Arabian Journal for Science and Engineering (2021) 46:1029–1051

https://doi.org/10.1007/s13369-020-04808-9

RESEARCH ARTICLE-ELECTRICAL ENGINEERING

Enhanced Sine–Cosine Algorithm for Optimal Planning of Distribution

Network by Incorporating Network Reconfiguration and Distributed
Generation
Usharani Raut1 · Sivkumar Mishra2

Received: 14 February 2020 / Accepted: 17 July 2020 / Published online: 12 August 2020
© King Fahd University of Petroleum & Minerals 2020

Abstract
One of the important directions of the current research trend in distribution network planning in the prevailing smart grid
scenario, is to explore various possibilities to enhance the performance of these networks without expanding the existing
infrastructure. This paper proposes an enhanced sine–cosine algorithm (ESCA) to obtain an optimally planned system by
simultaneous incorporation of network reconfiguration (NR) and DG allocation. In the proposed algorithm, the SCA is
enhanced with neighborhood search strategy and self-adapting levy mutation strategy to ensure proper balance between
exploration and exploitation during different reconfiguration phases. A multi-objective function is formulated considering
the reduction of total real power loss and annual operation costs with suitable weights without violating the system operating
constraints. The proposed algorithm is successfully experimented on 33- and 69-bus distribution system with four distinct
scenarios of NR and DG allocation, and its performance assessment is based on technical (total system active power loss
index, overall voltage stability index and voltage profile improvement index), economic (total system operation cost index)
and reliability (expected energy not supplied index) indices. As the computation of reliability index adds complexity to the
problem, a graph theory-based algorithm is proposed for its accurate calculation. The obtained results showed the effectiveness
of ESCA for solving simultaneous NR and DG allocation problem over other competitive algorithms, and its robustness is
confirmed through a detailed statistical analysis such as plotting of box plots, normality checking and two nonparametric
tests, namely Friedman ANOVA and Wilcoxon signed rank tests.

Keywords Distributed generations · Distribution network reconfiguration · Power loss · Operation cost · Reliability index

List of Symbols L limit Lower limit of the variable

Adjm Adjacency matrix levy Levy step length
Csw Switching cost λi Failure rate of node i
EENDI Expected energy not delivered index maxitr Maximum iterations
itr Current iteration nDG Total number of DGs
Ib,k K th branch current nsw Total number of switches
max
Ib,k Maximum value of K th branch current npop Population size
n bus Total number of buses
B Sivkumar Mishra OVSI Overall voltage stability index
[email protected]; [email protected] PL,i Active load demand at node i
Usharani Raut PDG,i Power output of ith DG
[email protected] Sin,i Initial status of ith switch
Sfn,i Final status of ith switch
1 Department of Electrical Engineering, International Institute
TSPL Total system power loss
of Information Technology, Bhubaneswar, Bhubaneswar,
Odisha 751003, India TSPLI Total system active power loss index
2 T S P L NRDG Total system power loss with NR and DG
Department of Electrical Engineering, Centre for Advanced
Post Graduate Studies, Biju Patnaik University of TSOC Total system operation cost
Technology, Rourkela, Odisha, India TSOCI Total system operation cost index

123
1030 Arabian Journal for Science and Engineering (2021) 46:1029–1051

ti Repair time of node i UVDA Uniform voltage distribution-based constructive

Ui Service unavailability of node i reconfiguration algorithm
Ulimit Upper limit of the variable VPE Voltage profile enhancement
VPII Voltage profile improvement index VSIM Voltage stability index maximization
wmx Maximum value of inertia weight w
wmn Minimum value of inertia weight w
Y CDG Annual cost of DG generated power 1 Introduction
Y Csub Annual cost of substation power
Amidst the gradual transformation of the power system into
smart grid, power distribution utilities are constantly facing
Abbreviations challenges due to huge increased quality energy demand.
APLM Active power loss minimization To meet this, these power utilities have started adopting
AELM Annual energy loss minimization novel optimal planning strategies to exploit the use of vari-
AELCM Annual energy loss cost minimization ous resources along with modern smart grid tools to ensure
APGEM Annual pollutant gas emission minimization economic, reliable and safety electric services to its con-
BVDM Bus voltage deviation minimization sumers. Thus, the major focus of current distribution system
BE Branch exchange planning research is in the direction of efficient use of exist-
CSCA Chaotic sine cosine algorithm ing infrastructure with proper planning, in order to ensure
CTLHSO Comprehensive teaching learning harmony the above-mentioned goals. Numerous approaches have been
search optimization available in the literature for techno-economic and reliability
DFS Depth fast search study of distribution systems such as NR, optimal integration
DG Distributed generation of various devices such as shunt capacitors, DGs, D-FACTS
DGA DG allocation devices (D-STATCOM, etc.) and simultaneous use of NR
DGICM DG installation cost minimization and DG allocation.
EDNR Electric distribution network reconfiguration Since long, NR has been extensively studied in the lit-
EGSA Enhanced gravitational search algorithm erature for its significant role to improve the operation and
ESCA Enhanced sine cosine algorithm control of distribution systems with multiple objectives, such
FF Firefly algorithm as minimizing power loss and maximizing the power quality
FWA Fireworks algorithm and reliability with several related indices [1]. Mendoza et
GA Genetic algorithm al. [2] formulated a micro-genetic algorithm-based reconfig-
HHO Harris Hawks optimization uration solution to show the best possible trade-off between
HSA Harmony search algorithm technical (power loss) and reliability indices. In 2013, Gupta
HTLBO Hybrid teaching learning-based optimization et al. [3] analyzed the NR on the basis of power quality and
ICSA Improved cuckoo search algorithm reliability indices by using an efficient genetic algorithm.
IRRO Improved raven roosting optimization The same problem is resolved in [4] with inclusion of an
IS-BPSO Improved selective binary particle swarm opti- additional objective of voltage sag by using quantum firefly
mization algorithm. In 2017, Nguyen et al. [5] suggested a runner root-
ISCA Improved sine cosine algorithm based reconfiguration algorithm, taking into account total
ISFL Improved shuffled frog leaping algorithm system power loss (TSPL), voltage variance, load balanc-
LLIM Line loading index minimization ing and number of switching. The authors in [6,7] addressed
MOCSOS Multi-objective chaotic symbiotic organisms the issues of NR considering the improvement of reliability
search algorithm indices. An enhanced gravitational search algorithm (EGSA)
MOCDE Multi-objective opposition-based chaotic differ- is proposed in [8] to find the best open switch location with
ential evolution simultaneous consideration of technical, economic and reli-
MSO Minimum switching operation ability indices.
NR Network reconfiguration Gradually, the insertion of different energy sources like
PMOSCA Pareto multi-objective sine cosine algorithm DGs, capacitors, D-STATCOMs, etc., has started getting
QOCSOS Quasi-oppositional-chaotic symbiotic attention for efficiency enhancement of distribution systems
organisms search in present smart grid environment. A number of articles have
RPLM Reactive power loss minimization been published in the literature considering the primary con-
SFA Selective firefly algorithm cerns of DG planners. In 2011, Khalesi et al. [9] presented a
SNRDG Simultaneous NR and DG allocation dynamic programming-based approach for TSPL and relia-
SFSA Stochastic fractal search algorithm bility enhancement by using DGs. Singh and Goswami [10]

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1031

developed an optimized planned framework through the use for maximization of the overall benefits considering several
of distributed generators considering service reliability, oper- other objectives.
ation efficiency, energy costs, power quality and security. Literature review [1] reveals that new efficient meta-
Ameli et al. [11] determined the optimal DG position and heuristic algorithms with enhanced searching features often
size by simultaneous consideration of operation, economic successfully handle these complex and large-scale distribu-
and reliable aspects. In 2016, Bohre [12] solved an opti- tion system optimization problems with multi-objectives and
mal DG allocation problem with the consideration of total constraints. Sine–cosine algorithm is a recently proposed
system active and reactive power loss, reliability and sensi- meta-heuristic algorithm with much less control parameters
tivity index. Muthukumar et al. [13] formulated an optimal [28] and has been applied successfully in various applications
capacitor planning problem based on the simultaneous con- of power system in both single- and multi-objective scenarios
sideration of TSPL, voltage variance and voltage stability [29–31]. SCA is first applied to EDNR problem in [32] for
index (VSI) with appropriate weights using a hybrid heuristic loss minimization, to simultaneous EDNR and DGA in [33]
technique. Recently, the authors in [14,15] attempted to solve for loss minimization and voltage stability index maximiza-
the optimal DG inclusion problem with the consideration of tion and to DGA problem with Pareto-based multi-objective
various technical objectives like TSPL, voltage variance and approach in [34] for certain conflicting objectives by the
VSI. The authors in [16] determined the optimal quantity, authors of this paper. Like other stochastic search-based opti-
location and size of DG by minimizing TSPL. The authors mization algorithms, SCA faces some challenges including
in [17,18] attempted to insert D-STATCOM for performance low convergence rate, difficulties in maintaining the diversity
enhancement in deregulated scenario. among the solutions and trapping in local optimum. Using
Current research focused on simultaneous NR and DG the past experience to deal with SCA, authors proposed an
allocation through the use of single and multi-objectives enhanced version of SCA in this work to solve simultane-
has resulted significant contributions toward technological ous EDNR and DGA problem to overcome the mentioned
advancement. In 2012, Rao et al. [19] solved the NR prob- shortcomings of SCA. This enhancement is threefold. Firstly,
lem with the incorporation of DG by minimizing TSPL via the method of linear decreasing inertia weight and exponen-
harmony search algorithm (HSA). Imran [20] presented net- tial decreasing conversion parameter is adopted to maintain
work reconfiguration with DG using fireworks algorithm by proper balance during exploration and exploitation phases.
minimizing TSPL and voltage deviation index (VDI). In [21], Secondly, a local exploitative mechanism via neighborhood
discrete artificial bee colony (DABC) algorithm is used to search is used to exploit promising search spaces and prevent
tackle the issues of simultaneous NR and DG by maximiz- being trapped into local solutions. Lastly, the mechanism of
ing system loadability. Hamida et al. [22] addressed the same levy mutation is adopted to enhance the global exploration
problem by using a Pareto multi-objective approach with capability of the algorithm. In order to assess the overall
the consideration of technical, economic and environmen- benefits in the present planning scenario, the problem is
tal objectives. Muthukumar and Jayalalitha [23] formulated solved in a multi-objective environment with simultaneous
a novel approach for integration of DG and capacitor with consideration of total system power loss (TSPL) and total
NR by minimizing TSPL. Moghaddam et al. [24] proposed system operation cost (TSOC). The outputs of ESCA are
an improved GSA to solve NR with DG in a practical test contrasted with recent competitive algorithms in terms of
distribution system by considering technical, economic and techno-economic and reliability indices.
stability indices. An adaptive cuckoo search-based approach The major contributions of the work are:
is utilized in [25] to determine the optimal network config-
uration and DG allocation by minimizing TSPL as well as – Development of ESCA capable of tackling the challenges
voltage stability deviation. The problem of NR with DG of of mixed integer and constrained optimization problems
varying power factor is solved in [26], considering TSPL, such as EDNR and DGA.
voltage variance and wastage of energy from renewable DG – A local search mechanism with the use of a neighborhood
sources like solar and wind. Recently, the problem of NR and search strategy and a levy mutation strategy is integrated
DG is solved in [27] by minimizing TSPL via stochastic frac- with the original SCA to enhance the exploitation, which
tal search algorithm (SFSA). For the sake of better analysis, a improves avoiding local optimal trapping and ensures
brief overview of recent DGA, EDNR and combined EDNR global solutions with higher optimization accuracy.
and DGA-based approaches as reported in the literature is – A multi-objective framework is formulated with simul-
presented in Table 1. The table shows that most of the recent taneous consideration of technical and economic objec-
works are based on technical objectives as far as simultane- tives.
ous EDNR and DGA are concerned, which motivates authors – The efficiency of the algorithm is verified by conducting
of this paper to take up this for further research and to explore four different cases of NR and DG allocation on IEEE 33
and 69-bus distribution systems.

