi-objective problem subject to opera-
tional and electric constraints. The objectives considered are
I
2
R loss reduction, voltage at buses to be kept within a spec-
ied range, and persistence of network radial structure, even
after reconguration in which all loads must be served. A load
ow program was developed, and the algorithm based on [20]
is used to compute the power ow. It is then integrated into the
GSA for determining the minimum loss RDS conguration.
The distribution network presented by Baran and Wu [21] is
used to demonstrate the reliability and efcacy of the proposed
algorithm.
This article is organized as follows. Section 2 describes the
load ow, and Section 3 delivers the mathematical model of
the problem. The GSA and its ability to solve the optimization
problem are discussed in Section 4. The test system and result
analysis are addressed in Sections 5 and 6, respectively. Finally,
the article is concluded in Section 7.
2. LOAD FLOW
To accommodate the present emerging domestic, industrial,
and commercial loads day by day, effective forecasting of the
RDS is essential. To ensure effective planning with load trans-
ferring, the load ow study of an RDS becomes of the utmost
signicant. Load owanalysis is concerned with describing the
operating state of an entire power system. NewtonRaphson
and fast decoupled load ow solution techniques are used to
solve well-behaved power systems; however, these are gener-
ally unsuitable for solving load ow for RDSs because of their
low X/R ratios of branches.
Asection of the RDShas a sending-end bus (the ith bus) and
a receiving-end bus (the jth bus). The line in connection with
these two sections has an impedance (Z = r + j x). The power
owthrough this line can be in both directions. The power ow
at the sending-end bus (S
i
= P
i
+ j Q
i
) is different from the
power ow at the receiving-end bus (S
j
= P
j
+ j Q
j
).
A load ow algorithm [20] solves the power balance equa-
tions at all buses and nds the corresponding voltage solution.
At load buses, the load ow algorithm will solve for the bus
voltage magnitude and phase angle. The known parameters at
a load bus are the received real and reactive powers. Hence, a
load ow must solve for the bus voltage magnitude in Eq. (1)
[20] and phase angle in Eq. (2) [20]:
V
2
j
=
_
r P
j
+ x Q
j

V
2
i
2
_
+
_
_
r P
j
+ x Q
j

V
2
i
2
_
2

_
r
2
+ x
2
_
_
P
2
j
+ Q
2
j
_
, (1)

j
=
i
sin
1
_
x P
j
r Q
j
V
i
V
j
_
. (2)
If the voltage magnitude and phase angle values are to be
computed for the receiving-end bus, the only variables needed
are the receiving-end bus real and reactive power values, the
sending-end bus voltage magnitude and phase angle value, and
the value of the line impedance connecting the two buses. All
values needed for the load bus calculations are easily attainable
in practice.
2.1. Load Flow Algorithmic Steps
Step 1: Read system data structure.
Step 2: Go to slack bus.
Step 3: Initialize P
acc
= 0 and Q
acc
= 0.
Step 4: Calculate P and Q for all buses.
Step 5: Calculate V
j
and
j
for all buses using Eqs. (1) and
(2).
Step 6: Determine P
loss
and Q
loss
for all lines.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

706 Electric Power Components and Systems, Vol. 42 (2014), No. 7
Step 7: Update P
acc
and Q
acc
using the formulas P
acc
=
P
loss
+ P
j
and Q
acc
= Q
loss
+ Q
j
.
Step 8: Go to the next bus and reprise Step 4 to 8 up until last
bus is reached.
Step 9: Check for convergence (Eq. (3)) and print the result;
else, go to Step 2.
2.1.1. Convergence Criteria.
In this load ow, it is checked whether the sum of powers
owing out of the lines connected to each bus equals (or equals
within a tolerable limit) the net power injected into that bus.
Mathematically, convergence criteria for the presented load
ow are given in Eq. (3) [20]:
(PG
i
PD
i
)
_
_

j
(V
i
V
j
Y
i j
cos(
i

j

i j
))
_
_
,
(QG
i
QD
i
)
_
_

j
(V
i
V
j
Y
i j
sin(
i

j

i j
))
_
_
.
(3)
3. MATHEMATICAL PROBLEM STATEMENT
3.1. Total Real and Reactive Power Loss
The real power loss in the line connecting the ith and jth bus
is given by
P
i j (loss)
=
[P
j
]
2
+[Q
j
]
2
[V
i
]
2
R
i j
, (4)
Q
i j (loss)
=
[P
j
]
2
+[Q
j
]
2
[V
i
]
2
X
i j
. (5)
The total real power loss of an RDS having n buses and n 1
branches is given by
P
T,loss
=
n

i =1
P
i j (loss)
, (6)
Q
T,loss
=
n

i =1
Q
i j (loss)
, (7)
3.2. Voltage Deviation Index (VDI)
The VDI [1] is calculated using the formula
VDI =
_

