Electrical Power and Energy Systems 63 (2014) 461–472

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

A novel integration technique for optimal network reconfiguration

and distributed generation placement in power distribution networks
A. Mohamed Imran a,⇑, M. Kowsalya a, D.P. Kothari b
a
School of Electrical Engineering, VIT University, Vellore 632014, Tamil Nadu, India
b
Director Research, Wainganga College of Engineering & Management, Nagpur, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a novel integration technique for optimal network reconfiguration and distributed
Received 1 October 2013 generation (DG) placement in distribution system with an objective of power loss minimization and volt-
Received in revised form 12 March 2014 age stability enhancement. Fireworks Algorithm (FWA) is used to simultaneously reconfigure and allo-
Accepted 1 June 2014
cate optimal DG units in a distribution network. FWA is a new swarm intelligence based optimization
algorithm which is conceptualized using the fireworks explosion process of searching for a best location
of sparks. The radial nature of the network is secured by generating proper parent node-child node path
Keywords:
of the network during power flow. Voltage Stability Index (VSI) is used to pre-identify the optimal can-
Distributed generation
Fireworks Algorithm
didate locations for DG installation. Six different scenarios are considered during DG placement and
Network reconfiguration reconfiguration of network to assess the performance of the proposed technique. Simulations are carried
Power loss minimization out on well-known IEEE 33- and 69-bus test systems at three different load levels. The simulated results
Voltage stability enhancement demonstrate well the performance and effectiveness of the proposed method.
Ó 2014 Elsevier Ltd. All rights reserved.

Introduction
In recent times, electric distribution systems are becoming large
and complex leading to higher system losses and poor voltage reg-
ulation. Studies indicate that almost 10–13% of the total power
generated is lost as I2R losses at the distribution level [1], which
in turn, causes increase in the cost of energy and poor voltage pro-
file along the distribution feeder. Therefore, it becomes important
to improve the reliability of the power transmission in distribution
networks. The most common methods used for voltage stability
enhancement and power loss reduction in power distribution sys-
tems are network reconfiguration and DG placement [2]. The con-
figuration of the network, placement and size of DG units should
be optimal to maximize the benefits and reduce their impact on
the power system. Thus the optimal integration between these
two problems becomes a significant and complex problem.
Network reconfiguration is the process of changing the topology
by altering the open/closed status of switches, so as to find a radial
operating structure that minimizes the losses and improves the
voltage stability while satisfying the operating constraints. Many
researchers solved the network reconfiguration problems using
different methods for the last two decades [2]. Authors in [3] were
⇑ Corresponding author. Mobile: +91 9962680869.
E-mail addresses: mohamedimran.a@vit.ac.in, imraneee20@gmail.com
(A. Mohamed Imran).
the first to solve the network reconfiguration problem for loss min-
imization using a branch and bound-type optimization technique.
Later on, several optimization algorithms have been developed for
loss minimization and/or voltage profile improvement. Authors in
[4] used a discrete artificial bee colony algorithm to find the max-
imum loadability point by optimizing the distribution network,
and also used continuation power flow along with graph theory
to compute power flow. In [5], authors solved the reconfiguration
problem assuming a series fault at a bus using Bacterial Foraging
Optimization Algorithm (BFOA). Authors in [6] proposed modified
honey bee mating optimization algorithm to investigate the net-
work reconfiguration problem with the consideration effect of
the renewable energy sources. In the very latest researches, a
new non-revisiting genetic algorithm had been proposed for solv-
ing the reconfiguration problem [7].
DG units are small generating plants connected directly to the
distribution network or on the customer site of the meter. The
number of DG units installed in the distribution system has been
increasing significantly, and their technical, economical, and envi-
ronmental impacts on power system are being analyzed. The most
critical factors that influence the technical and economic impacts
are type, size, and location of DG units in power system. Recently,
several studies have been carried out on impact of DG units [8–12].
Authors in [8,9] presented an analytical and improved analytical
expression for finding the optimal location and size of DG units

http://dx.doi.org/10.1016/j.ijepes.2014.06.011
0142-0615/Ó 2014 Elsevier Ltd. All rights reserved.
462 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472

Nomenclature

PLk real power load at bus k QLk reactive power load at bus k
Pk real power flowing out of bus k Qk reactive power flowing out of bus k
Pk+1,eff total effective real power load fed through the bus k + 1 Qk+1,eff total effective reactive power load fed through the bus
Ik equivalent current injected at node k k+1
Jk branch current in the line section between buses k and Vk voltage magnitude at bus k
k+1 DV max maximum voltage drop limit between buses 1 (substa-
Rk resistance of the line section between buses k and k + 1 tion) and k
PLoss ðk; k þ 1Þ real power loss in the line connecting buses k and Xk reactance of the line section between buses k and k + 1
k+1 PT; Loss total power loss of the system (base case)
PRT; Loss total power loss of the system after reconfiguration PDG total power loss of the system with DGs
T; Loss
Pmin
TDG minimum total real power generation limit Pmax maximum total real power generation limit
TDG
PDG,k real power supplied by DG at node k
PSS power supplied by the substation
PTDG total real power supplied by DG in the system
nd number of candidate buses for DG installation
Sk apparent power flowing in the line section between
Sk;max maximum power flow capacity limit of line section be-
buses k and k + 1
tween buses k and k + 1
V worst Worst voltage magnitude of the system.
nb total number of branches
n total number of buses

