Swarm and Evolutionary Computation 15 (2014) 58–65

Contents lists available at ScienceDirect

Swarm and Evolutionary Computation

journal homepage: [Link]/locate/swevo

Regular Paper

Optimal size and siting of multiple distributed generators

in distribution system using bacterial foraging optimization
Mohamed Imran A n, Kowsalya M
School of Electrical Engineering, VIT University, Vellore-632014, Tamil Nadu, India

art ic l e i nf o a b s t r a c t

Article history: Optimal location and size of distributed generation (DG) in the distribution system play a significant role
Received 20 April 2013 in minimizing power losses, operational cost and improving voltage stability. This paper presents a new
Received in revised form approach to find the optimal location and size of DG with an objective of minimizing network power
5 October 2013
losses, operational costs and improving voltage stability. Loss sensitivity factor is used to identify the
Accepted 8 December 2013
Available online 17 December 2013
optimal locations for installation of DG units. Bacterial Foraging Optimization Algorithm (BFOA) is used to
find the optimal size of DG. BFOA is a swarm intelligence technique which models the individual and
Keywords: group foraging policies of the Escherichia coli bacteria as a distributed optimization process. The technical
Distributed generation constraints of voltage and branch current carrying capacity are included in the assessment of the
Bacterial Foraging Optimization Algorithm
objective function. The proposed method has been tested on IEEE 33-bus and 69-bus radial distribution
Power loss minimization
systems with various load models at different load levels to demonstrate the performance and
Voltage stability
Operational cost minimization effectiveness of the technique.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction Raphson method of the load flow study was implemented in [7].
The objective was to lower down both the cost and loss more
The importance of DGs in future smart grid increases considering effectively. It also focuses on optimizing weighting factor, which
their impacts on the modern power system. The number of DG units balances the cost and the loss factors and helps to build up desired
installed in the distribution system has been increasing significantly, objectives with maximum potential benefit.
and their technical, economical, and environmental impacts on The artificial neural network based method was developed to
power system are being analyzed. Currently, the technical impacts find optimal DG size and locations due to complexity of multiple
of interest are voltage profile, power loss, power quality, reliability, DG concepts in [8]. Genetic Algorithm (GA) was used to determine
protection, power control and stability [1–3]. The most critical the optimal size and location of multiple DG units to minimize the
factors that influence the technical and economic impacts are type, system losses and power supply by the main grid, considering the
size, and location of DG units in the power system. The location and voltage limits at each bus of the system [9]. A combination of
size of DG unit should be optimal to maximize the benefits and Particle Swarm Optimization (PSO) and GA was introduced to find
reduce their impact on the power system. Inappropriate placement the optimal location of a fixed capacity of DG units installed in the
in some situations can reduce the benefits and even endanger the system to minimize the total power loss, to improve voltage profile
entire system operation [4]. Hence, optimal location and sizing of DG and voltage stability index [10]. The PSO algorithm was employed
units have fascinated many researchers working in the DGs. for a multiobjective index based approach to identify optimal size
An analytical and Improved Analytical (IA) expression for and location of multiple DGs in the distribution system with
finding the optimal location and size of DG units for reducing different load models [11].
power loss had been presented along with methodologies for Modal analysis and continuous power flow are used to determine
identifying the optimal location in [4,5]. A new index considering the candidate bus for DG placement by using the loading parameter
stable node voltages referred as Power Stability Index (PSI) to find as the comparison index for selecting the best places [12]. Authors
the most sensitive bus for DG placement was developed, and a justified that the best candidate for DG placement is different from
search algorithm was used for finding the optimum size of DG at the best location for reactive power compensation. The loss sensi-
optimum location to minimize power loss in [6]. A simple tivity factor (LSF) used in [13], finds the optimal candidate locations
conventional iterative search technique along with the Newton for reactive power compensation (i.e. capacitor placement) but not
an optimal site for DG installation.
n
Corresponding author. The above methods discussed have gained encouraging results in
E-mail address: mohamedimran.a@[Link] (Mohamed Imran A). finding the optimal location and size of DGs, but they also have some

2210-6502/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
[Link]
 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65 59

Nomenclature n total number of buses

Qk Reactive power load at bus k
Pk real power load at bus k Qk,k þ 1 reactive power flowing in the line between buses k
Pk,k þ 1 real power flowing in the line between buses k and and k þ1
k þ1 Qk þ 1,eff total effective reactive power supplied beyond the bus
Pk þ 1,eff total effective real power supplied beyond the bus kþ 1 k þ1
Vk voltage magnitude at bus k Ik equivalent current injected at node k
Jk,k þ 1 branch current in the line section between buses k and Jk,k þ 1,max maximum branch current limit of line section
k þ1 between buses k and k þ1
Vmin minimum voltage limits of the buses Vmax maximum voltage limits of the buses
ΔVmax maximum voltage drop limit between buses 1 and k Vworst worst voltage magnitude of the system
Rk,k þ 1 resistance of the line section between buses k and kþ 1 Xk,k þ 1 reactance of the line section between buses k and kþ 1
PLoss(k,k þ1) power loss in the line section between buses k PDG,TLoss total power loss of the system with DGs
and k þ1 without DGs P max
DGT maximum total power generation limit of the system
P min
DGT minimum total power generation limit of the system PDG,k power supplied from DG at the bus k
PDG,Loss(k,k þ1) power loss in the line section between buses k b total number of branches
and k þ1 with DGs SDG,T (MVA) total MVA supplied by DG in the system