NV B
i =1
(V
Li
V
L
)
2
N
. (8)
To enumerate the degree of limits violation imposed on
voltages at buses in an RDS, the VDI is well-dened, where
NVB is the number of buses that violate the recommended
voltage limits, and V
L
is the upper limit of the ith load bus.
In the course of reconguration, if the state of the system has
voltage limit violations, the anticipated solution must try to
lessen the VDI. When a branch is switched on and another is
switched out in a loop, the solution space is no longer con-
tinuous. The variable that denes the status of a branch as to
whether it is switched in/out adopts discrete states of zero or
one. Owing to the discontinuous and discrete nature of the
problem, classical techniques are rendered inappropriate, and
the practice of global search techniques is essential.
3.3. Objective Function for Network Reconguration
in RDS
The objective is to minimize the I
2
R losses in an RDS, thereby
enhancing the voltage prole of the system. This is attained
by nding out the best set of branches to be switched out such
that the subsequent RDS experiences the lowest I
2
R loss and
has the best voltage prole.
The mathematical model of the problem can be expressed
by the following expression:
minimize f =
n

i =1
P
i j (loss)
+
_

NV B
i =1
(V
Li
V
L
)
2
N
,
subject to V
min
|V
i
| V
max
. (9)
The rst term in the Eq. (9) represents the total I
2
R loss in
the system, and the second term denotes the VDI.
4. GSA
This article adopts a new search algorithmthe GSAas the
solution technique for feeder reconguration. This algorithm
was originally developed by Rashedi et al. [22] to address
various non-linear problems. Duman et al. [23] proposed the
GSA to decide the optimal settings for control variables of
the optimal power ow problem in power systems. To nd the
optimum emission dispatch, optimum fuel cost, best compro-
mising emission, and fuel cost, Mondal et al. [24] applied the
GSA to solve the optimization problem.
In this article, the GSA is applied to minimize feeder losses
by reconguring the RDS. It is formulated as a real power
loss minimization problem subject to operational and electric
constraints. The GSA is based on the law of gravity and mass
interactions. In this algorithm, the searcher agents are a group
of masses that act together with each other based on Newto-
nian gravity and the laws of motion. The algorithm considers
agents as objects consisting of different masses. Complete
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 707
agents move due to the gravitational attraction force acting
among them, and the advancement of the algorithm directs the
movements of all agents globally headed toward agents with
heavier masses. Every agent in the GSA is specied by four
parameters: the position of the mass in the dth dimension, in-
ertia mass, active gravitational mass, and passive gravitational
mass.
4.1. GSA Algorithmic Steps
Step 1: Initialization of the agents. Initialize the positions of N
numbers of agents randomly chosen within the given
search interval using Eq. (10):
X
i
=
_
x
1
i
, . . . , x
d
i
, . . . , x
n
i
_
for i = 1, 2, . . . , N.
(10)
Step 2: Fitness evolution and best tness computation for
each agent. Performthe tness evolution for all agents
at each iteration; also compute the best and worst t-
ness at each iteration dened for minimization prob-
lems in Eqs. (11) and (12):
best(t ) = min
j {1,...,N}
t
j
(t ), (11)
worst(t ) = max
j {1,...,N}
t
j
(t ). (12)
Step 3: Compute gravitational constant G. Compute gravita-
tional constant G at iteration t using Eq. (13):
G (t ) = G
o
e
(t /T)
. (13)
Step 4: Calculate the mass of the agents; calculate gravita-
tional and inertia masses for each agent at iteration t
by Eq. (11):
M
ai
= M
pi
= M
i i
= M
i
(i = 1, 2, . . . , N),
m
i
(t ) =
t
i
(t ) worst
i
(t )
best(t ) worst(t )
,
M
i
(t ) =
m
i
(t )

