Solar Energy 140 (2016) 219226

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

A comparison of heuristic optimization techniques for optimal

placement and sizing of photovoltaic based distributed generation in a
distribution system
S. Daud a, A.F.A. Kadir a, C.K. Gan a, A. Mohamed b, Tamer Khatib c,
a
b
c

Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka, Melaka 76100, Malaysia
Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Selangor 43000, Malaysia
Department of Energy Engineering and Environment, An-Najah National University, 97300 Nablus, Palestine

a r t i c l e

i n f o

Article history:
Received 13 September 2016
Received in revised form 28 October 2016
Accepted 7 November 2016

Keywords:
Photovoltaic
Distributed generation
Heuristic techniques
Optimization

a b s t r a c t
This paper presents the application of heuristic optimization techniques for determining optimal placement and sizing of photovoltaic-based distributed generation (PVDG) in a distribution system. The objective functions of the optimization problem, consider a real power loss, voltage deviation, average voltage
total harmonic distortion (THD) and system average voltage dip magnitude (SAVDM). Various heuristic
optimization techniques were applied and compared in order to determine the optimum placements
and sizing of PVDG in IEEE 69-bus radial distribution system. This paper proposed the improved gravitational search algorithm (IGSA) and its performances are compared with two other algorithms such as the
gravitational search algorithm (GSA) and particle swarm optimization (PSO). A comparison of the performances is also made using optimization techniques when PVDGs are fixed at critical buses. The test
results showed that IGSA outperformed GSA and PSO in finding the optimal PVDG locations and sizes.
2016 Published by Elsevier Ltd.

1. Introduction
Distributed generations (DGs) are small generating units that
are installed mainly, near the load centers and in strategic points
of the electric power distribution system. There are two types of
DGs which are inverter-based and non-inverter based DGs. Examples of technologies applied in DGs are micro-turbines, fuel cells,
wind and solar energy. DGs can be connected to satisfy consumers
local demand or supplying energy to the remaining of the electrical
system (Borges and Falcao, 2003). In order to sustain the benefits of
DG, the location, capacity, type and number of DGs must be optimum. Therefore, a feasibility study needs to be performed since a
power system may be affected by the DG installation. In Kadir
et al. (2014), a study was done to evaluate the power quality
impact of renewable type of DGs in a distribution system. Many
research works have been carried out to find the optimization
method for DG installation using analytical, numerical and
heuristic methods (Borges and Falcao, 2003).

Corresponding author.
E-mail addresses: saadahdaud@yahoo.com (S. Daud), fazliana@utem.edu.my
(A.F.A. Kadir), azah_mohamed@ukm.edu.my (A. Mohamed), t.khatib@najah.edu
(T. Khatib).
http://dx.doi.org/10.1016/j.solener.2016.11.013
0038-092X/ 2016 Published by Elsevier Ltd.

Genetic algorithm (GA) has been applied for determining the

optimal allocation of DGs in a meshed distribution system
(Akorede et al., 2011). A hybrid method which is a combination
of particle swarm optimization (PSO) and nonlinear optimal power
flow (OPF) was used to determine the optimum location and size of
multiple DGs (Dias et al., 2012). PSO is widely being used in DG
optimal allocation due to its effectiveness and simplicity and in
El-Zonkoly (2011), it is applied for determining optimal locations
and size of DGs with respect to various load models. Other heuristic optimization techniques applied to identify the location and
size of DGs are such as the ant colony optimization (ACO) in
Lingfeng and Singh (2008), and the harmony search algorithm in
Nekooei et al. (2013). Other research works on optimal allocation
of DGs, consider the use of analytical method in Hung and
Mithulananthan (2013), Hung et al. (2010) and Elsaiah et al.
(2014), which is said to be easily implemented and perform faster
than the other optimization methods (Georgilakis and
hatziargyriou, 2013). Numerical methods applied for the optimal
allocation of DGs are such as ordinal optimization, sequential
quadratic programming and nonlinear programming (Jabr and
Pal, 2009; Morales et al., 2011; Atwa and El-Saadany, 2011).
The objective of this paper is to identify the best optimization
method for determining the optimal location and sizing of photovoltaic distributed generation (PVDG) in a distribution system

220

S. Daud et al. / Solar Energy 140 (2016) 219226

and implementing a comparative study on three heuristic optimization techniques, namely, improved gravitational search algorithm (IGSA), PSO and GSA. The optimization problem considers
minimization of system real power loss, average voltage deviation,
average voltage total harmonic distortion (THDv) and system average voltage dip magnitude (SAVDM). A comparison is also made
with and without optimum location and capacity of PVDG in the
IEEE 69-bus radial distribution system. Methods such as NewtonRaphson load flow, harmonic load flow and method of fault position are combined with the heuristic optimization techniques to
obtain the best solution or optimum fitness function.