123
 1032

123
Table 1 A brief overview of recent approaches for DGA and EDNR
Publication year Planning methods Solving algorithm Test systems used Objectives considered

2019 [43] DGA HTLBO 33, 69 APLM, BVDM, VSIM

2019 [44] DGA MOCSOS 33, 69 APLM, BVDM, VSIM
2019 [45] DGA MOCDE 33, 69 APLM, BVDM, AELM
2020 [46] DGA QOCSOS 33, 69, 118 APLM, BVDM, VSIM
2020 [47] DGA IRRO 33, 69 APLM, RPLM,VPE,LLIM, VSIM, DGICM
2020 [48] DGA CSCA 33, 69 APLM, BVDM, VSIM
2020 [34] DGA PMOSCA 33, 69 APLM, VSIM, AELCM, APGEM
2019 [49] EDNR IS-BPSO 33, 94 APLM
2019 [50] EDNR SFA 33, 70, 84 APLM
2019 [51] EDNR Graph theory-based GA 10 APLM,DGICM
2019 [52] EDNR ICSA 33, 69, 119 APLM
2019 [40] EDNR and DGA CTLHSO 33, 69 APLM,VPE
2019 [53] EDNR and DGA IPSO, TLBO, JAYA 33, 69 APLM, VSIM
2019 [36] EDNR and DGA IEJAYA 33, 69 APLM, LE
2020 [54] EDNR and DGA ISFL 33, 69 APLM,MSO, BVDM
2020 [55] EDNR and DGA HHO 84 APLM, VSIM
2019 [56] EDNR and DGA Iterative 33, 118 ELR
2020 [33] EDNR and DGA ISCA 33, 69 APLM, VSIM
Arabian Journal for Science and Engineering (2021) 46:1029–1051
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1033

Fig. 1 Electrical equivalent of a

simple distribution line

– Results show a significant improvement in terms of PR2 + Q 2R

Ploss(R,R+1) = R ∗R (2)
TSPLI, TSOCI, EENDI, OVSI and VPII of the test sys- | V R |2
tems over existing algorithms such as UVDA, SFSA,
CTLHSO, FF, HSA and FWA. 2.2 Total System Operation Cost (TSOC)
– The robustness of the algorithms is established through a
detailed statistical analysis such as plotting of box plots, It includes the annual cost of total power drawn from sub-
normality checking and two nonparametric hypotheses station, annual cost of DG generated power and the cost of
tests, namely Friedman ANOVA and Wilcoxon signed switching. The TSOC is evaluated using Eq. (3):
rank tests.
T S OC = Y CDG + Y Csub + Csw (3)
The remaining portion is structured as follows: Sect. 2 where Y CDG , Y Csub and Csw represent the annual cost of
explains the multi-objective formulation of the problem. The DG generated power, annual cost of substation power and
details of the proposed algorithm are presented in Sect. 3, and switching costs, which are defined in Eqs. (4), (5) and (7),
the implementation of the proposed approach to the simulta- respectively.
neous NR and DG allocation problem is explained in Sect. 4.
The simulation results are presented and discussed in Sect. 5. n
DG
The overall conclusion of the work is presented in Sect. 6. Y CDG = PDG,i × CostDG × 8760 (4)
i=1

In Eq. (4), n DG is the number of DGs, PDG,i is the active

2 Multi-objective Problem Formulation power output of ith DG unit, CostDG is the cost of DG power
in USD/kWh.
The present work is considered in a multi-objective frame-
work with simultaneous minimization of power loss and Y Csub = Psub × Costsub × 8760 (5)
system operating costs. The mathematical formulation of
objectives is given in the following sub-sections: where Costsub is the substation power cost in USD/kWh and
Psub is the active power generated from the substation, which
2.1 Total System Power Loss (TSPL) is defined in Eq. (6):

The total real power loss of the whole distribution system 

n bus n
DG
Psub = PL,i + T S P L NRDG − PDG,i (6)
(TSPL) with NR and DG allocation is defined in Eq. (1) as
i=1 i=1
per [35]:
where n bus is the total number of buses, PL,i is the active

nbr load demand at node i, T S P L NRDG is the total power loss
T SPL = Ploss(R,R+1) (1) of the system with NR and DG.
R=1

nsw
Csw = Costsw,i × | Sin,i − Sfn,i | (7)
where nbr is the total number of branches and Ploss(R,R+1) is i=1
the power loss in the line section connected between Rth and
(R + 1)th node, which is determined using Eq. (2) (referring where nsw is the total number of switches, Costsw is the cost
Fig. 1). per switching in USD. Sin,i and Sfn,i are the initial and final

123
1034 Arabian Journal for Science and Engineering (2021) 46:1029–1051

status of ith switch. The status of the switches is assumed to – Node voltage limits:
be 0 and 1 corresponding to open and closed states, respec- The lower and upper operating limits are defined in Eq.
tively. (15):

2.3 Problem Formulation 0.9 ≤ Vnode,i ≤ 1.01, i = 1 to n bus (15)

The objective function (OF) based on simultaneous consid- where Vnode,i is the ith node voltage.
eration of TSPL and TSOC is expressed as per Eq. (8): – Branch current limits:

O F = Minimi ze(wt1 × T S P L + wt2 × T S OC) (8) | Ib,k |≤| Ib,k

max
|, k = 1, 2, . . . nbr (16)
| wt1 | + | wt2 |= 1 (9)
max are the kth branch current and max-
where Ib,k and Ib,k
where wt1 and wt2 are the weight coefficients which are imum allowable value of the kth branch current of the
determined on the basis of the individual impact and impor- system, respectively.
tance of the specific objective. In this work, both the – Radial configuration:
objectives are given equal importance (wt1 = wt2 = 0.5). This constraint is to ensure the radial structure of the sys-
tem. A depth fast search (DFS)-based technique as per
2.4 Constraints [36] is utilized to map out the path between predecessor
and discovered node using the user-defined MATLAB
The multi-objective function defined in Eq. (8) must be sub- function graphpred2path(). The system is assumed to be
jected to the following operational constraints, violation of radial (radialstatus=1), if the length of the path is 2; oth-
which may lead to invalid solutions to the problem. erwise, it is non-radial (radialstatus=0).
– Isolation:
– Power balance constraint: This restriction is intended to ensure the connection of
Equation (10) is being used to ensure a balance between each node with the substation node. If radial status of
supply and demand. the system is equal to 1, the connectivity among nodes is
verified as per [36].
n
DG 
n bus
Psub + PDG,i − PL,i − T S P L NRDG = 0 (10)
i=1 i=2
3 Proposed Enhance Sine–Cosine Algorithm
– DG position limits: (ESCA)
In view of the practicability, the constraint to position
the DG units is determined as per Eqs. (11) and (12), The ESCA makes improvement in the following three
respectively. aspects:

2 ≤ DG pos ≤ n bus (11) 1. Adoption of exponentially decreasing conversion param-

DG posi = DG posk , (i, k) ∈ n bus (12) eter and linearly decreasing inertia weight to ensure
proper balance between exploration and exploitation.
where DGpos refers to the position of DG units. 2. Inclusion of neighborhood search strategy to prevent
– DG size: local trapping.
The allowable size of the DGs is considered as per Eq. 3. The self-adaptive levy mutation strategy is used for suc-
(13) [20]: cessful exploration of search space. Hence, the local
development ability is improved.

n bus n
DG 
n bus
0.1 PL,k ≤ PDG,n ≤ 0.6 PL,k (13) 3.1 Conventional Sine–Cosine Algorithm
k=2 n=1 k=2
In the conventional SCA, the mathematical sine and cosine
– DG penetration limits: functions are used to update the initial population individual
This restriction is determined using Eq. (14): as described in Eq. (17):

itr + r n × sin(r n )× | r n × Y itr
best,n − Ym,n |, if r n 4 < 0.5
n itr
DG 
n bus itr+1
Ym,n =
Ym,n 1 2 3
PDG,n ≤ PL,k + T S P L NRDG (14) Yi,itrj + r n 1 × cos(r n 2 )× | r n 3 × Ybest,n
itr − Yi,itrj |, else
n=1 k=2 (17)