shortcomings such as computation time in solving real large-scale Branch current in the line section between buses k and kþ 1 is
systems, calculation efficiency and convergence. The effect of load calculated by using Kirchhoff's current law as
models on distributed generation planning in the distribution system J k;k þ 1 ¼ I k þ 1 þI k þ 2 ð2Þ
was investigated and stated that load models can significantly affect
the DG planning [14]. Most of the researches mentioned in the The above equation is generalized in matrix form by using the Bus
literature that they used DGs with unity power factor in the current Injection to Branch Current matrix (BIBC) [15]. Now the branch
distribution system with only constant power loads. Also most of current at each line can be calculated in a matrix form as follows:
the authors did not consider the operating cost of DGs, and achieved ½J ¼ ½BIBC½I ð3Þ
power loss minimization and voltage stability by installing multiple
DGs of large sizes. From Fig. 1, by Kirchhoff's voltage law, the voltage at the bus kþ1 can
The present work is aimed to develop a fast and novel com- be calculated as
putation technique to find the optimal location and size of multi- V k þ 1 ¼ V k –J k;k þ 1 ðRk;k þ 1 þ jX k;k þ 1 Þ ð4Þ
ple DGs for loss minimization, operational cost minimization and
to improve voltage stability in the radial distribution system with The power loss in the line section between buses k and kþ1 is
different types of loads. In this paper, LSF that determines the most computed as
!
sensitive bus for DG installation has been used. The LSF at each bus P 2k;k þ 1 þ Q 2k;k þ 1
is ranked in descending order and the top three sensitive buses are P Loss ðk; k þ 1Þ ¼ Rk;k þ 1 ð5Þ
j Vkj 2
selected as candidate location for DG installation. BFOA, a dis-
tributed optimization technique, is employed to minimize the The total power loss of the system, PTLoss, is determined by the
objective function by determining the optimal size of DG units at summation of losses in all line sections, which is given as
candidate locations. The advantage of exempting BFOA from the b
determination of optimal candidate location of DGs is to reduce P TLoss ¼ ∑ P Loss ðk; k þ 1Þ ð6Þ
the search space, computation time and to improve convergence k¼1

characteristics. The novelty of this work lies in the combined

technique used to identify the optimal placement and sizing of 2.2. Power loss with DG
multiple-DG units in order to minimize power losses, operational
costs and to improve voltage stability at different load models. The DG units installation at optimal location will lead to line loss
proposed method is tested on a well-known 33-bus and 69-bus reduction, improved voltage stability, peak demand saving,
test system, and the results obtained are statistically compared improved reliability and security. The power loss of a line section
with other classical methods. connecting between buses k and kþ 1 after DG installment is
computed as
!
P2 þQ 2DG;k;k þ 1
2. Problem formulation P DG;Loss ðk; k þ1Þ ¼ Rk;k þ 1 DG;k;k þ 1 ð7Þ
j Vkj 2
2.1. Power flow equations

Power flow in a distribution system is calculated by the

following set of recursive equations derived from the single line
diagram as shown in Fig. 1.
From Fig. 1, the equivalent current injected at node k is cal-
culated as


P k þ jQ k n
Ik ¼ ð1Þ
Vk Fig. 1. Sample distribution system.
60 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65

Total power loss in the system after DG installation is calculated by minimized subject to various operational constraints to satisfy the
the summation of the losses in all line sections of the system as electrical requirements of the distribution network. These con-
follows: straints are discussed as follows:
b
P DG;TLoss ¼ ∑ PDG;Loss ðk; k þ 1Þ ð8Þ Power balance constraints
k¼1
n n b
∑ PDG;k r ∑ P k þ ∑ P Loss;k;k þ 1 ð15Þ
k¼2 k¼2 k¼1
2.3. Power loss reduction with DG installation

Power loss index (ΔPLDG) [16] is the ratio of total power loss Thermal limits
with DG to the total power loss without DG and is written as j J k;k þ 1 j r j J k;k þ 1; max j ð16Þ
P
ΔPLDG ¼ DG;TLoss ð9Þ
P TLoss Voltage drop limits
Net power loss reduced with DG installation in the system can be j V 1  V k j r ΔV max ð17Þ
maximized by minimizing ΔPLDG.

