Ain Shams Engineering Journal (2018) 9, 403–414

Ain Shams University

Ain Shams Engineering Journal

www.elsevier.com/locate/asej
www.sciencedirect.com

ELECTRICAL ENGINEERING

Optimal allocation of remote control switches in

radial distribution network for reliability
improvement
Saheli Ray 1, Subhadeep Bhattacharjee 2, Aniruddha Bhattacharya *

Department of Electrical Engineering, National Institute of Technology Agartala, Tripura 799046, India

Received 21 April 2015; revised 5 September 2015; accepted 8 January 2016

Available online 23 April 2016

KEYWORDS Abstract This paper presents differential search algorithm in order to solve reliability opti-
Differential search mization problem of radial distribution network. Remote control switches have been opti-
algorithm; mally allocated to improve reliability at a compromised cost. A multi-objective problem
Multi-objective function; has been formulated and solved using differential search algorithm. The test systems consid-
Radial distribution system; ered in this paper are an 8 bus radial distribution network and a 33 bus radial distribution
Remote control switches; network. Simulation results obtained using differential search algorithm when applied to the
Repair time; test cases, have been compared with those obtained by particle swarm optimization. Differ-
Restoration time ential search algorithm has been found to provide superior results as compared to particle
swarm optimization.
Ó 2016 Production and hosting by Elsevier B.V. on behalf of Ain Shams University. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction has increased. As distribution system has the greatest contribu-

tion to the interruption of supply to a consumer [1]; hence,
Distribution system reliability has proved to be of great improving distribution system reliability is of serious concern
concern in the present days of power system operation. With in today’s power market. The enhancement of reliability
the deregulation of power system and enhanced competitive always incurs a cost as it involves some additional preventive
environment, the demand for uninterrupted quality power and corrective measures. So, the reliability improvement meth-
ods need to be adopted keeping in view the cost involved in the
* Corresponding author. Mobile: +91 9862469286. process. Failure rate, repair time and restoration time are some
E-mail addresses: [email protected] (S. Ray), subhadeep_bhattacharjee@ important parameters of defining reliability. Reducing the values
yahoo.co.in (S. Bhattacharjee), [email protected] of one or more of the above parameters can improve reliability
(A. Bhattacharya). considerably. Several approaches can be adopted to improve
1
Mobile: +91 8794772113.
2
Mobile: +91 9436582874.
reliability, out of which, the present authors have adopted opti-
Peer review under responsibility of Ain Shams University.
mal placement of remote control switch (RCS) in the radial dis-
tribution network. RCSs are devices, which can isolate or
connect a section of a network. Suitable locations of RCSs in
a network may reduce the time to resume power and thus
Production and hosting by Elsevier improve reliability. Placing one RCS at each segment of a
http://dx.doi.org/10.1016/j.asej.2016.01.001
2090-4479 Ó 2016 Production and hosting by Elsevier B.V. on behalf of Ain Shams University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
404 S. Ray et al.

namely average interruption frequency index (SAIFI) and

system average interruption duration index (SAIDI) as well
as the cost of equipments. Golestani and Tadayon [23] used
Linear Fragmented Particle Swarm optimization for optimal
switch placement in distribution system. Assis et al. [24] pro-
posed a memetic algorithm based optimization methodology
to sectionalizing, tie, manual, and automatic switches in distri-
bution networks. Amanulla et al. [25] used binary particle
swarm optimization-based search algorithm to find the
optimal status of the switches in order to maximize the reliabil-
ity and minimize the real power loss. Zou et al. [26] adopted
methods including feeder reconfiguration, recloser installation,
recloser replacement, and distributed generation (DG) installa-
tion to minimize system average interruption duration index
(SAIDI), an important reliability index. Brown et al. [27] used
sequential feeder method and a multi-objective genetic algo-
rithm (GA) together to solve the optimization of the feeder
addition problem in an islanded distribution system with
Figure 1 Eight bus network (Test case I).
DGs. Vitorino et al. [28] presented the application of an
improved genetic algorithm (IGA) to optimize simultaneously
network definitely improves reliability greatly, but at the same loss and reliability of a radial distribution system through a
time it may incur a high installation and maintenance cost, as process of network reconfiguration as an optimization. Zhang
the number of RCSs required is large. Hence, a compromise is et al. [29] proposed a reliability-oriented reconfiguration
required, and here lies the importance of optimal allocation of (ROR) method for improving distribution reliability and
RCSs. While adopting the present work, a number of literatures energy efficiency, based on interval analysis. Pfitscher et al.
have been reviewed in which similar type of work has been done. [30] presented a new methodology for automatic reconfigura-
Some of these are briefly discussed here. tion of distribution network, in order to improve network
An artificial intelligence technique with multi agent system performance indicators, such as losses and reliability.
was used by Bouhouras et al. [2] for performing cost/worth Kavousi-Fard and Akbari-Zadeh [31] proposed a multi-
assessment of reliability improvement in distribution networks. objective distribution feeder reconfiguration problem for relia-
Haifenga et al. [3] adopted Monte-Carlo simulation based bility enhancement as well as loss reduction. Raofat [32]
approach for providing a basis for using a parallel computing adopted a GA based method to allocate DGs and RCSs simul-
environment in power system reliability and cost evaluations. taneously in order to reduce energy loss and improve reliability
Switch allocation problem has been a topic of research interest considering multilevel load.
for decades and many studies have been performed [4–6]. Recently, Pinar Civicioglu [33] introduced a new algorithm
RCSs are gaining importance in reliability improvement named differential search (DS) algorithm to solve the problem
studies with the recent trend of automation. Some studies have of transforming geocentric cartesian coordinates into geodetic
been carried out in order to develop strategies for RCS without coordinates and compared its performance with classical
covering allocation of switches [7,8]. Allocation of switches has methods and other computational intelligence algorithms. DS
been considered in [9–12]. Optimal placement of switches and algorithm adopts the seasonal migration behavior of many
reclosers has been considered in [13–14]. Abiri-Jahromi et al. organisms where they shift from one habitat to a more efficient
[15] utilized mixed integer linear programming (MILP) for one, in terms of efficiency of food areas. The individual organ-
optimal placement of sectionalizing switches. Viotto Romero isms form a Superorganism which as a whole move toward more
et al. [16] proposed a dedicated Taboo Search (TS) algorithm efficient area. The effectiveness of DS algorithm has already
for optimal switch allocation in distribution systems for been compared with other algorithms such as artificial bee
automatic load transfer. Bernardon et al. [17] proposed a colony algorithm (ABC), self-adaptive differential evolution
methodology to consider the impact of RCS when computing algorithm (JDE), adaptive differential evolution algorithm
the reliability indices and the algorithm for multi-criteria deci- (JADE), strategy adaptation based differential evolution algo-
sion making to allocate these switches. Benavides et al. [18] rithm (SADE), differential evolution algorithm with ensemble
proposed a new iterated sample construction with path of parameters (EPSDE), gravitational search algorithm (GSA),
relinking (ISCPR) to solve distribution system switch alloca- particle swarm optimization (PSO) and covariance matrix
tion problem. Zheng et al. [19] studied the quantitative impact adaptation evolution strategy (CMA-ES). DS algorithm has
of automatic switches on the reliability of power distribution been found to solve the problem at a very high level of accuracy
systems. Esmaeilian and Fadaeinedjad [20] adopted a Binary [33]. Unlike other algorithms such as differential evolutionary
Gravitational Search Algorithm (BGSA) for network reconfig- algorithm (DE), JDE, and ABC, DS algorithm may simultane-
uration and capacitor placement in distribution system in ously use more than one individual during updating steps. An
order to improve reliability. Tippacon and Rerkpreedapong important advantage of DS algorithm over many other algo-
[21] adopted multiobjective ant colony optimization (MACO) rithms is that DS algorithm has no inclination to correctly
whereas Pombo et al. [22] adopted a memetic algorithm approach the best possible solution. Therefore, exploration
combining Non dominated Sorting Genetic Algorithm II ability of the algorithm is significantly improved compared to
(NSGA-II) with a local search algorithm for switch and reclo- many other existing algorithms. Hence, it may be proved to be
sure allocation in order to minimize the reliability indices a successful strategy for solution of multimodal functions.
Allocation of remote control switches 405