N
j =1
m
j
(t ).
(14)
Step 5: Calculate accelerations of the agents; compute the
acceleration of the ith agents at iteration t, Eq. (15):
a
d
i
(t ) =
F
d
i
(t )
M
i i
(t )
; (15)
F
d
i
(t ), the total force acting on the ith agent, is calcu-
lated from Eq. (16):
F
d
i
(t ) =

j Kbest, j =i
rand
j
F
d
i j
(t ). (16)
Step 6: Update velocity and positions of the agents; compute
velocity and position of the agents at the next iteration
(t + 1) using Eq. (17):
v
d
i
(t +1) = rand
i
v
d
i
(t ) +a
d
i
(t ),
x
d
i
(t +1) = x
d
i
(t ) +v
d
i
(t +1). (17)
Step 7: Repeat Steps 2 through 6 until iterations reach their
maximum limit; return the best tness computed at
the nal iteration as a global tness of the problemand
the positions of the corresponding agent at specied
dimensions as the global solution of that problem.
4.2. GSA Algorithm for Reconguration of RDS
An RDS can have several loops. Each loop has several
branches. Any one of these branches forming the loops must
be switched out such that the radial nature of the RDS is
maintained. The switching in/out of branches alters the ow
of power and changes the resulting kilowatt losses and volt-
age prole. The objective of the solution technique is to de-
termine that combination of branches such that the resulting
RDS yields the minimum kilowatt losses and the best voltage
prole.
Let vector X
i
= (x
1
i
, . . . , x
d
i
, . . . , x
n
i
) (i = 1, 2, . . ., N)
refer to the combination of branches switched out in the RDS
such that the ID number of branches switched out in the ith
loop is stored in variable X
i
. The GSA initially chooses NC
combinations as the starting guesses. The steps of the GSA to
reconguration problem of RDS are given in what follows.
4.2.1. Algorithmic Steps
Step 1: Initialization. Initialize the positions of the N number
of agents, which is the different switching statuses
randomly chosen within the given search interval us-
ing Eq. (10) and check for its radiality as given in
[20].
Step 2: Best tness computation for each combination of
switches. Perform the tness evaluation for all status
of switches at each iteration that is radial in structure
for the given systems; also compute the best and worst
tness, which is again the network conguration that
gives the least I
2
> R loss and the maximum I
2
R loss
at each iteration dened for minimization problems
in Eqs. (11) and (12).
Step 3: Attain gravitational constant G. Find gravitational
constant G at iteration t using Eq. (13).
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

