Electrical Power and Energy Systems 77 (2016) 314–326

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Optimum placement of shunt capacitors in a radial distribution system

for substation power factor improvement using fuzzy GA method
Srinivasa Rao Gampa, D. Das ⇑
Department of Electrical Engineering, Indian Institute of Technology, Kharagpur 721302, India

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

Article history: In this work a combination of fuzzy multiobjective and genetic algorithm (GA) based approach is pro-
Received 20 May 2014 posed for optimal shunt capacitor placement to improve the substation power factor near unity, reduce
Received in revised form 21 July 2015 the real power loss, and reduce the burden on the substation and to improve the voltage profile of the
Accepted 17 November 2015
distribution network. In order to obtain best nodes for capacitor placement, a sensitivity index based
on real power loss reduction and voltage profile improvement is considered. In the present work, an
attempt is made to make reactive current component drawn by distribution network through substation
Keywords:
is nearly zero such that power factor at the substation will be near unity. A fuzzy multiobjective function
Shunt capacitors
Fuzzy multiobjective
is formed considering substation reactive current component reduction, real power loss reduction,
Genetic algorithm branch current constraint limit, minimum and maximum voltage limit satisfaction. The fuzzy multiobjec-
Power factor improvement tive function is maximized using GA for obtaining the optimum sizing of fixed and switched shunt capac-
Energy loss reduction itors. Simulation results are shown to demonstrate the advantage of the proposed method compared to
optimal shunt capacitor placement based on annual energy savings method.
Ó 2015 Elsevier Ltd. All rights reserved.

Introduction
Shunt capacitors usage is a common practice to supply reactive
power in the distribution networks. Installation of shunt capacitors
reduces power losses, improves the power factor and feeder volt-
age profile. Therefore it is essential to find optimal location and
sizes of capacitors for gaining maximum benefits by the shunt
capacitor installation. Ng et al. [1] classified the capacitor place-
ment techniques available in the literature into four categories.
They are analytical, numerical programming, heuristic and artifi-
cial intelligence based techniques. In the early literature many
researchers proposed calculus based analytical techniques for
capacitor placement solution [2–6]. Initially, methods are devel-
oped based on the assumption of uniform conductor size and load-
ing throughout the feeder and later they are extended for different
conductor sizes and nonuniform loading. With the development of
powerful computational techniques numerical methods were
developed for the selection of optimal nodes and sizing of capaci-
tors considering all the operational constraints [7–13]. In order to
reduce the large search space required by the numerical methods
many authors have proposed heuristic methods to obtain nearly
optimal solution for capacitor placement problem [14–18]. For
⇑ Corresponding author.
E-mail address: ddas@ee.iitkgp.ernet.in (D. Das).
the last two decades artificial intelligence techniques are widely
used for identifying optimal locations and sizing of capacitors,
out of them fuzzy logic, genetic algorithms are more popular. Ng
et al. [19], Mekhamer et al. [20], Masoum et al. [21], Shi and Liu
[22] and Bhattacharya and Goswami [23] used fuzzy set theory
based on human experience and intuition for identification of opti-
mal nodes for capacitor placement and for optimal capacitor sizing,
they have used numerical methods and heuristic methods. Boone
and Chiang [24], Iba [25], Sundharajan and Pahwa [26], Delfanti
et al. [27], Das [28], Malik et al. [29] and Swarnakar et al. [30]
applied genetic algorithms for optimal allocation of capacitors.
Gallego et al. [31], De Souza et al. [32], Hsiao et al. [33], Das [34],
and Abul’Wafa [35] have used combination of fuzzy logic and
genetic algorithms for optimal allocation of capacitors in distribu-
tion systems. In the recent years many nature inspired based arti-
ficial intelligence techniques are developed and used for optimal
capacitor allocation problem. Huang et al. [36] have proposed
two stage immune algorithm based multi objective optimization
approach for solving shunt capacitor placement problem. Rao
[37] has used fuzzy and plant growth simulation algorithm to
determine optimal locations and size of the capacitor to improve
the voltage profile and reduce the active power losses. Singh and
Rao [38] have proposed a particle swarm optimization based
algorithm for obtaining the optimal size and locations of the capac-
itors utilizing the daily load curve. Ziari et al. [39] have proposed a

http://dx.doi.org/10.1016/j.ijepes.2015.11.056
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 315

Nomenclature

Iac
i current magnitude through ith branch after capacitor Ke energy cost in $/kW h
placement CC cost of capacitor in $/kVAr
IC i maximum current carrying capacity of ith branch CI capacitor installation cost in $/location
NB total number of nodes of the distribution system CRF capital recovery factor
F fuzzy multiobjective function NL total number of load levels
P Si power injected at substation before capacitor placement NCL total number of capacitor installation locations
at ith load level Ti total number of hours of ith load level duration in one
PQC
Si power injected at substation after capacitor placement year
at ith load level SCP annual economical savings due to shunt capacitor
QC j total capacity of shunt capacitors installed at jth rank placement
node