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1035

itr+1
where Ym,nitr is the nth dimension value of the mth individual Ybest,n = Ybest,n
itr
+ levy × θ (n) × Ybest,n
itr
(23)
at the current iteration itr.
itr
Ybest,n is nth dimension value of the targeted optimal solu- The levy step length (levy) and the control coefficient (θ )
tion in the current iteration itr. are determined as per Eqs. (24) and (27), respectively:
The purpose of random parameters r n 1 , r n 2 , r n 3 and r n 4 u×σ
is to ensure the search within the search space. The conver- Levy = 0.01 × 1
(24)
va
sion parameter r n 1 is considered as per Eq. (18):
where u and v are the random numbers in the range of
 max − itr 
r n1 =
itr
c (18) (0,1). a has a constant value between 1 and 2. σ is con-
maxitr sidered as per Eq. (25):
1
where c has a constant value. itr is the current iteration (1 + β) × sin( πβ
2 )
β

number, and maxitr is the maximum iteration number. σ = β−1

(25)
( 2 )
( 1+β
2 )×β ×2
r n 2 is in the range of [0,2π ]. r n 3 is in the range of [0,2].
The switching parameter r n 4 is used to control the transition where
between sine and cosine function, which is in the range of
[0,1]. (n) = (n − 1)! (26)
r ( j)
( −ε×itr
max )(1− rmax ( j) )
θ (i) = e itr (27)
3.2 ESCA
where
– Strategy 1 In strategy 1, Eq. (17) is modified as per
1  itr
npop
Eq. (19) by incorporating linear variation of inertia r ( j) =| Pbest,
itr
j − X i, j | (28)
weight (expressed in Eq. (20)) and exponential varia- npop
i=1
tion of conversion parameter r n 1 (expressed in Eq. (21)).


w(itr ) × Ym,n
itr
+ r n 1 × sin(r n 2 )× | r n 3 × Ybest,n
itr
− Ym,n
itr
|, if r n 4 < 0.5
itr+1
Ym,n = (19)
w(itr ) × Ym,n
itr
+ r n 1 × cos(r n 2 )× | r n 3 × Ybest,n
itr
− Ym,n
itr
|, otherwise

itr
w(itr ) = wmx − (wmx − wmn ) (20)
maxitr where npop is size of the population.

where wmx and wmn are maximum and minimum value rmax ( j) = max(P jitr ) − min(P jitr ) (29)
of inertia weight w.
itr
r n 1 (itr ) = c e maxitr (21)

– Strategy 2
This strategy enables diversity to be maintained at dif- 4 Implementation of the Proposed
ferent optimization phases and avoids the possibility of Algorithm to the Simultaneous NR and DG
local trapping. The update mode using this strategy is Allocation (SNRDG) Problem
expressed in Eq. (22):
⎧ The simultaneous NR and DG allocation (SNRDG) is a
⎪ w(itr) × Ym,n
itr
+rn 1 × sin(rn 2 )× | rn 3 × Ybest,n
itr
⎪
⎪ mixed integer problem, where the control variables are tie
⎪
⎨ ×(1+λs × uni f rnd(−1, 1)) − Ym,n itr
|, if rn 4 < 0.5
itr+1
Ym,n =
⎪
(22) switch position, DG position and DG size. The tie switch
⎪
⎪ w(itr) × Ym,n
itr
+ rn 1 × cos(rn 2 )× | rn 3 × Ybest,n
itr
⎪
⎩
×(1 + λs × uni f rnd(−1, 1)) − Ym,n |, otherwise
itr and DG positions are discrete variables, whereas DG size
is a continuous variable. Initial tie switches are generated
where unifrnd is a MATLAB command to generate uni- using a BE-based heuristic method [37]. Initial DG position
form distribution number within the specified boundary and size are generated following Eqs. (11), (12) and (13).
and λs is the step length used for effective search orien- The dimension of the solution vector depends on the number
tation. of DGs to be installed. Considering a IEEE 33-bus system
– Strategy 3 In this strategy, the re-formation of solutions having 5 open switches and 3 DGs, the solution vector can
with levy flight is defined in the following equations: be expressed as follows:

123
1036 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Fig. 2 Flowchart of proposed

ESCA to solve SNRDG
problem

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1037

p = [sw1 sw2 sw3 sw4 sw5 DG P1 DG P2 DG P3 23: f obj( poplbest ) = f obj( poplbest,updated )
24: end if
DGC1 DGC2 DGC3 ]
25: itr = itr + 1
26: End
where sw1 · · · sw2 are the open switches, DG P1 , . . . DG P3 27: end while
are the DG positions and DGC1 , . . . DGC3 are the DG sizes 28: Return Glbest obtained so far
in kW corresponding to their position. In this work, DG is
modeled as a biomass source, which is capable of injecting
only active power. The pseudocode of proposed ESCA to get 4.1 Performance Evaluation of the Proposed ESCA
optimal solution to SNRDG problem is given in Algorithm
1. The discrete and continuous variables are handled by the In order to evaluate the effectiveness of ESCA for solving
algorithm by using Eqs. (30) and (31), respectively: simultaneous NR and DG allocation problem, the following
indices are used.
Yn ew = mod(uint16(Y ), Ulimit ) (30)
4.1.1 Total System Active Power Loss Index (TSPLI)
where Y is a discrete variable and Ulimit is the upper limit of
the variable. Equation (32) is used to evaluate TSPLI of the system.

If (X i < L limit ) then X i = L limit T SPLI =

T S P L NRDG
(32)
(31) T S P L0
If (X i > Ulimit ) then X i = Ulimit
where T S P L NRDG is the total system active power loss (kW)
where X i is the continuous variable, Ulimit and L limit are the
with NR and DG, T S P L 0 is the total system active power
upper and lower limit of the variable.
loss (kW) in base case.
A detail flowchart of proposed ESCA to solve SNRDG
Lower value of this parameter indicates more positive
problem is shown in Fig. 2.
effect on the system in terms of TSPL.

Algorithm 1 Pseudocode of proposed ESCA based on neigh-

4.1.2 Total System Operation Cost Index (TSOCI)
borhood search and self adapting levy mutation
Input: The initial population
Output: Global optimal solution, Glbest
Equation (33) is used to evaluate TSOCI.
1: Begin
2: Read the distribution system data T S OCNRDG
T S OC I = (33)
3: Initialize the parameters of ESCA including maxitr , npop, wmx , T S OC0
wmn , c, λs , and β
4: Generate initial population (popl) following the constraints where T S OCNRDG is the total system operation cost (USD)
popl = [ popl1 , Popl2 , · · · , Popli , · · · , poplnpop ]
popli = [open switch location, DG location, DG si ze]
with NR and DG, T S OC0 is the system operation cost (USD)
5: Calculate the objective values of the population using the load flow in base case.
proposed in [36], fobj(popl) and find the best solution obtained so Lower value of this parameter indicates better impact on
far poplbest the system performance.
6: itr = 1
7: while itr ≤ maxitr do
8: Set Glbest = Poplbest 4.1.3 Expected Energy not Delivered Index (EENDI)
9: Determine w and r n 1 according to Eqs. (20) and (21), respec-
tively. In this work, the expected energy not delivered (EEND) is
10: Update r n 2 , r n 3 and r n 4
11: Use Eq. (22) to update popl and generate poplupdated considered among different reliability indices as it is impor-
12: Check all the operating constraints and bring the variables to the tant from the point of view of both the system operators and
limit the consumers. The EENDI is determined using Eq. (34):
13: Check the radial and connected status of poplupdated
14: if (It is radial and connected) then E E N DNRDG
15: status=1 E E N DI = (34)
16: Calculate the objective values of poplupdated and identify the E E N D0
best solution poplbest,updated
17: else where E E N DNRDG is the expected energy not delivered with
18: Go to step 10 NR and DG, E E N D0 is the expected energy not delivered
19: end if
in base case.
20: Mute poplbest,updated using Eq. (23) (strategy 3)
21: if f obj( poplbest,updated ) < f obj( poplbest ) then Lower value of this index indicates better reliability of the
22: poplbest = poplbest,updated system in the presence of NR and DG.

123
1038 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Calculation of Expected energy not delivered (EEND) 15: min=dist(i)