for reducing power loss along with methodologies for identifying to generate proper parent node child node path as in [17]. This
the optimal location. Authors in [10] proposed Particle Swarm path generation will ensure the radial nature of the system, also
Optimization (PSO) based technique to solve the optimal place- prevents the creation of unconnected branches or nodes, and for-
ment of different types of DGs for power loss minimization. In mation of mesh loops. Hence at each phase of distribution system
[11], authors proposed a new method which employs a modified reconfiguration, the power flow is carried out only after the gener-
teaching–learning based optimization algorithm to determine the ation of proper parent node child node path. For the sample system
optimal site and size of DG units. In [12], authors introduced a shown in Fig. 1, the generation of parent node child node path is
hybrid population-based algorithm with the combination of PSO illustrated in Table 1.
and gravitational search algorithm to determine the optimal place
and size of multi-DG units in distribution systems. More recently, Power flow equations
BFOA based approach had been proposed to solve the DG place-
ment problem at different load levels [13]. Distribution system power flow is calculated by the following
All the above researches [2–13] focussed only on the optimiza- set of recursive equations derived from the single line diagram
tion of either distribution network or the DG placement. However, shown in Fig. 2.
the objective was to minimize the power loss and/or to improve From Fig. 2, the equivalent current injected at node k is calcu-
the voltage stability, network reconfiguration usually did not take lated as
the DG units into consideration and vice versa. To benefit the
whole distribution network effectively, it is necessary to integrate
both the network reconfiguration and DG placement problems
[14,15]. In [16], authors have dealt network reconfiguration and
DG placement simultaneously using Harmony Search Algorithm 3 8 10
(HSA) based only on minimization of power losses.
The novelty of this paper is that it proposes recently developed
fireworks optimization algorithm for solving the distribution sys- 1 2 4 9
tem network reconfiguration together with DG placement for the
problem of power loss minimization and voltage stability enhance-
ment. The proposed technique makes use of VSI to pre-identify the
candidate bus locations for DG installation. Also the technique
monitors the radial nature of the network at all phases of reconfig- 6
uration by generating proper parent node-child node path during
power flow. The proposed technique has been validated on two
standard IEEE 33- and 69-bus test systems, and the results simu-
lated are compared with the results of other classical techniques 11
available in the literature to assess the performance and effective-
ness of the proposed method.

Problem formulation 5

Radial nature of the system

7
To monitor the radial nature of the distribution network, the
nodes of the distribution system are optimally ordered in order Fig. 1. Sample distribution system.
 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472 463

! !
Table 1 PRT;Loss þ PDGT;Loss
Illustration of parent node child node path generation. Minimize F ¼ min: þ DV D ð8Þ
PT;Loss
Parent node 1 2 2 2 5 5 4 4 9 9
Child node 2 3 4 5 6 7 8 9 10 11 Subjected to the following operating constraints.
Power conservation limits:
  X
n X
nb X
nd
PLk þ jQ Lk
Ik ¼ ð1Þ Pss ¼ P Lk þ PLoss ðk; k þ 1Þ  P DG;k ð9Þ
Vk k¼2 k¼1 k¼1

Branch current in the line section connecting buses k and k + 1 Voltage deviation limits:
is calculated by using Kirchhoff’s Current law as
jV 1  V k j 6 DV max ð10Þ
J k ¼ Ikþ1 þ Ikþ2 ð2Þ
Distribution line capacity limits:
The above equation is generalised in matrix form by using Bus
current Injection to Branch Current matrix (BIBC) [18]. Now the jSk j 6 jSk;max j ð11Þ
branch current at each line can be calculated in a matrix form as
follows: Distributed generation capacity limits:

½J ¼ ½BIBC  ½I ð3Þ Pmin max

TDG 6 P TDG 6 P TDG ð12Þ
By Kirchhoff’s voltage law, the voltage at the bus k + 1 can be
calculated as X
n X
n
where Pmin
TDG ¼ 0:1  PLk and Pmax
TDG ¼ 0:6  P Lk ð13Þ
V kþ1 ¼ V k  J k  ðRk þ jX k Þ ð4Þ k¼2 k¼2

The power loss in the line section connecting buses k and k + 1 Furthermore, the radial nature of distribution network must be
is computed as maintained, and all loads must be served. If any one of the above
! constraints is violated, the resultant solution will be rejected.
P2k þ Q 2k
PLoss ðk; k þ 1Þ ¼ Rk  ð5Þ
jV k j2
Voltage Stability Index for DG installation
The total power loss of the system is determined by the summa-
tion of losses in all the line sections, which is given as Voltage Stability Index (VSI) that identifies the most sensitive
bus is used to determine the candidate bus locations for DG instal-
X
nb
PT;Loss ¼ PLoss ðk; k þ 1Þ ð6Þ lation in the system [19]. The estimation of these candidate buses
k¼1 initially helps in reducing the search space significantly for the
optimization technique.
Consider a line section consisting an impedance of Rk + jXk and a
Voltage deviation index
load of Pk+1,eff + jQk+1,eff connected between k and k + 1 buses as
shown in Fig. 3.
One of the main advantages of the optimal network reconfigu-
From Fig. 3, the following equations can be written:
ration and DG installation is the fall in voltage deviation. This index
rebukes the configuration and size of DGs which gives the higher V k  V kþ1
voltage deviation. The voltage deviation index (DVD) can be Jk ¼ ð14Þ
Rk þ jX k
defined as
 
V1  Vk Pkþ1;eff  jQ kþ1;eff ¼ ðV kþ1 Þ  J k ð15Þ
DV D ¼ max 8 k ¼ 1; 2; . . . ; n ð7Þ
V1
By solving the above two equations, the VSI at bus k + 1 is
During network reconfiguration and DG installation, if the state derived and given as
of the system has voltage limit violations, the proposed technique
will try to minimize DVD closer to zero and thereby improves volt- VSIðk þ 1Þ ¼ jV k j4  4:0ðPkþ1;eff X k  Q kþ1;eff Rk Þ2
age stability and network performance.
 4:0ðPkþ1;eff Rk þ Q kþ1;eff X k ÞjV k j2 ð16Þ
Objective function of the problem Using the above equation, VSIs are computed from the power
flows for all the buses of the given system and the values are
The proposed objective function (F) of the problem is formu- arranged in ascending order. VSIs decide the sequence in which
lated to minimize the total power loss and voltage deviation index, the buses are to be considered for DG installation. The optimal size
which is given by of DGs at the candidate buses are determined by using FWA.

