Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018

ISSN 1583-1078 p. 249-270

Engineering, Environment

A new hybrid algorithm of particle swarm optimizer with grey wolves’

optimizer for solving optimal power flow problem

Aboubakr KHELIFI 1, Saliha CHETTIH 1and Bachir BENTOUATI 1*

1
LMSF Laboratory, Electrical Engineering Department, Laghouat 03000, Algeria,
E-mail(s): 1a.khelifi@lagh-univ.dz; 1s.chettih@lagh-univ.dz; 1*b.bentouati@lagh-univ.dz
*
Corresponding author, +213 666 81 29 21

Received: June 17, 2018 / Accepted: June 17, 2018 / Published: June 30, 2018

Abstract
The current trend of study is to hybridize two and more algorithms to gain the best
solution in the area of optimization problems. In this paper presents the recently
developed hybrid optimization technique named PSO-GWO combines the framework
of particle swarm optimization (PSO) with grey wolves optimization (GWO) to solve
the optimal power flow (OPF) problem. OPF is formulated as a nonlinear optimization
problem with conflicting objectives and subjected to both equality and inequality
constraints. The performance of this technique is deliberated and evaluated on the
standard IEEE 30-bus test system with a single objective and multi-objective cases
such as fuel cost minimization, Active power loss reduction, Voltage profile
improvement and Voltage stability enhancement, and is compared to approaches
available in the literature. The hybrid PSO-GWO provides better results compared to
the original PSO, GWO, and other techniques mentioned in the literature as shown in
the simulation results.
Keywords
Optimal power flow; Voltage stability; Active power loss; Emission; Constraints;
Hybrid PSO-GWO.

249
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
Introduction

The Optimal Power Flow (OPF) is a significant appliance for planning and operation
studies in the power system operator. OPF is a widely non-linear and non-convex
optimization problem, and this is more difficulty in practical applications in the large
number's presence of discrete variables. The goal of OPF is to provide the optimal settings of
the power system by improving objective function while meeting the equality and inequality
constraints [1], then this problem has been addressed by several researchers. The objective
functions, such as the minimization of total fuel cost, improvement of voltage stability index
and reduction of real power loss are considered individually in the literature for this study.
The problem of power flow is one of the fundamental problems in which the load and
the powers of the generator are given or corrected. The OPF has a long history in its
development, and it was primarily introduced by Carpentier in 1962, and the next
investigations on OPF in [2]. However, it took a long time to turn into an effective technique
that could be applied in daily use. The actual interest for OPF is focused on its capability to
solve for the optimal solution that has considered the security of the system. The optimal
power flow has been applied to regulate the active power outputs and voltages of the
generator, transformer tap settings, shunt reactors/capacitors and other controllable variables
to minimize the generator fuel cost, network active power loss, voltage stability index, while
keeping all the load bus voltages, generator reactive power outputs, network power flows, and
all other state variables in the power system within their secure and operational bounds. In its
most common problem formulation, the OPF is a non-convex, static, large-scale optimization
problem with both continuous and discrete control variables [3]. Even in the absence of non-
convex generator operating cost functions, prohibited operating zones (POZs) of generating
units, and discrete control variables, the OPF problem is a non-convex because of the
existence of the non-linear alternating current power flow equality constraints. The existence
of discrete control variables, such as transformer tap positions, switchable shunt devices,
phase shifters, further complicates the formulation and solution of the problem.
Different conventional optimization methods have been used to solve the OPF
problem. These involve Newton methods [4], linear programming [5], and quadratic
programming [6]. A comprehensive survey of different conventional optimization techniques
used to solve OPF problems was presented. Nevertheless, in practice, conventional techniques

250
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

suffer from some weakness. Some of its shortcomings through other things are: First, they do
not assure to find the global optimum, second, conventional techniques involve complex
computations with a long time, and they do not suitable for discrete variables [7].
During the last little decades, a lot of powerful meta-heuristics were developed.
Several of them have been implemented to the OPF problem with very successfully. various
of the modern implementations of meta-heuristics for the OPF problem are: Black Hole (BH)
[8], League Championship Algorithm (LCA) [9], Gravitational Search Algorithm (GSA)[10],
Artificial Bee Colony (ABC)[11], Group Search Optimization (GSO)[12], Imperialist
Competitive Algorithm (ICA) [13], Differential Search Algorithm (DSA) [14], Teaching
Learning Based Optimization (TLBO)[15], and Krill Herd Algorithm (KHA) [16], adaptive
clonal selection algorithm(ACSA) [17] .Though, because of changing objectives while
solving OPF problems, no algorithm is the greatest one to solve all the OPF problems.
Consequently, there is still a need for a novel algorithm, which can effectively solve the most
of OPF problems.
Particle Swarm Optimization (PSO) has an uncomplicated concept, simple to carry
out, Relative effectiveness to control parameters and computational adequacy[18] Although it
has many advantages, it obtain restricted in the local minimum, When dealing with severely
constrained problems because of limited local/global search abilities [19, 20]. Gray Wolf
Optimizer (GWO) is a powerful evolutionary algorithm newly developed by Mirjalili [20] it
has the capability to converge to the superior quality near-optimal solution and has preferable
convergence properties than other dominant techniques. In addition, GWO has a perfect
balance between exploration and exploitation that result in avoidance high of the local optima.
In this article, we introduce a new hybrid algorithm, named a PSO-GWO which is
constructed on incorporating PSO with GWO algorithms. The effectiveness of this technique
to the OPF problem with non-smooth cost functions like as fuel cost with prohibited zones,
piecewise quadratic cost function, fuel cost with valve-point effects is tested and analyzed on
the standard IEEE 30-bus test systems With various objective functions and is Compared to
the techniques mentioned in the literature. Investigational results on the OPF problem show
that the modern hybrid algorithm has preferable effectiveness in both convergence and global
excellent, compared with original PSO, GWO and other algorithms mentioned in the
literature. We confirm that the suggested method has perfect efficiency and capability to get a
solution of OPF problem. This suggested technique can optimize the number of various

251
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
objectives and can be useful for system operators in choosing a wise decision in implementing
the system performance.