2. Problem formulation
An optimization problem can be defined as the maximization of
an objective function while satisfying a number of equality and
inequality constraints. A multi-objective optimization is formulated for optimum PVDG placement and sizing by minimizing total
power loss, voltage deviation, average THDv and system average
voltage dip magnitude (SAVDM), which is expressed as follows:

F min cploss bTHDv vV dev gSAVDM

From Eq. (1), F can be defined as the fitness function, THDv is the
percentage of the average THDv, V dev is the percentage of voltage
deviation at all system buses, and SAVDM is the percentage of system average voltage dip magnitude. c is the coefficient of ploss , b is
the coefficient of THDv , v is the coefficient of V dev and g is the coefficient of SAVDM. The sum of the coefficient factor method is used
to decide the relative importance of the objectives in order to
obtain the best optimum solution. The coefficient factors are
assumed to be 0.25 each since all of the objectives are important
in order to install DGs with lower power loss, more stable voltage
and enhanced power quality.
The total real power loss, average THDv , V dev and SAVDM are
given by:

ploss

n
X
plossi

i1

Pm
THDv

V dev

i1 THDv i

V iref  V i
Vi

SAVDM bi1 

GSA was proposed as a solution to solve optimization problems

by Rashedi et al. (2009). The development of this algorithm is
based on the law of motion and Newtons law of gravity. The algorithm comprises of a collection of search agents that are considered
as objects that interact with each other through a gravitational
force. The performance of agents is measured by their masses
and a global movement due to gravitational force causes the
objects to move towards other objects with heavier masses. The
slow movement of heavier masses guarantees the utilization step
of the algorithm and corresponds to good solution. GSA consists
of four parameters that are position, inertial mass, active and passive gravitational mass. The position represents the solution, while
the gravitational and inertial masses are determined using a fitness
function. The optimum solution in the search spaces represented
by the heaviest mass (Sabri et al., 2013). The computational procedures of GSA consider the following equations:
 ith agents position can be given by:

X i x1i ; . . . ; xdi ; . . . ; xni ;

X ni

where
is the position of each ith agent, Size is the capacity of
PVDG, VC is the voltage control and Place is the location of the
PVDG.
 The update of gravitational constant (G) is given by:

Gt G0

T t
T

10

where Gt is the gravitational constant value at each time, t and

G0 is the gravitational constant value at the first quantuminterval of time, t0 . T is the total number of iterations.
 The update of mass (M) considering a weighting range between
0 and 1, corresponds to the following fitness:

fitnessi t  worstt
bestt  worstt

mi t
M i t Pn
j1 mj t

where n is the number of lines, m is the number of buses, V iref is the

reference voltage at bus i and V i is the actual voltage at bus i. VDAi is
the voltage dip amplitude at bus i. The bus voltage must be kept
within permissible operating range throughout the optimization
process and the total harmonic level at each bus must be less than
or equal to the maximum permissible level. The inequality constraints are expressed as follows:

V min 6 jV i j 6 V max

THDv i 6 THDv max

V min is the lower bound and V max is the upper bound of bus voltage
limits. jV i j is the RMS value of the i th bus voltage and THDv max Is the
maximum permissible level at each bus, which is 5%.