As DS algorithm has proved to be a new and effective As these indices do not take into account the number of
evolutionary algorithm [33], the present authors have adopted customers and load connected, the severity of the fault is not
this algorithm with a view to test its computational efficiency revealed by these indices. To get a clear picture of the severity
to solve a multi-objective function in order to enhance system of the fault, customer oriented indices are derived from the
reliability at a reduced cost. The objective of this paper was to basic indices. Among several customer oriented indices,
solve a multi-objective function in order to find a compromised expected energy not supplied (EENS) is the index of concern
solution both to enhance the reliability by optimal allocation in this work which is given by
of RCSs and to minimize the cost incurred. In most of the X
EENS ¼ L j Uj ð1Þ
previous work, where optimal placement of switches has been
considered, number of RCS has been taken as fixed. In some where
literature where number of RCS has been considered as
variable, multi-objective problem formulation has not been X
njo
X
nj1

Uj ¼ ki0 repi0 þ ki1 resi1 ð2Þ

considered. In the present paper, both number and position i0 ¼1 i1 ¼1
of RCS have been considered as variable and a multi-
objective function has been formulated. The outcome of the ki0 and repi0 denote the failure rate and repair time of i0th
proposed technique has been compared with a well known distributor segment, and nj0 denotes the total numbers of seg-
and widely used optimization technique, PSO. ments where the fault has occurred and power can be resumed
Section 2 of the paper provides a brief description of the to the jth load point only after repairing of those faults. ki1 and
function of RCS in radial distribution system and its impact resi1 denote the failure rate and switching time or restoration
on the reliability parameters. Section 3 describes mathematical time of i1th distributor segment, and nj1 denotes the total num-
formulation of the optimization problem. Section 4 presents bers of segments where the fault has occurred and power can
the DS algorithm and the steps involved to solve the optimal be restored to the jth load point through switching operation
RCS allocation problem in order to enhance distribution before repairing of those faults. Uj denote the annual outage
system reliability. Simulation studies are presented and dis- duration for jth load point. Lj is the average load connected
cussed in Section 5. The conclusion is drawn in Section 6. at jth load point.

2. RCS in radial distribution network and reliability indices 2.1. Logic for energy interruption duration calculation of a given
load point
With the recent trends of automation of distribution networks,
RCS is proved to be very convenient as its switching time is Distribution segments are branches of a distribution network.
very less. RCS may be sectionalizing-switch (normally closed) The failure rate (k), repair time (rep) and restoration time (res)
or tie-switch (normally open). In the present work, the RCS of distribution segments affect the reliability of load points.
considered for installation is normally closed type. In radial A load point experiences interruption of power for a failure
network, normally closed RCS can be operated to isolate a in any segment if either
faulty section from the rest of the network. The location of
RCS can contribute to enhance the reliability of a network (a) The segment is in the path between the source and the
to a great extent. load or
The basic reliability indices commonly used are failure rate, (b) The segment is not in the path between the source and
repair time, restoration time and outage duration. Failure rate the load but there is no fuse in between the segment
denotes the frequency of occurrence of failure. Repair time and the load point.
represents the time required to repair a faulty section after a
fault occurs. Restoration time represents the time required to After a failure occured in such a distribution segment, the
restore service after an interruption occurs. Outage duration time to resume power in the load point may be repair time
represents the annual duration of outage and is given either or restoration time. This can be selected using following
by the product of failure rate and repair time or by the product conditions:
of failure rate and restoration time, as applicable.
While RCS does not affect the failure rate, however, it can (i) If the segment is in the path between the source and the
have a considerable impact on the outage duration. Optimal load, time to resume power will be repair time.
placement of RCSs can reduce the outage duration to a consid- (ii) If the segment is not in the path between the source and
erable extent, thus improve the reliability. If there is a fault, the load and there is no RCS (remote control switch) in
located at downstream to the load point, and if there is no between the segment and the load point, time to resume
switch in between the fault and the load point of consideration, power will be repair time.
time to restore power to the load point will be equal to the time (iii) If, the segment is not in the path between the source and
needed to repair the fault i.e. repair time. On the other hand, a the load and there is at least one RCS in between the
switch in between the load point of consideration and fault segment and the load point, time to resume power will
location (downstream to the load point) can reduce this time be restoration time i.e. switching time.
to the operating time of the switch i.e. restoration time, as
the opening of switch will isolate the faulty segment from the EENS of a particular load point is obtained by the multipli-
healthy portion and power can be restored to the healthy cation of annual outage duration and load of that load point.
portion. The annual outage duration is obtained by the multiplication
406 S. Ray et al.