Distributed generation capacity limits

2.4. Voltage deviation index
P min max
DGT r P DGT r P DGT ð18Þ
One of the advantages of DG installation is the drop in voltage
where
deviation. This index rebukes the DG size which gives the higher
n n
voltage deviation. The voltage deviation index [16] can be defined P min max
DGT ¼ 0:1 ∑ P k and P DGT ¼ 0:6 ∑ P k ð19Þ
as k¼2 k¼2


V Vk
ΔV D ¼ max 1 8 k ¼ 1; 2; …; n ð10Þ
V1
If any one of the above constraints is violated, the resultant
During DG installation, if the state of the system has voltage
solution will be rejected.
limit violations, the proposed technique will try to minimize the
ΔVD closer to zero and thereby improves voltage stability and
network performance. 2.7. Load model

A load model is a mathematical representation of the relation-

2.5. Operational cost minimization
ship between a bus voltage (magnitude and frequency) and the
power (active and reactive) or current flowing into the bus load
The operational cost of DISCOs has two components. The first
[17]. The effect of load variation at node-k is represented as
cost is for real power supplied from the substation. This can be
follows:
minimized by minimizing the total power losses of the system.
β
The second cost is the cost of real power supplied by the DG P k ¼ ρP k;actual V k ð20Þ
installed. This cost can be minimized by minimizing the amount of
real power drawn from DG. Thus the total operating cost (TOC) β
Q k ¼ ρQ k;actual V k ð21Þ
that may be minimized is given as
TOC ¼ ðc1 P DG;TLoss Þ þ ðc2 P DGT Þ ð11Þ where β ¼0 for Constant Power (CP) load; 1 for Constant Current
(CC) load; and 2 for Constant Impedance (CI) load.
where c1 and c2 are the cost coefficients of real power supplied by The load factor ρ is a multiplier by which the load power
substation and DGs in $/kW. PDGT is the total real power drawn demands at all nodes are increased or decreased. Decisions and
from installed [Link] net operating cost (ΔOC) of DG reduced is results based on the constant power load model assumptions are
calculated as not technically feasible if it is employed on the system having non-
TOC constant power load behavior. Therefore, distribution systems
ΔOC ¼ ð12Þ with above mentioned three types of load at different load factors
c2 P max
DGT
are simulated to check the robustness and capability of the
The TOC of the system with DG installation can be minimized by proposed method for practical implementation.
minimizing ΔOC.

2.6. Objective function of the problem 3. Loss sensitivity factor for DG installation

The proposed objective function (F) of the problem is formu- Loss sensitivity factor (LSF) is computed to determine the
lated to minimize the power loss, voltage deviation and total locations for the DG installation [18]. The estimation of these
operating cost of the distribution system, which is given by candidate buses initially helps in reduction of the search space for
Minimize F ¼ minðα1 ΔPLDG þ α2 ΔV D þ α3 ΔOCÞ ð13Þ the optimization procedure.
For the system shown in Fig. 1, the LSF at the line section
where between k and kþ 1 buses can be determined with the help of
3 following equation:
∑ αq ¼ 1:0 4 αq A ½0; 1 ð14Þ
q¼1 ∂P Lineloss 2P k þ 1;eff Rk;k þ 1
LSFðk; k þ1Þ ¼ ¼ ð22Þ
∂P k þ 1;eff j Vkþ1j 2
These weights (αq) are intended to give the corresponding
importance to power loss, voltage deviation index and operating Using the above equation, LSFs are computed from the load
cost reduction of the objective function. The objective function is flows and values are arranged in a descending order for all the
 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65 61

line section of the given system. LSFs decide the sequence in Step 5: Chemotaxis loop: j¼jþ 1.
which the buses are to be considered for DG installation. The i For i¼1,2,…,S take a chemotactic step for bacterium i as
optimal size of DGs at the candidate buses are calculated by follows.
using BFOA. ii Compute the objective function, F(i,j,k,l).
Let F(i,j,k,l) ¼F(i,j,k,l) þFcc(θi(j,k,l), P(j,k,l)). Fcc is computed using
the following equation:
4. Overview of Bacterial Foraging Optimization Algorithm
s
F cc ðθ; Pðj; k; lÞÞ ¼ ∑ F cc ðθ; θ ðj; k; lÞ
i
BFOA is an efficient swarm intelligence based stochastic
i¼1
search technique developed recently by Passino [19]. In recent
s

p

∑ ðθ m  θ Þ2
years, BFOA has been applied to solve numerous optimization i
¼ ∑  dattrac tan t exp  wattrac tan t
problems in power systems [20,21] due to its ability in searching i¼1