X ni Size; VC; Place1 ; Size; VC; Place2 ;... ; Size;VC;Placen

mi t

for i 1; 2; . . . ; n

 In the meanwhile, the detailed position of each ith agent is

given by:

Pm

i1 VDAi
m

3. Improved gravitational search algorithm

11

12

where fitnessi t is the fitness value of the agent i a time t, and

worstt, and bestt are the maximum and minimum fitness values, respectively.
 The update of K best is given by:

K best K best

final

T t
100  K best
T


final

13

 The total force (F) is calculated by using the following

equations:

M ij  d
xj  xdi
Rij e
r
2
XD 

d
d

x

x
Rij X i ; X j 2
j
i
d1

F dij G

e small coefficient; 252 :

14
15
16

 As a stochastic approach, the total force on agent, i in the d

dimension a randomly weighted sum of the dth components of
the force exerted from other agents. The total force is given as

221

S. Daud et al. / Solar Energy 140 (2016) 219226

F di t

N
X
j2K best ;j i

randj F dij t

17

where randj is a random number in the interval between 0 and 1.

 The acceleration, a is formulated by:

adi

F di
Mi

18

 The formula for velocity,

d
i t

1 randi v

v is given by:

d
i t

d
i t

19

 The position, x is updated by using:

xdi t 1 xdi t v di t 1

20

21

where is the chaotic value, q is a control parameter with real

value range between 0 and 4, and t represents the number of iterations. By combining Eqs. (21) and (10), a new gravitational constant
formula is derived and described as follows,

Gt G0

T t
T

kki 1; if Mk > Mi and Rik 6 s

Otherwise it equal to 0.
agent to the agent kth.

F dij kki G

The execution of GSA depends on two opposite objectives:

exploration and exploitation. Exploration is the potential of
expanding the global investigation of the search space while
exploitation is the ability to define the optima around an optimum
solution. The algorithm must utilize exploration to prevent from
trapping in a local optimum. As the seeking process continues,
exploration fades out and exploitation fades in to enable the solution found to be superior (Kadir et al., 2014). GSA has some weaknesses in the searching process, such as the balance control
between exploration and exploitation in which more exploration
will cause premature convergence while more exploitation can
affect the convergence rate. In addition, the best agent is still
exploring the global space, even it is in its best position (Ibrahim
et al., 2012). Thus, to overcome these weaknesses, an IGSA is proposed to improve the quality of the solution and enable fastest
convergence rate. A typical chaotic system is employed to develop
the IGSA using the following equation:

t 1 qt1  t; 0 6 1 6 1

In GSA, high dependence on random variable in the calculation

might create less significant impact in the implementation of gravitational theory on the search algorithm. Therefore, Eq. (17) is
eliminated in order to decrease the reliance on randomize exploration process. Thus k can be obtained using the following
conditions,

22

Mi  Mj  d
xj  xdi
Rij 

23

s is the maximum distance of the ith

24

These equations result in effective acceleration on the ith agent

by the heavier kth agent.
4. Proposed method for optimal PVDG placement and sizing
In the proposed method for optimal PVDG placement and sizing, IGSA is employed together with the MATPOWER NewtonRaphson load flow, harmonic load flow and fault position method.
The MATPOWER Newton-Raphson load flow is used to calculate
power loss and voltage deviation which is part of the fitness
function of the IGSA. The fitness of average THDv is obtained from
harmonic load flow. The general steps of the harmonic load flow is
depicted in Fig. 1.
In this simulation, harmonic distortion is said to be injected by
the PVDG that acts as a harmonic producing device in the system.
The typical harmonic spectrum of inverter-based DG is obtained
from Kadir et al. (2014) and that the voltage THD for PVDG is
injected such that it does not exceed the voltage THD limit of 5%
as stated in the IEEE standard 519-1992.
The fault position method is used to determine the expected
number of sags by calculating the voltage at a given fault position.
It is based on the short-circuit simulations that are repeated at all
the system buses. If the fault voltage is below a preset threshold
value, it is considered as a voltage dip. In this paper, a complete
matrix formulation of fault-position method is applied for voltage
dip characterization. The dips are stored in dip matrices that contain only the dips in all the buses where faults occur. The worst

Fig. 1. Flow chart of the harmonic load flow.

222

S. Daud et al. / Solar Energy 140 (2016) 219226

case scenario of short-circuit phenomena considered is the threephase fault at all the buses in the system.
Voltage dip amplitude (VDA) which is the amplitude of the voltage dip is considered as the fault voltage approximated by the rectangular waveform placed at the minimum of the time evolution of
the RMS fault voltage (Carpinelli et al., 2007). VDA has to be kept
maximized in order to reduce the dip phenomenon. However, in
this optimization problem, all the objectives must be minimized
and therefore VDM term is being used instead. Fig. 2 illustrates
the different of VDA and VDM. The fault rate at the buses is
assumed to be 0.04 faults per year (Park et al., 2010). The stepby-step procedure of the FPM is shown in Fig. 3.