Material and method

Optimal Power Flow formulation

The OPF is a power flow problem that provides the optimal settings of the control
variables for specific settings of the load by means of reducing a predefined objective
function such as the cost of real power generation or transmission losses. OPF takes into
account the operating limits of the system and it can be mathematically formulated as a
nonlinear constrained optimization problem as follows, Eq. (1):
Minimize: J (𝑥, 𝑢) (1)
g  x ,u   0

Subject to: h  x ,u   0

Where: J (x, u) - objective function; h (x, u) - set of equality constraints; g(x, u) - set of
inequality constraints; U - the vector of control variables; X - the vector of state variables; the
control variables u and the state variables x of the OPF problem are explained in (2) and (3),
respectively.
Control variables:
These are the set of variables that can be regulated to satisfy the load flow equations
[21]. The set of control variables in the mathematical formulation of the OPF problem are:
PG: is the 𝑖-th active power bus generator (except swing generator).
VG: is the voltage magnitude at 𝑖-th PV bus (generator bus).
T: is the transformer tap setting.
QC: is the shunt VAR compensation.
The control variables U can be formulated as Eq. (2):

u  PG2  PGNG ,V G1 V GNG ,QC1 QC NC ,T1 T NT  (2)

Where: NC, NT and NG are the number of VAR compensator, the number of regulating
transformers and the number of generators respectively.
State variables:

252
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

These are the set of variables that report any unique state of the system [21]. The set of
state variables for mathematically formulated the OPF problem as follow:
PG1: is the generator active power at slack (or swing) bus.
VL: is the bus voltage of 𝑝-th load bus (PQ bus).
QG: reactive power generation of all generator units.
SL: transmission line loading (or line flow)
The state variables X can be formulated as Eq. (3):

x   PG1 ,V L1 ...V LNL ,QG1 ...QG NG , S l1 ...S l nl  (3)

Where: NL, and nl are the number of load buses and the number of transmission lines,
respectively.
Constraints:
The OPF constraints can be classified into equality and inequality constraints, as
explained in the next sections.
1. Equality constraints
The equality constraints that express the typical nonlinear power flow equations that
control the power system, presented as follows.
a) Real power constraints, Eq. (4):

PGi  PDi V i V j G ij cos  ij   B ij sin  ij   0

NB
(4)
j 1

b) Reactive power constraints, Eq. (5):

QGi  Q Di V i V j G ij sin  ij   B ij cos  ij   0

NB
(5)
j 1

Where: NB is the number of buses, PD and Q D are active and reactive load demands,

respectively,  ij   i   j is the difference in voltage angles between bus 𝑖 and bus 𝑗 G ij is

the transfer conductance and B ij is the susceptance between bus 𝑖 and bus 𝑗, respectively.

2. Inequality constraints:
The Inequality constraints that reflect operational of the system and the system's
physical limits presented as follows.
1. Generator constraints. For all generators comprising the slack: voltage, active and
reactive outputs shall be limited by their minimum and maximum limits as follows, Eq. (6-8):
V Gmin V Gi V Gmax i  NG
i i (6)

253
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI

PGmin  PGi  PGmax i  NG

i i (7)

QGmin  QGi  QGmax i  NG

i i (8)
2. Transformer constraints. Transformer tap settings must be limited to their
specified minimum and maximum limits as follows, Eq. (9):

T jmin  T j  T jmaxj  NT (9)

Shunt VAR compensator constraints. Shunt VAR compensator have to be limited

by their lower and upper limits as follows, Eq. (10):
QCmin
k
 QC k  QCmax
k
k  NC (10)

Security constraints:
These comprise the constraints of voltage magnitude at load buses and transmission
line loadings. The Voltage of each load bus has to be limited within its minimum and
maximum operating limits. Line flow through each transmission line must be limited by its
capacity limits. These constraints can be expressed as given follows, Eq. (11-12):
V Lmin
p
V L p V Lmax
p
p  NL (11)

S lq  S lmax
q
q  nl (12)

Where: V Lmin
p
and V Lmax
p
- represents lowest and the upper load voltage of i-th unit, S lq -

represents apparent power flow of i-th branch S lmax

q
- represents maximum apparent power

flow limit of i-th branch.