m¼1

s p
the promising areas of the solution space. The idea of BFOA is
þ ∑  hrepellant exp  wrepellant ∑ ðθm  θi Þ2 ð24Þ
based on the social foraging behavior of Escherichia coli bacteria i¼1 m¼1
present in human intestines. During foraging of a bacterium, the
locomotion is achieved by a set of tensile flagella. Flagella help
Let Flast ¼F(i,j,k,l)
an E. coli bacterium to tumble or swim, which are two basic
Tumble: generate a random vector Δ(i) with each element
operations (movements) performed by a bacterium at the time
Δm(i), m ¼1,2,...,p, is a random number in the range of [ 1, 1].
of foraging. When they rotate the flagella in the clockwise
iii Move θ ðj þ 1; k; lÞ using the following equation:
i
direction, each flagellum pulls on the cell. This results in the
moving of flagella independently and finally the bacterium
tumbles less frequently. Alternatively, in a harmful place it ΔðiÞ
θi ðjþ 1; k; lÞ ¼ θi ðj; k; lÞ þ cðiÞpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð25Þ
tumbles frequently to find a nutrient gradient. The four prime Δτ ðiÞΔðiÞ
steps of BFOA are chemotaxis, swarming, reproduction, and
elimination and dispersion. The detailed steps of BFOA can be
This results in a step of size c(i) in the direction of the tumble
referred from [22].
for bacterium i.
The uniqueness of BFOA lies in the chemotaxis process in which
iv Compute F(i,j þ1,k,l).
the bacterium moves by an amount of cΔ if the objective function
v Swim
value is reduced for new location. Otherwise, it remains in the
(a) Let m ¼ 0 (counter for swim length).
same position. This chemotactic process of BFOA has an edge over
(b) While mo Ns
the other classical optimization algorithms, especially in context to  Let m ¼ m þ1.
the convergence behavior of BFOA very near the global optima.  If F(i,jþ 1,k,l) oFlast ( if doing better), let Flast ¼ F(i,jþ1,k,l)
and move θ ðj þ 1; k; lÞ using Eq. (25) and use this θi to
i

4.1. Application of BFOA to find optimal DG sizes compute the new F(i,jþ1,k,l) as in (iv).
Else, let m ¼Ns. This is the end of the while statement.
The prime steps of BFOA applied to minimize the objective vi Go to next bacterium (iþ1) if iaS (i.e., go to (ii) to process the
function are as follows: next bacterium).
Step 6: If joNc, go to step 5.
Step 1: Initialization of following parameters: Step 7: Reproduction.
 p: Dimension of the search space, number of DGs to be placed  The S/2 bacteria with the highest F values die and the remain-
to minimize the objective function. The loss sensitivity factors ing S/2 bacteria with the best values split and placed at the
are calculated from the load flow and sorted in descending same location as their parents.
orders. The buses that are more sensitive are picked to install Step 8: If koNre, go to step 4.
DG units. Every dimension signifies a DG size at the selected Step 9: Elimination and dispersal: Select a best bacterium with
candidate bus (p ¼3). the lowest F and keep it for the next chemotactic step. For
 S: Total number of bacteria in the population (100). i¼1,2….S  1 with probability Ped, eliminate and disperse each
 N: Number of iterations (50). bacterium so that the number of bacterium in the population
 Nc: The number of chemotactic steps (4). remains constant. The dispersed bacteria give the brand new
 Ns: The swimming length (4). DG sizes at candidate buses.
 Nre: The number of reproduction steps (4). Step 10: If l oNed, then go to step 3.
 Ned: The number of elimination-dispersal events (2). Step 11: if qo N, then go to step 2, otherwise end. Evaluate the
 Ped: Elimination-dispersal probability (0.5). objective function for the final bacterium population. Deter-
 c(i): The size of the step taken in the random direction specified mine the bacterium which gives the minimum F and this
by the tumble (0.05*ones(S,1)). bacterium give the optimal sizes of DGs at candidate buses.
 θi : The initial random location of each bacterium and is Run the power flow with the obtained DG units sizes installed
given as at candidate buses and display the results.

θi ¼ ðθi1 ; θi2 ; θi3 Þ 8 i ¼ 1; 2; …; S ð23Þ These are the prime steps of BFOA implemented to minimize F.

where θ1, θ2, and θ3 are the sizes of DGs in kW at

candidate buses. 5. Simulation and numerical results
The base power loss and voltages at all nodes without DGs is
calculated from the load flow. To demonstrate the effectiveness of the proposed method using
Step 2: Iteration loop: q ¼q þ1 LSFs and BFOA, it is applied to two 12.66 kV IEEE test systems
Step 3: Elimination–dispersal loop: l ¼l þ1 consisting of 33 and 69 buses. To give more importance for
Step 4: Reproduction loop: k ¼k þ1 power loss minimization and voltage stability enhancement, the
62 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65