708 Electric Power Components and Systems, Vol. 42 (2014), No. 7
GSA parameters 33-Bus 69-Bus
N = number of agents 2000 2500
Maximum iterations 500 750
20 20
G
0
100 100
TABLE 1. GSA parameters
Step 4: Estimate the mass of the agents. Calculate gravita-
tional and inertia masses for each agent at iteration t
by Eq. (14).
Step 5: Evaluate accelerations of the agents; compute the
acceleration of the ith agent at iteration t using Eqs.
(15).
Step 6: Find the velocity and the position of agents; calculate
velocity and position of agents at the (t +1)th iteration
using Eq. (17).
Step 7: Repeat Steps 2 throught 6 until iterations reach their
maximum limit; return the best switching combina-
tion that ultimately gives the minimal I
2
Rlosses found
at the nal iteration as an optimal network topology
of the problem.
4.2.2. Parameters for GSA Algorithm.
To get an optimal solution using the GSA, the parameters in
Table 1 have been used to nd the optimum switching status
due to which the resulting RDS yields the minimum I
2
R loss
and the best voltage prole.
The performance of the GSA in network reconguration
of an RDS is estimated. Twenty independent trials have been
made, with 2000 agents and 500 iterations per trial for the
33-bus test system and 2500 agents and 750 iterations per trial
for the 69-bus test system. The value and gravitational time
constant G
0
are set to 20 and 100, respectively, in both cases.
The optimal results obtained for network reconguration using
these considerations in the algorithm are tabulated in Tables 2
and 3.
5. TEST SYSTEM
5.1. 33-Bus Test System
The I
2
R loss minimization by reconguration is executed on a
33-bus RDS (Figure 1), and the load data are given in [17]. The
test system consists of 33 buses, 32 lines, and 5 tie switches.
The rst bus is considered as the substation bus. Loads are
connected to all buses except the rst bus, which is the sub-
station bus. The total real power load and reactive power load
of this test system are 3715 kW and 2300 kVAR, respectively.
The substation voltage is 12.66 kV.
5.2. 69-Bus Test System
To validate the proposed method, the reconguration process
is carried out on a 69-bus RDS (Figure 2). The load and trans-
mission line details were given in [10]. The test systemconsists
of 69 buses, 68 lines, and 5 tie switches. The total real power
load and reactive power load of this test system are 3801.89
kW and 2694.1 kVAR, respectively. The substation voltage is
12.66 kV.
6. RESULT ANALYSIS
6.1. Case A
The results of the 33-bus test system are given in Case A. The
program is coded in MATLAB software (The MathWorks,
Natick, Massachusetts, USA), which is installed on an Intel
R
Core
TM
i5-2410M CPU @ 2.30 GHz with a setup memory of
4.00 GB and a 64-bit operating system (Dell Inspiron N5110).
Final conguration
Technique Initial
conguration RGA Heuristic OPF heurisic HSA BFOA GSA
Author Rao et al. Zhu Gomes et al. Gomes et al. Rao et al. Kumar and Proposed
[17] [28] [25] [16] [17] Jayabarathi [14] algorithm
Year 2011 2002 2005 2006 2011 2012 2013
Open switches s33, s34, s35,
s36, s37
s7, s9, s14,
s32, s33
s7, s9, s14,
s32, s37
s7, s9, s14,
s32, s37
s7, s10, s14,
s37, s36
s7, s9, s13,
s14, s32
s7, s14, s28,
s9, s32
Power loss (kW) 202.418 139.53 136.57 136.57 138.06 135.78 134.61
Loss reduction
(%)
31.07 32.53 32.53 31.793 32.92 33.49
Minimum voltage
(p.u.)
0.9237 0.9315 0.9502 0.9502 0.9342 0.9589 0.9604
TABLE 2. Case A: Result analysis
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 709
Final conguration
Technique Initial
conguration FMOA Meta-heuristic Meta-heuristic GSA
Author Savier and Savier and Swarnkar et al. HSA Proposed
Das [10] Das [10] [6] Rao et al. [7] algorithm
Year 2007 2007 2011 2013 2013
Open switches s69, s70, s71,
s72, s73
s14, s56, s61,
s69, s70
s69, s18, s13,
s56, s61
s14, s58, s61,
s69, s70
Power loss (kW) 224.894 99.59 99.59 99.35 98.57
Loss reduction (%) 55.72 55.72 55.82 56.17
Minimum voltage (p.u.) 0.9092 0.9483 0.9428 0.9428 0.9495
TABLE 3. Case B: Result analysis
Before reconguration, tie switches s33, s34, s35, s36, and
s37 are kept open. For the given total real power load of 3715
kW, the I
2
R loss is obtained as 202.418 kW in the base case.
The minimum voltage is registered as 0.9237 p.u. at the 18th
bus. The optimal network conguration for loss reduction is
achieved after applying the GSA, when all tie switches s33,
s34, s35, s36, and s37 are closed and sectionalizing switches
s7, s14, s28, s9, and s32 are now opened. As a result, the I
2
R
loss is reduced to 134.6104 kW from the base case of 202.418
kW, witnessing 67.808 kW of real power loss reduction. The
worst voltage is found to be 0.96044 p.u. at the 32nd bus, also
falling within the voltage limits.
In 2005 and 2006, Gomes et al. [16, 25] carried out re-
conguration of a 33-bus RDS using a heuristic algorithm
and reduced the I
2
R loss to 136.57 kW. In 2011, Rao et al.
[17] proposed the HSA and lessened the losses to 138.05 kW.
Kumar and Jayabarathi [14] suggested the BFOA in 2012 to
reduce the losses to 135.78 kW. In this article, the GSA is
proposed and loss reduction is 33.49%, where the percentage
of loss reduction is more when compared to previous results.
This proves the effectiveness of the proposed algorithm. The
results of Case A and comparison of the proposed algorithm
with other optimization techniques are tabulated in Table 2.
The optimally recongured structure of the RDS is drawn in
Figure 3. The voltage prole of the RDS before and after re-
conguration is shown in Figure 4.
6.2. Case B
The results of the 69-bus test system are given in Case B. The
program is coded in MATLAB software, which is installed in
FIGURE 1. Line diagram of 33-bus RDS.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