modified discrete particle swarm optimization technique for the Sensitivity analysis for the placement of capacitors
determination of rating and locations of fixed and switched capac-
itors. El-Fergany et al. [40] have proposed artificial bee colony In the present analysis, a new sensitivity index is used for the
based approach for allocation of static capacitors along the radial placement of capacitors in a distribution network. In the sensitivity
feeders of the distribution system. Hung et al. [41] classified analysis the load flow solutions are obtained by compensating the
devices capable of supplying reactive power only as type-2 kind total reactive load at every node and from the load flow solutions
of distributed generation. Aman et al. [42] presented a literature the active power loss reduction and maximum node voltage incre-
survey on optimal shunt capacitor placement. ment are calculated for all nodes of distribution system.
From the above literature survey, it can be seen that the Let us define a sensitivity index,
capacitor placement problem is aimed to obtain optimal locations
Sk ¼ LSIk  VSIk ð1Þ
and sizing of capacitors to reduce the active power losses, to
increase feeder capacity and to improve the voltage profile of where
the radial distribution system. In the present work a new sensi- Sk = Sensitivity index of kth node due to total reactive load
tivity analysis is proposed to find suitable nodes for the capacitor compensation.
placement problem. For obtaining optimum sizing of capacitors at LSIk = Loss sensitivity index of kth node due to total reactive
the optimum locations identified the objectives of minimizing load compensation.
reactive current drawn by the distribution network from the grid VSIk = Voltage sensitivity index of kth node due to total reactive
through substation, real power loss reduction, feeder capacity load compensation.
improvement and voltage profile improvement are considered
to improve the substation power factor, nearer to unity. Since The loss sensitivity index can be expressed mathematically as
all the objectives considered are non-commensurable in nature, follows:
the conventional approaches that optimizes a single objective
PLRk  PLRmin
function are not suitable for this problem. Therefore the fuzzy LSIk ¼ ð2Þ
approach is adopted for considering all the multiple objectives PLRmax  PLRmin
simultaneously.
The conventional calculus based techniques mainly depends PLRk ¼ PLB  PLk ð3Þ
upon the existence of derivatives of the single objective function where
and may lead toward the local optimum solution. The actual PLB = Real power loss without shunt capacitor compensation.
search space in majority of the practical cases is associated with PLk = Real power loss due to total reactive load compensation at
many discontinuous functions and hence the conventional opti- kth node.
mization methods are not suitable for finding optimal solutions PLRk = Real power loss reduction due to total reactive load com-
for non differentiable multiobjective functions. Genetic Algo- pensation at kth node.
rithms (GA) are probabilistic search techniques based on natural PLRmin = Minimum Real power loss reduction.
selection and genetics. The advantage of GA compared to other PLRmax = Maximum Real power loss reduction.
conventional techniques is it works with binary coding of param-
eters rather than parameters themselves and hence it is more The voltage sensitivity index can be expressed mathematically
efficient optimization technique for non differentiable multiobjec- as follows:
tive functions consisting both discrete and continuous variables.
GA proceeds in the direction of maximization of fitness function DVMk  DVMmin
VSIk ¼ ð4Þ
in the selected multi dimensional search space and hence the DVMmax  DVMmin
solution proceeds toward global optimum. Hence for the present e ac  V
DVMk ¼ max j V e bc j ð5Þ
i;k i;k
capacitor placement problem a genetic algorithm based fuzzy
multiobjective approach is used for optimum sizing of fixed and where
switched capacitors. DVMk = Maximum node voltage increment due to total reactive
The main motivation of this work is that the distribution net- load compensation at kth node.
work should not draw reactive power from the grid. Shunt capac- DVMmin = Minimum of Maximum node voltage improvement.
itors must meet the total reactive power load demand and reactive DVMmax = Maximum of Maximum node voltage improvement.
power loss of the distribution network. Therefore, in this paper, the e bc = Voltage at ith node before capacitor placement at kth
V i;k
approach is to make the substation power factor near unity, i.e. node.
drawing negligible reactive power from the grid. e ac = Voltage at ith node after capacitor placement at kth node.
Vi;k
316 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326

Significance of Eq. (1) is that due to reactive power compensa-
tion in the distribution network it is expected that the voltage pro-
file will improve and the loss will decrease. The loss sensitivity
values are normalized in between 0 and 1 considering the highest
loss reduction value as 1 and smallest loss reduction value to zero
[35,37]. Similarly the maximum node voltage increment values are
also normalized in between zero and one. The node corresponding
to the highest product of loss sensitivity index and voltage sensi-
tivity index will have highest sensitivity index and it is best suit-
able for capacitor placement.
In the present work two case studies, i.e., an 11 kV, 51 node
radial distribution network and a 12.66 kV, 69 node distribution
network are considered and the single line diagrams are shown
in Figs. 1 and 2 respectively. The line data and load data for the
51 node distribution system are given in Appendix A and for 69
distribution network the data are taken from [43]. The substation
(s/s) voltage is V e 1 ¼ ð1 þ j0:0Þ pu. Ranking of nodes is made
according to the descending order values of the sensitivity index
Sk calculated using Eq. (1) for optimal allocation of capacitors.
The ranking of the nodes for 51 node distribution system is
given in Table 1, and for 69 node distribution system the order of
ranking is given in Table 2. The plots for sensitivity index (Sk) val-
ues vs node number are shown in Figs. 3 and 4 for 51 node distri-
bution system and for 69 node distribution system respectively.
The nodes with top five ranks are selected as best locations for
capacitor placement for the both the examples considered for
capacitor placement.
Formation of fuzzy multiobjective function
In the present section fuzzy membership functions [44] are
developed for substation reactive current component, real power
loss reduction, branch current capacity, minimum voltage limit
and maximum voltage limit of the distribution system.

Table 1
Ranking of nodes of 51 node distribution system for capacitor placement.

Order of the Node Order of the Node Order of the Node

ranking of number ranking of number ranking of number
nodes nodes nodes
1 15 18 41 35 35
2 14 19 39 36 36
3 11 20 08 37 32
4 49 21 46 38 24
5 40 22 23 39 18
6 44 23 50 40 25
7 48 24 19 41 31
8 16 25 33 42 06
9 51 26 37 43 07
10 10 27 13 44 17
11 09 28 05 45 03
12 34 29 22 46 12
13 43 30 29 47 28
14 42 31 47 48 27
15 38 32 20 49 02
16 45 33 21 50 04
17 26 34 30 51 01
Fig. 1. 51 node distribution system network.

Table 2
Ranking of nodes of 69 node distribution system for capacitor placement.