Particle swarm optimization (PSO):
Particle swarm optimization, inspired by societal the conduct of birds is an
optimization technique based on swarm intelligence that suggested a population-based
research process by taking particles and moving them around in the search space for be given
the best solution for the problem. In PSO, particles change position in a multi-dimensional
search space, any particle adjusts its location pursuant to its own experiment and the
experiment of neighbouring particles, and take advantage of the best position encountered by
itself and its neighbours. The direction of the swarm of a particle is determined by all the
neighbouring particles of the particle and experiment of its history [22]. PSO is a non-
deterministic, stochastic optimization method and supplies a population-based search
operation for global optimization, Have the main advantage of easy to achieve and few

254
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

parameters to regulate. PSO involves two expressions A and B. Position and speed are
updated over the course of iteration from these follows Eq. (13-15):


v ijt 1  wv ijt  c1r1 Pbest t  X t
  c r Gbest
2 2
t
X t
 (13)

X t 1  X t v t 1 i=1,2….NP j=1,2….NG (14)

w w max

w max
w min
  iteration (15)
max iteration
Where: vijt+1, vijt is the speed of j-th member of i-th particle at iteration number, r1 and r2 are
two random values within [0, 1].

Grey wolf optimization (GWO)

Mirjalili et al are the first to propose a new algorithm called are Grey Wolf
Optimization [23], The technique was inspired by the popular conduct and hunting
mechanism of grey wolves in nature. In a group, the grey wolves follow so strong social
leadership hierarchy. The group leaders are male and female, named alpha (α). The second
level of grey wolves, which are subaltern wolves that assist leaders, are named beta (β). The
deltas (δ) are the grey wolves' third level that must submit to alphas and betas but controls
the omega. The downgrade of the grey wolf is omega (ω), which must capitulate to all other
wolves that govern. The GWO technique is presented in mathematical models as next [24].

Social hierarchy

In the mathematical pattern of the social hierarchy of grey wolves, alpha (α) is
considered the most suitable solution. As a result, the second preferable solution is called beta
(β) and the third preferable solution is called delta (δ) respectively. The remaining candidate
solutions are taken as omega (ω). In the GWO, the optimal (shooting) alpha, beta, and delta
are guided. The omega wolves must come behind these wolves [24].

Encircling prey

The grey wolves surround victim through the hunt. The encircling conduct can be
modelled mathematically as follows, Eq. (16-17):

255
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
r r r r
M  N  X t   X t  (16)
r r r r
X t  1  X r t   L  M (17)
r r r
Where: t - shows the existing iteration, L and N are coefficient vectors, X r is the position
r r
vector of the victim and X shows the position vector of a grey wolf. The vectors L and
r
N are calculated as follows, Eq. (18-19):
r r r r
L  2l  r1  l (18)
r r
N  2  r2 (19)
Where: elements of 'l' are linearly diminished from 2 to 0 through the course of iterations and
r1, r2 are arbitrary vectors in the cavity [0,1].

Hunting

Hunting is generally guided by alpha, beta and delta, which have a better
understanding of the potential location of the victim. The other search agents have to update
their positions based on the preferable search position of agent. The position of their agent
update can be expressed as follows Eq. (20-22) [25]:
r r r r
M   N1 X   X 
r r r r
M   N1 X   X  (20)
r r r r
M   N1 X   X 
r r r r
X 1  X   L1 M  
r r r r
X 2  X   L1 M  
(21)
r r r r
X 3  X   L1 M  

r r r
r X1X 2 X 3
X t  1  (22)
3
Search for prey and attacking prey

The 'L' is an arbitrary value in the interval [-2a, 2a].When | L | <1, the wolves are
forced to offensive the victim. Offensive the victim is the exploitation capability and
searching for a victim is the exploration capability. The random values of 'L' are employed to

256
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

force the search agent to move far from the victim. When | L | > 1, grey wolves are forced to
move far from the victim.

Hybrid PSO-GWO

Whoever the technique is used for optimization, it has advantages and disadvantages.
Exams were suggested for enhancement to achieve the best possible technique. Hybridization
techniques can help to find enhancement in order to the advantages and disadvantages of each
technique are compensated. Presently, the hybrid meta-heuristics have become more
deliberative in order to the best results for some optimization problems have been received
with hybrid methods. In this article, we have select two new meta-heuristics (PSO and GWO)
to solve the OPF problem, but like all techniques, PSO and GWO have advantages and
disadvantages. So as to combine the characteristics of the two techniques to draw their
benefits and have the preferable results, we hybridized them as appear in Figure 1.

Start

Initialize GWO and PSO parameters

Define Initial objective function

Generate initial population and find alpha, beta and delta wolves

Update the position of actual wolf by employing equation (19)

PSO algorithm

Obtain updated positions

Update l, L and N: calculate the fitness of all wolves

Update wolf positions of alpha, beta and delta

Stop condition satisfied

Optimal solution

Figure 1. Flowchart of the hybrid GWO-PSO algorithm

257
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
The PSO-GWO algorithm has been using to solve the OPF problem for exam system
and for many cases with various objective functions. The considered power systems networks
are the IEEE 30-bus test system network. The advanced software program is written in
MATLAB computing environment and used on a 2.20 GHz i7 personal computer. In our
study, the PSO-GWO population size or a number of stars is selection to be 50.