The proposed method to determine the optimal location and

capacity of PVDG is tested on the IEEE 69-bus radial distribution
system. The bus data of this test system is obtained from
Rugthaicharoencheep (2009). The total load of the system is
3.8 MW and 2.69 MVAr. The base power is set to 100 MW with
12.66 kV rated voltage. The minimum and maximum voltage limits
are 0.9 p.u and 1.05 p.u, respectively. There are some assumptions
made in the simulations which are described as follows:
 The simulation is conducted based on snapshot at the peak
load condition.
 The cost is not considered.
 The maximum penetration level of renewable DGs is 52%.
 The PVDG will inject active power only since PVDG is
renewable source.
 The maximum number of DG connected to the system is 2.
 The base case system does not consider harmonic source,
harmonic distortion and faults.
 30 independent runs are conducted to measure the frequency in reaching optimum solutions.
In the simulations, several parameters of IGSA, GSA and PSO are
set as follows:
 The initial gravity constant G0 is set to 100.
 Kbest is monotonously decreasing from the maximum value
of 1002.0%.
 s is set to 30% of the ith agent.
 Population size is set to 50.
The boundary constraints for the control parameters are set as
follows:

Fig. 2. Approximated voltage amplitude during dip phenomenon (Carpinelli et al.,

2007).

 0.5 MW 6 DG capacity 6 2.0 MW.

 Bus 2 6 DG location 6 Bus 69.
 0.99 p.u 6 DG voltage control 6 1.01 p.u.

Fig. 3. Flow chart of the fault position method (FPM).

223

S. Daud et al. / Solar Energy 140 (2016) 219226

The step-by-step procedure of the simulations is:

i. Input all the system information.
ii. Apply optimization techniques which are IGSA, GSA or PSO.
iii. Conduct load flow analysis using MATPOWER, harmonic
load flow and voltage dip analysis using th fault position
method.
iv. Calculate the objective functions.
v. Ensure the THDv level is within 5% and the bus voltage in the
range between 0.9 and 1.05 p.u.
vi. Determine the optimum solution based on the minimum
value of the objective function.
5. Results and discussion
The IEEE 69-bus radial distribution system is used as the test
system in this study. The performance of the proposed IGSA is
compared with PSO and GSA. The maximum iteration for the three
techniques is set at 300. In the comparative performance study,
five cases are being examined:
 Case I: Base case to identify the total real power loss and
voltage deviation before the installation of PVDGs.
 Case II: A single PVDG is installed at optimum capacity in
the test system.

 Case III: Two PVDGs are installed and sized optimally in the
test system.
 Case IV: A single PVDG is placed at Bus 61 with optimum
capacity in the system. Bus 61 is the weakest bus in the system from voltage collapse point of view. In Kamaruzzaman
and Mohamed (2015), the ranking of the weak buses starts
from the weakest bus, Bus 61, Bus 68, Bus 45 and Bus 17.
 Case V: Two PVDGs are placed at bus 61 and Bus 68 with
optimum capacity in the system.
5.1. Case I: base case
The test system is simulated using the MATPOWER software
program to determine the power loss and voltage deviation in
the system. The system is assumed to be free from harmonic distortion and no fault occurs in the system. The system power loss
at base case is found to be 0.23 MW and the base case average voltage deviation is 0.027%.
5.2. Case II: optimal placement and sizing of single PVDG
The optimization results for Case II are shown in Table 1 and the
convergence characteristics of IGSA, GSA and PSO are shown in
Fig. 4.

Table 2
Optimization results for Case III.
Table 1
Optimization results for Case II.