Order of the Node Order of the Node Order of the Node

ranking of number ranking of number ranking of number
nodes nodes nodes
1 61 24 54 47 28
2 64 25 09 48 41
3 59 26 67 49 63
4 21 27 66 50 60
5 65 28 14 51 58
6 12 29 13 52 57
7 11 30 22 53 56
8 50 31 46 54 47
9 62 32 45 55 44
10 49 33 53 56 42
11 18 34 48 57 38
12 17 35 52 58 32
13 16 36 34 59 31
14 24 37 20 60 30
15 08 38 06 61 25
16 69 39 33 62 23
17 68 40 40 63 19
18 27 41 39 64 15
19 26 42 35 65 05
20 51 43 43 66 04
21 55 44 37 67 03
22 10 45 29 68 02
23 07 46 36 69 01
Fig. 2. 69 node distribution system network.
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 317

value is assigned. If q is less than or equal to qmin then unity mem-

bership value is assigned and if q is greater than qmax, zero mem-
bership value is assigned.
From Fig. 5, lIQ can be written as:
8
>
<1 for q 6 qmin
lIQ ¼ ðqmax qÞ
ðqmax qmin Þ
for qmin < q 6 qmax ð7Þ
>
:
0 for q > qmax
In this work with the proposed method for shunt capacitors
placement it is aimed at reducing the reactive current component
at the substation to 5% of the base case value. For the values of q
less than or equal to 0.05 the fuzzy membership values are given
a value of unity and for the case of reactive current component
greater than or equal to base case value the fuzzy membership val-
Fig. 3. Plot of sensitivity index for capacitor placement of 51 bus system.
ues are considered as zero. Hence the values for qmin and qmax are
taken as 0.05 and 1.0 respectively. When the value of q is less than
qmax and greater than qmin a fuzzy membership value in between
zero and one is assigned.

Membership function for real power loss reduction (lPL)

The objective of this membership function developed is to

reduce the real power loss due to the compensation of reactive
power using shunt capacitors.
Let us define the real power loss index ‘x’ as,
ac
Ploss
x¼ bc
ð8Þ
Ploss
where
Plossbc = Real power loss of the distribution network before
Fig. 4. Plot of sensitivity index for capacitor placement of 69 bus system.
capacitor placement.
Plossac = Real power loss of the distribution network after capac-
Membership function for substation reactive current component itor placement.
reduction (lIQ)
Fig. 6 shows the membership function for real power loss
The main purpose of this membership function is to reduce sub- reduction (lPL). Eq. (8) indicates that for higher the values of x
station reactive current component and to improve the power fac- the membership value will be low because real power loss reduc-
tor at the substation of the distribution system. tion is low and if x has a lower value then the membership value
Let us define, will be high because real power loss reduction is high.
From Fig. 6, lPL can be written as:
Iac
q 8
q¼ ð6Þ >
<1 for x 6 xmin
Ibc
q
lPL ¼ ðxmax xÞ
ðxmax xmin Þ
for xmin < x 6 xmax ð9Þ
>
:
where 0 for x > xmax
q = Substation Reactive current component index.
In the present work with the proposed method for shunt capac-
Ibc
q = Reactive current component at substation before capacitor itors placement it is aimed at reducing the real power loss to 40% of
placement. the base case value. For the values of x less than or equal to 0.4 the
Iac
q = Reactive current component at substation after capacitor fuzzy membership values are given a value of unity and for the
placement. case of zero loss reduction and losses greater than base case value
the fuzzy membership values for real power loss reduction are con-
Fig. 5 shows the fuzzy membership function for substation reac- sidered as zero. Hence the values for xmin and xmax are taken as 0.4
tive current component reduction (lIQ). When the value of q
exceeds qmin and less than or equals to qmax, lower membership

Fig. 5. Membership function for substation reactive current component. Fig. 6. Membership function for active power loss reduction.
318 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326

and 1.0 respectively. When the value of x is less than xmax and 1.0 respectively. When the value of r is less than rmax and greater
greater than xmin a fuzzy membership value in between zero and than rmin a fuzzy membership value in between zero and one is
one is assigned. assigned.

Membership functions for minimum and maximum voltage limits

Membership function for current carrying capacity of branch (lR)
The main purpose of this membership function is to avoid the
branch current constraint violation.
Let us define,
The placement of shunt capacitors is aimed to maintain all the
node voltages to be in between specified minimum and maximum
voltage limits.
For this purpose two membership functions are used:

Iac Membership function for minimum voltage limit

Branch current capacity index ¼ i
;
IC i Fig. 8 shows the membership function for minimum voltage
for i ¼ 1; 2; . . . ; NB  1 ð10Þ limit (lV min ). If the minimum system voltage is less than ymin or
greater than ymax then zero membership value is assigned. If it is
Now we define, in between ymin and yp lower membership value is assigned and
  if it is lying in between yp and ymax, unity membership value is
Iac assigned.
r ¼ max i
; for i ¼ 1; 2; 3; . . . ; NB  1 ð11Þ
IC i From Fig. 8, we can write,
8
Fig. 7 shows the fuzzy membership function for branch current >
> 0 for y 6 ymin
carrying capacity (lR). When the value of r exceeds rmin and less > ðyy Þ
>
< min
for ymin < y < yp
than or equals to rmax, lower membership value is assigned. If r is lV min ¼ ðyp ymin Þ
ð13Þ
>
> 1 for yp 6 y 6 ymax
less than or equal to rmin then unity membership value is assigned >
>
:
and if r is greater than rmax, zero membership value is assigned. 0 for y > ymax
From Fig. 7, lR can be written as:
In the present work with the proposed method for shunt capac-
8 itors placement it is aimed at improving the minimum voltage to
>
<1 for r 6 rmin
0.95 pu value. For y values lies in between 0.95 and 1.0 the fuzzy
lR ¼ ðr max rÞ
ðr max r min Þ
for r min < r 6 r max ð12Þ membership values are given a value of unity and for the case of
>
:
0 for r > r max y values less than 0.925 and greater than 1.0 the fuzzy membership
values are considered as zero. For y values greater than 0.925 and
In the present work with the proposed method for shunt capac-
less than 0.95, membership values in between zero and one are
itor placement it is aimed at reducing maximum branch current to
assigned. Hence in the present work ymin ¼ 0:925; yp ¼ 0:95 and
40% of the base case value. For the values of r less than or equals to
ymax = 1.0 are considered for the fuzzy membership function devel-
0.4 the fuzzy membership values are given a value of unity and for
oped for minimum voltage limit.
the case of maximum branch current values greater than or equal
to base case values the fuzzy membership values are considered as
Membership function for maximum voltage limit
zero. Hence in the present work rmin and rmax are taken as 0.4 and
Fig. 9 shows the membership function for maximum voltage
limit (lV max ). If the maximum system voltage is less than zmin or
greater than zmax then zero membership value is assigned. If it is
in between zmin and zp unity membership value is assigned and if

Fig. 7. Membership function for maximum branch current carrying capacity index.