IEEE 30-bus test system

In order to illustrate the performance of the proposed PSO-GWO method, it has been
examined first on the standard IEEE 30-bus test system. The standard IEEE 30-bus system
selection in this paper has the next characteristics [26]: 6-generators at buses 1, 2, 5, 8,
11 and 13, 4-transformers with off-nominal tap ratio at lines 11, 12, 15 and 36, 9-
shunt VAR compensation buses at buses 10, 12, 15, 17, 20, 21, 23, 24 and 29.
In addition, line data, bus data, generator data, and lower and upper restriction for
control variables are presented in [27].
For this exam system, Six various cases have been studied with various objectives
and all the obtained results are outlined in Tables 1, 4 and 7. The first column of this table
appears the optimal control settings found here:
- PG1Through PG6 and VG1 through VG6 represent the powers and the voltages of
generator 1 through generator 6.
- T11, T12, T15 and T36 are the tap settings of transforms involved between lines 11,
12, 15 and 36.
- QC10, QC12, QC15, QC17, QC20, QC21, QC23, QC24 and QC29 represent
the shunt VAR compensations connected to buses 10, 12,15, 17, 20, 21, 23, 24 and 29.
Moreover, fuel cost ($/h), active power losses (MW), voltage deviation and L max
represent the total fuel cost of the system, the total active transmission losses, the deviation of
load voltages from 1and the index of stability, respectively. More description of these results
will be presented in the next sections.
Case 1: Minimization of generation fuel cost
The first case studied in this article is the basic case of minimizing the cost generation
fuel expressed by a quadratic function. Therefore, the objective function of this case is Eq.
(23):

258
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

NG
J   f i $ / h  (23)
i 1

Where: fi - is the fuel cost of the ith generator. Usually, the OPF generation fuel cost curve is
formulated by a quadratic function.
Hence, fi can be formulated as follows Eq. (24):


f i  ai  bi PGi  c i PG2i  (24)

Where: ai, bi, ci - are the element, the linear and the quadratic cost coefficients of the ith
generator, respectively. The values of these coefficients are presented in [27].
Figure 2 appears the trend of total fuel cost over iterations. It seems that the proposed
technique has good convergence characteristics. The optimal settings of control variables are
presented in Table 1. The total fuel cost obtained by the suggested PSO-GWO technique is
(799.1079 $/h). Compared to the original PSO, GWO the total fuel cost is significantly
decreased.
Using the identical conditions (limits of control variables, initial conditions, and
system data), the results obtained in Case 1 apply the PSO-GWO technique are compared to
other methods described in the literature as appears in Table 2. There is some proof, that the
suggested technique outperforms several methods used to solve the OPF problem by
decreasing of generation fuel cost. For instance, the results obtained by the PSO-GWO are
better than the ones obtained the GWO and PSO methods.
Case 2: Minimization of fuel cost and voltage deviation
Bus voltage is one of the most significant and considerable security and service quality
indices [27]. Reducing only the total cost in the OPF problem as in Case 1 may result in a
suitable Solution, but voltage profile may not be reasonable. Consequently, this case
purposes at minimizing fuel cost with a improve voltage profile by considering a dual
objective function.
The voltage profile is optimized by reducing the load bus voltage deviation (VD) from
1.0 p.u, the objective function, in this case, can be formulated as follows Eq. (25):
J  J cost wJvoltageDeviation (25)

Where: w - is an appropriate weighting factor, to be chosen by the user to accord a weight to

each of the two expressions of the objective function. In this case, w is selection as 100.
J cost and JVoltageDeviation are presented as follows Eq. (26-27):

259
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
NG
J cost   f i (26)
i 1

NL
JvoltageDeviation   V i  1 (27)
k 1

Results and discussion

The PSO-GWO technique has been utilized to search for the optimal solution of the
problem. The variations in the fuel cost and voltage deviation through the iterations are
outlined in Figure 3a and Figure 3b.
825
GWO
820 PSO
PSO-GWO
fuel cost ($/h)

815

810

805

800

795
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 2. Convergent curves of Case 1

The optimal settings of the control variables are presented in Table 1. Apply PSO-
GWO the fuel cost and the voltage deviation yielded are (803.5881$/h) and (0.1044 p.u.),
respectively. The voltage profile obtained by PSO-GWO is compared with other algorithms
as appears in Table 2. It is clear that the voltage profile is the least among all other
comparable methods. It is decreased from 1.6899 p.u. in the case 1 to 0.1044 p.u. in case 2,
Hence, in case 2, the fuel cost is slightly augmented by 0.55% compared to case 1.

260
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

804.9
GWO
804.7 PSO
PSO-GWO
804.5
fuel cost ($/h)

804.3

804.1

803.9

803.7

803.5
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 3a. Convergent curves of Case 2
0,6336
GWO
0,558 PSO
0,4824
PSO-GWO
VD (p.u)

0,4068

0,3312

0,2556

0,18

0,1044
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 3b. Convergent curves of Case 2

CASE 3: Minimization of fuel cost considering valve point effect

So as to have a realistic and greater effective modeling of generator cost functions, the
valve point–effect must be considered. The generating units with multi-valve steam turbines
display a major variation in the fuel-cost functions and output a ripple-like effect [30]. So as
to considered the valve-point effect of generating units, a modeled as a sinusoidal term is
added to the cost function. Thus, the objective function can be formulated as follow Table 1
and Table 2:

261
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
Table 1. Optimal settings of the control variables for case1 to case 3.
Control Case 1 Case 2 Case 3
variable PSO- GWO PSO PSO- GWO PSO PSO- GWO PSO
GWO GWO GWO
PG1 (MW) 176.57 176.66 177.84 177.52 176.09 178.16 200.17 200.08 200.00
PG2 (MW) 48.81 47.83 49.06 49.26 48.92 49.11 43.14 41.86 43.53
PG5 (MW) 21.61 21.19 21.48 21.77 21.59 21.64 18.32 18.49 18.63
PG8 (MW) 21.02 20.25 22.09 22.74 22.39 22.44 10.08 11.28 10.00
PG11 (MW) 11.78 12.91 10 10 12.21 10.04 10.00 10.01 10.00
PG13 (MW) 12.21 13.21 12 12 12 12 12.00 12.01 12.00
V1(p.u) 1.1 1.1 1.1 1.05 1.04 1.06 1.10 1.10 1.10
V2(p.u) 1.09 1.09 1.1 1.03 1.02 1.04 1.08 1.09 1.10
V5(p.u) 1.06 1.06 1.1 1.01 1.01 1.01 1.06 1.06 1.08
V8(p.u) 1.07 1.07 1.1 1.00 1.01 1.00 1.06 1.07 1.10
V11(p.u) 1.1 1.1 1.1 1.1 0.98 1.1 1.10 1.09 1.10
V13(p.u) 1.1 1.1 1.1 0.97 1.03 0.96 1.10 1.10 1.10
Qc10(Mvar) 1.80 0.38 0 0 0 0 5.00 0.18 0
Qc12(Mvar) 2 0.12 0 1.35 2.24e-06 5 0.25 0.23 5.00
Qc15(Mvar) 5 1.57 0 5 5 5 3.92 0.12 5.00
Qc17(Mvar) 5 0.09 0 0 5 5 3.53 0.22 0
Qc20(Mvar) 0.01 2.47 5 5 5 5 0.59 3.43 0
Qc21(Mvar) 4.87 1.49 5 5 5 5 5.00 2.68 5.00
Qc23(Mvar) 3.08 1.58 5 5 5 5 0.02 0.07 3.66
Qc24(Mvar) 5 1.79 5 4.93 5 5 5.00 1.49 5.00
Qc29(Mvar) 0.46 3.22 0 2.64 5 5 3.81 1.61 5.00
T6–9 1.04 1.04 1.1 1.1 0.99 0.96 1.02 1.08 0.99
T6–10 0.90 0.92 0.9 0.9 0.9 1.1 0.93 0.93 1.10
T4–12 1.07 1.02 1.02 0.92 1.02 0.9 1.01 1.06 1.10
T28–27 0.96 0.99 0.99 0.97 0.98 0.98 0.99 0.99 1.10
Fuel cost 799.1079 799.5063 800.4361 803.5881 803.5802 803.7326 829.8394 830.4784 831.6223
($/h)
VD 1.6899 1.3010 1.6857 0.1044 0.1151 0.1464 1.5136 0.9963 1.0404
L max 0.1283 0.1338 0.1294 0.1487 0.1491 0.1490 0.1321 0.1362 0.1453
Emission 0.3646 0.3642 0.3687 0.3677 0.3631 0.3695 0.4433 0.4426 0.4425
(ton/h)
ploss (MW ) 8.5944 8.6613 9.0431 9.8890 9.8007 9.9780 10.3089 10.3455 10.7552

Table 2. Comparison of the results obtained for Case 1to Case 3.

Case 1 Case 2 Case 3
Algorithms Fuel cost ($/h) Algorithms VD (pu) Algorithms Fuel cost ($/h)
PSO-GWO 799.1079 PSO-GWO 0.1044 PSO-GWO 829.8394
GWO 799.5063 GWO 0.1151 GWO 830.4784
PSO 800.4361 PSO 0.1464 PSO 831.6223
MGBICA [31] 801.1409 BHBO [8] 0.1262 ICBO[36] 830.4531
ABC [32] 800.660 DE [34] 0.1357 BSA[21] 830.7779
HSFLA-SA[33] 801.79 IEM [35] 0.1270

    
NG
f (x ,u )   ai  bi PGi  c i PG2i  d i  sin e i  PGmin
i
 PGi (28)
i 1

Where: d i and e i are the coefficients that show the valve-point loading effect.

262
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

In this case to arrive at a rise in cost than in case 1 with a conclusive value being
829.8394$/h, obtained by PSO-GWO. The optimal control variables obtained are shown in
Table 1 output outcome of a method used in our study are better than most of the results
revealed in past literature on the problem of OPF that is presented in table 2.

870
GWO
865
PSO
860
PSO-GWO
fuel cost ($/h)

855

850

845

840

835

830

825
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 4. Convergent curves of Case 3

Case 4: Minimization of fuel cost and enhancement of voltage stability

The prediction of voltage instability is a problem of paramount significance in power

systems. In [28] Kessel and Glavitch have developed a voltage stability index named L max

which is defined build on local indicators L j and it is presented by Eq. (29):

L max  max  L j , j  1, 2,..., NL (29)

Where: L j is the local indicator of bus j and it is given as follows Eq. (30)

j  1, 2,..., NL
NG
Vi
L j  1   H LG ji (30)
i 1 Vj

Where: H - matrix is produced by the partial inversion of Ybus. More specifics can be given in
[28].
The indicator L max varies between 0 and 1 where the lower the indicator, the more the

system stable. Thus, enhancing voltage stability can be obtained by the minimization of L max
the complete system [29]. Hence, the objective function can be formulated as Eq. (31):
 NG 
J  x ,u     ai  bi PGi  c i PG2i   Lmax  L max  (31)
 i 1 

263
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
Where: Lmax is a weighting factor chosen as 100 in this work. The results of the optimization

study are presented in Table 3 while the direction of convergence appears in Figure 5a and 5b.
It seems that the Lmax has been decreased from 0.1283 to 0.1251compared with CASE 1,
Hence the results obtained are compared with other algorithms as given in Table 4.