of failure rate and time to resume power (either repair time or ber of RCS and position of RCS have been considered as vari-
restoration time, as applicable), as presented in (2). Hence, the able while minimizing the objective function.
failure rate and repair time/restoration time of distribution
segments have a direct impact on the EENS of a load point. 4. Solution methodology using DS algorithm
With the increase of the failure rate and repair time/restoration
time, EENS also increases. DS algorithm simulates the Brownian-like random-walk
movement used by an organism to migrate [33]. Due to period-
3. Problem formulation ical climatic changes, many organisms show seasonal migra-
tion behavior where they shift from one habitat to a more
In this paper, the objective was to obtain the optimum number efficient one with respect to capacity and efficiency of food
and location of RCS in radial distribution system. Increasing areas. In the process of migration, the species undergoing
the number of RCS may reduce the EENS but at the same migration forms a Superorganism consisting of a large number
time, it may increase the cost involved. A multi-objective for- of individuals and the Superorganism changes its position
mulation is developed with a view to reduce the EENS cost toward more fruitful areas.
without excessive increase in RCS cost. Here, the target is to The artificial organisms (i.e. Xi, i = 1, 2, 3 . . . N) constituting
find a compromised solution such as to improve the reliability a Superorganism contain members equal to the size of the
(by reducing equivalent cost of EENS) without excessive problem (Xij, j = 1, 2, 3 . . . D). D is the size of the problem
increase in RCS cost. and N denotes the number of elements in a Superorganism.
The objective function to reduce EENS is In initial position, a member of an artificial organism is given
X
n by
J1 ¼ EENSj  C1  CPV1 ð3Þ Xij ¼ randðupj  lowj Þ þ lowj ð10Þ
j¼1
The Superorganism migrates toward global minimum and
where EENSj corresponds to the EENS of jth load point, n during this process, the members search for some randomly
corresponds to the total number of load points, C1 stands selected position suitable to stop over temporarily and on find-
for per unit cost of EENS ($/kW h) and CPV1 is the cumula- ing such position, the members of the artificial Superorganism
tive present value (CPV) of EENS cost. The CPV method con- immediately settle there and continue their migration from this
verts all costs and benefits of a plan during the lifecycle to the position onward.
first year of operation and thus helps to evaluate the total costs In order to discover site, randomly selected individuals
and benefits during the economic lifecycle of the equipments move toward the targets of donor [Xrandom-shuffling(i)]. The
[32]. CPV1 is calculated as follows: extent to which the change occurs is controlled by a scale
1  ðPV1 ÞEL value. The Stopoversite position is given as
CPV1 ¼ ð4Þ
1  PV1 Stopoversite ¼ Superorganism þ Scale  ðdonor
where  SuperorganismÞ ð11Þ
ð1 þ Iinf Þð1 þ LGÞ The members to participate in search process are selected
PV1 ¼ ; ð5Þ
ð1 þ Iint Þ by random process of specific structure. If any element goes
beyond the limits of habitat, the element is randomly deferred
EL is the economic lifetime of the equipments, Iinf is the infla- to another position. Software code of the algorithm of DS
tion rate, Iint is the interest rate and LG is the load growth rate. algorithm can be found in [34].
The objective function to reduce RCS cost is
J2 ¼ nRCS  ðCi þ Cm  CPV2 Þ ð6Þ 4.1. Sequential steps of DS algorithm

where nRCS denotes the total number of RCS present in the sys-
The stepwise DS algorithm is mentioned as follows:
tem. Ci stands for the installation cost and Cm stands for the
maintenance cost of each RCS. CPV2 is the cumulative present
value (CPV) of maintenance cost of RCS cost which is
expressed as follows: Required:
N: size of Superorganism, where i = {1, 2, 3, . . ., N}
1  ðPV2 ÞEL D: The dimension of the problem
CPV2 ¼ ð7Þ
1  PV2 G: No. of maximum generation.
(1) Initialize Superorganism, where Superorganism is termed as
where Artificial Organismi
(2) For i = 1: N
ð1 þ Iinf Þ
PV2 ¼ ð8Þ (3) yi = Evaluate objective function corresponding to Artificial
ð1 þ Iint Þ Organismi
(4) end for
Therefore, the overall objective function to represent multi-
(5) For cycle 1:G, do
objective formulation is expressed as
(6) Random shuffling of Superorganismi i.e
J ¼ w  J1 þ ð1  wÞ  J2 ð9Þ Donor = SuperorganismRandom_shuffling
(7) Calculate Scale = randg[2.rand1].(rand2  rand3)
where w is a weightage value assigned to a single objective, in
order to find the Pareto optimal solution. Here, both the num-
Allocation of remote control switches 407

(8) Stopoversite ¼ Superorganism þ Scale

ðdonor  SuperorganismÞ
(9) p1 = c1  rand4 and p2 = c2  rand5, where c1 and c2 are
for i = 1: N
control parameters EL
1Þ
(10) If rand6 < rand7, then Evaluate CPV1 ¼ 1ðPV
1PV1
(11) If rand8 < p1 Evaluate PV1 ¼ ð1þIð1þI
inf Þð1þLGÞ
int Þ
(12) r = rand (N, D) for j = 1: n
(13) for Counter1: N, do For each Artificial organismi identify the no. of network
(14) r (Counter1,:) = r (Counter1,:) < rand9
segments (nj0 ) having repair time (rep) & no. of network segments
(15) end for
(nj1 ) having restoration time (res) based on the logic mentioned in
(16) else
subsection 2.1
(17) R = ones (N, D)
for i0 = 1: nj0
(18) for Counter2 = 1: N, do
Urepj ¼ Urepj þ ki0 repi0
(19) r (Counter2, randi(D)) = r (Counter2, randi(D))
< rand10 end
(20) end for for i1 = 1: nj1
(21) end if Uresj ¼ Uresj þ ki1 resi1
(22) else end
(23) r = ones (N, D) Uj = Urepj þ Uresj
(24) for Counter3 = 1:N, do 1) EENSi = EENSi + Uj  Lj
(25) d = randi (D, 1, [p2.rand.D]) end
(26) for Counter4 = 1: size (d), do Evaluate J1i = ((8  EENSi  1  (340/8760))
(27) r (Counter3, d(Counter4)) = 0 + (8  EENSi  0.4  (5500/8760)) + (8  EENSi  0.3 
(28) end for (2920/8760)))  CPV1
(29) end for end
(30) end if for i = 1: N
(31) IndividualsI,J = rI,J > 0 | I 2 [1 N], J 2 [1 D] J2i = 0;
(32) StopoverSiteðindividualI;J Þ ¼ SuperorganismðindividualI;J Þ ð1þIinf Þ
Evaluate PV2 ¼ ð1þI int Þ
(33) If StopoverSite crosses the limits, set 2Þ
EL

StopoverSite ¼ randðupj  lowj Þ þ lowj . Evaluate CPV2 ¼ 1ðPV

1PV2

(34) Evaluate StopoverSite, yStopoverSite [nRows nColumn] = size(find(Artificial organismi ==1));

(35) Modify ySuperorganismi by yStopoverSitei if Evaluate J2i = nColumn  18,000 + nColumn  2000  CPV2
end
yStopoverSitei < ySuperorganismi .
(36) Artificial

StopoverSitei if yStopoverSitei < ySuperorganismi
organismi =
Artificial organismi ; else For the bi-objective function, evaluate the overall objective
function for each set of initially generated Artificial organismi
using (9).

Step 5: Perform steps 3 to 30 of Section 4.1.