Vk V k+1 V k+2

Jk P k + jQ k J k+1
th
k line P k+1 + jQ k+1

k k+1 k+2
P Lk + jQ Lk P Lk+1 + jQ Lk+1 P Lk+2 + jQ Lk+2

Fig. 2. Sample distribution feeder.

464 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472

Vk V k+1 At each iteration of FWA, among all the current sparks and fire-
works, the best location is always kept for the next explosion gen-
R k + jX k Jk eration. After that, n  1 fireworks are selected with some
k th line probabilities proportional to their distances to other locations.
The selection probability of a location is defined by the following
k k+1 equations:
X X
P k+1,eff +jQ k+1,eff Rðxi Þ ¼ dðxi ; xj Þ ¼ jjxi  xj jj ð23Þ
j2K j2K
Fig. 3. Simple distribution line.

Rðxi Þ
pðxi Þ ¼ P ð24Þ
j2K Rðxj Þ
Overview of fireworks optimization algorithm (FWA)

where K is the set of all current locations of both fireworks and

Fireworks Algorithm (FWA) is a swarm intelligence based sto-
sparks.
chastic search technique, recently developed by Tan and Zhu
It is observed that, the spark search process of FWA incorpo-
[20]. In recent years, FWA has been applied in solving number of
rates the structure of existing heuristics method. It preserves the
optimization problems [21] due to its ability to search the promis-
best location that is fairly near the desired optimum and selects
ing areas of the solution space. In FWA, the explosion process of a
n  1 locations from the two types of sparks generated with some
firework is considered as a search in the local space around a spe-
probability for the next explosion generation. This is found to be
cific point where the firework is set off through the sparks gener-
the uniqueness of FWA with other optimization algorithms like
ated in the explosion. At each generation, FWA selects some
Genetic Algorithm (GA). In GA, the new locations are selected only
quality points as fireworks, which generates a number of sparks
from two of the existing locations (the parents). In addition, FWA
to search the local space around them. The search process contin-
considers each dimension of a location independently while gener-
ues until at least one spark reaches a fairly desired optimum, or the
ating a new location of sparks, whereas GA cannot since it has to
stopping criterion is met.
maintain the gene structure. These properties of FWA increase
In FWA presented in [20], the number of sparks and the ampli-
accuracy and convergence in solving the complex conventional
tude of explosion for each firework xi is defined as follows:
problem.
ymax  f ðxi Þ þ n
si ¼ m  Pn ð17Þ
i¼1 ðymax  f ðxi ÞÞ þ n
Application of FWA for power loss minimization and voltage
^ f ðxi Þ  ymin þ n stability enhancement
Ai ¼ A  P n ð18Þ
i¼1 ðf ðxi Þ  ymin Þ þ n
^ The algorithmic framework of FWA to optimize the distribution
where m and A are control parameters, f(xi) is the value of objective network with simultaneous DG allocation are presented as follows.
function (fireworks) at location xi, ymax and ymin are the maximum Step 1: Initialization of the FWA parameters.
(worst) and minimum (best) value of the objective function among N: number of iterations (50).
the n fireworks, and n is a smallest constant in the computer to d: dimension of search space i.e. number of tie switches and
avoid zero-division-error. number of candidate buses for DG
To avoid overwhelming effects of splendid fireworks, bounds installation (5 + 3).
are defined for si, which is given in Eq. (19). n: number of locations (20).
8 m: parameter controlling total number of sparks (25).
< roundða:mÞ
> if si < am Â: maximum explosion amplitude (20).
Si ¼ roundðb:mÞ if si > bm; a < b < 1; ð19Þ Constants a = 0.08, b = 0.8 and ň = 5.
>
: Step 2: Randomly select n locations of sparks for fireworks as
roundðsi Þ otherwise
follows.
where the parameters a and b are constants. 2 3
x11 x12 x13   x1d
For a d-dimensional problem, the location of each spark xj gen-
6 x2 x22 x23   7
x2d
erated by xi can be obtained by randomly setting z directions 6 1 7
6 7
(z < d), and for each dimension k setting the component xkj 6      7
x¼6
6
7 ð25Þ
(1 6 j 6 Si, 1 6 k 6 z). There are two ways for setting xkj . For most 6      7 7
sparks, a displacement h = Ai. rand (1, 1) is added to xkj as 6 n 7
4 x1 xn2 xn3   xnd 5
xkj ¼ xkj þh ð20Þ

To keep the diversity, for a few specific sparks, an explosion
coefficient based on Gaussian distribution is applied to xkj and is
given as:
xkj ¼ xkj þ Gaussianð1; 1Þ
In both the ways, if the obtained new location falls out of the
search space, it is mapped to the search space according to the
Eq. (22),
xkj ¼ xkmin þ jxkj j % ðxkmax  xkmin Þ
The structure of above solution vector x for radial distribution
network optimization is expressed by ‘‘Arc No.(i)’’ and ‘‘SW.No.(i)’’
for each switch i. ‘‘Arc No.(i)’’ identifies the arc (branch) number
that contains ith open switch, and ‘‘SW.No.(i)’’ identifies the switch
ð21Þ that is normally open on Arc No.(i). For large distribution networks,
it is not efficient to represent every arc in the string, as its length
will be very long. Therefore, to memorize the radial configuration,
it is enough to number only the open switch positions [22]. For
solving network reconfiguration problem with simultaneous DG
allocation, the first part of the solution vector is taken as the num-
ð22Þ
ber of open switches in the distribution network, and the second
 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472 465

23 24 25 26 27 28 29 30 31 32 33

LP5 LP4
LP2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
LP1
LP3

19 20 21 22

Fig. 4. Single line diagram of 33-bus system.