Table 3. Optimal settings of the control variables for case 4 to case 6.

Control Case 4 Case 5 Case 6
variable PSO- GWO PSO PSO- GWO PSO PSO- GWO PSO
GWO GWO GWO
PG1 (MW) 177.13 178.06 177.9656 113.11 112.85 111.84 51.39 51.55 51.37
PG2 (MW) 49.22 49.64 49.1534 59.11 58.98 58.40 79.93 79.91 80.00
PG5 (MW) 21.99 21.99 21.3946 27.48 27.61 27.35 50.00 49.96 50.00
PG8 (MW) 18.60 19.01 21.9131 35.00 35.00 35.00 35.00 34.94 35.00
PG11 (MW) 12.28 11.51 10.0000 26.60 27.87 30.00 29.99 30.00 30.00
PG13 (MW) 12.98 12.14 12.0000 27.19 26.29 26.05 40.00 39.95 40.00
V1(p.u) 1.10 1.10 1.1000 1.10 1.10 1.10 1.10 1.10 1.10
V2(p.u) 1.09 1.09 1.1000 1.09 1.09 1.10 1.10 1.10 1.10
V5(p.u) 1.07 1.07 1.0744 1.07 1.07 1.08 1.08 1.08 1.10
V8(p.u) 1.07 1.08 1.0852 1.08 1.08 1.10 1.09 1.09 1.10
V11(p.u) 1.10 1.10 1.0624 1.10 1.10 1.10 1.10 1.10 1.10
V13(p.u) 1.10 1.10 1.1000 1.10 1.10 1.10 1.10 1.10 1.10
Qc10(Mvar) 5.00 1.81 0 2.89 2.24 5.00 0.01 3.44 5.00
Qc12(Mvar) 0.05 0.30 5.0000 0.96 3.62 0 0.47 0.85 5.00
Qc15(Mvar) 3.04 0.39 0 5.00 1.22 0 2.71 2.78 5.00
Qc17(Mvar) 0.02 1.73 5.0000 4.72 0.53 5.00 4.58 3.83 0
Qc20(Mbar) 5.00 0.10 5.0000 3.91 0.00 0 5.00 1.07 0
Qc21(Mvar) 0.36 4.87 0 0.27 1.33 5.00 0.50 3.70 5.00
Qc23(Mvar) 0.51 0.17 5.0000 1.82 1.74 0 4.27 3.86 0
Qc24(Mvar) 0.62 0.55 5.0000 5.00 0.03 0 5.00 1.02 5.00
Qc29(Mvar) 0.84 0.24 0 1.14 0.02 2.21 0.03 1.80 2.21
T6–9 0.99 0.98 0.9000 1.04 0.97 0.97 1.04 1.08 1.10
T6–10 0.91 0.93 1.1000 0.92 1.02 1.10 0.92 0.90 0.90
T4–12 0.99 0.98 1.0132 1.01 1.03 1.10 1.01 1.00 1.01
T28–27 0.94 0.94 0.9426 0.99 0.98 1.01 0.97 0.99 0.99
Fuel cost 799.7188 799.9043 800.3247 834.9486 835.9200 838.1687 966.9805 966.5853 967.3598
($/h)
VD 1.8169 1.7976 1.7615 1.6909 1.3259 1.2613 1.8087 1.6767 1.8015
L max 0.1251 0.1250 0.1255 0.1303 0.1327 0.1343 0.1281 0.1302 0.1282
Emission 0.3659 0.3690 0.3692 0.2426 0.2422 0.2406 0.2072 0.2072 0.2072
(ton/h)
ploss (MW ) 8.7897 8.9339 9.0267 5.0851 5.1929 5.2424 2.8946 2.9100 2.9732

Table 4. Comparison of the results obtained for Case 4 to Case 6.

Case 4 Case 5 Case 6
Algorithms L max Algorithms Emission(ton/h) Algorithms Ploss (MW)
PSO-GWO 0.1251 PSO-GWO 0.2426 PSO-GWO 2.8946
GWO 0.1250 GWO 0.2422 GWO 2.9100
PSO 0.1255 PSO 0.2406 PSO 2.9732
ABC [32] 0.1379 MOGWO [40] 0.245126 MSA[39] 3.1005

264
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

ARCBBO[37] 0.1369 NSGA-II[40] 0.3214 ABC[32] 3.1078

Gbest-ABC [38] 0.1370 ARCBBO[37] 3.1009
851
GWO
844
PSO
837 PSO-GWO
fuel cost ($/h)

830

823

816

809

802

795
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 5a. Convergent curves of Case 4
0.12
GWO
0.19 PSO
0.18
PSO-GWO

0.17
L-max

0.16

0.15

0,14

0,13

0.12
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 5b. Convergent curves of Case 4

CASE 5: Minimization of fuel cost and emission

Electrical power generation from conventional sources of energy emits hazardous
gases into the environment. The quantity of sulfur oxides (SOx), nitrogen oxides (NOx)
emission in tons per hr (t/h) is higher with the rise in generated power (in p.u. MW) next to
the relationship presented in Eq. (32).
 i PGi  
 
NG
f  x ,u   emission     i  i PGi   i PG2i  0.01  i e (32)

i 1 


Where: αi, βi, Υi, ωi and μi are all emission coefficients provided in [21]. The objective
function for this case is assumed by Eq. (33):
 NG 
f  x ,u     ai  bi PGi  c i PG2i   E  emission (33)
 i 1 
The weight factors are selected as E = 100 in this case.