Algorithm

GSA

PSO

IGSA

64
53
1.636
1.367
1.0072
0.994

38
63
1.393
0.511
1.0078
0.992

47
21
0.854
1.845
1.0049
1.003

Algorithm

GSA

PSO

IGSA

Optimization results
PVDG1 location
PVDG1 Size (MW)
PVDG1 voltage (p.u)

3
1.925
0.9960

15
0.518
1.0008

65
0.937
1.0002

Optimization results
PVDG1 location
PVDG2 location
PVDG1 size (MW)
PVDG2 size (MW)
PVDG1 voltage (p.u)
PVDG2 voltage (p.u)

Performances
Worst fitness
Best fitness
Average fitness
Standard deviation
Average computational time (s)

0.335
0.322
0.327
0.004
306.23

0.336
0.322
0.327
0.003
305.87

0.332
0.322
0.326
0.002
305.53

Performances
Worst fitness
Best fitness
Average fitness
Standard deviation
Average computational time (s)

0.388
0.313
0.358
0.022
317.209

0.379
0.305
0.350
0.024
306.065

0.411
0.286
0.361
0.027
306.000

Overall impacts
Losses (%)
Voltage deviation (%)
THDv (%)
SAVDM (%)

0.195
0.0178
0.148
0.927

0.194
0.0174
0.149
0.929

0.195
0.0173
0.149
0.929

Overall impacts
Losses (%)
Voltage deviation (%)
THDv (%)
SAVDM (%)

0.083
0.0095
0.226
0.934

0.117
0.0099
0.161
0.934

0.037
0.010
0.163
0.934

Fig. 4. Convergence characteristics for single PVDG placement in IEEE 69-bus system.

224

S. Daud et al. / Solar Energy 140 (2016) 219226

Fig. 5. Convergence characteristics when two PVDGs are placed in IEEE 69-bus system.

GSA, PSO and IGSA give similar best fitness value of 0.3222. The
optimal placement for PVDG by GSA is at bus 3 with capacity of
1.925 MW and 0.996 p.u. voltage. For PSO, bus 15 is the optimal
PVDG placement with capacity of 0.518 MW and 1.0008 p.u. voltage. However, the solution for IGSA is that bus 65 is the optimal
PVDG location with optimal size of 0.937 MW and 1.0002 p.u.
Table 3
Optimization results for Case IV.

voltage. The average computational time taken for the three optimization techniques are almost the same. In term of accuracy, IGSA
is considered most accurate because it gives the lowest standard
Table 4
Optimization results for Case V.
Algorithm

GSA

PSO

IGSA

61
68
0.919
0.914
1.003
0.994

61
68
1.873
1.029
0.992
1.009

61
68
1.332
1.676
0.994
1.007

Algorithm

GSA

PSO

IGSA

Optimization results
PVDG1 location
PVDG1 size (MW)
PVDG1 voltage (p.u)

61
0.931
0.995

61
1.607
1.0089

61
1.429
1.0066

Optimization results
PVDG1 Location
PVDG2 location
PVDG1 size (MW)
PVDG2 size (MW)
PVDG1 voltage (p.u)
PVDG2 voltage (p.u)

Performances
Worst fitness
Best fitness
Average fitness
Standard deviation
Average computational time (s)

0.353
0.349
0.351
0.0008
307.280

0.354
0.349
0.351
0.0010
358.734

0.352
0.349
0.350
0.0008
300.241

Performances
Worst fitness
Best fitness
Average fitness
Standard deviation
Average computational time (s)

0.396
0.382
0.388
0.0039
361.542

0.401
0.383
0.389
0.0048
348.904

0.397
0.381
0.389
0.0042
347.197

Overall impacts
Losses (%)
Voltage deviation (%)
THDv (%)
SAVDM (%)

0.109
0.0102
0.345
0.934

0.113
0.0102
0.341
0.934

0.110
0.0102
0.344
0.934

Overall impacts
Losses (%)
Voltage deviation (%)
THDv (%)
SAVDM (%)

0.058
0.0058
0.527
0.935

0.129
0.0049
0.462
0.935

0.093
0.0064
0.489
0.935

Fig. 6. Convergence characteristics when one PVDG is placed at bus 61 in IEEE 69-bus system.