Fig. 9. Membership function for maximum voltage limit.

Table 3
Annual load duration schedule.

Load level No of hrs/yr

0.4 1000
0.5 1000
0.6 1000
0.7 1000
0.8 1000
0.9 1000
1.0 2760
Fig. 8. Membership function for minimum voltage limit.
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 319

it is lying in between zp and zmax lower membership value is

assigned.
From Fig. 9, we can write,
8
>
> 0 for z < zmin
>
>
<1 for zmin 6 z 6 zp
lV max ¼ ðzmax zÞ ð14Þ
>
> for zp < z < zmax
>
> ðzmax zp Þ
:
0 for z P zmax
In the present work with the proposed method for shunt capac-
itor placement it is aimed at limiting the maximum voltage to
1.025 pu value. For z values lies in between 1.0 and 1.025 the fuzzy
Fig. 10. Load duration curve for 51 node distribution system.
membership values are given a value of unity and for the case of

Fig. 11. Flow chart for optimal shunt capacitor sizing using GA.
320 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326

z values greater than 1.05 and less than 1.0 the fuzzy membership obtained for all load levels upto the peak load level successively
values are considered as zero. For z values greater than 1.025 and by incorporating the optimally sized shunt capacitors upto previ-
less than 1.05, membership values in between zero and one are ous load level. Fig. 11 shows the flow diagram for optimal sizing
assigned. For z values greater than 1.025 and less than 1.05, mem- of shunt capacitors using GA.
bership values in between one and zero are assigned. Hence in the The optimum sizing of shunt capacitor values obtained using
present work zmin = 1.0, zp = 1.025 and zmax = 1.05 are considered GA are real values and they are adjusted to the nearest integer val-
for the fuzzy membership function developed for maximum ues of capacitances available in the market for incorporation into
voltage limit. the distribution system. The fixed capacitors and switched capaci-
The above membership functions developed for each objective tors obtained for all load levels are shown in Table 4 for 51 node
function above are added to form a single objective function distribution node network and in Table 6 for 69 node distribution
through weighting factors and is given by network. The total optimal shunt capacitors sizing obtained using
Maximize GA for performance improvement of the distribution system for
all load levels is shown in Table 5 for 51 node distribution system
F ¼ w1 lIQ þ w2 lPL þ w3 lR þ w4 lV min þ w5 lV max ð15Þ
and in Table 7 for 69 node distribution system. The GA parameters,
i.e., no of generations, population size, crossover rate, mutation
In Eq. (15), we have assumed that all the objectives have equal
rate used for obtaining optimum sizing of capacitor placement
importance and w1 = w2 = w3 = w4 = w5 = 1.0 is assumed.
are shown in Table 8.
The computational efficiency of proposed fuzzy GA methodol-
Optimal sizing of shunt capacitors using GA ogy is compared with energy savings method proposed in [30].
The algorithms are developed for both the methodologies using
In the present work binary coded Genetic algorithm (GA) [45] is MATLAB 2007a software and executed on a Pentium dual core
used for optimum sizing of shunt capacitors. In GA fitness function CPU with 3.2 GHz speed and 2 GB RAM computer. The computa-
is needed to be maximized for obtaining the optimum values of tional time for 51 node distribution system with energy savings
parameters of the objective function. Here the fuzzy multiobjective method [30] is 51 seconds and with proposed method is 35 sec-
function ‘F’ described by Eq. (15) is considered as fitness function onds. The computational time for 69 node distribution system with
for optimum sizing of capacitors using GA. Optimum shunt capac- energy savings method [30] is 89 s and with proposed method is
itor sizing is done in a step by step process starting from lowest 39 s. Hence it can be said that the proposed method is computa-
load level. There are seven load levels varying from 0.4 to 1.0 in tionally efficient compared to GA based energy savings method
a step of 0.1 are considered for the present analysis. The load dura- proposed in [30].
tion time for the corresponding load levels considered for the 51
node distribution system is shown in Table 3. The approximated
load duration curve is shown in Fig. 10.
First for 0.4 load level the optimum sizing of shunt capacitors is Table 6
done and they are used as fixed capacitors. The shunt capacitors Optimal capacitor sizing obtained using GA for 69 node distribution system.
obtained at 0.4 load level are incorporated into the distribution Type of capacitors Load level Reactive power of the capacity of
system and the additional switched capacitors required for 0.5 capacitors installed (kVAr)
level are optimized again using GA. Similarly the additional QC61 QC64 QC59 QC21 QC65
switched capacitors required for the remaining load levels are
Fixed capacitors 0.4 350 50 300 275 25
Switched capacitors 0.5 100 25 75 50 0
0.6 50 25 150 75 0
Table 4
0.7 125 0 25 75 0
Optimal capacitor sizing obtained using GA for 51 node distribution network.
0.8 100 0 100 75 25
Type of capacitors Load level Reactive power of the capacity of 0.9 100 150 0 0 100
capacitors installed (kVAr) 1.0 75 75 75 0 75