265
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
The results yielded after optimization applied the PSO-GWO technique are presented
in Table 3 and the trend of optimization is shown in Figure 6. The results appear that the
emission has been decreased from (0.3646 ton/h) to (0.2426 ton/h), Thus, the total fuel cost
has augmented from (799.1079$/h) to (834.9486$/h) i.e. by 4.29% compared with CASE 1,
and the results obtained are compared with other techniques as shown in Table 4.
914
GWO
904
PSO
894 PSO-GWO
fuel cost ($/h)

884

874

864

854

844

834
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 6. Convergent curves of Case 5

Case 6: Minimization of real power loss

In this case, the purpose of the OPF problem is to minimize power losses; the real
power loss to be minimized is formulated as follows, Eq. (34):

G ij V i 2 V j2  2V iV j cos  ij  
nl nl
f  x ,u   Ploss   (34)
i 1 j 1, j  i

Where:  ij   i   j is the difference in voltage angles between bus 𝑖 and bus 𝑗 and G ij is

transfer conductance.
The tendency to decrease the objective function of total real power transmission loss
using the PSO-GWO technique appears Figure 7. The optimal settings of the control variables
are presented in Table 3 in this case 6 by PSO-GWO result in real power losses of 2.8946
MW, better than all the results summarized in Table 4.

266
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

9
GWO
8 PSO
PSO-GWO
Ploss (MW) 7

2
0 50 100 150 200 250 300 350 400 450 500
Iteration
Figure 7. Convergent curves of Case 6

Conclusions

In this article, a new Hybrid technique called PSO-GWO is applied to solve OPF
problems. One test system and six cases have been studied in order to evaluate the
performance of the suggested technique. The OPF problem was reported as a non-linear
optimization problem with equality and inequality constraints. Where several objective
functions have been considered to minimize the fuel cost, to improve the voltage profile, and
to enhance the voltage stability. However, the yielded results have been compared to those
yielded using standard optimization techniques such as PSO, GWO. The essential conclusion
that can be extracted from this article is that the PSO-GWO is a very efficient and robust
technique for solving OPF problems. It has perfect convergence characteristics and can be
realized better effectiveness than some well-known optimization techniques. A comparison of
the results yielded from PSO-GWO and other techniques confirm the superiority of the
algorithm for the suggested PSO-GWO on stochastic methods in terms of solution efficiency
for the OPF problems.

References

1. Cain M., O'Neill R., Castillo A., History of optimal power flow and formulations, FERC
Staff Tech Pap; 2012. p. 1–36.

267
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
2. Hinojosa V.H., Araya R., Modelling a mixed-integer-binary small-population
evolutionary particle swarm algorithm for solving the optimal power flow problem in
electric power systems, Appl. Soft Comput. J., 2013, 13, p. 3839–3852.
3. Niknam T., Azizipanah-Abarghooee R., Narimani M.R., Reserve constrained dynamic
optimal power flow subject to valve-point effects, prohibited zones and multi-fuel
constraints, Energy, 2012, 47 (1):451e64.
4. Ambriz-Perez H., Acha E., Fuerte-Esquivel C.R., De La Torre A., Incorporation of a
UPFC model in an optimal power flow using Newton’s method, IEEE Proc Gener Transm
Distrib., 1998, 145: 336e44.
5. Momoh J.A., El-Hawary M.E., Adapa R., A review of selected optimal power literature to
1993. Part II: Newton, linear programming and interior point methods, IEEE Trans
Power Syst., 1999; 14:105e11.
6. Burchett R.C., Happ H.H., Vierath D.R., Quadratically convergent optimal power flow,
IEEE Trans Power Appar Syst., 1984; 103 (11), p. 3267–76.
7. Niknam T. , Raoul Narimani M., Jabbari M., Malekpour A.R., A modified shuffle frog
leaping algorithm for multi-objective optimal power flow, Energy, 2011, 36, p. 6420–
6432.
8. Bouchekara H.R.E.H., Optimal power flow using black-hole-based optimization
approach, Appl Soft Comput J., 2014, 24, p. 879–88.
9. Bouchekara H.R.E.H., Abido M.A., Chaib A.E., Mehasni R., Optimal power flow using
the league championship algorithm: a case study of the Algerian power system,
Energy Convers. Manag., 2014, 87, p. 58–70.
10. Bhattacharya A., Roy P.K., Solution of multi-objective optimal power flow using
gravitational search algorithm, IET Gener. Transm. Distrib., 2012, p. 751–763.
11. Khorsandi A., Hosseinian S.H., Ghazanfari A., Modified artificial bee colony algorithm
based on fuzzy multi-objective technique for optimal power flow problem, Electr. Power
Syst. Res., 2013, 95, p. 206–13.
12. Daryani N., Hagh M.T., Teimourzadeh S., Adaptive group search optimization
algorithm for multi-objective optimal power flow problem, Appl. Soft Comput., 2016,
38, p. 1012–1024.
13. Ghasemi M., Ghavidel S., Ghanbarian M.M., Massrur H.R., Gharibzadeh M., Application
of imperialist competitive algorithm with its modified techniques for multi-objective