710 Electric Power Components and Systems, Vol. 42 (2014), No. 7
FIGURE 2. Line diagram of 69-bus RDS.
an Intel
R
Core
TM
i5-2410M CPU @ 2.30 GHz with a setup
memory of 4.00 GB and a 64-bit operating system. Before
reconguration, switches s69, s70, s71, s72, and s73 are open,
and for the total real power load of 3801.89 kW, the I
2
R loss
is 224.894 kW in the base case. The minimum voltage is
registered as 0.9092 p.u. at the 65th bus. The optimal network
conguration for loss reduction is achieved after applying the
GSA, where tie switches s71, s72, and s73 are closed, and
sectionalizing switches s14, s58, and s61 are now opened. As
a result, the I
2
R loss is reduced to 98.5718 from 224.894 kW,
witnessing 126.3231 kW of real power loss reduction. The
worst voltage is found to be 0.9495 p.u. at the 61st bus.
Savier and Das in 2007 [10] carried out network recongu-
ration using a fuzzy multi-objective approach and reduced I
2
R
losses to 99.59 kW. In 2011, Swarnkar et al. [6] proposed a
meta-heuristic algorithm to answer the reconguration prob-
lem and lessen the losses to 99.59 kW. Rao et al. [7] in 2013
introduced a meta-heuristic HSA to do network recongura-
tion and decreased losses to 99.35. In this article, the GSA
is proposed and the loss reduction is 56.17%, where the per-
centage of loss reduction is more when compared to all results
arrived using various algorithms; this proves the efcacy of the
proposed algorithm. The results of Case B and the comparison
of the proposed algorithm with other optimization techniques
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 711
FIGURE 3. Optimally recongured topology of 33-bus RDS.
are tabulated in Table 3. The optimally recongured structure
of the RDS is drawn in Figure 5. The voltage proles of the
RDS before and after reconguration are shown in Figure 6.
6.3. Results Study of Network Reconguration
in the Presence of DG Units
Kumar and Navuri [26] determined the optimumpositions and
size of DG units with the objective of improving the voltage
prole and loss reduction in an RDS. The selection of can-
didate buses was carried out using a loss sensitivity factor
(LSF), whereas the estimation of optimal DG unit size was
by means of simulated annealing (SA). Gandomkar et al. [27]
lessened distribution losses by incorporating DG units to the
RDS through implementing a genetic algorithm to the TS ap-
proach. The effectiveness of the GSA is veried by installing
DG units in 33- and 69-bus RDSs. The scenario has been seg-
FIGURE 4. Comparison of voltage prole of 33-bus RDS.
mented into Item I and Item II. In both scenarios, DG units are
installed after network reconguration.
6.3.1. Item I
The I
2
R loss in the 33-bus test system after reconguration is
134.6104 kW. After the installation of DG to the recongured
network, the losses were reduced to 71.12 kW, presenting a loss
reduction of 64.86%. The candidate bus locations selected for
the installation of DG units are 19, 23, and 22.
6.3.2. Item II
In the 69-bus test system, I
2
R loss after reconguration is
98.57 kW. The losses were reduced to 50.25 kW after installa-
tion of DG units to the recongured network, witnessing a loss
reduction of 77.88%. The candidate bus locations selected for
the installation of DG units are 14, 40, and 48.
6.3.3. Performance of Proposed Algorithm
In 2013, Rao et al. [7] installed DG units after reconguration
to minimize I
2
R loss. The authors reduced the losses in 33-
and 69-bus test systems to 97.13 and 51.30 kW, respectively.
The results of the test systems in the establishment of DG
after reconguration and the results comparison of the pro-
posed algorithm with [7] are tabulated in Table 4. The voltage
proles of the 33- and 69-bus RDSs in the presence of a DG
unit after reconguration are shown in Figures 7 and 8 re-
spectively. The result comparison shown in Table 4 proves the
efcacy of the proposed algorithm.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