Table 2
Performance analysis of proposed method on 33-bus system at different load levels.

Scenario Item Load level

Light (0.5) Nominal (1.0) Heavy (1.6)
Base case (scenario I) Switches opened 33–34–35–36–37 33–34–35–36–37 33–34–35–36–37
PT,Loss (kW) 47.06 202.67 575.31
VWorst in p.u. 0.9583 0.9131 0.8529
(Bus no) (18) (18) (18)
Only reconfiguration (scenario II) Switches opened 07–14–09–32–28 07–14–09–32–28 07–14–09–32–28
PT,Loss (kW) 33.39 139.98 381.24
VWorst in p.u. 0.9714 0.9413 0.9027
(Bus no) (32) (32) (32)
% Loss reduction 29.04 30.93 33.73
Only DG installation (scenario III) Switches opened 33–34–35–36–37 33–34–35–36–37 33–34–35–36–37
DG size in MW (candidate bus) 0.2948 (14) 0.5897 (14) 0.9441 (14)
0.0947 (18) 0.1895 (18) 0.3013 (18)
0.5072 (32) 1.0146 (32) 1.6784 (32)
PT,Loss (kW) 21.37 88.68 238.07
VWorst in p.u. 0.9844 0.9680 0.9484
(Bus no) (30) (30) (29)
% Loss reduction 54.58 56.24 58.57
DG installation after reconfiguration (scenario IV) Switches opened 07–14–09–32–28 07–14–09–32–28 07–14–09–32–28
DG size in MW (candidate bus) 0.2991 (32) 0.5996 (32) 0.9590 (32)
0.1564 (33) 0.3141 (33) 0.4902 (33)
0.0836 (18) 0.1591 (18) 0.2434 (18)
PT,Loss 20.26 83.91 224.92
VWorst in p.u. 0.9810 0.9612 0.9363
(Bus no) (30) (30) (30)
% Loss reduction 56.82 58.59 60.90
Reconfiguration after DG installation (scenario V) Switches opened 07–34–09–32–28 07–34–09–32–28 07–34–09–32–28
DG size in MW (candidate bus) 0.2948 (14) 0.5897 (14) 0.9441 (14)
0.0947 (18) 0.1895 (18) 0.3013 (18)
0.5072 (32) 1.0146 (32) 1.6784 (32)
PT,Loss (kW) 16.56 68.28 181.29
VWorst in p.u. 0.9860 0.9712 0.9542
(Bus no) (29) (29) (29)
% Loss reduction 64.81 66.31 68.49
Simultaneous reconfiguration and DG installatio (scenario VI) Switches opened 07–14–10–32–28 07–14–11–32–28 07–14–11–32–28
DG size in MW (candidate bus) 0.2586 (32) 0.5367 (32) 0.9590 (32)
0.3218 (29) 0.6158 (29) 1.1901 (29)
0.2803 (18) 0.5315 (18) 1.0206 (18)
PT,Loss (kW) 16.22 67.11 172.97
VWorst in p.u. 0.9862 0.9713 0.9554
(Bus no) (14) (14) (14)
% Loss reduction 65.53 66.89 69.93

The bold value defines the significance of proposed method in terms of power loss minimization and Voltage profile improvement.
466 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472

Table 3
Comparison of performances for 33-bus system.

Scenario Item FWA HSA [16] GA [16] RGA [16]

Scenario II Switches opened 07–14–09–32–28 07–14–09–32–37 33–34–09–36–28 07–14–09–32–37
PT,Loss (kW) 139.98 138.06 141.60 139.46
% Loss reduced 30.93 31.88 30.15 31.20
VWorst (p.u) 0.9413 0.9342 0.9310 0.9315
Scenario III DG size (MW) 1.7937 1.7256 1.6044 1.777
PT,Loss (kW) 88.68 96.76 100.1 97.60
% Loss reduced 56.24 52.26 50.60 51.84
VWorst (p.u) 0.9680 0.9670 0.9605 0.9687
Scenario IV Switches opened 07–14–09–32–28 07–14–09–32–37 33–34–09–36–28 07–14–09–32–37
DG size (MW) 1.0727 1.0909 1.448 1.100
PT,Loss (kW) 83.91 97.13 98.36 98.23
% Loss reduced 58.59 52.07 51.46 51.53
VWorst (p.u) 0.9612 0.9479 0.9506 0.9479
Scenario VI Switches opened 07–14–11–32–28 07–14–10–32–28 07–34–10–32–28 07–12–09–32–27
DG size (MW) 1.6841 1.6684 1.9633 1.774
PT,Loss (kW) 67.11 73.05 75.13 74.32
% Loss reduced 66.89 63.95 62.92 63.33
VWorst (p.u) 0.9713 0.9700 0.9766 0.9691