The EEND at each node of the system is calculated as per 16: min_index=i
17: end if
Eq. (35) using Algorithm 2. 18: end for
19: u=min_index
E E N Di = PL,i ∗ Ui (35) 20: status(u)=1
21: for i=2:nbus do
22: if (∼status(i) && Adjm(u,i) && dist(u)+Adjm(u,i)< dist(i))
where PL,i is the active load demand of node i and Ui is the then
service unavailability of node i related to reparation time of 23: parent(i)=u
node i , which is the sum of preceding branch unavailability 24: dist(i)=dist(u)+Adjm(u,i)
times. 25: end if
26: end for
27: end for
Ui = λi ∗ ti (36) 28: pathsize=dist
29: ipathf(1)=1, ipathto(1)=1 {ipathf and ipathto are the initial and
where λi and ti are the failure rate [fail/km-year] and repair final memory locations of the pathline array}
30: for j=2:n bus do
time [h/fail] of each node i which are the sum of preceding 31: ipathf(j)=ipathto(j-1)+1
branch (branches connected between substation and node i) 32: ipathto(j)=ipathf(j)+pathsize(j)-1
failure rates and repair times, respectively. To determine the 33: end for
preceding branches of any node of the system, a graph theory- 34: pathline(1)=0
35: idx=2
based searching is used in Algorithm 2. 36: for pth=2:n bus do
The failure rate of each branch j is calculated using Eq. 37: while (parent(pth=-1) do
(37): 38: x=parent(pth)
39: for drow=1:nbr do
λmax − λmin 40: if (d(drow,1)==pth && dt(drow,2)==x)  (dt(drow,2)==pth
λj = (r j − rmax ) + λmax (37) && dt(drow,1)==x) then
rmax − rmin 41: pathline(idx)=dt(drow,3)
42: idx=idx+1
where λmax and λmin are the maximum and minimum failure 43: end if
rate, which are assumed to be 0.5 fail/year (for the branch 44: end for
with highest impedance) and 0.1 fail/year (for the branch 45: pth=x
46: end while
with lowest impedance). The repair time is taken 8 h/year. 47: end for
The EEND of the whole system is computed as per Eq. 48: Determine the failure rate (λ) of each branch using Eq. (37)
(38). 49: for i=2:nbus do
50: λ=0 {failure rate is initialized}
51: t=0 {repair time is initialized}
n bus
52: for j=ipathf(i):ipathto(i) do
EEND = E E N Di (38)
53: pathline(j)
i=2 54: for k=1:n bus -1 do
55: if a(k)==pathline(j) then
56: λ=λ+λ(k)
57: t=t+1.5
Algorithm 2 Pseudocode to determine EEND 58: U=λ*t
Input: Distribution data matrix and substation node (source) 59: end if
Output: The value of EEND 60: k=k+1
1: Read the distribution system matrix, d 61: end for
2: Enter any open switch set and update ’d’ as ’dt’. 62: end for
3: Generate Adjm 63: A(i)=λ, B(i)=t and C(i)=U
{if there is a connection between node u and v, Adjm(u,v)=Adjm(v,u)=1, 64: end for
else it is zero} 65: EEND=8*A(2:n bus )*PL
4: Enter the source node,1
5: dist(1)=0 {initialize the distance of source node, 1 from itself }
6: for i=1:n bus do
7: parent(1)=-1
8: dist(i)=999 {Assignment of distances of other nodes from source
node,1}
9: status(i)=0 {initialize the status of nodes as 0}
10: end for
11: for count=1:n bus do
12: min=999
13: for i=1:n bus do
14: if status==0 && dist(i) ≤ min then

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1039

4.1.4 Overall Voltage Stability Index (OVSI) Table 2 Performance of proposed ESCA for different values of λs and

The overall voltage stability (OVS) is an indicator of mainte- λs 33-bus 69.bus

nance of the voltage profile in stressed situations. The OVSI TSPL (kW) TSPL (kW)
DG NR NR + DG DG NR NR + DG
is evaluated by Eq. (39):
0.005 10 73.648 145.02 55.05 75.134 101.915 38.422
O V SNRDG
OV SI = (39) 0.005 30 73.397 142.6 53.966 74.748 100 37.914
O V S0 0.005 60 73.519 144.1 54.422 74.819 99.8 38.223
0.005 90 73.421 143.295 54.022 74.83 99.6 38.096
where O V S0 and O V SNRDG are the overall voltage stabil-
0.01 10 73.598 141 54.952 74.928 101.33 38.034
ity (OVS) for base configuration and that after incorporating
0.01 30 73.186 139.55 53.53 74.411 98.605 36.958
NR and DG simultaneously. Higher value of OVSI indicates
0.01 60 73.394 140 54.334 74.542 99.3 37.287
better enhancement in OVS.
0.01 90 73.193 139.976 53.665 74.476 99.6 37.113
The VSI at every node and OVS is determined by Eqs. (40)
0.02 10 73.99 145 55.137 75.211 101.915 40.28
and (41), respectively (referring Fig. 1) [20]:
0.02 30 73.566 142.6 54.233 75.115 99.609 39.32
 2 0.02 60 73.696 144.023 54.475 75.126 100 39.83
V S I (R + 1) = | V R4 | −4 PR+1,eff X R − Q R+1,eff R R 0.02 90 73.592 143 54.310 75.153 99.3 39.485
 
0.03 10 74.12 145.68 56.036 75.369 101.93 40.477
−4 PR+1,eff X R − Q R+1,eff R R | V R |2
0.03 30 73.598 142.603 55.059 75.172 99.609 39.73
(40) 0.03 60 73.723 143.5 55.523 75.269 99.8 39.919

n bus
0.03 90 73.605 142.7 55.169 75.199 99.3 40.101
OV S = V S I (i) (41)
0.05 10 73.601 14 55.005 75.077 101.938 38.159
i=2
0.05 30 73.204 140.277 53.64 74.576 99.609 37.38
4.1.5 Voltage Profile Improvement Index (VPII) 0.05 60 73.405 141 54.378 74.766 101.33 37.58
0.05 90 73.215 140.5 53.825 74.661 100 37.478
The VPII is defined as the ratio of the voltage profile index
(VPI) of the network with and without considering NR and
DG allocation. It is defined in Eq. (42) proposed ESCA is validated considering four different sce-
narios of NR and DG installation. In Scenario-1, the system
V P INRDG is considered without reconfiguration and DG. In Scenario-
V PI I = (42)
V P I0 2, the system is considered with inclusion of DG units only. In
Scenario-3, only reconfiguration of the system is considered.
The VPI of the system is evaluated as per Eq. (43): In Scenario-4, reconfiguration of the system is considered
in the presence of DG units. In each scenario, the system is
1 
n bus
V PI = V Pi (43) simulated in two distinct cases of single-objective and multi-
n bus objective optimization. In Case-1, only TSPL minimization
i=1
is considered, and in Case-2, simultaneous minimization of
V Pi is calculated as per the expression given in Eq. (44): TSPL and TSOC is considered.
The parameters of ESCA used in the simulation are max-
4(Vi − Vmin )(Vmax − Vi ) imum number of iterations, maxitr = 100, population size,
V Pi = i = 1, 2, . . . n bus (44)
(Vmax − Vmin )2 npop = 30, maximum inertia weight, wmx = 0.9, minimum
inertia weight, wmn = 0.4, c = 2, λs = 0.01, = 30,
where Vmax and Vmin are the maximum and minimum mag- β = 1.5. Among these parameters, the maxitr and npop are
nitude of voltage in p.u. set fixed for each scenario. The value of wmx is set at 0.9 in
Higher value of VPII indicates better voltage profile enhance- the early iteration to emphasize the global exploration and is
ment of the system. linearly decreased to wmn , which is set at 0.4 at the end of the
iteration to enhance the exploitation. The conversion param-
eter of r n 1 is initially set at c = 2 as per basic SCA [28] and
5 Results and Discussion then exponentially decreased with increase in iteration. The
optimum values of λs (λs = 0.01) and ( = 30) are selected
The proposed approach is experimented on IEEE 33-bus and by tuning to handle different test distribution systems in dif-
69-bus radial distribution systems. The effectiveness of the ferent scenarios. In each set of input (λs , ), 20 random trials

123
1040 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Fig. 3 Performance comparison Scenario−2 Scenario−3 Scenario−4

in different scenarios for 33-bus 1.5
system

Value of the index

1

0.5

0
TSPLI TSOCI OVSI EENDI VPII

(a) Comparison of performance indices in different scenarios of Case-1

Scenario−2 Scenario−3 Scenario−4

1.5
Value of the index

0.5

0
TSPLI TSOCI OVSI EENDI VPII

(b) Comparison of performance indices in different scenarios of Case-2

Fig. 4 Comparison of voltage Base case DG NR NR and DG

profiles in different cases for 33 1
bus system
Voltage magnitude (p.u)

0.98

0.96

0.94

0.92

0.9
0 5 10 15 20 25 30 35
Bus number
(a) Comparison of voltage profiles in different scenarios of case 1

Base case DG NR NR and DG

1
Voltage magnitude (p.u.)

0.98

0.96

0.94

0.92

0.9
0 5 10 15 20 25 30 35
Bus number
(b) Comparison of voltage profiles in different scenarios of case 2

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1041

Table 3 Performance analysis of 33-bus system in different scenarios

Scenarios Parameters Optimized indices
TSPL (Case-1) TSPL, TSOC (Case-2)

Scenario-2 (DG Installation) Open branches 33-34-35-36-37 33-34-35-36-37

DG Position/Size (kW) 30/928.66 13/789 24/826 14/627.90 25/854.73 29/976.93
TSPL (kW) 73.18 75.07
TSPLI 0.3611 0.3704
TSOC (USD) 1.327E6 1.328E6
TSOCI 0.9667 0.9674
OVS (p.u.) 29.73 29.48
OVSI 1.13 1.12
EEND (kWh/year) 7972.2 8664.8
EENDI 0.1767 0.1920
Min Vbus (p.u)/bus no. 0.9635/33 0.9597/33
VPII 1.4667 1.4464
Std 0.5 0.86
Scenario-3 (Reconfiguration) Open branches 7-9-14-32-37 7-10-14-32-28
TTSPL (kW) 139.55 140.7
TSPLI 0.6886 0.6943
TSOC (USD) 1.3506E6 1.3510E6
TSOCI 0.9839 0.9842
OVS (p.u.) 28.04 28.23
OVSI 1.0716 1.0787
EEND (kWh/year) 37428 34931
EENDI 0.8294 0.7740
Min Vbus (p.u)/bus no. 0.9378/32 0.9413/32
VPII 1.3213 1.3379
Std 0.6 0.85
Scenario-4 (Reconfiguration and DG Installation) Open branches 7-14-9-27-30 27-32-7-10-13
DG Position/Size (kW) 12/567.23 18/712.5 25/1190 15/649 21/590 29/1440
TSPL (kW) 53.53 57.18
TSPLI 0.2641 0.2822
TSOC (USD) 1.3205E6 1.3218E6
TSOCI 0.9620 0.9629
OVS (p.u.) 30.18 30.53
OVSI 1.15 1.16
EEND(kWh/year) 7338.2 7819.5
EENDI 0.1626 0.1733
Min Vbus (p.u)/bus no. 0.9651/31 0.9725/32
VPII 1.49 1.513
Std 0.8 1.2

are repeated and the average results are presented in Table 2. demand. The number of independent experiment taken for
The cost of DG (CostDG ) and substation power (Costsub ) per each scenario is 20 to obtain a global optimal solution. The
kWh is assumed 0.04 USD. The switching cost (costsw ) is proposed ESCA is coded in MatlabR2014a and run on a sys-
assumed 0.04 USD per switching. From technical and com- tem with 8 GB of RAM and intel(R) Core (TM) i5-7200U
mercial point of view, the total number of DGs is limited to 3 CPU (2.5 GHz).
and maximum capacity should not exceed 60% of total load