QC15 QC14 QC11 QC49 QC40

Fixed capacitors 0.4 25 50 125 125 300
Table 7
Switched capacitors 0.5 25 0 0 50 100
Optimal sizing of total capacitors installed for 69 node distribution system.
0.6 0 0 25 25 100
0.7 25 0 25 25 75 Load level Total available reactive power capacity of the capacitors
0.8 0 25 25 50 75 installed (kVAr)
0.9 0 0 50 25 100
1.0 75 0 75 0 0 QC61 QC64 QC59 QC21 QC65
0.4 350 50 300 275 25
0.5 450 75 375 325 25
0.6 500 100 525 400 25
Table 5 0.7 625 100 550 475 25
Optimal sizing of total capacitors installed for 51 node distribution network. 0.8 725 100 650 550 50
0.9 825 250 650 550 150
Load level Total available reactive power capacity of the capacitors 1.0 900 325 725 550 225
installed (kVAr)
QC15 QC14 QC11 QC49 QC40
0.4 25 50 125 125 300
Table 8
0.5 50 50 125 175 400
GA parameters used in the optimum sizing of shunt capacitors.
0.6 50 50 150 200 500
0.7 75 50 175 225 575 No of generations 100
0.8 75 75 200 275 650 Population size 200
0.9 75 75 250 300 750 Crossover rate 0.95
1.0 150 75 325 300 750 Mutation rate 0.001
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 321

Results and discussions
The effect of optimum capacitor placement on the performance
of the distribution system with the proposed method is analyzed
considering two examples of 51 node distribution network and
69 node distribution system. The proposed methodology is com-
pared with the methodology considering the objective of maximiz-
ing net annual savings proposed in [30].
The effect of optimum capacitor placement on substation real
power supply and reactive power supply for all the load levels is
shown in Tables 9 and 10. From Tables 9 and 10, it can be noticed
that the burden on the substation is reduced to the significant level
with the shunt capacitor placement in the distribution system.
From Table 9, it is seen that the real power supply reduction at
the substation in the case of lower load levels is almost same with
both methodologies considered for comparison for the both the
networks considered. But as the load level increases the real power
supply reduction at the substation is slightly higher in the case of
methodology proposed in [30] compared to the proposed fuzzy GA
methodology. From Table 10, it is seen that the reactive power sup-
ply reduction at the substation is much higher in the case of pro-
posed methodology as compared to the methodology proposed in
[30].
The effect on substation reactive current component and cur-
rent drawn at substation due to optimal shunt capacitor placement
is shown in Tables 11 and 12 respectively. It can be seen from
Table 11 that the reactive current component is reduced greatly
in the case of proposed methodology compared to the methodol-
ogy based on annual energy savings proposed in [30]. From
Table 12 it can be observed that the current drawn from the sub-
station is reduced significantly in the case of proposed fuzzy GA
based step by step shunt capacitor based methodology compared
to the annual energy savings methodology proposed in [30].
The real power loss comparison before and after capacitor
placement is shown in Figs. 12 and 13 for 51 node distribution sys-
tem and 69 node distribution system respectively. From Figs. 12
and 13, it can be observed that the real power losses are reduced
significantly with the placement of shunt capacitors compared to
base case. It can also be observed that the real power loss is almost
same at lower load levels both in the case of energy savings

Table 9
Real power drawn from substation before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors capacitors (Method in capacitors (Proposed of shunt capacitors capacitors (Method in capacitors (Proposed
(kW) [30]) (kW) method) (kW) (kW) [30]) (kW) method) (kW)
0.4 1004.30 1000.00 1001.49 1553.39 1543.59 1545.22
0.5 1261.73 1254.22 1257.39 1952.70 1936.99 1939.50
0.6 1521.91 1510.76 1515.14 2356.84 2335.26 2338.39
0.7 1784.95 1769.73 1775.16 2766.06 2733.95 2739.11
0.8 2050.98 2030.75 2038.13 3180.65 3137.43 3145.66
0.9 2320.13 2294.54 2303.56 3600.92 3545.98 3559.88
1.0 2592.56 2561.33 2571.86 4027.19 3959.86 3978.13

Table 10
Reactive power drawn from substation before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors capacitors (Method in capacitors (Proposed of shunt capacitors capacitors (Method in capacitors (Proposed
(kVAr) [30]) (kVAr) method) (kVAr) (kVAr) [30]) (kVAr) method) (kVAr)
0.4 644.11 214.81 15.46 1092.69 363.38 88.99
0.5 810.62 278.83 4.90 1370.85 413.95 114.90
0.6 979.49 319.40 20.85 1651.19 191.57 92.83
0.7 1150.83 437.18 38.63 1933.83 419.68 146.67
0.8 1324.73 506.61 33.63 2218.88 549.97 128.05
0.9 1501.30 628.57 30.54 2506.48 832.54 62.86
1.0 1680.68 803.18 54.62 2796.77 1117.51 49.50

Table 11
Substation reactive current component before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors (Method in capacitors (Proposed of shunt capacitors (Method in capacitors (Proposed
capacitors (A) [30] (A) method) (A) capacitors (A) [30]) (A) method) (A)
0.4 58.55 19.53 1.41 86.31 28.70 7.03
0.5 73.69 25.35 0.45 108.28 32.70 9.08
0.6 89.04 29.04 1.90 130.43 15.13 7.33
0.7 104.62 39.74 3.51 152.75 33.15 11.59
0.8 120.43 46.06 3.06 175.27 43.44 10.12
0.9 136.48 57.14 2.78 197.98 65.76 4.97
1.0 152.80 73.02 4.97 220.91 88.27 3.91
322 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326

Table 12
Substation current component before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors (Method in capacitors (Proposed of shunt capacitors (Method in capacitors (Proposed
capacitors (A) [30]) (A) method) (A) capacitors (A) [30]) (A) method) (A)
0.4 108.46 92.94 91.06 150.02 125.26 122.26
0.5 136.33 116.80 114.31 188.46 156.46 153.47
0.6 164.53 140.38 137.75 227.31 185.08 184.85
0.7 193.07 165.72 161.41 266.59 218.48 216.67
0.8 221.96 190.27 185.31 306.33 251.60 248.68
0.9 251.23 216.28 209.43 346.55 287.71 281.24
1.0 280.88 244.03 233.86 387.29 325.00 314.25

Fig. 12. Real power loss with and without capacitor placement of 51 node network.

Fig. 13. Real power loss with and without capacitor placement for 69 node network.