268
 Leonardo Electronic Journal of Practices and Technologies Issue 32, January-June 2018
ISSN 1583-1078 p. 249-270

optimal power flow problem: a comparative study, INF Sci (Ny), 2014, 281, p. 225–47.
14. Bouchekara HRE-H, Abido M.A., Optimal power flow using differential search
algorithm, Electr Power Components Syst., 2014, 42, p. 1683–99.
15. Bouchekara H.R.E.H., Abido M.A., Boucherma M., Optimal power flow using teaching-
learning-based optimization technique, Electr. Power Syst. Res., 2014, 114, p. 49–59.
16. Mukherjee A., Mukherjee V., Solution of optimal power flow using chaotic krill herd
algorithm, Chaos, Solitons Fract., 2015, 78, p. 10–21.
17. Rao B.S., Vaisakh K., Multi-objective adaptive clonal selection algorithm for solving
optimal power flow considering multi-type FACTS devices and load uncertainty,
Appl. Soft Comput., 2014, 23, p. 286–297.
18. Mirjalili S., Lewis A., Adaptive gbest-guided gravitational search algorithm, Neural
Comput Appl., 2014, 25 (7–8), p. 1569–1584.
19. Dhillon J.S., Kothari D.P., Power system optimization, 2nd edn. PHI, New Delhi, 2010.
20. Mirjalili S., Mirjalili V., Lewis A., Grey wolf optimizer, Adv. Eng. Softw., 2014, 69, p.
46–61.
21. Chaib A.E., Bouchekara H.R.E.H., Mehasni R., Abido M.A., Optimal power flow with
emission and non-smooth cost functions using backtracking search optimization
algorithm, Int. Journal of Electrical Power & Energy Systems, 2016, 81, p. 64-77.
22. Yohannes M.S., Solving economic load dispatch problem using particle swarm
optimization technique, International Journal of Intelligent Systems and Applications
(IJISA), 2012, 4 (12), 12.
23. Mirjalili S., Mirjalili S.M., and Lewis A., Grey wolf optimization, Advances in
Engineering Software, 2014, 69, p. 46-61.
24. Sharma S., Mehta S, Chopra N., Economic load dispatch using grey wolf optimization,
International Journal Of Engineering Research and Applications(IJERA),ISSN:2248-
9622, 2015, 5 (4).
25. Vikram K.K., A novel hybrid PSO–GWO approach for unit commitment problem, Neural
Computing and Applications, Springer, 2015.
26. Abido M.A., Optimal power flow using particle swarm optimization, Int. J. Electr. Power
Energy Syst., 2002, 24 (7), p. 563–571.
27. Abou El Ela A.A., Abido M.A., Optimal power flow using differential evolution
algorithm, Electr. Power Syst. Res., 2010, 80 (7), p. 878–885.

269
A new hybrid algorithm of particle swarm optimizer with Grey wolves optimizer for solving optimal power flow
problem
Aboubakr KHELIFI, Saliha CHETTIH and Bachir BENTOUATI
28. Kessel P., Glavitsch H., Estimating the voltage stability of a power system, IEEE Trans
Power Deliv., 1986, 1, p. 346–54.
29. Abou El Ela A.A., Abido M.A., Spea S.R., Optimal power flow using differential
evolution algorithm, Electr. Power Syst. Res., 2010, 80, p. 878–85.
30. Hardiansyah H., A modified particle swarm optimization technique for economic load
dispatch with valve-point effect, Int. J. Intell Syst. Appl., 2013, 5, p. 32–41.
31. Ghasemi M., Ghavidel S., Ghanbarian M.M., Gitizadeh M., Multi-objective optimal
electric power planning in the power system using Gaussian bare-bones imperialist
competitive algorithm, Inf. Sci., 2015, 294, p. 286–304.
32. Adaryani M.R., Karami A., Artificial bee colony algorithm for solving multi-objective
optimal power flow problem, Int. J. Electr. Power Energy Syst., 2013, 53, p. 219–230.
33. Niknam T., Narimani M.R., Azizipanah-Abarghooee R., A new hybrid algorithm for
optimal power flow considering prohibited zones and valve point effect, Energy
Convers. Manage., 2012, 58, p. 197–206.
34. Vaisakh K., Srinivas L.R. Evolving ant direction differential evolution for OPF with non-
smooth cost functions, Engineering Applications of Artificial Intelligence, 2011, 24, p.
426–436.
35. Bouchekara H.R.E.H., Abido M.A., Chaib A.E., Optimal power flow using an improved
electromagnetism-like mechanism method, Electric Power Components and Syst., 2016.
36. Bouchekara H.R.E.H., Chaib A.E., Abido M.A., El-Sehiemy R.A., Optimal power flow
using an improved colliding body’s optimization algorithm, Applied Soft Computing,
2016, 42, p. 119-131.
37. Kumar A.R., Premalatha L., Optimal power flow for a deregulated power system using
adaptive real coded biogeography-based optimization, Electr. Power Energy Syst., 2015,
73, p. 393–399.
38. Roy R., Jadhav H.T., Optimal power flow solution of power system incorporating
stochastic wind power using Gbest guided artificial bee colony algorithm, Electrical
Power and Energy Systems, 2015, 64, p.562–578.
39. Mohamed A.A.A., Mohamed Y.S., El-Gaafary A.A., Hemeida A.M., Optimal power flow
using moth swarm algorithm, Electric Power Systems Research, 2017, 142, p. 190-206.
40. Dilip L., Bhesdadiya R., Jangir P., Optimal power flow problem solution using multi-
objective grey wolf optimizer algorithm, Springer Nature Singapore, Pte. Ltd., 2018.

270