S. Daud et al. / Solar Energy 140 (2016) 219226

225

Fig. 7. Convergence characteristics when two PVDGs are placed at bus 61 and bus 68 in IEEE 69-bus system.

deviation of 0.002. Fig. 4 shows the graph plot of iteration versus

fitness function for three algorithm which are the GSA, PSO and
IGSA. From the figure, it can be identify that all the algorithms have
approximate value at the beginning of the iteration. The fitness
function improves as the number of iterations increases until it
reaches a constant after certain number of iterations.
5.3. Case III: optimal placement and sizing of two PVDGs
The summary of the results for Case III are tabulated in Table 2
and the convergence characteristics of the three optimization techniques are depicted in Fig. 5. In Case III, when two PVDGs are
installed in the system, IGSA performs better than GSA and PSO
and it gives the lowest best fitness function value of 0.286. The best
candidates for PVDG placement are at bus 47 and bus 21 with the
PV capacities of 0.854 MW and 1.845 MW, respectively. The graph
plotted in Fig. 5 shows that the IGSA provides minimum fitness
value compared with PSO and GSA. The fitness value improves
with the number of iterations. IGSA and GSA reach constant value
after about 25 iterations while PSO reaches constant value after 60
iterations.
5.4. Case IV: single PVDG installed at the critical bus
Case IV depicts the performances of each technique when PVDG
is located at the critical or weakest bus in the system which is bus 61.
The location of PVDG is fixed at bus 61 and the algorithms will search
for the optimum size considering the fitness function. The optimization results are shown in Table 3 and the convergence characteristic
is shown in Fig. 6. The best fitness values of the three optimization
techniques are the same and IGSA gives the fastest average elapsed
time compared to PSO and GSA. The graph in Fig. 6 shows the
improvement fitness values of IGSA, GSA and PSO technique as the
number of iterations increases. For this case, all the algorithms
reaches the same minimum values which is 0.349.
5.5. Case V: two PVDGs installed at the two critical buses
In Case V, two PVDGs are installed at two critical buses of 61
and bus 68. The optimization results of Case V are tabulated in
Table 4 while Fig. 7 depicts the convergence characteristic of the
three optimization techniques. The results showed that IGSA performs better and the fastest to obtain the optimum solution compared to GSA and PSO. Fig. 7 gives the graph plot of iteration and
fitness function for GSA, PSO and IGSA at two critical buses in
the system. The figures show that IGSA obtains minimum fitness

value than PSO and GSA. IGSA is more reliable with the optimum
fitness function of 0.3781.
From the results of all cases, it can be shown that installing
PVDG at the optimum location and capacity can minimize system
losses and therefore contribute to a more dependable power distribution system. It is noted that IGSA gives better performance compared to PSO in terms of computational time and more accurate
solutions. The results of the IGSA is better when PVDGs are placed
at the critical buses because power loss and voltage deviation are
greatly reduced. However, the average THD are slightly higher
for these cases (case IV and V) and therefore not appropriate for
power quality improvement.
6. Conclusion
A comparative study has been carried out to evaluate the performance of IGSA, GSA and PSO in determining the optimal placement and sizing of PVDG in a distribution system. IGSA is an
improved heuristic optimization technique that has several advantages such as adaptive learning rate, better execution and fast convergence in finding the optimal solution. The objective function of
the optimization considers total real power loss, average voltage
deviation, average THDv and average voltage dip magnitude. From
the results, IGSA outperformed PSO and GSA with better and accurate solution in a faster time. The number of iterations play an
important role in order to obtain a stabilized and most accurate
values. All of the cases plotted the improvement values of the fitness function as the number of iteration increases. This paper also
shows that it is better to search for an optimum locations and sizes
of PVDG in the system, instead of fixing the locations based on a
certain issue such as the voltage collapse point of view.
Acknowledgments
This research work was supported by Univetsiti Teknikal
Malaysia Melaka, Malaysia, under research Grant No. RAGS/2014/
TK03/FKE/B00048, and the Ministry of Education of Malaysia. The
first author would like to gratefully acknowledge the financial support provided by Universiti Teknikal Malaysia Melaka under UTeM
Zamalah Scheme.
References
Akorede, M.F., Hizam, H., Aris, I., Kadir, M.Z.A.A., 2011. Effective method for optimal
allocation of distributed generation units in meshed electric power systems. IET
Gener. Transm. Distrib. 5, 276287.