part is the number of candidate buses chosen for DG installation. DG3 are sizes of DG units (kW) to be installed at candidate buses
Thus the solution vector for simultaneous reconfiguration and respectively.
DG installation is formed as follows: Step 3: Fireworks loop q = q + 1
Set off n fireworks at the selected n locations. i.e. for the
2 Open Switches DGSizes
3 selected n locations, objective function F(xi) is calculated by the
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{ zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{ power flows.
6 1 1 1 1 1 7
6 SW 1 SW 2 SW 3 SW 4 SW 5 DG11 DG12 DG13 7 Step 4: For each firework F(xi) 8 i ¼ 1; 2; . . . ; n
6 7
6 SW 2
SW 2 SW 3 SW 4 SW 25
2 2 2 2
DG1 DG2 DG3 7 2 27
Calculate the number of sparks that the firework yields: Si,
6 1
6 7 ð26Þ according to Eq. (19).
x¼6         7
6 7 Obtain the location of Si sparks of the firework xi using displace-
6         7
6 7
6 7 ment as in Eqs. (20) and (22);
4 SW n1 SW n2 SW n3 SW n4 SW n5 DGn1 DGn2 DGn3 5 Step 5: Generation of specific sparks loop k = k + 1.
Randomly select a firework xj and generate a specific spark for
the firework using Gaussian explosion method as in Eqs. (21) and
where SW1, SW2, SW3, SW4 and SW5 are open switches in the loops (22).
LP1 to LP5 formed corresponding to the tie switches; DG1, DG2 and Step 6: If k 6 ň, go to step 5.

33 34 35

32

31 53 54 55 56 57 58 59 60 61 62 63 64 65
47 48 49 50
30

29 LP4 66 67
28 LP5
LP2

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

36 LP1 LP3

37

38 52 51 68 69

39

40 41 42 43 44 45 46

Fig. 5. Single line diagram of 69-bus system.

 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472 467

Step 7: Evaluate the quality of all the above locations and select
the best location that gives minimum F and keep it for the next
explosion generation.
Step 8: Randomly select n  1 locations from the two types of
sparks and the current fireworks according to the probability given
in Eq. (24).
Step 9: if q 6 N, go to step 3. Otherwise end and display the
results of best location.
These are the prime steps of FWA employed to optimize the dis-
tribution network with simultaneous DG allocation for power loss
minimization and voltage stability enhancement.
Results and discussions
To demonstrate and examine the applicability of the proposed
technique in solving the network reconfiguration and installation
of DG units simultaneously using FWA, it is applied to two
12.66-kV test systems consisting of 33 and 69 buses. For both
the test systems, the substation voltage is considered as 1 p.u.,
and all the tie and sectionalizing switches are considered as candi-
date switches for solving reconfiguration problem. The maximum
number of DGs installed for the given test systems is limited to
three since it is studied that the rate of improvement of percentage
loss reduction decreases when the candidate locations increases
more than three at all load levels [16,23]. However, the proposed
technique can be implemented for any number of DGs. The FWA
parameters initialized in the above section is taken common for
both the test systems. In the simulation of the test systems, six dif-
ferent scenarios at three different load levels: 0.5 (light), 1.0 (nom-
inal), and 1.6 (heavy) are considered to analyze the superiority and
performance of the proposed method.
Scenario I: The system is without reconfiguration and DG units
(base case).
Scenario II: The system is optimally reconfigured by the avail-
able sectionalizing and tie switches.
Scenario III: Optimal size of DG units installed at the candidate
buses of the system.
Scenario IV: Optimal size of DG units is installed after reconfig-
uration of the network.
Scenario V: The system is optimally reconfigured in presence of
DG units installed (i.e. reconfiguring the distribution network
with the distributed energy resources).

Table 4
Performance analysis of proposed method on 69-bus system at different load levels.

Scenario Item Load Level

Light (0.5) Nominal (1.0) Heavy (1.6)
Base case (scenario I) Switches opened 69-70-71-72-73 69-70-71-72-73 69-70-71-72-73
PT,Loss (kW) 51.60 224.96 652.42
VWorst in p.u. 0.9567 0.9092 0.8445
(Bus no) (65) (65) (65)
Only reconfiguration (scenario II) Switches opened 69-70-14-56-61 69-70-14-56-61 69-70-14-56-61
PT,Loss (kW) 23.61 98.59 267.08
VWorst in p.u. 0.9754 0.9495 0.9165
(Bus no) (61) (61) (61)
% Loss reduction 54.24 56.17 59.06
Only DG installation (scenario III) Switches opened 69-70-71-72-73 69-70-71-72-73 69-70-71-72-73
DG size in MW (candidate bus) 0.2067 (65) 0.4085 (65) 0.6537 (65)
0.5903 (61) 1.1986 (61) 1.9177 (61)
0.1076 (27) 0.2258 (27) 0.3613 (27)
PT,Loss (kW) 19.05 77.85 206.49
VWorst in p.u. 0.9871 0.9740 0.9568
(Bus no) (62) (62) (62)
% Loss reduction 60.08 65.39 68.35
DG installation after reconfiguration (scenario IV) Switches opened 69-70-14-56-61 69-70-14-56-61 69-70-14-56-61
DG size in MW (candidate bus) 0.4920 (61) 1.0014 (61) 1.6060 (61)
0.1115 (62) 0.2145 (62) 0.3176 (62)
0.0716 (64) 0.1425 (64) 0.2274 (64)
PT,Loss 10.78 43.88 115.98
VWorst in p.u. 0.9860 0.9720 0.9544
(Bus no) (61) (61) (61)
% Loss reduction 79.11 80.49 82.22
Reconfiguration after DG installation (scenario V) Switches opened 69-70-12-58-61 69-70-12-58-61 69-70-12-58-61
DG size in MW (candidate bus) 0.2067 (65) 0.4085 (65) 0.6537 (65)
0.5903 (61) 1.1986 (61) 1.9177 (61)
0.1076 (27) 0.2258 (27) 0.3613 (27)
PT,Loss (kW) 9.77 39.69 104.04
VWorst in p.u. 0.9881 0.9763 0.9614
(Bus no) (61) (61) (61)
% Loss reduction 81.06 82.35 84.05
Simultaneous reconfiguration and DG installation (scenario VI) Switches opened 69-70-13-56-63 69-70-13-55-63 69-70-13-57-63
DG size in MW (candidate bus) 0.5715 (61) 1.1272 (61) 1.8176 (61)
0.1555 (62) 0.2750 (62) 0.5095 (62)
0.2129 (64) 0.4159 (65) 0.6342 (65)
PT,Loss (kW) 9.58 39.25 102.97
VWorst in p.u. 0.9905 0.9796 0.9685
(Bus no) (61) (61) (61)
% Loss reduction 81.43 82.55 84.21

