Energy Conversion and Management 59 (2012) 86–95

Contents lists available at SciVerse ScienceDirect

Energy Conversion and Management

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

Optimal power flow using gravitational search algorithm

Serhat Duman a, Uğur Güvenç a,⇑, Yusuf Sönmez b, Nuran Yörükeren c
a
Department of Electrical Education, Technical Education Faculty, Duzce University, Duzce, Turkey
b
Department of Electrical Technology, Gazi Vocational College, Gazi University, Ankara, Turkey
c
Department of Electrical Engineering, Engineering Faculty, Kocaeli University, Kocaeli, Turkey

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

Article history: In this paper, gravitational search algorithm (GSA) is proposed to find the optimal solution for optimal
Received 3 March 2011 power flow (OPF) problem in a power system. The proposed approach is applied to determine the optimal
Received in revised form 4 February 2012 settings of control variables of the OPF problem. The performance of the proposed approach examined and
Accepted 27 February 2012
tested on the standard IEEE 30-bus and 57-bus test systems with different objective functions and is com-
pared to other heuristic methods reported in the literature recently. Simulation results obtained from the
proposed GSA approach indicate that GSA provides effective and robust high-quality solution for the OPF
Keywords:
problem.
Optimal power flow
Gravitational search algorithm
Ó 2012 Elsevier Ltd. All rights reserved.
Optimization
Power systems

1. Introduction researchers. The OPF problem has been solved by using traditional
and evolutionary based algorithms. Conventional optimization
Optimal power flow (OPF) has become one of the most impor- techniques such as interior point method, linear programming,
tant problems and commonly studied subjects for optimal opera- nonlinear programming and quadratic programming have been
tion and planning processes of modern power systems. Recently, implemented to solve the OPF problem [6–11]. However, the dis-
the problem of OPF has received much attention by many research- advantage of these techniques is that it is not possible to use these
ers. The OPF is the fundamental tool that enables electric utilities techniques in practical systems because of nonlinear characteris-
to specify economic operating and secure states in power systems. tics such as valve point effects, prohibited operating zones and
Main objective of the OPF problem is to optimize a chosen objec- piecewise quadratic cost function. Therefore, it becomes necessary
tive function such as fuel cost, piecewise quadratic cost function, to improve the optimization methods that are capable of overcom-
fuel cost with valve point effects, voltage profile improvement, ing these disadvantages and handling such difficulties [12]. Lately,
voltage stability enhancement, through optimal adjustments of many population-based optimization techniques have been used
the power system control variables while at the same time satisfy- to solve complex constrained optimization problems. These tech-
ing various system operating such as power flow equations and niques have been increasingly applied for solving power system
inequality constraints [1,2]. The equality constraints are as nodal optimization problems such as economic dispatch, optimal reac-
power balance equations, while the inequality constraints are as tive power flow and OPF in decades. Some of the population-based
the limits of all control or state variables. The control variables in- methods have been proposed for solving the OPF problem success-
volve the tap ratios of transformer, the generator real powers, the fully, such that genetic algorithm [13], improved genetic algorithm
generator bus voltages and the reactive power generations of VAR [14], tabu search [4], particle swarm [15], differential evolution
sources. State variables involve the generator reactive power out- algorithm [16], simulated annealing [17], evolutionary program-
puts, the load bus voltages and network line flows [3,4]. In general ming [18].
the OPF problem is a large-scale, highly constrained nonlinear non- Anitha et al. presented a new variation of particle swarm opti-
convex optimization problem. mization algorithm to solve the OPF problem with IEEE 30-bus sys-
Dommel and Tinney firstly presented the formulation optimal tem. The obtained results of proposed approach are compared with
power flow [5]. Then this topic has been handled by many tabu search (TS), simulated annealing (SA), evolutionary program-
ming (EP), improved evolutionary programming (IEP) and particle
swarm optimization (PSO) methods [19]. Dutta and Sinha used PSO
⇑ Corresponding author.
algorithm to solve voltage stability constrained multi-objective
E-mail addresses: [email protected] (S. Duman), ugurguvenc@duzce.
edu.tr (U. Güvenç), [email protected] (Y. Sönmez), [email protected]
OPF problem. The proposed method successfully applied for IEEE
(N. Yörükeren). 30, IEEE 57 and IEEE 118-bus systems by the authors [20]. Bakirtzis

0196-8904/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.enconman.2012.02.024
 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95 87