4.2. Steps of DS algorithm as applied to the present problem
Step 6: If any variable j of StopoverSite (generated using
(11)) crosses the respective limits, set the value of that vari-
The sequential steps of the DS algorithm applied to find opti- able as StopoverSitei;j ¼ randðupj  lowj Þ þ lowj .
mum number and location of RCS of a radial distribution sys-
tem are as follows.
The upper limit of variable j is 1 and the lower limit is 0.
Required: N: size of Superorganism, where i = {1, 2, 3, . . .,
These are discrete variables which can have values of either 0
N}, D: The dimension of the problem, G: No. of maximum
or 1. It is assumed that a maximum of one RCS can be
generation.
installed at each distribution segment. The total number of
variables is equal to the total number of distribution segment
Step 1: Read input data: Lj, repi, resi, ki , inflation rate (Iinf),
in a system. A zero represents no RCS in a particular segment
interest rate (Iint), load growth rate (LG), economic lifetime
and 1 represents presence of RCS in that segment.
of equipments (EL), total number of load points (n), the DS
algorithm parameters like control parameters c1 and c2 etc.
Step 7: Evaluate objective function of StopoverSite i.e.
Step 2: Initialize the value of w = 0, where w is the weigh-
yStopoverSite as given by Eqs. (3) or (6) or (9), as performed
tage factor.
in Step 4.
Step 3: Initialize Superorganism by generating Artificial
Step 8: Modify y Superorganismi by y StopoverSitei if objective function
organism which contains either 0 or 1. 0 represents no
of StopoverSitei is less than cost function of Superorganismi.
RCS and 1 represents presence of RCS in a distribution seg-
Step 9: 
ment. Generate Artificial organism using (10), where upj and
StopoverSitei if y StopoverSitei < y Superorganismi
lowj are 1 and 0 respectively. Artificial organismi = Artificial organismi ; else
Step 4: Evaluate the objective function J1i and J2i using (3) Step 10: In case of the bi-objective problem, increment the
and (6) respectively for each initially generated Artificial value of w in steps of 0.1 and repeat steps 3–8. Repeat the
organismi set as per following steps: process until the value of w reaches 1.
408 S. Ray et al.

Step 11: Best compromise solution- the algorithm described

Table 2 Peak loads of different load points (Test case I).
above generates the non-dominated set of solutions known
as the Pareto-optimal solutions. The decision-maker (power Load point Load (kW)
system operator) may have imprecise or fuzzy goals for 1 0
each objective function. To aid the operator in selecting 2 1000
an operating point from the obtained set of Pareto- 3 700
optimal solutions, the fuzzy logic theory is applied to each 4 400
objective function to obtain a fuzzy membership function as 5 500
given below [35]: 6 300
7 200
8 8 150
>
> 1 if Ji 6 Jmin
< max i
Ji Ji
lJi ¼ if Jmin < Ji < Jmax ð12Þ
> Jmax
>
Jmin i i
:
i i

0 if Ji P Jmax
i
Table 3 Different load levels.
The best non-dominated objective function can be found
Load level Duration (h) % of peak load
when (13) is a maximum, where the normalized sum of objec-
tive function values for all objectives is highest: 1 340 1
PQ k 2 5500 0.4
lJ 3 2920 0.5
lk ¼ PM i¼1 PQ i k ð13Þ
k¼1 i¼1 lJi

where Q denotes the total number of individual objective func-

segment 2, and so on. The failure rate and repair time of
tion in (9), and M is the number of non-dominated solutions.
the segments are considered as in Table 1. The restoration
After completion of the process, the best solution of the prob-
time has been considered to be 5 min. Table 2 gives the
lem is obtained.
peak loads of different load-points as considered in the
present work. Three different load levels are considered
5. Results and discussions
as in Table 3. The per kilowatt cost of EENS is considered
to be 5$ for all the load levels; and the installation and
The DS algorithm has been implemented on two test systems maintenance cost of one RCS has been considered to be
and its performance has been compared with PSO for verifying 18,000$ and 2000$ respectively. The interest rate, inflation
its feasibility for solving optimization problems of distribution rate and load growth rate have been considered as 0.05,
system reliability. The algorithms have been coded in 0.08 and 0.05 respectively.
MATLAB software (version 7.10.0) on a processor of specifi- 2) Test Case II: A 33-bus radial test system as shown in
cation Intel (R) Core (TM) i7-2600 CPU 3.40 GHz with Fig. 2 has been considered, where there is a circuit
2 GB RAM. breaker (CB) at the beginning of the network and fuses
at the starting point of each lateral branch. The test sys-
5.1. Description of the test system tem consists of 32 distribution segments and 33 loads
points. Like the earlier case, the RCSs are considered
1) Test Case I: An 8-bus radial test system as shown in to be allocated at the beginning of any distribution seg-
Fig. 1 has been considered. The system consists of seven ment, except first distribution segment of branches as
feeder segments and contains a circuit breaker at the there is a fuse. Numbering of distribution segment is
beginning of the network. The RCSs are considered to done as in the previous test case. The loads, failure rate,
be allocated at the beginning of any distribution seg- repair time and restoration time of distribution segments
ment. Numbering of distribution segments is done in have been considered same as presented in [32]. Here
the following manner: distribution segment preceding also, three different load levels are considered as in
load point 2 is numbered as distribution segment 1; Table 3.The per kilowatt cost of EENS is considered
preceding load point 3 is numbered as distribution to be 8$ and the installation and maintenance cost of
one RCS has been considered to be 18,000$ and 2000$
respectively. The interest rate, inflation rate and load
Table 1 Failure rate and repair time of different segments growth rate have been considered same as in test case I.
(Test case I).
Distribution segment Failure rate, k (f/ Repair time, rep
no. yr.) (h) 5.2. Comparative study
1 0.4 10
5.2.1. Solution quality
2 0.2 9
3 0.3 12 Tables 4 and 5 present the best objective function values and
4 0.5 20 the corresponding number and location of RCS as obtained
5 0.2 15 by DS algorithm and PSO for the 8 bus distribution system
6 0.1 8 for minimizing EENS cost and RCS cost respectively. It is
7 0.1 12 obvious that minimizing RCS cost leads to no RCS installa-
tion in the network. Therefore, no improvement is possible
 Allocation of remote control switches 409

Figure 2 Thirty-three bus network (Test case II).

Table 4 Best results for EENS cost minimization obtained using different methods (Test case I).
Method No. of RCS position (distribution Objective Corresponding RCS Total cost ($) (EENS cost Simulation
RCS segment no.) function cost ($) + RCS cost) time (s)
EENS cost
($)
DS 6 2–7 1,414,165 328,860 1,743,025 1.0402
PSO 6 2–7 1,414,165 328,860 1,743,025 1.3245
Bold signifies the best results in terms of quality of solution and computational efficiency.