The bold value defines the significance of proposed method in terms of power loss minimization and Voltage profile improvement.
468 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472

1
0.995

Voltage Magnitude (p.u)

0.99
0.985
0.98
0.975 Scenario I
Scenario II
0.97 Scenario III
Scenario IV
0.965
Scenario V
0.96 Scenario VI

0.955
0 10 20 30 40 50 60 70
Bus No

Fig. 6. Comparison of voltage profiles of 69 bus system at light load conditions.

1
0.99
0.98
Voltage Magnitude (p.u)

0.97
0.96
0.95
0.94 Scenario I
Scenario II
0.93 Scenario III
Scenario IV
0.92
Scenario V
0.91 Scenario VI

0.9
0 10 20 30 40 50 60 70
Bus No

Fig. 7. Comparison of voltage profiles of 69 bus system at nominal load conditions.

1.02

0.98
Voltage Magnitude (p.u)

0.96

0.94
Scenario I
0.92
Scenario II

0.9 Scenario III

Scenario IV
0.88 Scenario V
Scenario VI
0.86

0.84
0 10 20 30 40 50 60 70
Bus No

Fig. 8. Comparison of voltage profiles of 69 bus system at heavy load conditions.

Scenario VI: System with simultaneous optimal network recon-
figuration and DG allocation.
An analytical software tool has been developed in MATLAB
environment for generating parent node child node path, to carry
out power flow, calculate voltages and power losses.
33-Bus test system
The line, load and tie line data of this test system are taken from
[24]. It consists of five tie lines and 32 sectionalize switches. The
normally closed sectionalize switches are 1–32 and the open tie
switches are 33–37. The single line diagram of 33-bus system with
the loops (LP1 to LP5) formed corresponding to each tie switch is
shown in Fig. 4. The total real and reactive power loads of the sys-
tem are 3.72 MW and 2.3 MVAr, respectively.
VSI for all nodes of the system is computed from power flow to
identify the candidate bus location of DGs for scenarios III, IV and
VI. After computing VSI at all buses, they are sorted in ascending
order and ranked. Then the top three locations which are more
sensitive are selected to install DG units in the system. The candi-
date locations selected for scenario III, IV and VI are given in
 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472 469

Table 2. The performance of the proposed method at three different
load levels is presented in Table 2.
From Table 2, it is observed that at nominal load, base case
power loss (kW) in the system is 202.67, which is reduced to
139.98, 88.68, 83.91, 68.28 and 67.11 using scenarios II, III, IV, V
and VI, respectively. The percentage power loss reduction for sce-
nario II to VI is 30.93, 56.24, 58.59, 66.31 and 66.89, respectively.
Similarly, at light and heavy load conditions, the percentage power
loss reduction for scenario II to VI is 29.04, 54.58, 56.82, 64.81 and
65.53; 33.73, 58.57, 60.90, 68.49 and 69.93 respectively. It is also
observed from Table 2 that, at all load levels, the minimum voltage
magnitude of the system is improved impressively in all the sce-
narios. At light, nominal and heavy loading conditions, the mini-
mum voltage magnitude (p.u.) is improved from 0.9583, 0.9131
and 0.8529 to 0.9862, 0.9713 and 0.9554 in scenario VI. It is
observed that at all the three loading conditions, the power loss
reduced and minimum voltage improved using scenario VI is the
highest, which proves the superiority of the proposed technique.
However, at all the three loading conditions,
the improvement in percentage power loss reduction and min-
imum voltage magnitude is almost the same in all the scenarios.
Scenario V reduces the power loss almost equal to scenario VI
but the size of DGs installed is higher than scenario VI. This reveals
that the simultaneous network reconfiguration and allocation of

12
Power Loss (kW)

10
8
6
4
2
0
Scenario VI
Scenario V 70
60
Scenario IV 50
Scenario III 40
30
Scenario II 20
10
Scenario I 0
Scenario Line No

Fig. 9. Comparison of power loss at each line of 69 bus system at light load conditions.

50
Power Loss (kW)

40

30

20

10

0
Scenario VI
Scenario V
60 70
Scenario IV 50
Scenario III 30 40
Scenario II 10 20
Scenario I 0
Scenario Line No

Fig. 10. Comparison of power loss at each line of 69 bus system at full load conditions.
Power Loss (kW)

150

100

50

0
Scenario VI
Scenario V 70
Scenario IV 50 60
Scenario III 30 40
Scenario II 10 20
Scenario I 0 Line No
Scenario

Fig. 11. Comparison of power loss at each line of 69 bus system at heavy load conditions.
470 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472

Table 5
Comparison of performances for 69-bus system.