712 Electric Power Components and Systems, Vol. 42 (2014), No. 7
FIGURE 5. Optimally recongured topology of 69-bus RDS.
FIGURE 6. Comparison of voltage prole of 69-bus RDS.
FIGURE 7. Comparison of voltage prole of 33-bus RDS in
the presence of DG.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 713
Rao et al. (2013) [7], meta-heuristic
HSA Proposed algorithm (GSA)
Scenario 33-Bus 69-Bus 33-Bus 69-Bus
Base case Open switches s33, s34, s35,
s36, s37
s69, s70, s71,
s72, s73
s33, s34, s35,
s36, s37
s14, s58, s61,
s69, s70
Power loss (kW) 202.67 225 202.418 224.894
Minimum voltage
(p.u.)
0.9131 0.9092 0.9237 0.9495
Item I (only
reconguration)
Open switches s7, s14, s9,
s32, s37
s69, s18, s13,
s56, s61
s7, s14, s28,
s9, s32
s14, s58, s61,
s69, s70
Power loss (kW) 138.06 99.35 134.61 98.57
Minimum voltage
(p.u.)
0.9342 0.9428 0.9604 0.9495
Loss reduction (%) 31.88 55.85 33.49 56.17
Item II (DG
installation after
reconguration)
Open switches s7, s14, s9,
s32, s37
s69, s18, s13,
s56, s61
s7, s14, s28,
s9, s32
s14, s58, s61,
s69, s70
Size of DG in kW
(bus number)
269 (32)
161 (31)
661 (30)
1067 (61)
353 (60)
426 (58)
694 (19)
1998 (23)
1066 (22)
357 (14)
139 (40)
784 (48)
Power loss (kW) 97.13 51.3 71.12 50.25
Minimum voltage
(p.u.)
0.9479 0.9619 0.9772 0.9625
Loss reduction (%) 52.07 77.2 64.86 77.66
TABLE 4. Performance of GSA in the presence of DG
FIGURE 8. Comparison of voltage prole of 69-bus RDS in
the presence of DG.
7. CONCLUSION
This article proposes a GSA for optimal reconguration of an
RDS to achieve the best voltage prole and minimal kilowatt
losses. AVDI is established herein that computes the deviancy
of load bus voltages from the recommended limits. This in-
dex is minimized in the proposed algorithm to improve power
quality. The RDS reconguration is a discontinuous solution
space problem with discrete zeroone variables and an objec-
tive that requires the determination of the best combination of
feeders in the RDS to be switched out so the resulting RDS
gives the minimumkilowatt losses and the best voltage prole.
The GSA technique is found particularly suitable for solving
optimization problems with discontinuous solution space and
objectives when the global optimum is desired. The optimal
solution in the GSA is obtained by movement of agents in the
search space, and its direction is based on the overall force
of all other agents. Therefore, the search direction toward the
optimal solution is effective in this algorithm. The proposed
method is tested on established 33- and 69-bus RDSs. The
results obtained demonstrate that the GSA method optimally
recongures the RDS, minimizing the kilowatt losses and ob-
taining the best voltage prole.
REFERENCES
[1] Venkatesh, B., and Ranjan, R., Optimal radial distribution sys-
temreconguration using fuzzy adaptation of evolutionary pro-
gramming, Int. J. Elect. Power Energy Syst., Vol. 25, No. 10,
pp. 775780, December 2003.
[2] Sahoo, N. C., and Prasad, K., A fuzzy genetic approach for
network reconguration to enhance voltage stability in radial
distribution systems, Energy Convers. Manag., Vol. 47, No.
1819, pp. 32883306, November 2006.
[3] AbulWafa, A. R., A new heuristic approach for optimal re-
conguration in distribution systems, Elect. Power Syst. Res.,
Vol. 81, No. 2, pp. 282289, February 2011.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