123
1042 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Table 4 Comparative
Methods Parameters Scenario-2 Scenario-3 Scenario-4
performance of ESCA with
other competitive algorithms for ESCA Open branches 33-34-35-36-37 7-9-14-32-37 7-14-9-27-30
33-bus system
DG Size (kW) 2543.7 – 2469.7
TSPLI 0.3611 0.6866 0.2641
TSOCI 0.9670 0.9839 0.9620
OVSI 1.13 1.07 1.15
EENDI 0.1767 0.8294 0.1626
VPII 1.4667 1.3213 1.49
Min Vbus (p.u) 0.9635 0.9378 0.9651
UVDA [39] Open switches 33-34-35-36-37 7-9-14-32-37 7-10-13-27-32
DG Size (kW) 2731 – 2689
TSPLI 0.3662 0.6886 0.2827
TSOCI 0.9672 0.9839 0.9629
OVSI 1.134 1.0716 1.166
EENDI 0.2468 0.8294 0.1649
VPII 1.46 1.3213 1.51
Min Vbus (p.u) 0.962 0.9378 0.976
SFSA [27] Open switches 33-34-35-36-37 7-9-14-32-37 7-9-14-27-30
DG Size (kW) 2924.8 – 2796.7
TSPLI 0.3525 0.6886 0.2615
TSOCI 0.9666 0.9839 0.9618
OVSI 1.1486 1.0716 1.16
EENDI 0.071 0.8294 0.1224
VPII 1.48 1.3213 1.5
Min Vbus (p.u) 0.9687 0.9378 0.9719
CTLHSO [40] Open switches 33-34-35-36-37 7-9-14-32-37 11-28-31-33-34
DG Size (kW) 2924.9 – 2959.5
TSPLI 0.3526 0.6886 0.2503
TSOCI 0.9666 0.9839 1.7167
OVSI 1.1486 1.0716 1.1731
EENDI 0.0710 0.8294 0.0968
VPII 1.4864 1.3213 1.5105
Min Vbus (p.u) 0.9687 0.9378 0.9734
FF [41] Open switches 33-34-35-36-37 – 8-9-28-32-33
DG Size (kW) – – 1773.8
TSPLI – – 0.3649
TSOCI – – 0.9671
OVSI – – 1.146
EENDI 0.1083
VPII 1.49
Min Vbus (p.u) – – 0.9735
HSA [19] Open switches 33-34-35-36-37 7-14-9-32-37 7-4-10-32-28
DG Size (kW) 1725.6 – 1668.4
TSPLI 0.4774 0.6866 0.3602
TSOCI 0.9952 0.9839 0.9892
OVSI 1.12 1.07 1.14
EENDI 0.0119 0.8294 0.1374
VPII 1.4575 1.3213 1.49
Min Vbus (p.u) 0.9670 0.9378 0.9704

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1043

Table 4 continued
Methods Parameters Scenario-2 Scenario-3 Scenario-4

FWA [20] Open switches 33-34-35-36-37 7-14-9-32-28 7-14-11-32-28

DG Size (kW) 1793.7 – 1684.1
TSPLI 0.4376 0.6907 0.3311
TSOCI 0.9709 0.9840 0.9655
OVSI 1.13 1.08 1.14
EENDI 0.0937 0.7737 0.2161
VPII 1.46 1.34 1.4864
Min Vbus (p.u) 0.9680 0.9413 0.9713

95 Table 5 Normality tests of proposed ESCA for 33-bus system

TSPL (kW)

90 Scenarios p value
85 Shapiro–Wilk Kolmogorov–
test Smirnov test
80

75 Scenario-2 (only DG 1.7344E−4 1.2174E−3

installation)
ESCA SCA UVDA HSA FWA
Scenario-3 (Network 3.7123E−3 0.001523
(a) Box plot for Scenario-2 reconfiguration)
141 Scenario-4 (DG installation 2.7129E−4 0.0161
and Network
reconfiguration)
TSPL (kW)

140.5

140

139.5 The performance of the proposed algorithm is evaluated

ESCA SCA UVDA HSA FWA
for nominal load, in both single- and multi-objective opti-
(b) Box plot for Scenario-3 mization cases and the results are shown in Table 3. It is
75
seen that in Case-1, the TSPL (kW) is reduced to 73.18,
139.55 and 53.53 using Scenarios 2, 3 and 4, respectively.
TSPL (kW)

70
This results in an improvement in TSOC (USD), OVS (p.u.),
65 EEND (kWh/year) of {1.327E6, 1.35E6, 1.3205E6}, {29.73,
60 28.04, 30.18} and {7972.2, 37428, 7338.2} for scenarios
55
2, 3 and 4, respectively. In case-2, the values of TSPL
ESCA SCA UVDA HSA FWA
(kW), TSOC (USD), OVS (p.u.) and EEND (kWh/year)
are {75.07, 140.7, 57.18}, {1.328E6, 1.351E6, 1.3218E6},
(c) Box plot for Scenario-4 {29.48, 28.23, 30.53} and {8664.8, 34931, 7819.5}, respec-
Fig. 5 Box plot in Case-1 of 33-bus system under different Scenarios tively for scenario-2 to 4. Findings show that in both cases all
indices are significantly improved in scenario-4 compared to
scenario-2 and scenario-3. Also, the minimum voltage and
VPII are found to be improved impressively in scenario-4
5.1 IEEE 33-Bus System than scenario-2 and 3.
The performance of the algorithm in terms of TSPLI,
This system has 32 sectionalizing switches and 5 tie switches. TSOCI, OVSI, EENDI and VPII for all scenarios is presented
The detailed line and load data of the system are considered in Fig. 3. Scenario-4 is found to give better compromised
as per [38]. The base kV and base kVA are taken as 12.66 and results than scenario-2 and 3 for both single- and multi-
100000, respectively. The total active, reactive and apparent objective cases. Also, the voltage profiles for different
load demand for base configuration are 3715 kW, 2300 kVAr scenarios of case-1 and case-2 are plotted in Fig. 4. Scenario-
and 4369.4 kVA, respectively. The values of TSPL (kW), 4 is found to give better enhancement in both the cases.
TSOC (USD), OVS (p.u.), EEND (kWh/year), minVbus (p.u.) Since no direct comparisons are available for case-2, so
and VPI (p.u.) for base case are 202.66, 1.3727E6, 26.17, in case-1, the simulation results obtained by ESCA are com-
4.5129E4, 0.9131 and 0.6452, respectively. pared to those from UVDA [39], SFSA [27], CTLHSO [40],

123
1044 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Table 6 Results of
Tests Methods Scenarios
nonparametric tests of different
Scenario-2 Scenario-3 Scenario-4
meta-heuristic methods for
33-bus system Friedman-ANOVA test ESCA 1.24 1.74 1.68
SCA 2.58 2.61 2.47
UVDA 3.69 3.97 3.13
HSA 3.51 3.19 2.94
FWA 2.41 2.49 2.31
Wilcoxon signed rank test ESCA versus SCA 2.32E−4 2.17E−3 1.29E−4
ESCA versus UVDA 1.82E−4 1.54E−4 2.27E−4
ESCA versus HSA 1.31E−3 1.76E−3 1.43E−4
ESCA versus FWA 1.27E−3 1.53E−3 1.41E−4

FF [41], HSA [19] and FWA [20]. The corresponding results TSPL only) of all scenarios. For better visualization, 20 TSPL
are presented in Table 4. It is observed that in scenario-2, values obtained from each technique are displayed using the
the TSPLI obtained from ESCA is 0.3611, which is close to box plot as shown in Fig. 5. Here, the lines extending paral-
UVDA (TSPLI = 0.3662), much better than HSA (TSPLI = lel from the boxes are whiskers. The lower and upper side of
0.4774) and FWA (TSPLI = 0.4376), slightly higher than the whisker represents lowest 25% and highest 25% of the
SFSA (TSPLI = 0.3525) and CTLHSO (TSPLI = 0.3526). data. The box represents the interquartile range, which will
The total DG rating required by ESCA is 2543.7 kW, which be 2σ (where σ is the standard deviation of the data). The
is 187.3 kW lower than UVDA, 381.1 kW lower than SFSA, middle line of the box is the median and ‘+’ represents the
381.2 kW lower than CTLHSO, 818.1 kW higher than HSA outliers. The lower σ value in all scenarios for ESCA rep-
and 750 kW higher than FWA. Thus, the difference in TSPLI resents smaller deviation in the results. Hence, the proposed
value obtained from different algorithms is significant as algorithm is more robust to handle the problem in different
large DG rating incurs higher DG installation cost. How- scenarios than other algorithms.
ever higher DG rating helps to reduce EEND with slightly To check the normality of the results obtained from differ-
improved VPII and OVSI. Since TSPL minimization is the ent algorithms, two hypotheses tests are performed, namely
only objective of this scenario, the performance of ESCA is Shapiro–Wilk test and Kolmogorov–Smirnov test.
better than other algorithms. In scenario-3, ESCA optimized The hypotheses are defined as follows:
the network configuration with open switches 7-9-14-32-37. NH: Tested data are derived from a normally distributed
The obtained TSPLI from ESCA is 0.6866, which is same as population.
that from UVDA, SFSA, CTLHSO, HSA and slightly lower AH: Tested data are not derived from a normally dis-
than FWA, where the open switches of the optimized network tributed population.
are 7-9-14-32-28. It is observed that change in opening of The evaluation involves comparison of p value (prob-
any switch among 32 and 28, slightly alters the final result. ability of rejection of null hypothesis) with the α-value
In scenario-4, the value of TSPLI is 0.2641 from ESCA, (confidence level). The p value less than α signifies rejec-
while that from UVDA is 0.2827, nearly same as that from tion of null hypothesis. Therefore, the tested data are not
SFSA (TSPLI = 0.2615), slightly higher than from CTLHSO normally distributed. In this work, both the normality tests
(TSPLI = 0.2503), much lower than from FF (TSPLI = are conducted using statistical software SYSTAT 12 for the
0.3649), HSA (TSPLI = 0.3602) and FWA (TSPLI = 0.3311). α of 0.05 and the p values obtained by different scenarios are
The improvement in TSPLI for CTLHSO is significant as the presented in Table 5. It is observed that the p values obtained
total DG capacity is 489.8 kW higher than ESCA. This results for all scenarios are less than α, which signifies that the data
in much higher TSOCI, slightly improved OVSI, EENDI and are not derived from the normally distribution curves.
VPII. For better performance comparison of different meth-
ods, two nonparametric tests, namely Friedman-ANOVA and
Wilcoxon signed rank tests, are also conducted, and the
5.1.1 Statistical Analysis results are reported in Table 6. In Friedman-ANOVA test,
the lowest value of ESCA for all scenarios indicates its supe-
In this work, for a fare performance comparison of ESCA riority over existing techniques. Wilcoxon signed rank test
with other algorithms such as SCA, UVDA, HSA and FWA, is performed for a pair-wise comparison of different algo-
each algorithm is run 20 times in Case-1 (minimization of