et al. utilized enhanced genetic algorithm for solving OPF. The x can be represented as:
authors tested their approach on IEEE 30-bus system and the three
area IEEE RTS-96 which is a 73 bus, 120-branch system [21]. Basu xT ¼ ½PG1 ; V L1 . . . V LNPQ ; Q G1 . . . Q GNPV ; Sl1 . . . S1NTL  ð4Þ
offered differential evolution algorithm (DE) to solve OPF problem where NPV defines the number of voltage controlled buses; NPQ and
with flexible alternating current transmission systems (FACTSs) NTL depict the number of PQ buses, the number of transmission
devices. The author has proposed to minimize generator fuel cost lines respectively.
with FACTS devices such as thyristor-controlled series capacitor In a similar way, the vector of control variables u can be ex-
(TCSC) and thyristor-controlled phase shifter (TCPS) in IEEE 30- pressed as:
bus systems [22]. Basu used multi-objective differential evolution
algorithm to solve the OPF problem with FACTS devices in IEEE uT ¼ ½PG2 . . . PGNG ; V G1 . . . V GNG ; Q C1 . . . Q CNC ; T 1 . . . T NT  ð5Þ
30-bus and IEEE 57-bus systems. The results were compared with I. PG defines the active power output of generators at PV bus.
the literature by the author [23]. Madhad et al. investigated effi- II. VG depicts the terminal voltages at generation bus bars.
cient parallel genetic algorithm applied to the OPF problem for III. QC represents the output of shunt VAR compensators.
large-scale system with shunt FACTS. In this study, authors are pre- IV. T stands for the tap setting of the tap regulating
sented three test systems IEEE 30-bus, IEEE 118-bus and 15 gener- transformers.
ation units with prohibited zones and compared with results of
other the literature [24]. where NT and NC define the number of tap regulating trans-
One of the recently improved heuristic algorithms is the gravi- formers and number of shunt VAR compensators, respectively.
tational search algorithm (GSA), which is based on the Newton’s
law of gravity and mass interactions. GSA has been verified high 2.1. Constraints
quality performance in solving different optimization problems
in the literature [25–29]. The most substantial feature of GSA is 2.1.1. Equality constraints
that gravitational constant adjusts the accuracy of the search, so g Defines equality constraints representing typical load flow
it is to speed up solution process [30–32]. Furthermore, GSA is equations:
memory-less, it works efficiently like the algorithms with memory
[31,32]. In this paper, a newly developed heuristic optimization P
NB  
PGi  PDi  V i V j Gij cosðdi  dj Þ þ Bij sinðdi  dj Þ ¼ 0 ð6Þ
called GSA method is proposed to solve the OPF problem which j¼1
is formulated as a nonlinear optimization problem with equality
and inequality constraints in a power system. The objective func- P
NB  
tions are minimization of fuel cost such as quadratic cost function, Q Gi  Q Di  V i V j Gij sinðdi  dj Þ þ Bij cosðdi  dj Þ ¼ 0 ð7Þ
j¼1
piecewise quadratic cost function, cost function with valve point
effect, improvement of the voltage profile and improvement of where Vi and Vj are the voltages of ith and jth bus respectively, PGi
voltage stability in both nominal and contingency situations. The and QGi the active and reactive power of ith generator, PDi and QDi
performance of the proposed approach is sought and tested on the demand of active and reactive power of ith bus and Gij, Bij and
the standard IEEE 30-bus and 57-bus test systems. Obtained simu- dij are the conductance, susceptance and phase difference of volt-
lation results demonstrate that the proposed method provides very ages between ith and jth bus and NB is the total number of buses.
remarkable results for solving the OPF problem. The results have
been compared to those reported in the literature. 2.1.2. Inequality constraints
The rest of the paper is organized as follows: Section 2 defines h defines inequality constraints that contain:
the mathematical formulation of optimal power flow problem
and in Section 3 the proposed approach GSA is presented. Section i. Generator constraints: Generator voltage, active and reac-
4 presents the results of simulation and compares techniques in tive outputs ought to be restricted by their lower and upper
the literature to solve the case studies of optimal power flow prob- limits as follows:
lems with IEEE 30-bus and 57-bus test systems. Finally, conclusion
V min max
Gi 6 V Gi 6 V Gi ; i ¼ 1; 2; . . . ; NPV
of the implementation of the proposed approach is illustrated in
Section 5. Pmin max
Gi 6 P Gi 6 P Gi ; i ¼ 1; 2; . . . ; NPV ð8Þ
Q min
Gi 6 Q Gi 6 Q max
Gi ; i ¼ 1; 2; . . . ; NPV
2. Mathematical problem formulation of OPF
where V min
and
Gi V max
are the minimum and maximum gener-
Gi

The OPF is a nonlinear optimization problem. The essential goal ator voltage of ith generating unit; P min
Gi and Pmax
Gi the mini-
of the OPF is to minimize the settings of control variables in terms mum and maximum active power output of ith generating
of a certain objective function subjected to various equality and unit and Q min max
Gi and Q Gi are the minimum and maximum reac-
inequality constraints. In general, the OPF problem can be mathe- tive power output of ith generating unit.
matically formulated as follows: ii. Transformer constraints: Transformer tap settings ought to
be restricted by their lower and upper limits as follows:
Min Fðx; uÞ ð1Þ
T min
i 6 T i 6 T max
i ; i ¼ 1; :2; . . . ; NT ð9Þ
subject to gðx; uÞ ¼ 0 ð2Þ
hðx; uÞ 6 0 ð3Þ where T min
and
i T max
define minimum and maximum tap set-
i
tings limits of ith transformer.
where F is the objective function to be minimized, x and u are vec- iii. Shunt VAR compensator constraints: Shunt VAR compensa-
tors of dependent and control variables respectively. x is the vector tors ought to be restricted by their lower and upper limits
of dependent variables including: as follows:
I. Generator active power output at slack bus PG1.
Q min max
Ci 6 Q Ci 6 Q Ci ; i ¼ 1; 2; . . . ; NC ð10Þ
II. Load bus voltage VL.
III. Generator reactive power output QG. where Q min
Ciand Q max
define minimum and maximum Var
Ci
IV. Transmission line loading Sl. injection limits of ith shunt capacitor.
88 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95

iv. Security constraints: These contain the constraints of volt- where Mi is the mass of the object i, Mj the mass of the object j, G(t)
age magnitudes at load buses and transmission line load- the gravitational constant at time t, e a small constant and Rij(t) is
ings. Voltage of each load bus ought to be restricted by the Euclidian distance between i and j objects defined as follows:
their lower and upper operating limits. Line flow through
Rij ðtÞ ¼ kX i ðtÞ; X j ðtÞk2 ð17Þ
each transmission line ought to be restricted by their capac-
ity limits. These constraints can be mathematically formu- The total force acting on the ith agent ðF di ðtÞÞ is calculated as
lated as follows: follows:
P
N
V min
Li 6 V Li 6 V max
Li ; i ¼ 1; 2; . . . ; NPQ ð11Þ F di ðtÞ ¼ randj F dij ðtÞ ð18Þ
j2kbest;j–i

Sli 6 Smax
li ; i ¼ 1; 2; . . . ; NTL ð12Þ where randj is a random number between interval [0, 1] and kbest is
where V min
and V max
minimum and maximum load voltage of ith the set of first K agents with the best fitness value and biggest mass
Li Li
unit. Sli defines apparent power flow of ith branch. Smax defines max- In order to find the acceleration of the ith agent, at t time in the
li
imum apparent power flow limit of ith branch. dth dimension law of motion is used directly to calculate. In accor-
The inequality constraints of dependent variables contain load dance with this law, it is proportional to the force acting on that
bus voltage magnitude; real power generation output at slack agent, and inversely proportional to the mass of the agent. adi ðtÞ
bus, reactive power generation output and line loading may be in- is given as follows:
cluded into an objective function as quadratic penalty function F di ðtÞ
method. In this method, a penalty factor multiplied with the square adi ðtÞ ¼ ð19Þ
M ii ðtÞ
of the disregard value of dependent variable is added to the objec-
tive function and any unfeasible solution obtained is declined. Moreover, the searching strategy on this notion can be defined
Mathematically, penalty function can be expressed as follows: to find the next velocity and next position of an agent. Next veloc-
ity of an agent is defined as a function of its current velocity added
P
NPV
2 P
NPQ
2 to its current acceleration. Hence, the next position and next veloc-
J mod ¼ F i ðPGi Þ þ kP ðPG1  Plim
G1 Þ þ kV ðV Li  V lim
Li Þ
i¼1 i¼1
ð13Þ ity of an agent can be computed as follows:
P
NPV
2 P
NTL
þ kQ ðQ Gi  Q lim
Gi Þ þ kS ðSli  Smax
li Þ2 v di ðt þ 1Þ ¼ randi  v di ðtÞ þ adi ðtÞ ð20Þ
i¼1 i¼1