Scenario Item FWA HSA [16] GA [16] RGA [16]

Scenario II Switches opened 69–70–14–56–61 69–18–13–56–61 69–70–14–53–61 69–17–13–55–61
PT,Loss (kW) 98.59 99.35 103.29 100.28
% Loss reduced 56.17 55.85 54.08 55.42
VWorst (p.u) 0.9495 0.9428 0.9411 0.9428
Scenario III DG size (MW) 1.8329 1.7732 1.9471 1.7868
PT,Loss (kW) 77.85 86.77 88.5 87.65
% Loss reduced 65.39 61.43 60.66 61.04
VWorst (p.u) 0.9740 0.9677 0.9687 0.9678
Scenario IV Switches opened 69–70–14–56–61 69–18–13–56–61 69–70–14–53–61 69–17–13–55–61
DG size (MW) 1.3584 1.8448 1.7422 1.6396
PT,Loss (kW) 43.88 51.30 54.53 52.34
% Loss reduced 80.49 77.20 75.76 76.73
VWorst (p.u) 0.9720 0.9619 0.9401 0.9611
Scenario VI Switches opened 69–70–13–55–63 69–17–13–58–61 10–15–45–55–62 10–16–14–55–62
DG size (MW) 1.8182 1.8718 2.0292 2.0654
PT,Loss (kW) 39.25 40.30 46.20 44.23
% Loss reduced 82.55 82.08 73.38 80.32
VWorst (p.u) 0.9796 0.9736 0.9727 0.9742

DG units (scenario VI) is better than other scenarios considered in each tie switch is shown in Fig. 5. The total real and reactive power
terms of quality of solutions. loads of the system are 3.80 MW and 2.69 MVAr, respectively.
In order to illustrate the performance of the proposed method, Similar to 33 bus test system, this test system is also simulated
the performance of FWA is compared with the results of HSA for six scenarios at three load levels and the results are presented
[16], GA [16] and Refined Genetic Algorithm (RGA) [16] available in Table 4. The total power loss (kW) at light, nominal and heavy
in the literature and is presented in Table 3. From the table, it is load conditions is reduced from 51.60, 224.96 and 652.42 to
perceived that at all scenarios, the performance of the FWA is bet- 9.58, 39.25 and 102.97 in scenario VI respectively. Table 2 and
ter than HSA, GA and RGA in terms of power loss minimization and Table 4 show the capability and effectiveness of the proposed
voltage stability enhancement. method in finding the optimal solutions for all the scenarios con-
sidered. From Table 4, it is also observed that scenario VI is more
69-Bus test system effective in reducing total power loss and improving minimum
voltage compared to other scenarios.
This is a large scale radial distribution system with 68 section- The voltage profiles of all six scenarios at light, nominal and
alizing switches (1–68) and five tie switches (69–73). The configu- heavy load conditions are compared and shown in Figs. 6–8,
ration, line, load and tie line data are taken from [25].The single respectively. From these figures, it is seen that the improvement
line diagram with the loops (LP1 to LP5) formed corresponding to in voltage profile of the system in scenario VI is better at all load
82.55