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1045

Table 7 Performance of 69-bus system in different scenarios

Scenarios Parameters Optimized Indices
TSPL (Case-1) TSPL, TSOC (Case-2)

Scenario-2 (DG Installation) Open branches 69-70-71-72-73 69-70-71-72-73

DG Position/Size (kW) 12/760.4 62/760.4 61/760.4 12/760.4 63/660.4 64/760.4
TSPL (kW) 74.4 79.04
TSPLI 0.3307 0.3513
TSOC (USD) 1.3583E6 1.3599E6
TSOCI 0.9838 0.9850
OVS (p.u.) 65.53 65.39
OVSI 1.0667 1.0645
EEND (kWh/year) 18702 17222
EENDI 0.2809 0.2587
Min Vbus (p.u)/bus no. 0.9717/65 0.9705/61
VPII 1.147 1.144
Std 0.5 0.8
Scenario-3 (Reconfiguration) Open branches 14-55-61-69-70 62-14-56-69-70
TSPL (kW) 98.6 99.6
TSPLI 0.4382 0.4427
TSOC (USD) 1.3363E6 1.3366E6
TSOCI 0.9679 0.9681
OVS (p.u.) 64.58 64.73
OVSI 1.0513 1.0537
EEND (kWh/year) 52470 51735
EENDI 0.7881 0.7771
Min Vbus (p.u)/bus no. 0.9495/61 0.9483/61
VPII 1.12 1.125
Std 0.7 0.8
Scenario-4 (Reconfiguration and DG Installation) Open branches 12-19-69-63-57 14-55-62-69-70
DG Position/Size (kW) 11/436 61/1300 65/461.6 60/810 61/797.8 65/573.8
TSPL (kW) 36.95 40
TSPLI 0.1642 0.1778
TSOC (USD) 1.3452E6 1.3158E6
TSOCI 0.9774 0.9531
OVS(p.u.) 66.39 66.47
OVSI 1.08 1.082
EEND (kWh/year) 13226 11182
EENDI 0.1987 0.1680
Min Vbus (p.u)/bus no. 0.9774/61 0.9813
VPII 1.1576 1.1604
Std 1.31 1.5

rithms. Table 6 shows that the p value obtained by different 5.2 IEEE 69-Bus System
methods in different scenarios is less than the chosen α, indi-
cating that the difference in mean value of the results obtained This system has 68 sectionalizing switches and 5 tie switches.
by different algorithms is statistically significant. The line and load data are considered as per [42]. The base
kV and base kVA are taken as 12.66 and 10000, respectively.
The base active, reactive and apparent load of the system

123
1046 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Fig. 6 Performance comparison Scenario−2 Scenario−3 Scenario−4

in different scenarios for 69-bus 1.4

system 1.2

Value of the index

1

0.8

0.6

0.4

0.2

0
TSPLI TSOCI OVSI EENDI VPII

(a) Comparison of performance indices in different scenarios of Case-1

Scenario−2 Scenario−3 Scenario−4

1.4

1.2
Value of the index

0.8

0.6

0.4

0.2

0
TSPLI TSOCI OVSI EENDI VPII

(b) Comparison of performance indices in different scenarios of Case-2

Fig. 7 Comparison of voltage Base case DG NR NR and DG

profiles in different cases for 69 1
bus system
Voltage magnitude (p.u.)

0.98

0.96

0.94

0.92

0.9
0 10 20 30 40 50 60 70
Bus number
(a) Comparison of voltage profiles in different scenarios of Case-1

Base case DG NR NR and DG

1
Voltage magnitude (p.u.)

0.98

0.96

0.94

0.92

0.9
0 10 20 30 40 50 60 70
Bus number
(b) Comparison of voltage profiles in different scenarios of Case-2

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1047

Table 8 Comparative
Methods Parameters Scenario-2 Scenario-3 Scenario-4
performance of ESCA with
other meta-heuristic algorithms ESCA Open branches 69-70-71-72-73 14-55-61-69-70 12-19-69-63-57
for the 69-bus system
DG Size (kW) 2281.2 – 2197.6
TSPLI 0.3307 0.4382 0.1642
TSOCI 0.9838 0.9679 0.9744
OVSI 1.0667 1.0513 1.08
EENDI 0.2809 0.7881 0.1987
VPII 1.147 1.12 1.157
Min Vbus (p.u) 0.9717 0.9495 0.9774
UVDA [39] Open switches 69-70-71-72-73 14-58-61-69-70 14-58-63-69-70
DG Size (kW) 2431 3900.6 2683
TSPLI 0.3228 0.4383 0.1650
TSOCI 1.578 0.9679 1.63
OVSI 1.07 1.05 1.08
EENDI 0.2289 0.7881 0.1849
VPII 1.15 1.13 1.16
Min Vbus (p.u) 0.9688 0.9495 0.9816
SFSA [27] Open switches 69-70-71-72-73 14-55-61-69-70 14-56-61-69-70
DG Size (kW) 2626.2 – 2461.6
TSPLI 0.3088 0.4384 0.1563
TSOCI 0.9825 0.9679 0.9739
OVSI 1.08 1.05 1.08
EENDI 0.1399 0.7881 0.1309
VPII 1.16 1.12 1.16
Min Vbus (p.u) 0.9790 0.9495 0.9810
CTLHSO [40] Open switches 69-70-71-72-73 14-56-61-69-70 14-56-61-69-70
DG Size (kW) 2625.4 – 2469.1
TSPLI 0.3085 0.4384 0.1563
TSOCI 1.626 0.9679 1.57
OVSI 1.08 1.05 1.08
EENDI 0.1402 0.7881 0.1295
VPII 1.16 1.127 1.16
Min Vbus (p.u) 0.9790 0.9495 0.9813
FF [41] Open switches – – 12-19-57-61-69
DG Size (kW) – – 1937.1
TSPLI – – 0.1791
TSOCI – – 0.9531
OVSI – – 1.07
EENDI – 0.2153
VPII – 1.15
Min Vbus (p.u) – – 0.9816
HSA [19] Open switches 69-70-71-72-73 69-18-13-56-61 69-17-13-58-61
DG Size (kW) 1773.2 – 1844.8
TSPLI 0.3856 0.4416 0.1791
TSOCI 0.9870 0.9917 0.9762
OVSI 1.0534 1.0505 1.0759
EENDI 0.2837 0.7441 0.2289
VPII 1.13 1.122 1.155
Min Vbus (p.u) 0.9677 0.9428 0.9736

123
1048 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Table 8 continued
Methods Parameters Scenario-2 Scenario-3 Scenario-4

FWA [20] Open switches 69-70-71-72-73 69-70-14-56-61 69-70-13-55-63

DG Size (kW) 1832.9 – 1818.2
TSPLI 0.3461 0.4382 0.1745
TSOCI 0.9847 0.99 0.9749
OVSI 1.06 1.0534 1.0752
EENDI 0.2474 0.7881 0.2665
VPII 1.14 1.127 1.1561
Min Vbus (p.u) 0.9740 0.9495 0.9796

are 3802 kW, 2694.6 kVAr and 4660.2 kVA, respectively. 86

84
The base values of TSPL (kW), TSOC (USD), OVS (p.u.),

TSPL (kW)
82
EEND (kWh/year), MinVbus (p.u.) and VPI (p.u.) are 225, 80
1.3806E6, 61.43, 66577, 0.9092 and 0.8560, respectively. 78

Like 33-bus system, the performance of proposed ESCA 76

74
for 69-bus system is also evaluated in both single- and multi- 72
objective optimization cases for nominal load and the results ESCA SCA UVDA HSA FWA

are shown in Table 7. It is observed that in case-1, the TSPL (a) Box plot for Scenario-2
(kW) is reduced to 74.4, 98.6 and 36.95 using scenario-
99.6
2, 3 and 4, respectively. This results in improvement in
99.4
TSOC (USD), OVS (p.u.), EEND (kWh/year) of {1.3583E6,
TSPL (kW)

99.2
1.3363E6, 1.3452E6}, {65.53, 64.58, 66.39} and {18702,
99
52470, 13226} for scenario-2, 3 and 4, respectively. So,
98.8
SNRDG (scenario-4) has significant impact on the perfor-
98.6
mance of the system than only DG (scenario-2) and only
ESCA SCA UVDA HSA FWA
NR (scenario-3). Also, the minimum voltages in all cases are
above minimum threshold. Hence, the quality of solution in (b) Box plot for Scenario-3
scenario-4 is better compared to scenario-2 and 3. 41
Also from Table 7, it is seen that in scenario-2, TSPL is 40.5
40
79.04 kW from case-2, which is 2.1 kW higher than case-1
TSPL (kW)