where kP, kV, kQ and kS are defined as penalty factors. xlim is the limit xdi ðt þ 1Þ ¼ xdi ðtÞ þ v di ðt þ 1Þ ð21Þ
value of the dependent variable x and given as;
( where v d
and
i ðtÞ xdi ðtÞ
are the velocity and position of an agent at t
lim
xmax ; x > xmax time in d dimension, respectively. randi is a random number in
x ð14Þ the interval [0, 1]. It is to give a randomized feature to the search.
xmin ; x < xmin
The gravitational constant, G, which is initialized randomly at
the starting, will be decrease according to time to control the
3. Gravitational search algorithm search accuracy.
G is a function of the initial value (G0) and time (t):
The gravitational search algorithm (GSA) is one of the newest
stochastic search algorithm developed by Rashedi et al. [32]. This
algorithm, which is based on Newtonian laws of gravity and mass Generate initial population
interaction, has a great potential to be a break-through optimiza-
tion method. In this algorithm, agents are taken into consideration
as objects and their performances are measured by their masses.
Evaluate fitnees for each agent
Every object represents a solution or a part of a solution to the
problem. All these objects attract each other by the gravity force,
and this force causes a global movement of all objects towards
the objects with heavier masses. Due to the heavier masses have Update the G,best and worst of the population
higher fitness values; they describe good optimal solution to the
problem and they move slowly than lighter ones representing
worse solutions. In GSA, each mass has four particulars: its posi-
Calculate the M and a for each agent
tion, its inertial mass (Mii), its active gravitational mass (Mai) and
passive gravitational mass (Mpi). The position of the mass equaled
to a solution of the problem and its gravitational and inertial
masses are specified using a fitness function [30–32]. Update the velocity and position
At the beginning of the algorithm the position of a system are
described with N (dimension of the search space) masses

X i ¼ ðx1i . . . xdi . . . xni Þ for i ¼ 1; 2; . . . ; N ð15Þ

No
Meeting end of criterion ?
where n is the space dimension of the problem and defines the xdi
position of the ith agent in the dth dimension.
Initially, the agents of the solution are defined randomly and
Yes
according to Newton gravitation theory, a gravitational force from
mass j acts mass i at the time t is specified as follows:
Return best solution
M i ðtÞxMj ðtÞ d
F dij ðtÞ ¼ GðtÞ ðxj ðtÞ  xdi ðtÞÞ ð16Þ
Rij ðtÞ þ e Fig. 1. The flow chart of the GSA.
 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95 89

29
27 28

30 26 25

23 24

15 18 19

17 20
21
14 16 22

13 12 10

11 9

1
3 4 8
6

2 5

Fig. 2. Single line diagram of IEEE 30-bus test system.

GðtÞ ¼ GðG0 ; tÞ ð22Þ bestðtÞ ¼ min fitj ðtÞ ð26Þ

j2f1;...;mg

GðtÞ ¼ G0 eaT
t
ð23Þ worstðtÞ ¼ max fitj ðtÞ ð27Þ
j2f1;...;mg

where a is a user specified constant, t the current iteration and T is For a maximization problem:
the total number of iterations. bestðtÞ ¼ max fitj ðtÞ ð28Þ
j2f1;...;mg
The masses of the agents are computed using fitness evaluation.
The heavier mass of an agent, the more influential is that agent, worstðtÞ ¼ min fitj ðtÞ ð29Þ
j2f1;...;mg
concerning the solution it represents. It is notable that as the New-
ton’s law of gravity and law of motion refer a heavy mass has a In order to solve the optimization problem with GSA, at the
higher pull on power and moves slower. The masses are updated beginning of the algorithm every agent is placed at a certain point
as follows: of the search space which specifies a solution to the problem at
every unit of time. Then according to Eqs. (20) and (21), the agents
M ai ¼ M pi ¼ M ii ¼ M i ; i ¼ 1; 2; . . . ; N are recovered and their next positions are computed. Other param-
fiti ðtÞ  worstðtÞ eters of the algorithm like the gravitational constant G, masses M
mi ðtÞ ¼ ð24Þ and acceleration a are computed via Eqs. 22, 23, 24, 25, and (19)
bestðtÞ  worstðtÞ
respectively, and are updated every cycle of time. The flow diagram
mi ðtÞ
M i ðtÞ ¼ N ð25Þ of the GSA is shown in Fig. 1 [32,33].
P
mj ðtÞ
j¼1 3.1. Application of gravitational search algorithm to the OPF problem

where fiti(t) represents the fitness value of the agent i at time t, and In this section, a new heuristic optimization algorithm based on
the best(t) and worst(t) in the population respectively specify the Newton’s law of gravity and mass interactions for solving the OPF
strongest and the weakest agent with regard to their fitness route. problem is described as follows:
For a minimization problem: Step 1. Search space identification.
90 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95

Table 1
Best control variables settings for different test cases.

Control Case 1: Case 2: Voltage Case 3: Voltage Case 4: Voltage stability Case 5: Piecewise Case 6: Quadratic cost
variables Quadratic cost profile stability enhancement during contingency quadratic cost function with valve point
settings (p.u.) function improvement enhancement condition function loadings
P1 1.75749826 1.73320940 1.77043235 1.84916647 1.39999960 1.99599439
P2 0.48165537 0.49263900 0.44918179 0.52336057 0.54926800 0.51946406
P5 0.21381724 0.21567799 0.23962118 0.21451240 0.24995982 0.15000000
P8 0.21561405 0.23274500 0.14170295 0.11408846 0.30249474 0.10000000
P11 0.12417360 0.13774500 0.17213382 0.10388632 0.19583234 0.10000001
P13 0.12510199 0.11964300 0.16008895 0.12487544 0.20218508 0.12000003
V1 1.086235 1.026900 1.098204 1.077534 1.049999 1.099002
V2 1.046685 1.009980 1.087066 1.039057 1.009321 1.018042
V5 1.035570 1.014280 1.093797 1.039118 1.014509 1.052247
V8 1.076962 1.008680 1.089740 1.098241 1.034569 0.950000
V11 1.077452 1.050289 1.099999 1.100000 0.950429 0.963430
V13 1.099999 1.016340 1.100000 1.100000 1.003616 0.950702
T11 0.939297 1.071330 0.900001 0.900000 1.100000 0.909096
T12 1.006593 0.900000 0.900000 0.900000 1.099763 0.918200
T15 0.907372 0.996500 0.900000 0.900002 1.099998 0.925648
T36 0.921855 0.973200 1.006570 0.918952 1.079138 0.945975
QC10 0.02190333 0.04143700 0.04999998 0.05000000 – –
QC12 0.05000000 0.03562000 0.04999748 0.05000000 – –
QC15 0.00000000 0.05000000 0.04999585 0.05000000 – –
QC17 0.02715239 0.00000000 0.05000000 0.05000000 – –
QC20 0.00000672 0.05000000 0.04998984 0.05000000 – –
QC21 0.00000000 0.05000000 0.04998921 0.04999998 – –
QC23 0.00000593 0.05000000 0.04997742 0.05000000 – –
QC24 0.00000000 0.04983700 0.04999937 0.00000000 – –
QC29 0.00000000 0.02588000 0.03723178 0.00000000 – –
Fuel cost ($/h) 798.675143 804.314844 806.601315 801.183476 646.848066 929.7240472
Power loss 0.08386049 0.09765939 0.09916103 0.09588964 0.06573957 0.15145848
(p.u.)
Voltage 0.872862 0.093269 0.900000 0.868275 0.822802 0.577974
deviations
Lmax 0.130759 0.135776 0.116247 0.093073 0.141948 0.156484