Table 13
Substation power factor before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors capacitors (Method in capacitors (Proposed of shunt capacitors capacitors (Method in capacitors (Proposed
(lag) [30] (lag) method) (lag) (lag) [30]) (lag) method) (lag)
0.4 0.8418 0.9777 0.9999 0.8179 0.9734 0.9983
0.5 0.8413 0.9762 0.9999 0.8185 0.9779 0.9982
0.6 0.8409 0.9784 0.9999 0.8190 0.9967 0.9992
0.7 0.8405 0.9708 0.9998 0.8196 0.9884 0.9986
0.8 0.8400 0.9703 0.9999 0.8201 0.9850 0.9992
0.9 0.8396 0.9645 0.9999 0.8207 0.9735 0.9998
1.0 0.8391 0.9542 0.9998 0.8214 0.9624 0.9999
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 323

Table 14
Minimum voltage magnitude before and after placement of shunt capacitors.

Load 51 node distribution system 69 node distribution system

level
Before placement After placement of shunt After placement of shunt Before placement After placement of shunt After placement of shunt
of shunt capacitors (Method in capacitors (Proposed of shunt capacitors (Method in capacitors (Proposed
capacitors (pu) [30]) (pu) method) (pu) capacitors (pu) [30]) (pu) method) (pu)
0.4 V 16 V 16 V 16 V 65 V 65 V 65
min = 0.9650 min = 0.9740 min = 0.9803 min = 0.9656 min = 0.9731 min = 0.9767
0.5 V 16
min = 0.9559 V 16
min = 0. 9664 V 16
min = 0.9753 V 65
min = 0.9567 V 65
min = 0.9662 V 65
min = 0.9708
0.6 V 16 = 0.9466 V 16 = 0. 9597 V 16 = 0.9693 V 65 = 0.9476 V 65 = 0.9634 V 65
min min min min min min = 0.9652
0.7 V 16
min = 0.9373 V 16
min = 0.9520 V 16
min = 0.9640 V 65
min = 0.9383 V 65
min = 0. 9545 V 65
min = 0.9586
0.8 V 16 = 0.9277 V 16 = 0. 9445 V 16 = 0.9592 V 65 = 0.9288 V 65 = 0. 9454 V 65
min min min min min min = 0.9529
0.9 V 16 = 0.9180 V 16 = 0. 9363 V 16 = 0.9538 V 65 = 0.9191 V 65 = 0. 9361 V 64
min min min min min min = 0.9498
1.0 V 16
min = 0.9081 V 16
min = 0. 9277 V 16
min = 0.9506 V 65
min = 0.9092 V 65
min = 0. 9266 V 64
min = 0.9451

Fig. 14. Voltage profile comparison at 40% load level for 51 node network.

Fig. 15. Voltage profile comparison at peak load level for 51 node network.

method [30] and proposed method but the real power losses are
slightly high in the case of proposed method compared to energy
savings method. In the case of proposed method eventhough the
real power loss of the distribution system is slightly high, from
Table 12, it can be observed that the total current drawn at the sub-
station is significantly less and hence the real power loss will be
greatly reduced beyond substation and the grid voltage profile will
be improved.
The substation power factor improvement due to shunt capac-
itor placement is shown in Table 13. From Table 13 it can be
observed that the power factor improvement is much better in
the case of proposed method as compared to the energy savings
method [30] and in the case of proposed fuzzy GA method the
power factor is almost improved to unity. The operation of substa-
tion at unity power factor using shunt capacitors will be helpful to
reduce the on-peak operating cost, deferral of network upgrada-
tion, reduction of real power losses, and lower transmission and
distribution costs. This kind of operation will help the grid in much
better way because the network is almost not drawing any reactive
power from the grid.
324 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326

Fig. 16. Voltage profile comparison at 40% load level for 69 node network.

Fig. 17. Voltage profile comparison at peak load level for 69 node network.

The voltage profile improvement is shown in Table 14 and from
Table 14, it can be observed that in the case of proposed method
the minimum voltage improvement is much better compared to
all other cases considered.
The voltage profile improvement due to shunt capacitor place-
ment at 40% lower load level is shown in Figs. 14 and 16 and the
voltage profile improvement at peak load level is shown in Figs. 15
and 17 for both 51 node and 69 node distribution systems respec-
tively. From Figs. 14–17 it can be observed that the proposed fuzzy
GA method gives the best voltage profile improvement compared
to energy savings method [30] at lower and higher load levels for
both the examples of the distribution systems considered.
Annual economical savings due to optimal capacitor placement
N
ir ðir þ 1Þ
CRF ¼ N
ð17Þ
ðir þ 1Þ  1
Here the value of N is taken as 20 and ir is taken as 10%. For a value
of the energy cost Ke = $0.096/kW h, the capacitor purchase cost
CC = $6/kVAr and the capacitor installation cost CI = $1000/location,
the calculated annual economical savings (SCP) obtained by the
installation of shunt capacitor placement are 8870 dollars in the
case of 51 node distribution system and 24,200 dollars in the case
of 69 node distribution system. Hence it can be said that annual
economical benefits can also be obtained with the installation of
shunt capacitors in addition to the improvement of the performance
of the distribution system.

The annual economical savings due to the effect of optimal Conclusions

shunt capacitor placement using the proposed fuzzy GA method
into the distribution system can be obtained by considering the In the present work a GA based fuzzy multiobjective approach
energy savings at the substation. has been proposed for improving the power factor of the substa-
The annual economical savings (SCP) can be given as: tion and voltage profile of the distribution system. A new sensitiv-
( ) ! ity analysis is used for identifying optimal nodes for capacitor
XNL   X
NCL
placement. A fuzzy multiobjective function is formed with the
SCP ¼ K e PSi  PQC
Si T i  CRF C C QC j þ ðC I  NCLÞ ð16Þ
objectives of reduction of substation reactive power component
i¼1 j¼1
current, the real power loss reduction, to avoid branch current con-
where CRF is the capital recovery factor and is defined as the annual straint violation and to satisfy minimum and maximum voltage
loan payment on the borrowed amount for N years at the rate of limit conditions. The optimum sizing of shunt capacitors is
interest of ir and is given as: obtained using GA by maximizing the fuzzy multiobjective
 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326 325