39.5
by compromising in EENDI which is 232.2 kWh/year lower 39
38.5
than case-1. In scenario-3, 1 kW increase in TSPL results in 38
735 kWh/year reduction in EEND. In scenario-4, 3.05 kW 37.5
37
rise in TSPL results in 2044 kWh/year reduction in EEND.
ESCA SCA UVDA FWA HSA
For better visualization, all the indices obtained from dif-
ferent scenarios of both case-1 and case-2 are shown in Fig. 6 (c) Box plot for Scenario-4
and a comparison of voltage profiles is shown in Fig. 7. From
Fig. 8 Box plot in Case-1 of 69-bus system of different scenarios
both the figures, it is observed that the outcomes are better
compromised in scenario-4 with improved voltage profile.
Since no direct results for case-2 are available in the lit-
erature, so the performance of the algorithm in case-1 for kW lower with 345 kW higher DG size and in CTLHSO,
this system is verified by comparing with other algorithms, TSPL is 5 kW lower with 244.2 kW higher DG size. How-
namely UVDA [39], SFSA [27], CTLHSO [40], FF [41], ever, higher DG size incurs high operational cost (TSOC),
HSA [19] and FWA [20] and the results are shown in Table 8. reduced EEND and slightly improved OVS and VPI. Hence,
It is observed that in scenario-2, TSPLI obtained from ESCA in this scenario, the output results of ESCA with lower DG
is 0.3307, while it is 0.3228 from UVDA, 0.3088 from rating is better than other algorithms. In scenario-3, TSPLI
SFSA, 0.3085 from CTLHSO, 0.3856 from HSA and 0.3461 obtained from ESCA, UVDA, SFSA, CTLHSO and FWA
from FWA. Though UVDA, SFSA, CTLHSO shows slight is almost same with little variation in decimal locations due
improvement in TSPLI, in UVDA, TSPL is 1.77 kW lower to change in opening of any switch among 55, 56 and 58.
with about 149.8 kW higher DG size, in SFSA, TSPL is 4.96 However, the value obtained by ESCA is much lower than
that of the HSA, where two open switch positions are dif-

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1049

Table 9 Normality tests of proposed ESCA for 69-bus system parameters are selected for the proposed algorithm as
Scenarios p value shown in Table 2 to implement different scenarios of NR
Shapiro–Wilk Kolmogorov– and DG allocation.
test Smirnov test – The time complexity of ESCA is higher than that of con-
ventional SCA due to integration of neighborhood search
Scenario-2 (only DG 6.954E−4 0.0141
installation) and levy mutation but both of them are same order of
Scenario-3 (Network 2.7962E−3 0.01783 magnitude.
reconfiguration)
Scenario-4 (DG installation 3.921E−4 0.0298
and Network 6 Conclusion
reconfiguration)
In this work, ESCA is successfully implemented to address
the issues of simultaneous NR and DG allocation for both 33-
and 69-bus distribution systems. The inclusion of exponential
ferent. In scenario-4, the TSPLI obtained by ESCA is better variation in conversion parameter, neighborhood search and
than other algorithms except SFSA and CTLHSO. But for self-adapting levy mutation helps in enhancing the search-
both the algorithms, TSPL is 1.8 kW lower with nearly 270 ing capability of the algorithm. Four different scenarios of
kW higher DG size. This results in much higher TSOC with NR and DG allocation are considered to show the effective-
improved EEND. ness of the algorithm in both single- and multi-objective
For this test system also the TSPL values obtained in Case- cases. The results show that in case of simultaneous NR
1 of different scenarios by different techniques are displayed and DG allocation (scenario-4), the quality of solutions in
using box plot as shown in Fig. 8. For better understanding terms of TSPLI, TSOCI, OVSI, EENDI and VPII is bet-
of comparative performance analysis of different algorithms, ter than considering only DG installation (scenario-2) and
normality followed by nonparametric tests is also conducted only NR (scenario-3). The robustness of the proposed algo-
and the corresponding results are presented in Tables 9 and rithm is assessed from the standard deviation (std) reported
10, respectively. The obtained results indicate the superiority in Tables 3 and 7, by solving the problem 20 times indepen-
of ESCA over other algorithms. dently for each scenario of different cases, and its statistical
significance is verified from two nonparametric hypothe-
5.3 Limitations of ESCA ses testing, namely Friedman-ANOVA and Wilcoxon signed
rank tests. The results established the superiority of ESCA
Although the proposed ESCA is more effective and robust in over other well-known competitive algorithms in solving NR
solving complex, nonlinear, mix integer optimization prob- problems in the presence of DGs.
lem like the simultaneous NR and DG allocation, it has
certain limitations, which are listed below:

– The ESCA needs to be properly tuned for obtaining

global solution. After several run, best tuned algorithm

Table 10 Results of
Tests Methods Scenarios
nonparametric tests of different
Scenario-2 Scenario-3 Scenario-4
meta-heuristic methods for
69-bus system Friedman-ANOVA test ESCA 1.64 2.03 2.14
SCA 2.75 2.89 2.95
UVDA 3.24 3.83 3.93
HSA 2.96 3.17 3.44
FWA 2.82 2.41 2.84
Wilcoxon signed rank test ESCA versus SCA 2.62E−4 1.97E−3 2.29E−4
ESCA versus UVDA 1.59E−4 1.34E−4 2.05E−4
ESCA versus HSA 1.65E−3 1.86E−3 1.53E−4
ESCA versus FWA 1.62E−4 1.73E−3 1.39E−4

123
1050 Arabian Journal for Science and Engineering (2021) 46:1029–1051

Compliance with Ethical Standards 14. Nguyen, T.P.; Vo, D.N.: A novel stochastic fractal search algorithm
for optimal allocation of distributed generators in radial distribution
systems. Appl. Soft Comput. 70, 773–796 (2018)
Conflict of interest The authors declare no conflict of interest regarding 15. Khoa, T.H., Nallagownden, P.; Baharudin, Z.; Dieu, V.N: One
the publication of this paper. rank cuckoo search algorithm for optimal placement of multiple
distributed generators in distribution networks. In: TENCON 2017-
Human and Animal Rights This paper does not include any studies on 2017 IEEE Region 10 Conference, pp. 1715–1720. IEEE (2017)
human or animal participants by any of the authors. 16. Nguyen, T.P.; Vo, D.N.; et al.: Improved stochastic fractal search
algorithm with chaos for optimal determination of location, size,
and quantity of distributed generators in distribution systems. Neu-
ral Comput. Appl. 31(11), 7707–7732 (2019)
17. Ghatak, S.R.; Sannigrahi, S.; Acharjee, P.: Comparative perfor-
mance analysis of DG and DSTATCOM using improved PSO based
References on success rate for deregulated environment. IEEE Syst. J. 12(3),
2791–2802 (2017)
1. Mishra, S.; Das, D.; Paul, S.: A comprehensive review on power 18. Ghatak, S.R.; Sannigrahi, S.; Acharjee, P.: Multi-objective
distribution network reconfiguration. Energy Syst. 8(2), 227–284 approach for strategic incorporation of solar energy source, bat-
(2017) tery storage system, and dstatcom in a smart grid environment.
2. Mendoza, J.E.; Lopez, M.E.; Coello Coello, C.A.; Lopez, E.A.: IEEE Syst. J. 13(3), 3038–3049 (2018)
Microgenetic multiobjective reconfiguration algorithm consider- 19. Srinivasa Rao, R.; Ravindra, K.; Satish, K.; Narasimham, S.V.L.:
ing power losses and reliability indices for medium voltage Power loss minimization in distribution system using network
distribution network. IET Gener. Transm. Distrib. 3(9), 825–840 reconfiguration in the presence of distributed generation. IEEE
(2009) Trans. Power Syst. 28(1), 317–325 (2012)
3. Gupta, N.; Swarnkar, A.; Niazi, K.R.: Distribution network recon- 20. Mohamed Imran, A.; Kowsalya, M.; Kothari, D.P.: A novel integra-
figuration for power quality and reliability improvement using tion technique for optimal network reconfiguration and distributed
genetic algorithms. Int. J. Electr. Power Energy Syst. 54, 664–671 generation placement in power distribution networks. Int. J. Electr.
(2014) Power Energy Syst. 63, 461–472 (2014)
4. Shareef, H.; Ibrahim, A.A.; Salman, N.; Mohamed, A.; Ling Ai, 21. Aman, M.M.; Jasmon, G.B.; Mokhlis, H.; et al.: Optimum tie
W.: Power quality and reliability enhancement in distribution sys- switches allocation and DG placement based on maximisation of
tems via optimum network reconfiguration by using quantum firefly system loadability using discrete artificial bee colony algorithm.
algorithm. Int. J. Electr. Power Energy Syst. 58, 160–169 (2014) IET Gener. Transm. Distrib. 10(10), 2277–2284 (2016)
5. Nguyen, T.T.; Nguyen, T.T.; Truong, A.V.; Nguyen, Q.T.; Phung, 22. Hamida, I.B.; Salah, S.B.; Msahli, F.; Mimouni, M.F.: Optimal
T.A.: Multi-objective electric distribution network reconfiguration network reconfiguration and renewable DG integration considering
solution using runner-root algorithm. Appl. Soft Comput. 52, 93– time sequence variation in load and DGs. Renew. Energy 121, 66–
108 (2017) 80 (2018)
6. López, J.C.; Lavorato, M.; Rider, M.J.: Optimal reconfiguration 23. Muthukumar, K.; Jayalalitha, S.: Integrated approach of network
of electrical distribution systems considering reliability indices reconfiguration with distributed generation and shunt capacitors
improvement. Int. J. Electr. Power Energy Syst. 78, 837–845 (2016) placement for power loss minimization in radial distribution net-
7. Fathy, A.; El-Arini, M.; El-Baksawy, O.: An efficient methodology works. Appl. Soft Comput. 52, 1262–1284 (2017)
for optimal reconfiguration of electric distribution network consid- 24. Mahboubi-Moghaddam, E.; Narimani, M.R.; Khooban, M.H.;
ering reliability indices via binary particle swarm gravity search Azizivahed, A.; et al.: Multi-objective distribution feeder recon-
algorithm. Neural Comput. Appl. 30(9), 2843–2858 (2018) figuration to improve transient stability, and minimize power loss
8. Narimani, M.R.; Vahed, A.A.; Azizipanah-Abarghooee, R.; Javid- and operation cost using an enhanced evolutionary algorithm at the
sharifi, M.: Enhanced gravitational search algorithm for multi- presence of distributed generations. Int. J. Electr. Power Energy
objective distribution feeder reconfiguration considering reliability, Syst. 76, 35–43 (2016)
loss and operational cost. IET Gener. Transm. Distrib. 8(1), 55–69 25. Nguyen, T.T.; Truong, A.V.; Phung, T.A.: A novel method based
(2014) on adaptive cuckoo search for optimal network reconfiguration
9. Khalesi, N.; Rezaei, N.; Haghifam, M.-R.: Dg allocation with appli- and distributed generation allocation in distribution network. Int.
cation of dynamic programming for loss reduction and reliability J. Electr. Power Energy Syst. 78, 801–815 (2016)
improvement. Int. J. Electr. Power Energy Syst. 33(2), 288–295 26. Tuladhar, S.R.; Singh, J.G.; Ongsakul, W.: Multi-objective
(2011) approach for distribution network reconfiguration with optimal DG
10. Singh, R.K.; Goswami, S.K.: Multi-objective optimization of dis- power factor using NSPSO. IET Gener. Transm. Distrib. 10(12),
tributed generation planning using impact indices and trade-off 2842–2851 (2016)
technique. Electr. Power Compon. Syst. 39(11), 1175–1190 (2011) 27. Tran, T.T.; Truong, K.H.; Vo, D.N.: Stochastic fractal search algo-
11. Ameli, A.; Farrokhifard, M.; Ahmadifar, A.; Haghifam, M.-R.: Dis- rithm for reconfiguration of distribution networks with distributed
tributed generation planning based on the distribution company’s generations. Ain Shams Eng. J. 11(2), 389–407 (2020)
and the DG owner’s profit maximization. Int. Trans. Electr. Energy 28. Mirjalili, S.: SCA: a sine cosine algorithm for solving optimization
Syst. 25(2), 216–232 (2015) problems. Knowl. Based Syst. 96, 120–133 (2016)
12. Bohre, A.K.; Agnihotri, G.; Dubey, M.: Optimal sizing and sit- 29. Rizk-Allah, R.M.; El-Sehiemy, R.A: A novel sine cosine approach
ting of DG with load models using soft computing techniques in for single and multiobjective emission/economic load dispatch
practical distribution system. IET Gener. Transm. Distrib. 10(11), problem. In: 2018 International Conference on Innovative Trends
2606–2621 (2016) in Computer Engineering (ITCE), pp. 271–277. IEEE (2018)
13. Muthukumar, K.; Jayalalitha, S.; Sureshkumar, K.; Sakthivel, A.; 30. Saeidi, M.; Niknam, T.; Aghaei, J.; Zare, M.: Multi-objective coor-
Balamurugan, K.; Ramasamy, M.: Application of hybrid heuristic dination of local and centralized volt/var control with optimal
technique for optimal shunt capacitors planning problem in radial switch and distributed generations placement. J. Intell. Fuzzy Syst.
distribution network. Soft. Comput. 23(21), 11227–11245 (2019) 36(6), 6605–6617 (2019)