Step 2. Generate initial population between minimum and max- G is set using in Eqs. (22) and (23), where G0 is set to 100, a is set
imum values of the control variables. to 10 and T is the total number of iterations. Maximum iteration
Step 3. Calculate value of fitness function of each agent for the numbers are 200 for all case studies. The software was written in
OPF problem. MATLAB 2008a computing environment and applied on a
Step 4. Update G(t), best(t), worst(t) and Mi(t) for i = 1, 2, . . . , N. 2.63 GHz Pentium IV personal computer with 512 MB-RAM. In
Step 5. Calculation of the total force in different directions. the following, the simulation results are presented:
Step 6. Calculation of acceleration and velocity.
Step 7. Updating agents’ position. 4.1.1. Case 1: quadratic cost function
Step 8. Repeat Steps 3–7 until the stop criteria is reached. The generator cost characteristics are defined as quadratic cost
Step 9. Stop. function of generator power output and the objective function
selected.

4. Numerical results P
NG P
NG
J¼ F i ðPGi Þ ¼ ðai þ bi PGi þ ci P2Gi Þ ð30Þ
i¼1 i¼1
4.1. IEEE 30-bus test system

Proposed GSA algorithm has been implemented to solve the

810
OPF problems. In order to test the efficiency and robustness of
GSA
the proposed GSA approach based on Newtonian physical law of
808
gravity and law of motion which is tested on standard IEEE 30-
bus test system shown in Fig. 2. The line, bus data, generator data
806
Fuel Cost ($/hr)

and the minimum and maximum limits for the control variables
are given in Appendix A.
804
Test system has six generators at the buses 1, 2, 5, 8, 11 and 13
and four transformers with off-nominal tap ratio at lines 6–9, 6–10,
802
4–12 and 28–27. In addition, buses 10, 12, 15, 17, 20, 21, 23, 24 and
29 were selected as shunt VAR compensation buses [15]. The total
800
system demand is 2.834 p.u. at 100 MVA base. The maximum and
minimum voltages of all load buses are considered to be 1.05–0.95
798
in p.u. The proposed approach has been applied to solve the OPF 0 20 40 60 80 100 120 140 160 180 200
problem for different cases with various objective functions. In Iteration
each case, 50 test runs were performed for solving the OPF problem
using the GSA approach. Fig. 3. Convergence of GSA for Case 1.
 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95 91

Table 2
Comparison of the simulation results for Case 1.

Methods Fuel cost ($/h) Simulation

times
Min Average Max
GSA 798.675143 798.913128 799.028419 10.7582
BBO [34] 799.1116 799.1985 799.2042 11.02
DE [1] 799.2891 NA NA NA
PSO [15] 800.41 NA NA NA
Improved GA 800.805 NA NA NA
[14]
MDE [2] 802.376 802.382 802.404 23.25
Gradient Method 804.853 NA NA 4.324
[36]
EADDE [39] 800.2041 800.2412 800.2784 3.32
EADHDE [40] 800.1579 NA NA NA
Enhanced GA 802.06 NA 802.14 76
[21] Fig. 4. System voltage profile.

where Fi and PGi are the fuel cost of the ith generator and the output
Table 3
of the ith generator, respectively. ai, bi and ci are the cost coefficients Comparison of the simulation results for Case 2.
of the ith generator and NG is the number of total generator. The
Methods Voltage profile improvement along with fuel Simulation times
values of cost coefficients are given in Table A1. The optimum con-
cost
trol parameter settings of GSA algorithm are given in Table 1. The
Min Average Max
minimum fuel cost obtained from the proposed approach is
798.675143 $/h. GSA is less by 0.054617%, 0.076812% compared GSA 0.093269 0.093952 0.094171 11.5873
to previously reported best results 799.1116 $/h, 799.2891 $/h BBO [34] 0.1020 0.1105 0.1207 13.23
DE [1] 0.1357 NA NA NA
respectively. Fig. 3 shows the convergence of GSA for minimum fuel PSO [15] 0.0891 NA NA NA
cost solution. The result obtained from the proposed GSA algorithm
was compared to the other methods in the literature. The results of
this comparison are given in Table 2. The average cost obtained
from the GSA for this case is 798.913128 $/h with a maximum cost 4.1.3. Case 3: voltage stability enhancement
of 799.028419 $/h. From Table 2, it can be seen that the results ob- Voltage stability is interested with the ability of a power system
tained from the proposed approach are better than those reported to maintain constantly acceptable bus voltage at each bus in the
in the literature. power system under nominal operating conditions. A system expe-
riences a state of voltage instability when the system is being sub-
4.1.2. Case 2: voltage profile improvement jected to a disturbance, rise in load demand, change in system
Bus voltage is one of the most significant safety and service configuration may lead to progressive and uncontrollable reduce
qualification indices. Considering only cost-based objective in the in voltage. Consequently, enhancement of voltage stability of a sys-
OPF problem may result in a feasible solution, however voltage tem is a significant parameter of power system operation and plan-
profile may not be reasonable. Hence, in the present case a twofold ning. Voltage stability can be defined via minimizing the voltage
objective function is taken in consideration to minimize the fuel stability indicator L-index values of each bus of a power system
cost and enhance voltage profile by minimizing all the load bus [1,34].
deviations from 1.0 per unit [1]. The objective function can be de- The L-index of a bus specifies the proximity of voltage collapse
scribed as follows: condition of that bus. L-index Lj of bus jth is defined as follows [37]

P
NG P
NPQ  
 P V i
NPV
J¼ ðai þ bi PGi þ ci P2Gi Þ þ g jV i  1:0j ð31Þ Lj ¼ 1  F ji  where j ¼ 1; 2; . . . ; NPQ ð32Þ
i¼1 i¼1 V i¼1 j

where g is a suitable weighting factor, to be selected by the user.