226

S. Daud et al. / Solar Energy 140 (2016) 219226

Atwa, Y.M., El-Saadany, E.F., 2011. Probabilistic approach for optimal allocation of
wind-based distributed generation in distribution systems. IET Renew. Power
Gener. 5, 7988.
Borges, C.L.T., Falcao, D.M., 2003. Impact of distributed generation allocation and
sizing on reliability, losses and voltage profile. In: IEEE Bologna Power Tech
Conference Proceedings. 1-5.
Carpinelli, G., Caramia, P., Perna, C.D., Varilone, P., Verde, P., 2007. Complete matrix
formulation of fault-position method for voltage-dip characterization. IET
Gener. Transm. Distrib. 1, 5664.
Dias, B.H., Oliveira, L.W., Gomes, F.V., Silva, I.C., Oliveira, E.J., 2012. Hybrid heuristic
optimization approach for optimal distributed generation placement and sizing.
In: IEEE Power and Energy Society General Meeting, pp. 16.
Elsaiah, S., Benidris, M., Mitra, J., 2014. Analytical approach for placement and sizing
of distributed generation on distribution systems. IET Gener. Transm. Distrib. 8,
10391049.
El-Zonkoly, A.M., 2011. Optimal placement of multi-distributed generation units
including different load models using particle swarm optimization. IET Gener.
Transm. Distrib. 5, 760771.
Georgilakis, P.S., hatziargyriou, N.D., 2013. Optimal distributed generation
placement in power distribution networks: models, methods, and future
research. IEEE Trans. Power Syst. 28, 34203428.
Hung, Duong Quoc, Mithulananthan, N., 2013. Multiple distributed generator
placement in primary distribution networks for loss reduction. IEEE Trans. Ind.
Electron 60, 17001708.
Hung, Duong Quoc, Mithulananthan, N., Bansal, R.C., 2010. Analytical expressions
for DG allocation in primary distribution networks. IEEE Trans. Energy Convers.
25, 814820.
Ibrahim, A., Mohamed, A., Shareef, H., 2012. A novel quantum-inspired binary
gravitational search algorithm in obtaining optimal power quality monitor
placement. J. Appl. Sci. 12, 822830.

Jabr, R.A., Pal, B.C., 2009. Ordinal optimization approach for locating and sizing of
distributed generation. IET Gener. Transm. Distrib. 3, 713723.
Kadir, A.F.A., Mohamed, A., Shareef, H., Ibrahim, A.A., Khatib, T., Elmenreich, W.,
2014. An improved gravitational search algorithm for optimal placement and
sizing of renewable distributed generation units in a distribution system or
power quality enhancement. J. Renew. Sustain. Energy 6. 033112-1.
Kadir, A.F.A., Khatib, T., Elmenreich, W., 2014. Integrating photovoltaic system in
power system: power quality impacts and optimal planning challenges. Int. J.
Photoenergy 2014, 17.
Kamaruzzaman, Z.A., Mohamed, A., 2015. Static voltage stability analysis in a
distribution system with high penetration of photovoltaic generation. Przeglad
Elektrotechniczny 2015, 113117.
Lingfeng, W., Singh, C., 2008. Reliability-constrained optimum placement of
reclosers and distributed generators in distribution networks using ant colony
system algorithm. IEEE Trans. Syst. Man Cybern. Part C Appl. Rev. 38, 757764.
Morales, J.L., Nocedal, J., Wu, Y., 2011. A sequential quadratic programming
algorithm with an additional equality constrained phase. IMA J. Numerical
Anal., 129
Nekooei, K., Farsangi, M.M., Nezamabadi-pour, H., Lee, K.Y., 2013. An improved
multi-objective harmony search for optimal placement of DGs in distribution
systems. IEEE Trans. Smart Grid 4, 557567.
Park, C.H., Hong, J.H., Jang, G., 2010. Assessment of system voltage sag performance
based on the concept of area of severity. IET Gener. Transm. Distrib. 4, 683693.
Rashedi, E., Nezamabadi-pour, H., Saryazdi, S., 2009. GSA: a gravitational search
algorithm. Inf. Sci. (Ny) 179, 22322248.
Rugthaicharoencheep, N., 2009. Feeder reconfiguration with dispatchable distributed
generators in distribution system by tabu search. In: Proceeding of the 44th
International Universities Power Engineering Conference. UPEC, pp. 15.
Sabri, N.M., Puteh, M., Mahmood, M.R., 2013. A review of gravitational search
algorithm. Int. J. Adv. Comput. Appl. 5, 139.