weighting factors α1, α2, and α3 used in the objective function are tivity factors of all buses are computed, and then they are ranked.
taken as 0.5, 0.4 and 0.1 respectively. The cost coefficients c1 and c2 The top three locations which are more sensitive are chosen to
used in TOC evaluation are taken as 4$/kW and 5$/kW for both the install DG units in the system. The candidate buses chosen for DG
test systems. The cost coefficient c2 is taken slightly higher than c1 unit's installation from LSFs are 14, 18, and 32 and the optimal size
considering the installation cost and maintenance cost of DGs. In of DGs installed at these buses is determined by BFOA. To analyze
the simulation of the test systems, different load models such as the performance of the proposed method on a 33-bus system with
CP at three different load factors: 0.5 (light), 1.0 (full load), 1.6 different types of loads at different load factors, the optimal DG
(heavy load), CC and CI at full load condition are considered to size at candidate locations, minimum voltage magnitude, total
analyze the efficiency of the proposed method. The maximum power loss and percentage power loss reduction before and after
number of DGs installed for the given test systems is limited to DG installation is shown in Table 1. It is observed that the
three since the rate of improvement of percentage loss reduction percentage loss reduction is almost the same even when the load
decreases when the candidate locations are more than three for all increases from light to heavy and for different load models. The
load models [23,24]. Also the TOC of the system will increase voltage profile also has been progressively improved for all types
as the number of DGs installed increases. However, the proposed of loads. This proves that the proposed technique is very accurate
technique can be implemented for any number of DGs. in finding the optimal size and location of DGs for the system with
For effective comparison and to study the superiority of the all types of loads. Also from Table 1, the values of power loss index,
proposed technique, the DGs operating at two different power voltage deviation index, TOC and computation time are also found
factors are simulated. The BFOA parameters initialized in the above to be impressive.
section is common for both the test systems. An analytical soft- The simulated results of BFOA are compared with the results
ware tool has been developed in MATLAB environment to run load obtained by GA, PSO, combined GA/PSO and simulated annealing
flow, calculate TOC, and to identify optimal location and size of (SA). The parameters chosen for GA, PSO, GA/PSO and SA are the
DG units. same as in [10,13]. For effective comparison, the test system is
simulated with DGs operating at two different power factors and
are tabulated in Table 2. It is observed that the optimal size of DGs
5.1. 33-Bus test system
obtained by BFOA approach is significantly small as compared to
the other classical approaches with great increase in the percen-
This test system is a 33-bus radial distribution system with 32
tage of power loss and TOC reduction. But in the case of the SA
branches. The line and load data of this test system are from [25].
method, the total power loss is reduced to 82.03 kW with the
The single line diagram of 33-bus system is shown in Fig. 2.
TOC of 12,666.6$ i.e. 2718.5$ greater than the BFOA, also the
The total real and reactive power loads of the system are 3.72 MW
minimum voltage magnitude improved by SA after DG installation
and 2.3 MVAr, respectively. The total power losses for the base
is 0.9676 p.u. which is lower than the proposed method. Similarly
case calculated from load flow with the loads CP(0.5), CP(1.0),
at 0.866 power factor, the TOC minimized by the BFOA is 9743.9$
CP(1.6), CC(1.0) and CI(1.0) are 48.78 kW, 210.98 kW, 603.36 kW,
which is 3342.4$ less than the TOC of the SA method. This
180.29 kW and 154.62 kW respectively. From load flow, loss sensi-
indicates that the proposed technique accurately predicts the
optimal location and size of DGs with the maximum percentage
22
power loss reduction, minimum operating cost and improved
21
voltage stability.
20 The performances of the classical methods on this test system
19 26 27 28 29 30 31 32 33 are statistically compared and are shown in Fig. 3. It is perceived
that the classical methods GA, PSO and GA/PSO highly reduce the
1
voltage deviation index, but the TOC and power loss index are
higher than the BFOA. The SA method greatly reduces the power
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 loss index, but the TOC and voltage stability index are found to be
23 greater than the BFOA. From Table 2 and Fig. 3. it is witnessed that
the performance of the BFOA is better compared to the above
24
methods in terms of quality of solutions.
25
In order to illustrate the convergence of BFOA, the convergence
Fig. 2. Single line diagram of 33-bus radial distribution system. characteristic of BFOA for this test system is compared with

Table 1
Performance analysis of the proposed method on 33-bus system at different load models.

Parameters CP load CC load (1.0) CI load (1.0)

Light (0.5) Full (1.0) Heavy (1.6)

Without DG With DG Without DG With DG Without DG With DG Without DG With DG Without DG With DG

DG size (MW) (bus) 0.3434 (14) 0.6521 (14) 0.9645 (14) 0.6023 (14) 0.5229 (14)
0.0781 (18) 0.1984 (18) 0.4423 (18) 0.1836 (18) 0.1996 (18)
0.4927 (32) 1.0672 (32) 1.7767 (32) 1.3095 (32) 0.9102 (32)
Power loss (kW) 48.78 21.61 210.98 89.90 603.36 243.63 180.29 78.49 154.62 68.87
% Power loss reduction 55.70 57.38 59.62 56.46 55.45
ΔPLDG 0.4429 0.4216 0.4037 0.4353 0.4454
Vworst (p.u.) (bus) 0.9540 (18) 0.9844 (29) 0.9038 (18) 0.9705 (29) 0.8360 (18) 0.9545 (30) 0.9120 (18) 0.9723 (29) 0.9194 (18) 0.9745 (29)
ΔVD 0.0460 0.0156 0.0962 0.0295 0.1640 0.0455 0.0880 0.0277 0.0806 0.0255
TOC ($) 4657 9948 16,892 10,790 8438
Computation time (s) 4.2 4.9 5.3 4.7 4.5
 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65 63

Table 2
Comparison and performance of 33-bus system with DGs at different power factors.