Value of g in this case is chosen as 100 [1,34]. The proposed method F ji ¼ ½Y 1 1 ½Y 2  ð33Þ
has been applied to search for the optimal solution of the problem.
The obtained optimal settings of the control variables from the pro- where NPV is number of PV bus and NPQ is number of load bus. Y1
posed GSA method are given in Table 1. Voltage profile in this case and Y2 are the sub-matrices of the system YBUS obtained after sep-
is compared to that of Case 1 as shown in Fig. 4. It is quite obvious arating the PQ and PV bus bars parameters as described in the fol-
that voltage profile is enhanced compared to that of Case 1. It is de- lowing equation
creased from 0.872862 p.u in Case 1 to 0.093269 p.u in Case 2. The     
result obtained from the proposed approach reduces 89.314576% in IPQ Y1 Y2 V PQ
¼ ð34Þ
this case. The comparison results are shown in Table 3. The average IPV Y3 Y4 V PV
voltage profile obtained from the GSA for this case is 0.093952 with
a maximum voltage profile of 0.094171. From the results in Table 3, L-index is computed for all the PQ bus. Lj is represents no load
it is clear that the results obtained from the GSA method are better case and voltage collapse conditions of bus j in the range of zero
than those reported in the literature. and one, respectively. Hence, a global system indicator L describing
Abou El Ela et al. [1], though fuel cost for Case 1 was 799.2891 $/ the stability of the complete system is given as follows:
h, there were load bus voltage violations. Approximately, for volt-
L ¼ maxðLj Þ; where j ¼ 1; 2; . . . ; NPQ ð35Þ
ages of load buses were up to 1.06, 1.07 and 1.08 p.u. It is clear that
this may be observed in Fig. 4 of Abou El Ela et al. [1]. The maxi- Lower value of L represents more stable system. In the OPF
mum acceptable voltage magnitudes at all load buses are 1.05 problem, inaccurate tuning of control variable settings may be in-
p.u for IEEE 30-bus test system [35]. crease the value of L-index, that is, may decrease the voltage stabil-
92 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95

Table 4 Table 5
Comparison of the simulation results for Case 3. Comparison of the simulation results for Case 4.

Methods Voltage stability improvement along with fuel Simulation times Methods Voltage stability improvement along with fuel Simulation times
cost cost
Min Average Max Min Average Max
GSA 0.116247 0.120538 0.12284 13.6378 GSA 0.093073 0.096531 0.099846 11.4926
BBO [34] 0.1104 0.1186 0.1214 16.29 DE [1] 0.1347 NA NA NA
DE [1] 0.1219 NA NA NA
PSO [15] 0.1246 NA NA NA

Table 6
Comparison of the simulation results for Case 5.
ity margin of a system [34]. In order to improve the voltage stabil- Methods Voltage stability improvement along with fuel Simulation times
ity and move the system far from the voltage collapse point, the cost
following objective function can be defined: Min Average Max
P
NG GSA 646.848066 646.896273 646.938163 10.2716
J¼ ðai þ bi P Gi þ ci P2Gi Þ þ gðmaxðLj ÞÞ ð36Þ BBO [34] 647.7437 647.7645 647.7928 11.94
i¼1
DE [1] 650.8224 NA NA NA
where g is a weighting factor to be selected by the user. Value of g PSO [15] 647.69 647.73 647.87 NA
MDE [2] 647.846 648.356 650.664 37.05
in this case is chosen as 6000. Applying the OPF problem to the pro-
posed technique importantly decrease the value of Lmax according
to Cases 1 and 2. This state can be seen from Table 1. The simulation
results obtained from the proposed GSA technique are compared to
662
other heuristic techniques in the literature. The comparison results
GSA
are shown in Table 4. From the Table 4 it is clear that in GSA ap- 660
proach, Lmax is 0.116247 which is reduce 4.637407%, 6.703852%
than in comparison to DE, PSO algorithms, respectively and is more 658
than the BBO algorithm.
Fuel Cost ($/hr)

656

4.1.4. Case 4: voltage stability enhancement during contingency

654
condition
A contingency state is simulated as outage of line (2–6). Voltage 652
stability is substantially interested in this case. Hence, in order to
enhance of voltage stability is considered as objective function 650

Eq. (36). The results obtained from this case are given in Table 1.
648
The comparison results are shown in Table 5. The proposed GSA
algorithm is indicated to improve of voltage stability in the contin- 646
gency condition. It appears that in the proposed approach Lmax is 0 20 40 60 80 100 120 140 160 180 200

0.093073 which is less 30.903489% than DE algorithm, for voltage Iteration

stability enhancement.
Fig. 5. Convergence of GSA for Case 5.

4.1.5. Case 5: piecewise quadratic fuel cost functions

In power system operation conditions, many thermal generat- cost in the OPF problem. The best fuel cost result obtained from the
ing units may be supplied with multiple fuel sources like coal, nat- GSA technique is compared with other techniques in Table 6. The
ural gas and oil. The fuel cost functions of these units may be convergence of GSA algorithm for the OPF problem with minimum
dissevered as piecewise quadratic fuel cost functions for different fuel cost is shown Fig. 5. From the results in Table 6, it can be seen
fuel types [1]. The cost coefficients for these units are given Table that the minimum fuel cost is 646.848066 $/h, with an average
A2. The fuel cost coefficients of other generators have the same val- cost of 646.896273 $/h and a maximum cost of 646.938163 $/h
ues as of Case 1 condition. The cost characteristics of generators 1 which are less in comparison to reported results in the literature.
and 2 are defined as follows:
8 4.1.6. Case 6: quadratic cost curve with valve point loadings
>
>
> ai1 þ bi1 PGi þ ci1 P2Gi Pmin
Gi 6 P Gi 6 P Gi1 In this case, the generating units of buses 1 and 2 are considered
>
<
ai2 þ bi2 PGi þ ci2 P2Gi PGi1 6 PGi 6 PGi2 to have the valve-point effects on their characteristics. Fuel cost
FðPGi Þ ¼ ð37Þ coefficients of these generators are taken from [39]. The cost coef-
>
>...
>
> ficients for these units are given in Table A3. The fuel cost coeffi-
:
aik þ bik PGi þ cik P2Gi PGik1 6 PGi 6 Pmax
Gi cients of remaining generators have the same values as of Case 1.
where aik, bik and cik are cost coefficients of the ith generator for fuel The cost characteristics of generators 1 and 2 are described as
type k. Objective function can be described as: follows:
   NG 
P
2 P F i ðPGi Þ ¼ ai þ bi PGi þ ci P2Gi
J¼ aik þ bik PGi þ cik P2Gi þ ai þ bi P Gi þ ci P2Gi ð38Þ  
 