Table 5 Best results for RCS cost minimization obtained using different methods (Test case I).
Method No. of RCS position (distribution Objective Corresponding EENS Total Cost ($) (EENS cost Simulation
RCS segment no.) function cost ($) + RCS cost) time (s)
RCS cost ($)
DS 0 – 0 4,198,200 4,198,200 1.0013
PSO 0 – 0 4,198,200 4,198,200 1.2872
Bold signifies the best results in terms of quality of solution and computational efficiency.

6 4
x 10 x 10
1.54 6

DS DS
PSO PSO
1.52 5

1.5
4
RCS Cost ($)
EENS Cost ($)

1.48
3
1.46
2
1.44

1
1.42

1.4 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
No. of Iteration No. of Iteration

Figure 3 Convergence characteristic for EENS cost minimiza- Figure 4 Convergence characteristic for RCS cost minimization
tion obtained by DS and PSO for Test case I. obtained by DS and PSO for Test case I.
410 S. Ray et al.

Table 6 Best results for multi-objective problem obtained using different methods (Test case I).
Method Variables Objective function Total ($) (EENS cost + RCS cost)
No. of RCS RCS position (distribution segment no.) EENS cost ($) RCS cost ($)
DS 2 2, 4 1,820,000 109,620 1,929,620
PSO 2 3, 4 1,980,200 109,620 2,089,820
Bold signifies the best results in terms of quality of solution and computational efficiency.

Table 7 Comparison of solution among different methods after 50 trials (Multi-objective problem) (Test case I).
Methods Minimum ($) Maximum ($) Average ($) Simulation time (s) No. of hits to optimum solution
DS 1,929,620 1,929,620 1,929,620 10.0924 50
PSO 2,089,820 2,341,640 2,104,929 13.1625 47
Bold signifies the best results in terms of quality of solution and computational efficiency.

6
5 x 10
x 10 6.72
4
DS
DS PSO
3.5 PSO 6.7

3 6.68
EENS Cost ($)
RCS Cost ($)

2.5
6.66

2
X=1980200 6.64
Y=109620
1.5
X=1820000
6.62
1 Y=109620

6.6
0.5

6.58
0 0 10 20 30 40 50 60 70 80 90 100
1 1.5 2 2.5 3 3.5 4 4.5
6 No. of Iteration
EENS Cost ($) x 10

Figure 6 Convergence characteristic for EENS cost minimiza-

Figure 5 Pareto optimal front for EENS cost and RCS cost
tion obtained by DS and PSO for Test case II.
obtained by DS and PSO for Test case-I.

from reliability point of view. Hence, for minimizing RCS cost, cost and RCS cost are obtained by the multi-objective formu-
same result is obtained by DS and PSO. Similarly, EENS cost lation. The results show that EENS cost and total cost are
minimization will lead to installation of RCS in almost all the quite less with DS than that obtained by PSO keeping the
segments. The solutions obtained by DS and PSO are same RCS cost same. A comparison of solutions for the multi-
which is quite logical. Figs. 3 and 4 show the respective conver- objective formulation obtained using DS algorithm and PSO
gence characteristics of minimizing EENS cost and minimizing is drawn in Table 7. The Pareto optimal front obtained by
RCS cost for the eight bus network (test case I). The results DS algorithm and PSO for simultaneous minimization of
show that minimum output is same in both the cases, but EENS cost and RCS cost for test case I is shown in Fig. 5.
the solutions converge to final results comparatively faster Table 8 presents the best objective function values and the
using DS compared to PSO, which is evident from Figs. 3 corresponding number and location of RCS as obtained by DS
and 4. Table 6 presents the best objective function values algorithm and PSO for the 33 bus distribution system (test case
and the corresponding number and location of RCS as II) for minimizing EENS cost and Fig. 6 shows the conver-
obtained by DS algorithm and PSO for the multi-objective for- gence characteristics. The results are same with both DS and
mulation of test case I. Best non-dominated values of EENS PSO, and RCS is installed in almost all the segments, except

Table 8 Best results for EENS cost minimization obtained using different methods (Test case II).
Method No. of RCS position (distribution segment no.) Objective function Corresponding Total Cost ($) (EENS Simulation
RCS RCS cost ($) cost + RCS cost) time (s)
EENS cost ($)
DS 27 2–11, 13–17, 19–21, 23–24, 26–32 6,591,615 1,479,900 8,071,515 3.7720
PSO 27 2–11, 13–17, 19–21, 23–24, 26–32 6,591,615 1,479,900 8,071,515 4.9021
Bold signifies the best results in terms of quality of solution and computational efficiency.
Allocation of remote control switches 411

Table 9 Best results for RCS cost minimization obtained using different methods (Test case II).
Method No. of RCS position (distribution segment no.) Objective function Corresponding Total Cost ($) (EENS Simulation
RCS EENS cost ($) cost + RCS cost) time (s)
RCS cost ($)
DS 0 – 0 8,271,100 8,271,100 2.0054
PSO 0 – 0 8,271,100 8,271,100 3.0352
Bold signifies the best results in terms of quality of solution and computational efficiency.

5
x 10
7 Table 11. As the multi-objective function provides a compro-
DS
PSO
mised solution, it may be proved to be more realistic than
6
the single objective formulation of minimizing either EENS
5
cost or RCS cost. The Pareto optimal front for simultaneous
minimization of EENS cost and RCS cost, in case of DS algo-
RCS Cost ($)

4 rithm and PSO is shown in Fig. 8. The Pareto optimal front

presents a smoother characteristic in case of DS algorithm
3 than in PSO.
2
5.2.2. Computational efficiency
1 Time taken by DS algorithm to reach the minimum solution
for EENS cost minimization is 1.0402 s for test case I and
0
0 10 20 30 40 50 60 70 80 90 100 3.7720 s for test case II. Whereas, PSO takes 1.3245 s and
No. of Iteration 4.9021 s for EENS minimization of test case I and test case
II respectively. For RCS cost minimization, DS algorithm
Figure 7 Convergence characteristic for RCS cost minimization takes 1.0013 s and PSO takes 1.2872 s for test case I. For test
obtained by DS and PSO for Test case II. case II, DS algorithm takes 2.0054 s and PSO takes 3.0352 s
for RCs cost minimization. PSO takes Time taken by DS algo-
those where fuses are installed. Table 9 shows the results of rithm to reach the best solution for the multi-objective formu-
minimizing RCS cost by DS and PSO, and Fig. 7 shows the lation of test case I is 10.0924 s and that for test case II is
corresponding convergence characteristic. Here also, due to 34.7621 s. For same objective formulation, time required by
obvious reasons as discussed in the earlier test case, the results PSO to reach best solution is 13.1625 s and 41.3287 s for test
obtained by DS and PSO are same. The convergence charac- case I and test case II respectively. These are quite prominent
teristics as shown in Figs. 6 and 7 reveal a faster convergence from Tables 4, 5, 7–9 and 11. The results show superior com-
of DS algorithm as compared to PSO. Table 10 presents the putational efficiency of DS algorithm.
best objective function values and the corresponding number
and location of RCS as obtained by DS algorithm and PSO 5.2.3. Robustness
for the multi-objective formulation. The results show that Performance of any heuristic algorithm cannot be judged by
though EENS cost obtained by DS algorithm is somewhat the results of a single run. Normally their performance is
more, the RCS cost is quite less compared to that obtained judged after running the programs of those algorithms for sev-
by PSO. A comparison of solutions for the multi-objective for- eral numbers of trials. Many numbers of trials should be made
mulation obtained using DS algorithm and PSO is presented in to obtain a useful conclusion about the performance of the