function. From the simulation results it is observed that the pro- References
posed method is advantageous in improving the substation power
factor and voltage profile compared to the shunt capacitor place- [1] Ng HN, Salama MMA, Chikhani AY. Classification of capacitor allocation
techniques. IEEE Trans Power Deliv 2000;15:387–92.
ment based on annual energy savings method. In the case of pro- [2] Neagle NM, Samson DR. Loss reduction from capacitors installed on primary
posed fuzzy GA method the reactive current component became feeders. AIEE Trans Part 1/1 1956;75:950–9.
negligible and the substation power factor improved almost to [3] Schmill JV. Optimum size and location of shunt capacitors on distribution
feeders. IEEE Trans Power Apparatus Syst 1965;84:825–32.
unity at all the load levels with the incorporation of optimum shunt [4] Bae YG. Analytical method of capacitor allocation on distribution primary
capacitors into the distribution system. The voltage profile has been feeders. IEEE Trans Power Apparatus Syst 1978;97(4):1232–8.
improved to much better level compared to other cases and real [5] Chiang HD, Wang JC, Cockings O, Shin HD. Optimal capacitor placements in
distribution systems: Part 1: a new formulation and the overall problem. IEEE
power losses are reduced to significant value. In addition to that a
Trans Power Deliv 1990;5(2):634–42.
significant amount of annual savings has also been achieved. [6] Chiang HD, Wang JC, Cockings O, Shin HD. Optimal capacitor placements in
distribution systems: Part 2: solution algorithms and numerical results. IEEE
Trans Power Deliv 1990;5(2):643–9.
[7] Duran H. Optimum number, location, and size of shunt capacitors in radial
Appendix A distribution feeder: a dynamic programming approach. IEEE Trans Power
Apparatus Syst 1968;87(9):1769–74.
See Table A. [8] Lee SH, Grainger JJ. Optimum placement of fixed and switched capacitors on
primary distribution feeders. IEEE Trans Power Apparatus Syst
1981;100:345–52.
[9] Grainger JJ, Lee SH. Optimum size and location of shunt capacitors for
Table A reduction of losses on distribution feeders. IEEE Trans Power Apparatus Syst
1981;100:1105–18.
51 node radial distribution system data.
[10] Ponnavaikko M, Rao KSP. Optimal choice of fixed and switched shunt
Br Send. Recev. R (X) X (X) Recev. end IC capacitors on radial distribution feeders by the method of local variations.
no end end (A) IEEE Trans Power Apparatus Syst 1983;102:1607–15.
PL (kW) QL (kVAr) [11] Kaplan M. Optimization of number, location, size, type and control setting of
1 1 2 0.2740 0.3560 40 30 324 shunt capacitors on radial distribution feeders. IEEE Trans Power Apparatus
Syst 1984;103(9):2659–65.
2 2 3 0.1370 0.1780 60 40 324
[12] Baran ME, Wu FF. Optimal capacitor placement on radial distribution systems.
3 3 4 0.3288 0.4272 20 10 324
IEEE Trans Power Deliv 1989;4(1):725–34.
4 4 5 0.1096 0.1424 80 60 324 [13] Baran ME, Wu FF. Optimal sizing of capacitors placed on radial distribution
5 5 6 0.5400 0.4356 38 18 234 systems. IEEE Trans Power Deliv 1989;4(1):735–43.
6 6 7 0.3600 0.2904 20 15 234 [14] Chis M, Salama MMA, Jayaram S. Capacitor placement in distribution systems
7 7 8 0.3600 0.2904 60 40 234 using heuristic search strategies. IEE Proc Gener Transm Distrib 1997;144
8 8 9 0.7200 0.5808 70 45 234 (3):225–30.
9 9 10 2.7320 0.7792 60 35 115 [15] Deng Y, Ren X, Zhao C, Zhao D. A heuristic and algorithmic combined approach
10 10 11 2.0490 0.5844 80 50 115 for reactive power optimization with time varying load demand in distribution
11 11 12 2.0490 0.5844 10 5 115 systems. IEEE Trans Power Syst 2002;17(4):1068–72.
12 12 13 0.9562 0.2727 25 15 115 [16] Pires DF, Martins AG, Antunes CH. A multiobjective model for VAR planning in
13 13 14 1.0928 0.3117 55 45 115 radial distribution networks based on tabu search. IEEE Trans Power Syst
14 14 15 1.5026 0.4286 120 80 115 2005;20(2):1089–94.
[17] Raju MR, Murthy KVSR, Ravindra K. Direct search algorithm for capacitive
15 15 16 3.0052 0.8571 40 25 115
compensation in radial distribution systems. Int J Electr Power Energy Syst
16 3 17 2.7320 0.7792 35 25 115
2012;42(1):24–30.
17 17 18 0.8196 0.2338 60 30 115
[18] Hamouda A, Sayah S. Optimal capacitors sizing in distribution feeders using
18 18 19 1.3660 0.3896 80 50 115 heuristic search based node stability indices. Int J Electr Power Energy Syst
19 19 20 1.3660 0.3896 60 35 115 2013;46:56–64.
20 20 21 2.0490 0.5844 50 30 115 [19] Ng HN, Salama MMA, Chikhani AY. Capacitor allocation by approximate
21 21 22 1.5026 0.4286 50 30 115 reasoning: fuzzy capacitor placement. IEEE Trans Power Syst 2000;15
22 4 23 1.6392 0.4675 80 60 115 (1):393–8.
23 23 24 1.7758 0.5065 45 25 115 [20] Mekhamer SF, Soliman SA, Moustafa MA, El-Hawary ME. Application of fuzzy
24 24 25 1.0928 0.3117 38 18 115 logic for reactive power compensation of radial distribution feeders. IEEE
25 25 26 0.8196 0.2338 78 48 115 Trans Power Syst 2003;18(1):206–13.
26 26 27 0.5464 0.1558 16 8 115 [21] Masoum MAS, Jafarian A, Ladjevardi M, Fuchs EF, Grady WM. Fuzzy approach