i¼1 i¼3 þ di sin ei Pmin Gi  P Gi  where i
The proposed GSA approach is applied to this case considering ¼ 1 and 2 ð39Þ
the upper limit of voltage magnitude at bus 1 as 1.05 and no shunt
VAR compensation buses [1,15]. The results obtained optimal set- where ai, bi, ci, di and ei are cost coefficients of the ith generating
tings of control variables for this case study are given Table 1, unit.
which shows that the GSA has best solution for minimizing of fuel Objective function can be defined as:
 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95 93

Table 7 Table 8
Comparison of the simulation results for Case 6. Best control variables settings for IEEE 57-bus test system.

Methods Voltage stability improvement along with fuel Simulation times Control variables GSA Control variables GSA
cost settings (p.u.) settings (p.u.)
Min Average Max P1 1.42369 T24–25 1.05921
P2 0.92630 T24–25 0.99921
GSA 929.7240472 930.9246338 932.0487291 9.8374
P3 0.45318 T24–26 0.92201
BBO [34] 919.7647 919.8389 919.8876 11.15
P6 0.72355 T7–29 0.93243
MDE [2] 930.793 942.501 954.073 41.85
P8 4.64743 T34–32 1.08828
P9 0.84999 T11–41 1.03902
P12 3.63951 T15–45 1.04318
   V1 1.05941 T14–46 1.02494
P
2
 
J¼ ai þ bi PGi þ ci P2Gi þ di sin ei Pmin
Gi  P Gi  V2 1.05759 T10–51 0.95425
i¼1 V3 1.06000 T13–49 0.92897
 NG  V6 1.06000 T11–43 1.09942
P
þ ai þ bi PGi þ ci P2Gi ð40Þ V8 1.05999 T40–56 0.96948
i¼3 V9 1.05999 T39–57 1.06200
V12 1.04590 T9–55 1.09388
The obtained optimal settings of control variables from the pro- T4–18 0.90000 QC18 0.15243
posed method are given in Table 1. Table 7 gives the comparison of T4–18 0.90000 QC25 0.14403
the proposed GSA technique other heuristic techniques reported in T21–20 0.90856 QC53 0.15102

the literature. It is quite obvious that the minimum fuel cost ob- Fuel cost ($/h) 41695.8717
tained from the proposed approach is 929.7240472 $/h with an
average cost of 930.9246338 $/h and a maximum cost of
932.0487291 $/h, which is less than MDE algorithm and is more
Table 9
than BBO algorithm. But the sum of real power of generating units Comparison of the simulation results for IEEE 57-bus
was given as 294.464 MW in BBO approach and real power loss test system.
was 12.18 MW whereas load was 283.4 MW. So power generation
Methods Fuel cost ($/h)
is not matching load plus losses. This approach did not meet the
load demand for this case. The variation of the total fuel cost is BASE-CASE [39] 51347.86
MATPOWER [39] 41737.79
shown in Fig. 6. The results obtained confirm the ability of the pro- EADDE [39] 41713.62
posed GSA approach to find accurate OPF solutions in this case
GSA 41695.8717
study.

4.2. IEEE 57-bus test system

4
x 10
4.176
To evaluate the effectiveness and efficiency of the proposed GSA GSA

approach in solving larger power system, a standard IEEE 57-bus 4.175

test system is considered. The IEEE 57-bus test system consist of
4.174
80 transmission lines, seven generators at the buses 1, 2, 3, 6, 8,
Fuel Cost ($/h)

9 and 12, and 15 branches under load tap setting transformer 4.173
branches. The shunt reactive power sources are considered at
buses 18, 25 and 53. The total load demand of system is 4.172
1250.8 MW and 336.4 MVAR. The bus data, the line data and the
4.171
cost coefficients and minimum and maximum limits of real power
generations are taken from Refs. [41,42]. The maximum and mini-
4.17
mum values for voltages of all generator buses and tap setting
transformer control variables are considered to be 1.1–0.9 in p.u. 4.169
0 50 100 150 200 250 300 350 400
The maximum and minimum values of shunt reactive power
Iteration

Fig. 7. Convergence of GSA for IEEE 57-bus test system.

948
GSA
946

944 Table A1
Generator cost coefficients for Case 1 [35,36].
942
Bus no. Cost coefficients
Fuel Cost ($/hr)

940
a b c
938
1 0.00 2.00 0.00375
936 2 0.00 1.75 0.01750
5 0.00 1.00 0.06250
934 8 0.00 3.25 0.00834
932
11 0.00 3.00 0.02500
13 0.00 3.00 0.02500
930

928
0 20 40 60 80 100 120 140 160 180 200
sources are 0.0 and 0.3 in p.u. The maximum and minimum values
Iteration
for voltages of all load buses are 1.06 and 0.94 in p.u form MAT-
Fig. 6. Convergence of GSA for Case 6. POWER [42], respectively. In simulation process, minimization of
94 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95

Table A2 Table A6
Generator cost coefficients for Case 5 [38]. The limits of the control variables [36].