714 Electric Power Components and Systems, Vol. 42 (2014), No. 7
[4] Kashem, M. A., Jasmon, G. B., and Ganapathy, V., A new
approach of distribution system reconguration for loss mini-
mization, Int. J. Elect. Power Energy Syst., Vol. 22, No. 4, pp.
269276, May 2000.
[5] Mena, A. J. G., and Garca, J. A. M., Comments on a new
heuristic approach for optimal reconguration in distribution
systems, Elect. Power Syst. Res., Vol. 83, No. 1, pp. 264265,
February 2012.
[6] Swarnkar, A., Gupta, N., and Niazi, K. R., A novel codi-
cation for meta-heuristic techniques used in distribution net-
work reconguration, Elect. Power Syst. Res., Vol. 81, No. 7,
pp. 16191626, July 2011.
[7] Rao, R. S., Ravindra, K., Satish, K., and Narasimham, S. V.
L., Power loss minimization in distribution system using net-
work reconguration in the presence of distributed genera-
tion, IEEE transactions on power systems, Volume 28, Issue 1,
pp. 317325, February 2013.
[8] Nasiraghdam, H., and Jadid, S., Optimal hybrid PV/WT/FC
sizing and distribution system reconguration using multi-
objective articial bee colony (MOABC) algorithm, Solar En-
ergy, Vol. 86, No. 10, pp. 30573071, October 2012.
[9] Swarnkar, A., Gupta, N., and Niazi, K. R., Adapted ant colony
optimization for efcient reconguration of balanced and un-
balanced distribution systems for loss minimization, Swarm
Evolut. Computat., Vol. 1, No. 3, pp. 129137, September 2011.
[10] Savier, J. S., and Das, D., Impact of network reconguration
on loss allocation of radial distribution systems, IEEE Trans.
Power Delivery, Vol. 22, No. 4, pp. 24732480, October 2007.
[11] Kashem, M. A., Ganapathy, V., and Jasmon, G. B., Ageometri-
cal approach for network reconguration based loss minimiza-
tion in distribution systems, Int. J. Elect. Power Energy Syst.,
Vol. 23, No. 4, pp. 295304, May 2001.
[12] Abdelaziz, A. Y., Osama, R. A., and Elkhodary, S. M., Distri-
bution systems reconguration using ant colony optimization
and harmony search algorithms, Elect. Power Compon. Syst.,
Vol. 41, No. 5, pp. 537554, February 2013.
[13] Wu, Y.-K., Lee, C.-Y., Liu, L.-C., and Tsai, S.-H., Study
of reconguration for the distribution system with distributed
generators, IEEE Trans. Power Delivery, Vol. 25, No. 3,
pp. 16781685, July 2010.
[14] Kumar, K. S., and Jayabarathi, T., Power system recongu-
ration and loss minimization for a distribution systems using
bacterial foraging optimization algorithm, Int. J. Elect. Power
Energy Syst., Vol. 36, No. 1, pp. 1317, March 2012.
[15] Li, K. K., Chung, T. S., Chen, G. G. J., and Tang, G. Q., A
tabu search approach to distribution network reconguration
for loss reduction, Elect. Power Compon. Syst., Vol. 32, No. 6,
pp. 571585, June 2010.
[16] Gomes, F., Carneiro Jr., S., Pereira, J. L. R., Vinagre, M. P.,
Garcia, P. A. N., and Ramos de Araujo, L., A new distribution
system reconguration approach using optimum power ow
and sensitivity analysis for loss reduction, IEEE Trans. Power
Syst., Vol. 21, No. 4, pp. 16161623, November 2006.
[17] Rao, R. S., Narasimham, S. V. L., Raju, M. R., and Rao, A. S.,
Optimal network reconguration of large-scale distribution
system using harmony search algorithm, IEEE Trans. Power
Syst., Vol. 26, No. 3, pp. 10801088, August 2011.
[18] Gupta, N., Swarnkar, A., and Niazi, K. R., Reconguration
of distribution systems for real power loss minimization using
adaptive particle swarm optimization, Elect. Power Compon.
Syst., Vol. 39, No. 4, pp. 317330, February 2011
[19] Rosseti, G. J. S., de Oliveira, E. J., de Oliveira, L. W., Silva
Jr., I. C., and Peres, W., Optimal allocation of distributed gen-
eration with reconguration in electric distribution systems,
Elect. Power Syst. Res., Vol. 103, pp. 178183, October 2013.
[20] Venkatesh, B., and Ranjan, R., Data structure for radial distri-
bution systemload owanalysis, IEEEProc. Generat. Transm.
Distribut., Vol. 150, No. 1, pp. 101106, January 2003.
[21] Baran, M. E., and Wu, F. F., Network reconguration in dis-
tribution systems for loss reduction and load balancing, IEEE
Trans. Power Delivery, Vol. 4, No. 2, pp. 14011407, April
1989.
[22] Rashedi, E., Nezamabadi-pour, H., and Saryazdi, S., GSA:
A gravitational search algorithm, Info. Sci., Vol. 79, No. 13,
pp. 22322248, June 2009.
[23] Duman, S., Guvenc, U., Sonmez, Y., and Yorukeren, N., Op-
timal power ow using gravitational search algorithm, Energy
Conversion Manag., Vol. 59, pp. 8695, July 2012.
[24] Mondal, S., Bhattacharya, A., and nee Dey, S. H., Multi-
objective economic emission load dispatch solution using grav-
itational search algorithm and considering wind power pen-
etration, Int. J. Elect. Power Energy Syst., Vol. 44, No. 1,
pp. 282292, January 2013.
[25] Gomes, F. V., Carneiro, S., Pereira, J. L. R., Vinagre, M. P., and
Garcia, P. A. N., A new heuristic reconguration algorithm for
large distribution systems, IEEE Trans. Power Syst., Vol. 20,
No. 3, pp. 13731378, August 2005.
[26] Kumar, I. S., and Navuri, P. K., An efcient method for optimal
placement and sizing of multiple distributed generators in a
radial distribution systems, Distrib. Generat. Alternat. Energy
J., Vol. 27, No. 3, pp. 5271, July 2012.
[27] Gandomkar, M., Vakilian, M., and Ehsan, M., Agenetic-based
tabu search algorithm for optimal DG allocation in distribu-
tion networks, Elect. Power Compon. Syst., Vol. 33, No. 12,
pp. 13511362, February 2007.
[28] Zhu, J. Z., Optimal reconguration of electrical distribution
network using the rened genetic algorithm, Elect. Power Syst.
Res., Vol. 62, No. 1, pp. 3742, May 2002.
BIOGRAPHIES
Y. Mohamed Shuaib received his B.E in electrical and elec-
tronics engineering from University of Madras in 1994 and
his M.E. in power system engineering from Annamalai Uni-
versity, Tamilnadu, India, in 2002. He is a research scholar
in Jawaharlal Nehru Technological University, Hyderabad, In-
dia, and since 2003 has been working as an assistant professor
(selection grade) in the Department of Electrical and Elec-
tronics Engineering, B.S. Abdur Rahman University, Chennai.
His current research includes capacitor placement and recon-
guration in radial distribution networks in the presence of
distributed generation.
M. Surya Kalavathi received her B.Tech and M.Tech fromSri
Venkateswara University, Tirupathi, Andhra Pradesh, India, in
1988 and 1992, respectively. She obtained her doctoral degree
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4