Method PDG,TLoss (kW) % Loss reduction Vworst (p.u.) (bus) DG location DG size (MW) SDG,T (MVA) Power factor TOC ($)

GA [10] 106.30 49.61 0.9809(25) 11 1.5000 2.9942 Unity 15396.2

29 0.4228
30 1.0714
PSO [10] 105.35 50.06 0.9806(30) 13 0.9816 2.9881 Unity 15361.9
32 0.8297
8 1.1768
GA/PSO [10] 103.40 50.99 0.9808(25) 32 1.2000 2.9880 Unity 15353.6
16 0.8630
11 0.9250
SA [13] 82.03 61.12 0.9676(14) 6 1.1124 2.4677 Unity 12666.6
18 0.4874
30 0.8679
BFOA 89.90 57.38 0.9705(29) 14 0.6521 1.9176 Unity 9948.1
18 0.1984
32 1.0672
SA [13] 26.72 87.33 0.9826(25) 6 1.1976 2.9975 0.866 13086.3
18 0.4778
30 0.9205
BFOA 37.85 82.06 0.9802(29) 14 0.6798 2.2153 0.866 9743.9
18 0.1302
32 1.1085
1.3795

1.3765

1.3757

0.55
1.6
1.1726

GA PSO GA/PSO SA BFOA

1.135

1.4 0.5
Objective Function (F)
0.8914

0.8731

1.2
0.45
1
0.5038

0.4993

0.8
0.4261

0.4
0.3888
0.49

0.6
0.1794
0.1266

0.35
0.4
0.0324

0.0295

0.0198
0.0194

0.0192
0.0191

0.0174

0.2 0.3
0
GA PSO GA/PSO SA BFOA SA @ BFOA @ 0.25
0.866 p.f 0.866 p.f 0 5 10 15 20 25 30 35 40 45 50
Methods Voltage Deviation Index Power Loss Index Net Operating Cost No of Iterations

Fig. 3. Comparison of performances for 33-bus system. Fig. 4. Comparison of convergence characteristics of objective function for 33-bus
system.

the above classical algorithms and is shown in Fig. 4. It can be been reduced for all types of loads. The voltage stability and the
observed that for BFOA, only 14 iterations are required to converge minimum voltage magnitude have also been greatly improved. The
to the best solution. Also BFOA shows a stable and quick conver- values of power loss index, voltage deviation index and TOC are also
gence with a global searching ability in finding the optimal DG found to be effectively minimized.
sizes. From Figs. 3 and 4, it is clear that the BFOA has outperformed The voltage profile curves of the system with different types of
PSO, a standard real-coded GA, SA and a hybrid algorithm based on loads are compared as shown in Fig. 6. It is observed that the
GA and PSO in a statistically meaningful fashion. voltage magnitude at all buses of the system with different types
of loads is highly improved after DG (unity p.f) installations at
5.2. 69-Bus system optimum locations. The power loss in each branch of the system is
compared with and without DG installations as shown in Fig. 7.
This is a large scale radial distribution system with 69 buses and From Fig. 7, it is observed that the proposed scheme prominently
68 branches. The line and load data of this test system are from [25]. reduces the loss at each line significantly for all types of loads.
The single line diagram of this system is shown in Fig. 5. The total The performance of the 69-bus system with DGs at different
real and reactive power loads of the system are 3.80 MW and power factors is also compared with the results of classical
2.69 MVAr, respectively. The total real power losses for the base methods similar to the 33-bus system and the results are tabulated
case calculated from load flows with the loads CP(0.5), CP(1.0), CP in Table 4. It is perceived that the classical methods GA, PSO and
(1.6), CC(1.0) and CI(1.0) are 51.59 kW, 224.98 kW, 652.42 kW, GA/PSO significantly improve the voltage profile, but the TOC and
188.72 kW and 158.88 kW, respectively. From loss sensitivity fac- percentage power loss reduction are higher than the BFOA. Since
tors, the top three buses 27, 61, and 65 which are found to be more the optimal candidate location for DG placement is also deter-
sensitive are chosen to install DG units in the system. Then the mined by the classical algorithms, the computation time (more
optimal size of DGs at these candidate buses is calculated by BFOA than 25 s) and the number of iterations (more than 45) taken to
in less computation time. The performance of the proposed method converge to the best solution found to be high whereas the
on this test system such as optimal DG size at optimal locations, proposed technique takes only 6.8 s and 15 iterations. From
minimum voltage magnitude, total power loss and percentage Table 4, it is witnessed that for both the power factor, BFOA
power loss reduction before and after DG installation is shown in reduces the total power loss, TOC and voltage deviation index
Table 3. It is observed that minimum 64% of total power loss has better than the SA method. This reveals the superiority and
64 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65

36 37 38 39 40 41 42 43 44 45 46

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

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

28
29
30
52 51 67 66 68 69
31

32 33 34 35

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

Table 3
Performance analysis of the proposed method on 69-bus system at different load models.