82.08
80.49

90
80.32
76.73
75.76
77.2

73.38

80
Percentage Power Loss Reduced (%)

65.39

61.43

61.04
60.66

70
56.17

55.85

55.42
54.08

60

50

40

30

20

10

0
Scenario II Scenario III Scenario IV Scenario VI
Scenarios FWA HSA [16] GA [16] RGA [16]

Fig. 12. Comparison of percentage power loss reduced by FWA with other classical techniques.
 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472
0.07
0.0589
0.0572
0.0572
0.06
0.0505
Voltage Deviation Index ( V D)
0.05
0.04
0.026
0.03
0.02
0.01
0
Scenario II
Scenarios
Fig. 13. Comparison of voltage deviation index reduced by FWA with other classical techniques.
levels. It is also observed that the fall in minimum voltage
magnitude is least with increase of load from light to heavy in
scenario VI.
The power loss in each line for all scenarios at three different
loading conditions are shown in Figs. 9–11, respectively. It is
observed that the power loss at each line has been reduced drasti-
cally with the increase of load from light to heavy in scenario VI
compared with other scenarios.
The performance of FWA on 69-bus system is also compared
with the results of other classical methods similar to 33-bus sys-
tem, and the results are presented in Table 5. For effective compar-
ison, the percentage power loss and voltage deviation index
reduced by FWA are compared graphically as shown in Figs. 12
and 13. From Table 5 and Figs. 12, 13, it is clear that the FWA
has outperformed HSA, a standard real-coded GA and RGA in terms
of power loss minimization and voltage stability enhancement in
all the scenarios.
Conclusion
In this paper, the application of integrating the network recon-
figuration and DG placement for the problem of minimization of
power loss and voltage deviation using FWA has been detailed
and investigated. Proposed method maintains the radial nature of
the network and proper current flow direction through-over all
reconfiguration phases, by generating proper parent node child
node path and forming BIBC matrix during power flow. The pro-
posed method is tested on 33- and 69-bus test systems at three dif-
ferent load levels. The results obtained clearly indicates that
scenario VI (network reconfiguration with simultaneous DG instal-
lation) is found to be more effective in minimizing the power loss
and improving the voltage profile compared to the other scenarios
considered. However, the proposed method is very efficient in find-
ing the optimal solution for all the scenarios. The simulated results
are compared with the results of HSA, GA and RGA available in the
literature. The computational results showed that the performance
of the FWA is better than HSA, GA and RGA in all scenarios. The
proposed method can be easily applied and adapted to any large
scale radial distribution networks.
471
0.0599
0.0389
0.0381
0.0323
0.0322
0.0313
0.0273
0.0264
0.028
0.0258
0.0204
Scenario III Scenario IV Scenario VI
FWA HSA[16] GA [16] RGA [16]
Acknowledgement
This research work was supported by Vellore Institute of Tech-
nology University, Vellore, India.
References
[1] Ng H, Salama M, Chikhani A. Classification of capacitor allocation techniques.
IEEE Trans Power Deliver 2000;15(1):387–92.
[2] Kalambe S, Agnihorti G. Loss minimization techniques used in distribution
network: bibliographical survey. Renew Sustain Energy Rev 2014;29:184–200.
[3] Merlin A, Back H. Search for a minimal loss operating spanning tree
configuration in an urban power distribution system. In: Proceeding in 5th
power system computation conf. (PSCC), Cambridge, U.K.; 1975:1–18.
[4] Aman MM, Jamson GB, Bakar AHA, Mokhili H. Optimum network
reconfiguration based on maximization of system loadability using
continuation power flow theorem. Int J Electr Power Energy Syst
2014;54:123–33.
[5] Sathish kumar K, Jayabharathi T. Power system reconfiguration and loss
minimization for an distribution systems using bacterial foraging optimization
algorithm. Int J Electr Power Energy Syst 2012;36:13–7.
[6] Niknam Taher, Fard Abdollah Kavousi, Seifi Alireza. Distribution feeder
reconfiguration considering fuel cell/wind/photovoltaic power plants. Renew
Energy 2012;37:213–25.
[7] Wang Chun, Gao Yuanhai. Determination of power distribution network
configuration using non-revisiting genetic algorithm. IEEE Trans Power Syst
2013;28(4):3638–48.
[8] Acharya Naresh, Mahat Pukar, Mithulananthan N. An analytical approach for
DG allocation in primary distribution network. Int J Electr Power Energy Syst
2006;28:669–78.
[9] Hung DQ, Mithulananthan N. Multiple distributed generators placement in
primary distribution networks for loss reduction. IEEE Trans Ind Elect
2013;60(4):1700–8.
[10] Kansal Satish, Kumar Vishal, Tyagi Barjeev. Optimal placement of different
type of DG sources in distribution networks. Int J Electr Power Energy Syst
2013;53:752–60.
[11] García Juan Andrés Martín, Mena Antonio José Gil. Optimal distributed
generation location and size using a modified teaching–learning based
optimization algorithm. Int J Electr Power Energy Syst 2013;50:65–75.
[12] Tan Wen Shan, Hassan Mohammad Yusri, Rahman Hasimah Abdul, Abdullah
Md Pauzi, Hussin Faridah. Multi-distributed generation planning using hybrid
particle swarm optimisation-gravitational search algorithm including voltage
rise issue. IET Gener Transm Distrib 2013;7(3):929–42.
[13] Mohamed Imran A, Kowsalya M. Optimal size and siting of multiple
distributed generators in distribution system using bacterial foraging
optimization. Swarm Evol Comput 2014;15:58–65.
[14] Tan Sicong, Xu Jian-Xin, Panda Sanjib Kumar. Optimization of distribution
network incorporating distributed generators: an integrated approach. IEEE
Trans Power Syst 2013;28(3):2421–32.
472 A. Mohamed Imran et al. / Electrical Power and Energy Systems 63 (2014) 461–472
[15] Georgilakis Pavlos S, Hatziargyriou Nikos D. Optimal distributed generation
placement in power distributed networks: models, methods, and future
research. IEEE Trans Power Syst 2013;28(3):3420–8.
[16] Srinivasa Rao R, Ravindra K, Satisk K, Narasimham SVL. Power loss
minimization in distribution system using network reconfiguration in the
presence of distributed generation. IEEE Trans Power Syst 2013;28(1):317–25.
[17] Thukaram D, Wijekoon Banda HM, Jerome Jovitha. A robust three phase power
flow algorithm for radial distribution systems. Elect Power Syst Res
1999;50:227–36.
[18] Teng Jen-Hao. A direct approach for distribution system load flow solutions.
IEEE Trans Power Deliver 2003;18(3):882–7.
[19] Chakravorty M, Das D. Voltage stability analysis of radial distribution
networks. Int J Electr Power Energy Syst 2001;23:129–35.
[20] Tan Ying, Zhu Yuanchun. Fireworks algorithm for optimization. Adv Swarm
Intell 2010;6145:355–64.
[21] Zheng Yu-Jun, Song Qin, Chen Sheng-Yong. Multiobjective fireworks
optimization for variable-rate fertilization in oil crop production. Appl Soft
Comput 2013;13:4253–63.
[22] Zhu JZ. Optimal reconfiguration of electrical distribution network using the
refined genetic algorithm. Elect Power Syst Res 2002;62:37–42.
[23] Mistry Khyati D, Roy Ranjit. Enhancement of loading capacity of distribution
system through distributed generator placement considering techno-
economic benefits with load growth. Int J Electr Power Energy Syst.
2014;54:505–515.
[24] Baran ME, Wu FF. Network reconfiguration in distribution systems for loss
reduction and load balancing. IEEE Trans Power Deliver 1989;4(2):1401–9.
[25] Savier JS, Das D. Impact of network reconfiguration on loss allocation of radial
distribution systems. IEEE Trans Power Deliver 2007;22(4):2473–80.