Table 10 Best results for multi-objective problem obtained using different methods (Test case II).
Method Variables Objective function Total ($) (EENS cost + RCS cost)
No. of RCS RCS position (distribution segment no.) EENS cost ($) RCS cost ($)
DS 7 2,6,20,24,27,30,32 6,948,400 383,680 7,332,080
PSO 10 2,3,7,17,20,24,27,29,31,32 6,833,100 548,110 7,381,210
Bold signifies the best results in terms of quality of solution and computational efficiency.

Table 11 Comparison of solution among different methods after 50 trials (Multi-objective problem) (Test case II).
Methods Minimum ($) Maximum ($) Average ($) Simulation time (s) No. of hits to optimum solution
DS 7,332,080 7,332,080 7,332,080 34.7621 50
PSO 7,381,210 7,383,813 7,381,574 41.3287 43
Bold signifies the best results in terms of quality of solution and computational efficiency.
412 S. Ray et al.

5
x 10 algorithm. As DS algorithm is a stochastic optimization tech-
18
DS
nique, randomness is obvious and many trials are required to
16 PSO obtain the optimum result. In this study, 50 trial runs were car-
14 ried out to obtain each result in order to take into considera-
tion the stochastic nature. An algorithm is said to be robust,
12
RCS Cost ($)

if it gives consistent result during these trial runs. Table 7

10 shows a comparison of the solution of DS algorithm and
X=6833100 PSO for the multi-objective formulation for test case I, where
8 Y=548110
former one proves to be better in terms of robustness. Table 11
6
shows the comparison of the solution for test case II. Here
4 also, DS algorithm outperforms PSO in terms of robustness.
2
X=6948400
Y=383680
Both Tables 7 and 11 reveal that out of 50 numbers of trials,
the number of hits to reach the minimum solution is 100%
0
6.4 6.6 6.8 7 7.2 7.4 7.6 7.8 8 8.2 8.4
6
using DS algorithm, which signifies robustness of the
EENS Cost ($) x 10 algorithm.
Therefore, the above results establish the enhanced ability
Figure 8 Pareto optimal front for EENS cost and RCS cost
of DS algorithm to achieve superior quality solutions, in a
obtained by DS and PSO for Test case II.
computational efficient and robust way for solving reliability
problem.

Table 12 Influence of DS parameters on multi-objective problem value (After 50 trials) (Test case II).
Superorganism size c2 c1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
20 0.1 7,386,230 7,386,010 7,385,010 7,385,310 7,386,230 7,386,432 7,386,784 7,386,880 7,386,910
0.2 7,385,830 7,385,710 7,384,620 7,384,220 7,385,010 7,386,002 7,386,184 7,386,620 7,386,790
0.3 7,384,030 7,382,520 7,381,210 7,382,100 7,383,420 7,385,010 7,385,172 7,386,130 7,386,290
0.4 7,385,810 7,385,680 7,384,590 7,384,320 7,385,640 7,385,920 7,386,032 7,386,560 7,386,384
0.5 7,386,010 7,386,230 7,385,340 7,385,720 7,386,124 7,386,280 7,386,124 7,387,430 7,389,512
0.6 7,386,318 7,386,250 7,385,172 7,385,800 7,385,382 7,386,390 7,386,594 7,388,620 7,390,670
0.7 7,386,760 7,386,552 7,386,020 7,385,924 7,386,090 7,386,512 7,386,580 7,388,732 7,394,824
0.8 7,387,020 7,387,520 7,386,980 7,389,928 7,390,040 7,394,512 7,397,580 7,397,732 7,399,610
0.9 7,387,202 7,387,670 7,387,080 7,389,728 7,391,040 7,392,214 7,396,580 7,398,054 7,399,400
50 0.1 7,354,210 7,354,024 7,346,022 7,343,332 7,343,990 7,344,172 7,352,110 7,353,032 7,355,814
0.2 7,353,040 7,352,708 7,345,062 7,342,042 7,343,750 7,347,390 7,353,108 7,353,820 7,356,124
0.3 7,349,172 7,337,120 7,332,080 7,336,930 7,340,124 7,344,324 7,348,324 7,350,178 7,353,142
0.4 7,350,036 7,338,420 7,334,390 7,336,982 7,339,570 7,345,880 7,350,270 7,350,928 7,353,784
0.5 7,351,036 7,347,330 7,340,280 7,339,320 7,347,520 7,347,920 7,353,572 7,355,010 7,359,262
0.6 7,354,432 7,350,738 7,348,176 7,341,572 7,345,390 7,348,378 7,356,732 7,356,430 7,360,480
0.7 7,358,920 7,356,180 7,355,278 7,351,048 7,354,028 7,360,310 7,362,490 7,368,362 7,370,120
0.7 7,360,120 7,360,002 7,359,172 7,363,520 7,365,124 7,366,234 7,370,020 7,378,314 7,381,210
0.9 7,360,040 7,359,890 7,359,102 7,363,012 7,364,124 7,367,318 7,369,320 7,377,210 7,380,320
100 0.1 735,410 7,354,108 7,346,042 7,343,298 7,343,986 7,344,190 7,352,120 7,353,048 7,355,818
0.2 7,353,036 7,352,708 7,345,062 7,342,088 7,343,764 7,347,390 7,353,110 7,353,862 7,356,124
0.3 7,349,180 7,337,120 7,332,080 7,336,930 7,340,132 7,344,280 7,348,500 7,350,202 7,353,150
0.4 7,350,038 7,338,470 7,334,400 7,336,872 7,339,480 7,345,630 7,350,180 7,350,788 7,353,792
0.5 7,351,036 7,347,342 7,340,120 7,339,320 7,347,520 7,347,814 7,353,580 7,355,000 7,359,270
0.6 7,354,428 7,350,738 7,348,172 7,341,580 7,345,398 7,348,370 7,356,740 7,356,420 7,360,474
0.7 7,358,916 7,356,180 7,355,278 7,351,040 7,354,036 7,360,300 7,362,540 7,368,354 7,370,120
0.8 7,360,120 7,360,002 7,359,158 7,363,500 7,365,130 7,366,232 7,370,002 7,378,310 7,381,210
0.9 7,360,040 7,359,896 7,359,102 7,363,010 7,364,122 7,367,322 7,369,300 7,377,210 7,380,320
200 0.1 7,354,210 7,354,022 7,346,020 7,343,336 7,343,990 7,344,180 7,352,100 7,353,032 7,355,812
0.2 7,353,040 7,352,708 7,345,068 7,342,042 7,343,756 7,347,394 7,353,110 7,353,820 7,356,124
0.3 7,349,174 7,337,120 7,332,080 7,336,928 7,340,122 7,344,324 7,348,326 7,350,178 7,353,140
0.4 7,350,034 7,338,422 7,334,390 7,336,980 7,339,578 7,345,878 7,350,270 7,350,930 7,353,786
0.5 7,351,038 7,347,328 7,340,280 7,339,320 7,347,522 7,347,930 7,353,570 7,355,010 7,359,260
0.6 7,354,430 7,350,742 7,348,174 7,341,572 7,345,396 7,348,374 7,356,732 7,356,428 7,360,482
0.7 7,358,920 7,356,178 7,355,276 7,351,048 7,354,030 7,360,308 7,362,488 7,368,366 7,370,108
0.8 7,360,120 7,360,000 7,359,174 7,363,518 7,365,124 7,366,232 7,370,024 7,378,308 7,381,210
0.9 7,360,040 7,359,840 7,359,090 7,363,020 7,364,176 7,367,310 7,369,300 7,377,208 7,380,316
Bold signifies the best results in terms of quality of solution and computational efficiency.
Allocation of remote control switches 413