27 27 28 1.0928 0.3117 18 10 115 for optimal placement and sizing of capacitor banks in the presence of
harmonics. IEEE Trans Power Deliv 2004;19(2):822–9.
28 28 29 0.2732 0.0779 40 30 115
[22] Shi J, Liu Y. Fuzzy evaluation based multi-objective reactive power
29 6 30 0.7020 0.4774 40 30 208
optimization in distribution networks. In: Fuzzy systems and knowledge
30 30 31 0.6480 0.4406 20 15 208
discovery. Lecture notes in computer science, vol. 3613. Berlin,
31 31 32 0.6480 0.4406 30 20 208 Heidelberg: Springer Publications; 2005. p. 11–9.
32 32 33 0.6480 0.4406 36 26 208 [23] Bhattacharya SK, Goswami SK. A new fuzzy based solution of the capacitor
33 33 34 0.5400 0.3672 50 40 208 placement problem in radial distribution system. Expert Syst Appl 2009;36
34 34 35 0.3240 0.2203 27 18 208 (3):4207–12.
35 35 36 0.3888 0.2644 33 16 208 [24] Boone G, Chiang HD. Optimal capacitor placement in distribution systems by
36 36 37 0.4320 0.2938 42 22 208 genetic algorithm. Int J Electr Power Energy Syst 1993;l15(3):155–62.
37 37 38 0.5940 0.4039 55 30 208 [25] Iba K. Reactive power optimization by genetic algorithm. IEEE Trans Power
38 38 39 0.7020 0.4774 44 26 208 Syst 1994;9(2):685–92.
39 7 40 1.9124 0.5454 80 70 115 [26] Sundharajan S, Pahwa A. Optimal selection of capacitors for radial distribution
40 40 41 3.0052 0.8571 60 30 115 systems using a genetic algorithm. IEEE Trans Power Syst 1994;9
41 41 42 2.4588 0.7013 45 30 115 (3):1499–507.
[27] Delfanti M, Granelli GP, Marannino P, Montagna M. Optimal capacitor
42 42 43 2.1856 0.6234 48 28 115
placement using deterministic and genetic algorithms. IEEE Trans Power
43 43 44 2.1856 0.6234 68 38 115
Syst 2000;15(3):1041–6.
44 44 45 0.6830 0.1948 77 23 115 [28] Das D. Reactive power compensation for radial distribution networks using
45 9 46 0.9562 0.2727 60 30 115 genetic algorithm. Int J Electr Power Energy Syst 2002;24(7):573–81.
46 46 47 1.0245 0.2922 40 20 115 [29] Haghifam MR, Malik OP. Genetic algorithm based approach for fixed and
47 47 48 1.2294 0.3506 45 45 115 switchable capacitors placement in distribution systems with uncertainty and
48 48 49 1.7758 0.5065 70 50 115 time varying loads. IET Gener Transm Distrib 2007;1(2):244–52.
49 49 50 1.6392 0.4675 30 20 115 [30] Swarnkar A, Gupta N, Niazi KR. Optimal placement of fixed and switched shunt
50 50 51 1.3660 0.3896 35 30 115 capacitors for large-scale distribution systems using genetic algorithms. In:
Innovative smart grid technologies conference Europe; 2010. p. 1–8.
326 S.R. Gampa, D. Das / Electrical Power and Energy Systems 77 (2016) 314–326
[31] Gallego RA, Monticelli AJ, Romero R. Optimal capacitor placement in radial
distribution networks. IEEE Trans Power Syst 2001;16(4):630–7.
[32] De Souza BA, Alves HN, Ferreira HA. Micro genetic algorithms and Fuzzy logic
applied to the optimal placement of capacitor banks in distribution networks.
IEEE Trans Power Syst 2004;19(2):942–7.
[33] Hsiao YT, Chen CH, Chien CC. Optimal capacitor placement in distribution
systems using a combination fuzzy-GA method. Electr Power Syst Res 2004;26
(7):501–8.
[34] Das D. Optimal placement of capacitors in radial distribution system using a
fuzzy-GA method. Int J Electr Power Energy Syst 2008;30(6–7):361–7.
[35] Abul’Wafa AR. Optimal capacitor placement for enhancing voltage stability in
distribution systems using analytical algorithm and fuzzy-real coded GA. Int J
Electr Power Energy Syst 2014;55:246–52.
[36] Huang TL, Hsiao YT, Chang CH, Jiang JA. Optimal placement of capacitors in
distribution systems using an immune multiobjective algorithm. Int J Electr
Power Energy Syst 2008;30(3):184–92.
[37] Rao RS. Optimal capacitor allocation for loss reduction in distribution system
using fuzzy and plant growth simulation algorithm. World Acad Sci Eng
Technol 2010;44:402–8.
[38] Singh SP, Rao AR. Optimal allocation of capacitors in distribution systems
using particle swarm optimization. Int J Electr Power Energy Syst 2012;43
(1):1267–75.
[39] Ziari I, Ledwich G, Ghosh A. A new technique for optimal allocation and sizing
of capacitors and setting of LTC. Int J Electr Power Energy Syst 2013;46:250–7.
[40] El-Fergany AA, Abdelaziz AY. Capacitor placement for net saving maximization
and system stability enhancement in distribution networks using artificial bee
colony based approach. Int J Electr Power Energy Syst 2014;54:235–43.
[41] Hung DQ, Mithulananthan N, Bansal RC. Analytical expressions for DG
allocation in primary distribution networks. IEEE Trans Energy Convers
2010;25(3):814–20.
[42] Aman MM, Jasmon GB, Bakar AHA, Mokhlis H, Karimi M. Optimum shunt
capacitor placement in distribution system: a review and comparative study.
Renew Sust Energy Rev 2014;30:429–39.
[43] Chakravorty M, Das D. Voltage stability analysis of radial distribution
networks. Int J Electr Power Energy Syst 2001;23(2):129–35.
[44] Zadeh LA. Fuzzy sets. Inf Control 1965;8:338–53.
[45] Goldberg DE. Genetic algorithms in search, optimization and learning. A
book. Reading (MA): Addison-Wesley; 1989.