Bus no. From MW To MW Cost coefficients Control variables Min Max

a b c P1 50 200
P2 20 80
1 50 140 55.00 0.70 0.0050
P5 15 50
140 200 82.5 1.05 0.0075
P8 10 35
2 20 55 40.00 0.30 0.0100 P11 10 30
55 80 80.00 0.60 0.0200 P13 12 40
V1 0.95 1.1
V2 0.95 1.1
V5 0.95 1.1
V8 0.95 1.1
Table A3
V11 0.95 1.1
Generator cost coefficients for Case 6 [38].
V13 0.95 1.1
Bus no. P min P max Cost coefficients T11 0.90 1.1
Gi Gi
T12 0.90 1.1
a b c d e
T15 0.90 1.1
1 50 200 150.00 2.00 0.0016 50.00 0.0630 T36 0.90 1.1
2 20 80 25.00 2.50 0.0100 40.00 0.0980 QC10 0.00 5.0
QC12 0.00 5.0
QC15 0.00 5.0
QC17 0.00 5.0
QC20 0.00 5.0
Table A4 QC21 0.00 5.0
Load data [35,36]. QC23 0.00 5.0
Bus Load Bus Load Bus Load QC24 0.00 5.0
no. no. no. QC29 0.00 5.0
P Q P Q P Q
(p.u.) (p.u.) (p.u.) (p.u.) (p.u.) (p.u.)
1 0.000 0.000 11 0.000 0.000 21 0.175 0.112 where Fi and PGi are the fuel cost of the ith generator and the output
2 0.217 0.127 12 0.112 0.075 22 0.000 0.000 of the ith generator, respectively. ai, bi and ci are the cost coefficients
3 0.024 0.012 13 0.000 0.000 23 0.032 0.016 of the ith generator and NG is the number of total generator. The ob-
4 0.076 0.016 14 0.062 0.016 24 0.087 0.067 tained optimal settings of control variables from the proposed
5 0.942 0.190 15 0.082 0.025 25 0.000 0.000
6 0.000 0.000 16 0.035 0.018 26 0.035 0.023
method are given in Table 8. Comparison of the proposed GSA tech-
7 0.228 0.109 17 0.090 0.058 27 0.000 0.000 nique other heuristic technique reported in the literature, base case
8 0.300 0.300 18 0.032 0.009 28 0.000 0.000 and MATPOWER is given in Table 9. The convergence characteristic
9 0.000 0.000 19 0.095 0.034 29 0.024 0.009 of the best fuel cost result obtained from the GSA approach is shown
10 0.058 0.020 20 0.022 0.007 30 0.106 0.019
in Fig. 7.
From the results in Table 9, it is quite clear that the best fuel
quadratic cost function is considered as objective function and also cost result obtained from the proposed GSA technique is
to test the performance of the proposed GSA approach. In this test 41695.8717 $/h, which is less in comparison to reported best result
system, 50 test runs were performed for solving the OPF problem the literature.
using the GSA method. The objective function is described as
follow: 5. Conclusion
P
NG P
NG
J¼ F i ðPGi Þ ¼ ðai þ bi PGi þ ci P2Gi Þ ð41Þ In this paper, one of the recently improved heuristic algorithms
i¼1 i¼1 which are the gravitational search algorithm was demonstrated

Table A5
Line data [35,36].

Line no From bus To bus R (p.u.) X (p.u.) B (p.u.) Tap settings Line no From bus To bus R (p.u.) X (p.u.) B (p.u.) Tap settings
1 1 2 0.0192 0.0575 0.0264 – 22 15 18 0.1070 0.2185 0.0000 –
2 1 3 0.0452 0.1852 0.0204 – 23 18 19 0.0639 0.1292 0.0000 –
3 2 4 0.0570 0.1737 0.0184 – 24 19 20 0.0340 0.0680 0.0000 –
4 3 4 0.0132 0.0379 0.0042 – 25 10 20 0.0936 0.2090 0.0000 –
5 2 5 0.0472 0.1983 0.0209 – 26 10 17 0.0324 0.0845 0.0000 –
6 2 6 0.0581 0.1763 0.0187 – 27 10 21 0.0348 0.0749 0.0000 –
7 4 6 0.0119 0.0414 0.0045 – 28 10 22 0.0727 0.1499 0.0000 –
8 5 7 0.0460 0.1160 0.0102 – 29 21 22 0.0116 0.0236 0.0000 –
9 6 7 0.0267 0.0820 0.0085 – 30 15 23 0.1000 0.2020 0.0000 –
10 6 8 0.0120 0.0420 0.0045 – 31 22 24 0.1150 0.1790 0.0000 –
11 6 9 0.0000 0.2080 0.0000 1.078 32 23 24 0.1320 0.2700 0.0000 –
12 6 10 0.0000 0.5560 0.0000 1.069 33 24 25 0.1885 0.3292 0.0000 –
13 9 11 0.0000 0.2080 0.0000 – 34 25 26 0.2544 0.3800 0.0000 –
14 9 10 0.0000 0.1100 0.0000 – 35 25 27 0.1093 0.2087 0.0000 –
15 4 12 0.0000 0.2560 0.0000 1.032 36 28 27 0.0000 0.3960 0.0000 1.068
16 12 13 0.0000 0.1400 0.0000 – 37 27 29 0.2198 0.4153 0.0000 –
17 12 14 0.1231 0.2559 0.0000 – 38 27 30 0.3202 0.6027 0.0000 –
18 12 15 0.0662 0.1304 0.0000 – 39 29 30 0.2399 0.4533 0.0000 –
19 12 16 00945 0.1987 0.0000 – 40 8 28 0.0636 0.2000 0.0214 –
20 14 15 0.2210 0.1997 0.0000 – 41 6 28 0.0169 0.0599 0.0065 –
21 16 17 0.0824 0.1932 0.0000 –
 S. Duman et al. / Energy Conversion and Management 59 (2012) 86–95 95