Shuaib et al.: Optimal Reconguration in Radial Distribution System Using Gravitational Search Algorithm 715
from Jawaharlal Nehru Technological University, Hyderabad,
and her post-doctoral from Carnegie Mellon University, USA.
She is presently the professor in the Department of Electrical
and Electronics Engineering at Jawaharlal Nehru Technologi-
cal University Hyderabad College of Engineering, Kukatpally,
Hyderabad, India. She has published 16 research papers and
is presently guiding ve Ph.D. scholars. Her research inter-
ests include simulation studies on transients of different power
system equipment.
C. Christober Asir Rajan was born in 1970 and received his
B.E. (distn.; electrical and electronics) and his M.E. (distn.;
power systems) from Madurai Kamaraj University, Madurai,
India, in 1991 and 1996, respectively. He received his post-
graduate degree in DI.S. (distn.) from Annamalai University,
Chidambaram, in 1994. He received his Ph.D in power sys-
tem from Anna University, Chennai, India (20012004). He
is currently working as an associate professor in the Electrical
Engineering Department at Pondicherry Engineering College,
Pondicherry, India. He has published technical papers in inter-
national and national journals and conferences. His areas of
interest are power system optimization, operational planning,
and control.
D
o
w
n
l
o
a
d
e
d

b
y

[
N
a
t
i
o
n
a
l

I
n
s
t
i
t
u
t
e

o
f

T
e
c
h
n
o
l
o
g
y

-

W
a
r
a
n
g
a
l
]

a
t

0
8
:
2
4

2
5

J
u
n
e

2
0
1
4