Parameters CP load CC load (1.0) CI load (1.0)

Light (0.5) Full (1.0) Heavy (1.6)

Without DG With DG Without DG With DG Without DG With DG Without DG With DG Without DG With DG

DG size (MW) (Bus) 0.1889 (27) 0.2954 (27) 0.4954 (27) 0.2994 (27) 0.2970 (27)
0.2434 (65) 0.4476 (65) 0.5667 (65) 0.3882 (65) 0.2738 (65)
0.6460 (61) 1.3451 (61) 1.8827 (61) 1.3095 (61) 1.2687 (61)
Power loss (kW) 51.59 18.342 224.98 75.238 652.42 205.67 188.72 64.279 158.88 54.838
% Power loss reduction 64.45 66.55 68.47 65.94 65.48
ΔPLDG 0.3555 0.3344 0.3152 0.3406 0.3452
Vworst (p.u.) (bus) 0.9567 (65) 0.9908 (61) 0.9092 (65) 0.9808 (61) 0.8445 (65) 0.9529 (61) 0.9174 (65) 0.9835 (17) 0.9247 (65) 0.9846 (17)
ΔVD 0.0433 0.0092 0.0908 0.0192 0.1555 0.0471 0.0826 0.0165 0.0753 0.0154
TOC ($) 5465 10741 15,546 10,242 9416
Computation time (s) 6.3 6.8 7.2 6.5 6.4

Comparison of Voltage magnitude with and without DGs Comparison of Line Losses with and without DGs

150
1
Voltage Magnitude

With DG Without DG
Power Loss (kW)

0.95 100
(p.u)

0.9 50
With DG Without DG
CI(1.0) 60 70
CC(1.0) 50 0
CP(1.6) 30 40 CI(1.0) 70
20 CC(1.0) 50 60
CP(1.0) 10 CP(1.6) 30 40
CP(0.5) 0 Bus No CP(1.0) 10 20
CP(0.5) 0
Type of Load Line No
Type of Load
Fig. 6. Comparison of voltage magnitude of 69-bus system with different types
Fig. 7. Comparison of line losses of 69-bus system with different types of loads.
of loads.

effectiveness of the proposed technique in finding the optimal bus systems. Also implementation on these systems with different
location and size of DGs for the system with all types of loads in types of loads at different load levels is an added advantage of the
less computational time. proposed scheme. From simulated results, it is studied that due to
the placement of optimal DG size at optimum locations, the power
losses at each line, TOC are reduced and voltage profile at all buses
6. Conclusion are improved substantially. The simulated results obtained using
BFOA is compared with the results of GA, PSO, GA/PSO and SA
In this paper, a new fast combined approach that uses LSF and methods. The results show that the performance of the proposed
BFOA to find the optimum locations and optimal size of DGs has technique on minimization of loss, TOC and voltage stability
been proposed. The proposed scheme is tested on 33-bus and 69- enhancement with DGs at different power factors is better than
 Mohamed Imran A, Kowsalya M / Swarm and Evolutionary Computation 15 (2014) 58–65 65

Table 4
Comparison and performance of 69-bus system with DGs at different power factors.

Method PDG,TLoss (kW) % Loss reduction Vworst (p.u.) (bus) DG location DG size (MW) SDG,T (MVA) Power factor TOC ($)

GA [10] 89.0 60.44 0.9936(57) 21 0.9297 2.9974 Unity 15,343.0

62 1.0752
64 0.9925
PSO [10] 83.2 63.02 0.9901(65) 61 1.1998 2.9879 Unity 15,272.3
63 0.7956
17 0.9925
GA/PSO [10] 81.1 63.95 0.9925(65) 63 0.8849 2.9880 Unity 15,264.4
61 1.1926
21 0.9105
SA [13] 77.1 65.73 0.9811(61) 18 0.4204 2.1813 Unity 11,214.9
60 1.3311
65 0.4298
BFOA 75.23 66.56 0.9808(61) 27 0.2954 2.0881 Unity 10,741.4
65 0.4476
61 1.3451
SA [13] 16.26 92.77 0.9885(61) 18 0.5498 2.3757 0.866 10,352.0
60 1.1954
65 0.3122
BFOA 12.90 94.26 0.9896(64) 27 0.3781 2.3587 0.866 10,265.1
65 0.3285
61 1.3361