and applied to solve optimal power flow problem. The OPF prob- [15] Abido MA. Optimal power flow using particle swarm optimization. Electr
Power Energy Syst 2002;24:563–71.
lem is formulated as a nonlinear optimization problem with equal-
[16] Varadarajan M, Swarup KS. Solving multi-objective optimal power flow using
ity and inequality constraints in power systems. In this study, differential evolution. IET Gener Transm Distrib 2008;2(5):720–30.
different objective functions were considered to enhance the volt- [17] Roa-Sepulveda CA, Pavez-Lazo BJ. A solution to the optimal power flow using
age profile, also to enhance the voltage stability in both nominal simulated annealing. Electr Power Energy Syst 2003;25:47–57.
[18] Somasundaram P, Kuppusamy K, Kumudini Devi RP. Evolutionary
and contingency conditions, to minimize the fuel cost such as qua- programming based security constrained optimal power flow. Electr Power
dratic cost function, piecewise quadratic cost function, cost func- Syst Res 2004;72:137–45.
tion with valve point effect. The proposed GSA approach were [19] Anitha M, Subramanian S, Gnanadass R. FDR PSO-based transient stability
constrained optimal power flow solution for deregulated power industry.
tested and investigated on the IEEE 30-bus and 57-bus test sys- Electr Power Compo Syst 2007;35(11):1219–32.
tems. This approach was successfully and influentially performed [20] Dutta P, Sinha AK. Voltage stability constrained multi-objective optimal power
to find the optimal settings of the control variables of test system. flow using particle swarm optimization. In: First international conference on
industrial and information systems (ICIISs) 2006, Sri Lanka, 8–11 August 2006.
The simulation results proved the robustness and superiority of the p. 161–6.
proposed approach to solve the OPF problem. The results obtained [21] Bakirtzis AG, Biskas PN, Zournas CE, Petridis V. Optimal power flow by
from the GSA technique were compared to other methods previ- enhanced genetic algorithm. IEEE Trans Power Syst 2002;17(2):229–36.
[22] Basu M. Optimal power flow with FACTS devices using differential evolution.
ously reported in the literature. The comparison verifies the influ- Electr Power Energy Syst 2008;30:150–6.
entially of the proposed GSA approach over stochastic techniques [23] Basu M. Multi-objective optimal power flow with FACTS devices. Energy
in terms of solution quality for the OPF problem. Convers Manage 2011;52:903–10.
[24] Madhad B, Srairi K, Bouktir T. Optimal power flow for large-scale power
system with shunt FACTS using efficient parallel GA. Electr Power Energy Syst
2010;32:507–17.
Appendix A
[25] Li C, Zhou J. Parameters identification of hydraulic turbine governing system
using improved gravitational search algorithm. Energy Convers Manage
See Tables A1–A6. 2011;52:374–81.
[26] Zibanezhad B, Zamanifar K, Nematbakhsh N, Mardukhi F. An approach for web
services composition based on QoS and gravitational search algorithm. In
References Proceedings of the innovations in information technology conference, 2010. p
340–4.
[27] Hassanzadeh HR, Rouhani M. A multi-objective gravitational search algorithm.
[1] Abou El Ela AA, Abido MA, Spea SR. Optimal power flow using differential
In: Proceedings of the communication systems and network conference, 2010.
evolution algorithm. Electr Power Syst Res 2010;80:878–85.
p. 7–12.
[2] Sayah S, Zehar K. Modified differential evolution algorithm for optimal power
[28] Balachandar SR, Kannan K. A meta-heuristic algorithm for set covering
flow with non-smooth cost functions. Energy Convers Manage
problem based on gravity. Int J Comput Math Sci 2010;4:223–8.
2008;49:3036–42.
[29] Duman S, Güvenç U, Yörükeren N. Gravitational search algorithm for economic
[3] He S, Wen JY, Prempain E, Wu QH, Fitch J, Mann S. An improved particle swarm
dispatch with valve-point effects. Int Rev Electr Eng 2010;5(6):2890–5.
optimization for optimal power flow. In: International conference on power
[30] Ceylan O, Ozdemir A, Dag H. Gravitational search algorithm for post-outage
system technology (POWERCON), Singapore; 21–24 November 2004. p. 1633–
bus voltage magnitude calculations. In: International universities’ power
7.
engineering conference, Wales (UK); 31 August–3 September, 2010.
[4] Abido MA. Optimal power flow using tabu search algorithm. Electr Power
[31] Rashedi E, Nezamabadi-pour H, Saryazdi S. BGSA: binary gravitational search
Compo Syst 2002;30(5):469–83.
algorithm. Nat Comput 2010;9:727–45.
[5] Dommel HW, Tinney TF. Optimal power flow solutions. IEEE Trans Power
[32] Rashedi E, Rashedi E, Nezamabadi-pour H, Saryazdi. GSA: a gravitational
Appar Syst 1968;87(5):1866–76.
search algorithm. Inform Sci 2009;179:2232–48.
[6] Yan X, Quantana VH. Improving an interior point based OPF by dynamic
[33] Abarghouei AA. A novel solution to traveling salesman problem using fuzzy
adjustments of step sizes and tolerances. IEEE Trans Power Syst
sets, gravitational search algorithm, and genetic algorithm. M.Sc. Thesis,
1999;14(2):709–17.
Faculty of Computer Science and Information Systems, Malaysia Technology
[7] Habiabollahzadeh H, Luo GX, Semlyen A. Hydrothermal optimal power flow
University; April 2010. p. 52–7.
based on combined linear and nonlinear programming methodology. IEEE
[34] Bhattacharya A, Chattopadhyay PK. Application of biogeography-based
Trans Power Appar Syst PWRS 1989;4(2):530–7.
optimisation to solve different optimal power flow problems. IET Gener
[8] Burchet RC, Happ HH, Vierath DR. Quadratically convergent optimal power
Transm Distrib 2011;5(1):70–80.
flow. IEEE Trans Power Appar Syst PAS 1984;103:3267–76.
[35] Alsac O, Stott B. Optimal load flow with steady-state security. IEEE Trans
[9] Momoh JA, Adapa R, El-Hawary ME. A review of selected optimal power flow
Power Appar Syst PAS 1974;93(3):745–51.
literature to 1993. I. Nonlinear and quadratic programming approaches. IEEE
[36] Lee KY, Park YM, Ortiz JL. A united approach to optimal real and reactive power
Trans Power Syst 1999;14(1):96–104.
dispatch. IEEE Trans Power Appar Syst PAS 1985;104(5):1147–53.
[10] Momoh JA, El-Hawary ME, Adapa R. A review of selected optimal power flow
[37] Kessel P, Glavitsch H. Estimating the voltage stability of a power system. IEEE
literature to 1993. II. Newton, linear programming and interior point methods.
Trans Power Deliv PWRD 1986;1(3):346–54.
IEEE Trans Power Syst 1999;14(1):105–11.
[38] Yuryevich J, Wong KP. Evolutionary programming based optimal power flow
[11] Huneault M, Galina FD. A survey of the optimal power flow literature. IEEE
algorithm. IEEE Trans Power Syst 1999;14(4):1245–50.
Trans Power Syst 1991;6(2):762–70.
[39] Vaisakh K, Srinivas LR. Evolving ant direction differential evolution for OPF
[12] Roy PK, Ghoshal SP, Thakur SS. Biogeography based optimization for multi-
with non-smooth cost functions. Eng Appl Artif Intell 2011;24:426–36.
constraint optimal power flow with emission and non-smooth cost function.
[40] Vaisakh K, Srinivas LR. ‘‘Genetic evolving ant direction HDE for OPF with non-
Expert Syst Appl 2010;37:8221–8.
smooth cost functions and statistical analysis. Expert Syst Appl
[13] Deveraj D, Yegnanarayana B. Genetic algorithm based optimal power flow for
2011;38:2046–62.
security enhancement. IEE Proc Gener Transm Distrib 2005;152(6):899–905.
[41] The IEEE 57-Bus Test System. <http://www.ee.washington.edu/research/pstca/
[14] Lai LL, Ma JT, Yokoyama R, Zhao M. Improved genetic algorithms for optimal
pf57/pg_tca57bus.htm>.
power flow under normal and contingent operation states. Int J Electr Power
[42] MATPOWER. <http://www.ee.washington.edu/research/pstca>.
Energy Syst 1997;19(5):287–92.