123
Arabian Journal for Science and Engineering (2021) 46:1029–1051 1051

31. Attia, A.-F.; El Sehiemy, R.A.; Hasanien, H.M.: Optimal power 45. Kumar, S.; Mandal, K.K.; Chakraborty, N.: Optimal DG placement
flow solution in power systems using a novel sine–cosine algorithm. by multi-objective opposition based chaotic differential evolu-
Int. J. Electr. Power Energy Syst. 99, 331–343 (2018) tion for techno-economic analysis. Appl. Soft Comput. 78, 70–83
32. Raut, U.; Mishra, S.: Power distribution network reconfigura- (2019)
tion using an improved sine–cosine algorithm-based meta-heuristic 46. Truong, K.H.; Nallagownden, P.; Elamvazuthi, I.; Vo, D.N.: A
search. In: Soft Computing for Problem Solving, pp. 1–13. Springer quasi-oppositional-chaotic symbiotic organisms search algorithm
(2019) for optimal allocation of DG in radial distribution networks. Appl.
33. Raut, U.; Mishra, S.: An improved sine–cosine algorithm for Soft Comput. 88, 106067 (2020)
simultaneous network reconfiguration and DG allocation in power 47. Nagaballi, S.; Kale, V.S.: Pareto optimality and game theory
distribution systems. Appl. Soft Comput. 92, 106293 (2020) approach for optimal deployment of DG in radial distribution sys-
34. Raut, U.; Mishra, S.: A new Pareto multi-objective sine cosine algo- tem to improve techno-economic benefits. Appl. Soft Comput. 92,
rithm for performance enhancement of radial distribution network 106234 (2020)
by optimal allocation of distributed generators. Evol. Intell. (2020). 48. Selim, A.; Kamel, S.; Jurado, F.: Efficient optimization technique
https://doi.org/10.1007/s12065-020-00428-2 for multiple DG allocation in distribution networks. Appl. Soft
35. Beenish Sultana, M.W.; Mustafa, U.S.; Bhatti, A.R.: Review on Comput. 86, 105938 (2020)
reliability improvement and power loss reduction indistribution 49. Pegado, R.; Ñaupari, Z.; Molina, Y.; Castillo, C.: Radial distribu-
system via network reconfiguration. Renew. Sustain. Energy Rev. tion network reconfiguration for power losses reduction based on
66, 297–310 (2016) improved selective BPSO. Electr. Power Syst. Res. 169, 206–213
36. Raut, U.; Mishra, S.: An improved Elitist–Jaya algorithm for (2019)
simultaneous network reconfiguration and DG allocation in power 50. Gerez, C.; Silva, L.I.; Belati, E.A.; Sguarezi Filho, A.J.; Costa,
distribution systems. Renew. Energy Focus 30, 92–106 (2019) E.C.M.: Distribution network reconfiguration using selective firefly
37. Raut, U.; Mishra, S.: A fast heuristic network reconfiguration algo- algorithm and a load flow analysis criterion for reducing the search
rithm to minimize loss and improve voltage profile for a smart space. IEEE Access 7, 67874–67888 (2019)
power distribution system. In: 2017 International Conference on 51. Souifi, H.; Kahouli, O.; Abdallah, H.H.: Multi-objective distribu-
Information Technology (ICIT), pp. 85–90. IEEE (2017) tion network reconfiguration optimization problem. Electr. Eng.
38. Rao, S.; Narasimham, S.V.L.; Ramalinga Raju, M.; Srinivasa Rao, 101(1), 45–55 (2019)
A.: Optimal network reconfiguration of large-scale distribution sys- 52. Nguyen, T.T.; Nguyen, T.T.: An improved cuckoo search algorithm
tem using harmony search algorithm. IEEE Trans. Power Syst. for the problem of electric distribution network reconfiguration.
26(3), 1080–1088 (2011) Appl. Soft Comput. 84, 105720 (2019)
39. Bayat, A.; Bagheri, A.; Noroozian, R.: Optimal siting and sizing 53. Rawat, M.S.; Vadhera, S.: Heuristic optimization techniques for
of distributed generation accompanied by reconfiguration of dis- voltage stability enhancement of radial distribution network with
tribution networks for maximum loss reduction by using a new simultaneous consideration of network reconfiguration and DG
uvda-based heuristic method. Int. J. Electr. Power Energy Syst. 77, sizing and allocations. Turk. J. Electr. Eng. Comput. Sci. 27(1),
360–371 (2016) 330–345 (2019)
40. Quadri, I.A.; Bhowmick, S.: A hybrid technique for simultaneous 54. Siahbalaee, J.; Rezanejad, N.; Gharehpetian, G.B.: Reconfiguration
network reconfiguration and optimal placement of distributed gen- and DG sizing and placement using improved shuffled frog leaping
eration resources. Soft Comput. 24, 1–22 (2019) algorithm. Electr. Power Compon. Syst. 47(16–17), 1475–1488
41. Badran, O.; Mokhlis, H.; Mekhilef, S.; Dahalan, W.; Jallad, J.: (2019)
Minimum switching losses for solving distribution nr problem with 55. Diaaeldin, I.M.; Abdel Aleem, S.H.E.; El-Rafei, A.; Abdelaziz,
distributed generation. IET Gener. Transm. Distrib. 12(8), 1790– A.Y.; Ćalasan, M.: Optimal network reconfiguration and distributed
1801 (2017) generation allocation using Harris Hawks optimization. In: 2020
42. Savier, J.S.; Das, D.: Impact of network reconfiguration on loss 24th International Conference on Information Technology (IT), pp.
allocation of radial distribution systems. IEEE Trans. Power Deliv. 1–6. IEEE (2020)
22(4), 2473–2480 (2007) 56. Hesaroor, K.; Das, D.: Annual energy loss reduction of distribution
43. Quadri, I.A.; Bhowmick, S.; Joshi, D.: A hybrid teaching–learning- network through reconfiguration and renewable energy sources.
based optimization technique for optimal DG sizing and placement Int. Trans. Electr. Energy Syst. 29(11), e12099 (2019)
in radial distribution systems. Soft. Comput. 23(20), 9899–9917
(2019)
44. Saha, S.; Mukherjee, V.: A novel multiobjective chaotic symbiotic
organisms search algorithm to solve optimal DG allocation prob-
lem in radial distribution system. Int. Trans. Electr. Energy Syst.
29(5), e2839(2019)

123