Table 13 Effect of superorganism size on multi-objective problem (Test case II).

Superorganism size No. of hits to minimum value Simulation time (s) Minimum ($) Maximum ($) Average ($)
20 47 29.1002 7,381,210 7,399,610 7,382,314
50 50 34.7621 7,332,080 7,332,080 7,332,080
100 48 38.2804 7,332,080 7,337,120 7,332,281
200 50 41.5520 7,332,080 7,332,080 7,332,080
Bold signifies the best results in terms of quality of solution and computational efficiency.

5.3. Tuning of parameters for DS algorithm [4] Miu KN, Chiang HD, Yuan B, Darling G. Fast service
restoration for large-scale distribution systems with priority
customers and constraints. In: 20th int. conf. power industry
To get optimum solution using DS algorithm, it is necessary to computer applications, 11–16 May, Columbus, OH; 1997. p. 3–9.
get proper values of parameters c1 and c2. Moreover, value of [5] Billinton R, Jonavithula S. Optimal switching device placement in
the cost function may vary with the Superorganism size also. radial distribution systems. IEEE Trans. Power Delivery 1996;11
For different values of these parameters, minimum value of (3):1646–51.
cost function is evaluated for the multi-objective function of [6] Celli G, Pilo F. Optimal sectionalizing switches allocation in
test case II. For a single value of one parameter, the other distribution networks. IEEE Trans. Power Delivery 1999;14
parameters have been varied for their all possible combina- (3):1167–72.
tions. The results of the tuning procedure are presented in [7] Cox PW. Self-healing networks: performance improvement by
automated switching algorithm. In: Proceedings of the 20th
Table 12.
international conference on electricity distribution (CIRED 2009);
Too large or small value of Superorganism size may not be
2009. p. 1–8.
capable to get the minimum value of cost. For each Superor- [8] Wagner T. Impact of remote controlled switches on distribution
ganism size of 20, 50, 100 and 200, 50 trials have been run. grid recovering process. In: Royal Institute of Technology (KTH),
After a rigorous tuning procedure, it has been found that the school of electrical engineeriung, Master Degree Project; 2010.
best value is obtained with c1 = 0.3 and c2 = 0.3 with a Super- [9] Carvalho PMS, Ferreira LAFM, da Silva AJC. A decomposition
organism size 50. For Superorganism size more than 50, there is approach to optimal remote controlled switch allocation in
no improvement in the result. Moreover, beyond Superorgan- distribution systems. IEEE Trans. Power Delivery 2005;20
ism size of 50, simulation time also increases. The variation of (2):1031–6.
results obtained by DS algorithm with variation in Superor- [10] Asr FT, Kazemi A. Determining optimum location of automated
switches in distribution network, EPEC2008. In: Proceedings of
ganism size for test case II is shown in Table 13.
the electric power conference, Canada; 6–7 October, 2008. p. 1–6.
[11] Bernardona DP, Sperandioa M, Garciaa VJ, Russi J, Canha LN,
6. Conclusion Abaide AR, Daza EFB. Methodology for allocation of RCS in
distribution networks based on a fuzzy multi-criteria decision
In this paper, DS algorithm has been successfully implemented making algorithm. Electr Power Syst Res 2011;81(2):414–20.
to find optimum number and location of RCS in a radial dis- [12] Bernardon DP, Sperandio M, Garcia VJ, Canha LN, Abaide AR,
Boeck Daza EF. AHP decision-making algorithm to allocate
tribution feeder. The performance of DS algorithm has been
remotely controlled switches in distribution networks. IEEE
compared with that of PSO algorithm. Both single objective Trans. Power Delivery 2011;26(3):1884–92.
and multi-objective formulations are considered and multi- [13] Wesz da Silva LG, Pereira RAF, Mantovani JRF. Optimized
objective formulation proves to provide more realistic solution allocation of sectionalizing switches and protection devices in
set, by compromising reliability improvement with the cost distribution networks by using a reactive taboo search algorithm.
incurred. Analyses of all the simulation results reveal that In: 18th international conference on electricity distribution, Turin;
the performance of DS algorithm in all respect is better in 6–9 June 2005. p. 1–5.
comparison with the PSO. Thus, DS algorithm may be consid- [14] da Silva LGW, Pereira RAF, Mantovani JRS. Optimized alloca-
ered as an efficient tool to solve multi-objective reliability opti- tion of sectionalizing switches and control and protection devices
mization problems. In future, DS algorithm may be tried to for reliability indices improvement in distribution systems. In:
2004 IEEE lPES Transmission & Distribution Conference &
solve much more complex reliability optimization problem,
Exposition: Latin America; 2004. p. 51–56.
considering optimal placement of both RCSs and distributed [15] Abiri-Jahromi A, Fotuhi-Firuzabad M, Parvania M, Mosleh M.
generators together. Optimized sectionalizing switch placement strategy in distribution
systems. IEEE Trans. Power Delivery 2012;27(1):362–70.
References [16] Viotto Romero ME, Wesz da Silva LG, Mantovani JRS. Optimal
switch allocation for automatic load transfer in distribution
[1] Billinton R, Allan RN. Reliability evaluation of power system. substations. In: IEEE Trondheim PowerTech, 19–23 Jun., Trond-
2nd ed. New York, USA: Springer; 1996. heim; 2011. p. 1–7.
[2] Bouhouras AS, Labridis DP, Bakirtzis AG. Cost/worth assess- [17] Bernardona DP, Sperandioa M, Garciaa VJ, Russia J, Canhab
ment of reliability improvement in distribution networks by LN, Abaideb AR, Dazac EFB. Methodology for allocation of
means of artificial intelligence. Int J Electr Power Energy Syst RCS in distribution networks based on a fuzzy multi-criteria
2010;32(5):530–8. decision making algorithm. Electr Power Syst Res 2011;81
[3] Haifenga G, Asgarpoor Sohrab. Parallel Monte Carlo simulation (2):414–20.
for reliability and cost evaluation of equipment and systems. [18] Benavides AJ, Ritt M, Buriol LS, Franca PM. An iterated sample
Electr Power Syst Res 2011;81(2):347–56. construction with path relinking method: application to switch
414 S. Ray et al.