[10] M.H. Moradi, M. Abedini, A combination of genetic and particle swarm

the methods compared. From simulated results, it can be con- optimization for optimal DG location and sizing in distribution system, Electr.
cluded that the proposed technique is very accurate in finding the Power Energy Syst. 34 (2012) 66–74.
[11] A.M. El-Zonkoly, Optimal placement of multi-distributed generation units
optimum solutions and this work can also be implemented for all
including different load models using particle swarm optimization, Swarm
systems with n number of buses and all types of loads. Evol. Comput. 1 (2011) 50–59.
[12] M. Ettehadi, H. Ghasemi, S. Vaez-Zadeh, Voltage stability based DG placement
in distribution networks, IEEE Trans. Power Deliv. 28 (1) (2013) 171–178.
Acknowledgment [13] Satish Kumar Injeti, N. Prema Kumar, A novel approach to identify optimal
access point and capacity of multiple DGs in a small, medium and large scale
radial distribution systems, Electr. Power Energy Syst. 45 (2013) 142–151.
This research work was supported by the Vellore Institute of [14] Devender Singh, R.K. Misra, Deependra Singh, Effect of load models in
Technology University, Vellore, India. The authors would like to distributed generation planning, IEEE Trans. Power Syst. 22 (4) (2007)
2204–2212.
thank the reviewers for their valuable comments and recommen- [15] Jen-Hao Teng, A direct approach for distribution system load flow solutions,
dations to improve the quality of the paper. IEEE Trans. Power Deliv. 18 (3) (2003) 882–887.
[16] Deependra Singh, Devender Singh, K.S. Verma, Multiobjective optimization for
DG planning with load models, IEEE Trans. Power Syst. 24 (1) (2009) 427–436.
References [17] Load representation for dynamic performance analysis [of power systems].
IEEE Transactions on Power Systems, 8 (2) (1993) pp. 472, 482. [Link]
[1] H.L. Willis, W.G. Scott, Distributed Power Generation: Planning and Evaluation, org/10.1109/59.260837.
Marcel Dekker, New York, 2000. [18] K. Prakash, M. Sydulu, Particle swarm optimization based capacitor placement
[2] T. Ackerman, G. Anderson, L. Soder, Distributed generation: a definition, Electr. on radial distribution systems, in: Proceedings of IEEE Power Engineering
Power Syst. Res. 57 (2001) 195–204. Society General Meeting, 2007, pp. 1–5.
[3] Viral Rajkumar, D.K. Khathod, Optimal planning of distributed generation [19] K.M. Passino, Biomimicry of bacterial foraging for distributed optimization and
systems in distribution system: a review, Renew. Sustain. Energy Rev. 16 control, IEEE Trans. Control Syst. 22 (3) (2002) 52–67.
(2012) 5146–5165. [20] Ambarish Panda, M. Tripathy, Optimal power flow solution of wind integrated
[4] Acharya Naresh, Mahat Pukar, N. Mithulananthan, An analytical approach for power system using modified bacteria foraging algorithm, Electr. Power
DG allocation in primary distribution network, Electr. Power Energy Syst. 28 Energy Syst. 54 (2014) 306–314.
(2006) 669–678. [21] K. Sathish Kumar, T. Jayabharathi, Power system reconfiguration and loss
[5] D.Q. Hung, N. Mithulananthan, Multiple distributed generators placement in minimization for an distribution systems using bacterial foraging optimization
primary distribution networks for loss reduction, IEEE Trans. Ind. Electron. 60 algorithm, Electr. Power Energy Syst. 36 (2012) 13–17.
(4) (2013) 1700–1708. [22] Sambarta Dasgupta, Swagatam Das, Abraham Ajith, Biswas Arijit, Adaptive
[6] M.M. Aman, G.B. Jasmon, H. Mokhlis, A.H.A. Bakar, Optimal placement and computational chemotaxis in bacterial foraging optimization: an analysis, IEEE
sizing of a DG based on new power stability index and line losses, Electr. Trans. Evol. Comput. 13 (4) (2009) 919–941.
Power Energy Syst. 43 (2012) 1296–1304. [23] R. Srinivasa Rao, K. Ravindra, K. Sathish, S.V.L. Narasimham, Power loss
[7] Sudipta Ghosh, S.P. Ghosal, Saradindu Ghosh, Optimal sizing and placement of minimization in distribution system using network reconfiguration in the
distributed generation in a network system, Electr. Power Energy Syst. 32 presence of distributed generation, IEEE Trans. Power Syst. 28 (1) (2013)
(2010) 849–856. 317–325.
[8] Faruk Ugranli, Engin Karatepe, Multiple-distributed generation under load [24] Juan Andres Martin Garcia, Antonia Jose Gil Mena, Optimal distributed
uncertainty and different penetration levels, Electr. Power Energy Syst. 46 generation location and size using a modified teaching–learning based
(2013) 132–144. optimization algorithm, Electr. Power Energy Syst. 50 (2013) 65–75.
[9] R.K. Singh, S.K. Gowsami, Optimumal location of distributed generations based [25] N.C. Sahoo, K. Prasad, A fuzzy genetic approach for network reconfiguration to
on nodal pricing for profit, loss reduction and voltage rise issue, Electr. Power enhance voltage stability in radial distribution systems, Energy Convers.
Energy Syst. 32 (2010) 637–644. Manage. 47 (2006) 3288–3306.