allocation in electrical distribution networks. Comput. Oper. Res. [34] Civicioglu P. <http://www.pinarcivicioglu.com/ds.htmlS>; 2011
2013;40:24–32. [accessed 02.10.11].
[19] Zheng H, Cheng Y, Gou B, Frank D, Bern A, Muston WE. [35] Granelli GP, Montagna M, Pasini GL, Maranino P. Emission
Impact of automatic switches on power distribution system constrained dynamic dispatch. Electr Power Syst Res
reliability. Electr Power Syst Res 2012;83:51–7. 1992;24:55–64.
[20] Esmaeilian HR, Fadaeinedjad R. Distribution system efficiency
improvement using network reconfiguration and capacitor allo-
cation. Int J Electr Power Energy Syst 2015;64:457–68. Saheli Ray did her B.E. in Electrical Engg.
[21] Tippachon W, Rerkpreedapong D. Multiobjective optimal place- from National Institute of Technology,
ment of switches and protective devices in electric power distri- Agartala, India (Formerly Tripura Engineer-
bution systems using ant colony optimization. Electr Power Syst ing College) in 2006, M.Tech in Power and
Res 2009;79:1171–8. Energy Systems from National Institute of
[22] Pombo AV, Murta-Pina J, Pires VF. Multiobjective planning of Technology, Silchar, India, in 2009. She is
distribution networks incorporating switches and protective currently pursuing Ph.D. in the Department
devices using a memetic optimization. Reliab Eng Syst Saf of Electrical Engineering at National Institute
2015;136:101–8. of Technology, Agartala, India. She is also
[23] Golestani S, Tadayon M. Optimal switch placement in distribu- working as Assistant Professor in the
tion power system using linear fragmented particle swarm Department of Electrical Engineering of NIT
optimization algorithm preprocessed by GA. In: 8th international Agartala, India. Her areas of interest include Power system reliability,
conference on the European energy market (EEM), 25–27 May, Power and Energy systems.
Zagreb, Croatia; 2011. p. 537–42.
[24] Assis LS, González JFV, Usberti FL, Lyra C, Cavellucc C, Zuben
Subhadeep Bhattacharjee did his BE (Electri-
FJV. Switch allocation problems in power distribution systems.
cal) from Tripura Engineering College (Pre-
IEEE Trans. Power Syst. 2015;30(1):246–53.
sently National Institute of Technology,
[25] Amanulla B, Chakrabarti S, Sing SN. Reconfiguration of power
Agartala, India) in 2001, Master of Electrical
distribution systems considering reliability and power loss. IEEE
Engineering (MEE) in 2003 and Ph.D.
Trans Power Delivery 2012;27(2):918–26.
(Engineering) in 2009 from Jadavpur Univer-
[26] Zou K, Keerthipala WWL, Perera S. SAIDI minimization of a
sity, Kolkata, India. He is currently working
remote distribution feeder. In: Proceedings of the 2007 Aus-
as Assistant Professor in Electrical Engineer-
tralasian universities power engineering conference (AUPEC
ing Department of National Institute of
2007). Perth, Australia; 9–12 December, 2007. p. 342–46.
Technology, Agartala, India. He is a Life
[27] Brown HE, Suryanarayanan S, Natarajan SA, Rajopadhye S.
Member of Indian society for Technical
Improving reliability of islanded distribution systems with dis-
Education (ISTE), Solar Energy Society of India (SESI), Member of
tributed renewable energy resources. IEEE Trans Smart Grid
Institution of Electronics and Telecommunication Engineers (India)
2012;3(4):2028–38.
(IETE), and Chartered Engineer (C.Eng.) from The Institution of
[28] Vitorino RM, Jorgeb HM, Neves LP. Loss and reliability
Engineers (India) (IEI). His research field is Energy and Renewable
optimization for power distribution system operation. Electr
Power, Distributed Power Generation and Hybrid Power System.
Power Syst Res 2013;96:177–84.
[29] Zhang P, Li W, Wang S. Reliability-oriented distribution network
reconfiguration considering uncertainties of data by interval Aniruddha Bhattacharya did his B.Sc. Engg. in
analysis. Int J Electr Power Energy Syst 2012;34(1):138–44. Electrical Engg. from Regional Institute of
[30] Pfitscher LL, Bernardon DP, Canha LN, Montagner VF, Garcia Technology, Jamshedpur, India, in 2000, and
VJ, Abaide AR. Intelligent system for automatic reconfiguration M.E.E. & Ph.D in Electrical power system from
of distribution network in real time. Electr Power Syst Res Jadavpur University, Kolkata, India, in 2008
2013;97:84–92. and 2011 respectively. His working experience
[31] Kavousi-Fard A, Akbari-Zadeh M-R. Reliability enhancement includes Siemens Metering Limited, India;
using optimal distribution feeder reconfiguration. Neurocomput- Jindal Steel & Power Limited, Raigarh, India;
ing 2013;106:1–11. Bankura Unnyani Institute of Engineering,
[32] Raoofat M. Simultaneous allocation of DGs and remote control- Bankura, India; and Dr. B. C. Roy Engineering
lable switches in distribution networks considering multilevel load College, Durgapur, India. He is currently an
model. Int J Electr Power Energy Syst 2011;33(8):1429–36. Assistant Professor in Electrical Engineering Department, National
[33] Civicioglu P. Transforming geocentric Cartesian coordinates to Institute of Technology, Agartala, India. His areas of interest include
geodetic coordinates by using differential search algorithm. power system load flow, optimal power flow, economic load dispatch,
Comput Geosci 2012;46:229–47. and soft computing applications to power system problems.