Received: 29 January 2018 Revised: 12 September 2018 Accepted: 18 September 2018

DOI: 10.1002/etep.2743

RESEARCH ARTICLE

An improved moth‐flame optimization algorithm for

solving optimal power flow problem

Mahrous A. Taher1 | Salah Kamel2,3 | Francisco Jurado4 | Mohamed Ebeed5

1
Hydro Power Plants Company, Aswan,
Summary
Egypt
2
State Key Laboratory of Power
This paper proposes an improved moth flame optimization (IMFO) algorithm
Transmission Equipment and System to effectively solve the optimal power flow (OPF) problems. The concept of
Security and New Technology, Chongqing moth flame optimization (MFO) is inspired from the movement of moth
University, Chongqing, China
3 towards the moon direction. IMFO is mainly based on the concept of MFO
Electrical Engineering Department,
Faculty of Engineering, Aswan University, with modifying the path of moths in new spirals around the flame. Standard
Aswan 81542, Egypt IEEE 30‐bus, IEEE 57‐bus and IEEE 118‐bus test systems are used to validate
4
Department of Electrical Engineering, and prove the efficiency and robustness of IMFO algorithm. The validation
University of Jaén, 23700 EPS Linares,
Jaén, Spain
of the proposed algorithm is based on 15 case studies in terms of different sin-
5
Department of Electrical Engineering, gle and multi‐objective functions: fuel cost minimization, gas emission reduc-
Faculty of Engineering, Sohag University, tion, active power loss minimization, voltage profile improvement, and
Sohag, Egypt
voltage stability enhancement. The simulation results of the proposed algo-
Correspondence rithm are compared with those obtained by other well‐known optimization
Francisco Jurado, Department of techniques. The obtained results demonstrate the capability and robustness
Electrical Engineering, University of Jaén,
23700 EPS Linares, Jaén, Spain.
of IMFO algorithm to solve OPF problems. The results reveal that IMFO algo-
Email: [email protected] rithm is capable of finding precise and better OPF solutions compared with the
other techniques. A comparison among the convergence characteristics of
IMFO technique and the other techniques proves the prevalence of IMFO to
attain the optimal power flow solution with fast convergence.

KEYWORDS
improved moth flame optimization, metaheuristic, optimal power flow, power system optimization

Nomenclature: ai; bi; ci, cost coefficient of the i‐th generator; Dij, distance of i‐th moth with respect to j‐th flame; di; ei, coefficients reflecting the
valve‐point effect; gi(x, u), equality constraints; Gij, Bij, conductance and susceptance of the admittance matrix; hi(x, u), inequality constraints; KG,
KQ,KV,KS, penalty factors of active, reactive power, voltage, and apparent power; Lj, voltage stability local indicator of bus j; N, population size (the
number of individuals); NB, number of busses; NC, number of shunt compensators; NG, number of generators; NPV, number of generation busses;
NPQ, number of load busses; NTL, number of transmission lines; NT, the number of regulating transformers; PG1, active power generation of slack
G1 ; PG1 , active power generation limits of slack bus; PGi ; PGi , active power generation limits of bus i; PG, PD, active power generation and
bus; Pmin max min max

load demand, respectively; Ploss, Qloss, active and reactive power transmission losses,; QC, shunt VAR compensation; QG, QD, reactive
power generation and load demand; Qmin Gi ; QGi , reactive power generation limits of the shunt VAR; Sli; Smax, apparent power flow of ith line
max

and its maximum; T min i ; T max

i , upper and lower limits of regulating transformer i,; u, vector of the independent variables or the control
Li ; V Li , upper and lower limits of voltage magnitude load bus i,; VLi, voltage magnitude at load bus i; VD, load bus voltage
variables; V min max

deviation; VGi, voltage magnitude at PV busses; x, vector of dependent variables or state variables; Yij, admittance matrix between bus i and
bus j; γi, βi, αi, ζi,and λi, coefficients of the ith generator emission; θ, polar angle; δij, phase angle difference between busses i and j; λLmax,
weighting factor of the Lmax with cost term; λi, weighting factor of the emission with the cost term; λVD, weighting factor of the VD term
with the cost term

Int Trans Electr Energ Syst. 2019;29:e2743. wileyonlinelibrary.com/journal/etep © 2018 John Wiley & Sons, Ltd. 1 of 28
https://doi.org/10.1002/etep.2743
2 of 28 TAHER ET AL.

1 | INTRODUCTION

The optimal power flow (OPF) is considered one of the nonlinear and complex optimization problems in power system
used for planning or operation stages.1,2 OPF is an important tool able to enhance power system efficiency, improve
voltage profile, improve stability index, and reduce the total gas emissions in thermal power plants (eg, minimization
of nitrogen oxide [NOx]).3 OPF regulates both continuous and discrete control variables to optimize certain objective
functions by satisfying the operating constraints.4,5 It is possible to merge one or more OPF advantages to gain compro-
mise best solutions from both of technical requirements and economical demands such as adaptive genetic algorithms
with adjusting population size6 and evolving ant direction hybrid differential evolution (DE) algorithm.7 Different
conventional optimization techniques, such as nonlinear programming,8-12 quadratic programming,12 and linear pro-
gramming,13,14 have been proposed to solve OPF problem. These techniques have a demerit in the regime of handling
of multi‐objective nonlinear functions, where it is not possible to identify the global optima.15 In addition, they need
complicated mathematical calculations which need long time, and they fail to deal with discrete variables.16 Stochastic
algorithms can be executed by random operators to find out the global optima in the search space. Simultaneously, it
avoids falling in local optima and gives different solutions at each run.17 Evolutionary algorithms are considered as
stochastic algorithms18 that search for global optimum solutions in the search space. The set of these solutions are
called candidate solutions which can be enhanced by the successive iterations. The iterations are executed until
reaching accurate global optima. This process is done without stagnation at local optima. The other category of optimi-
zation processes is the single or individual‐based optimization techniques which depend upon guessing a solution.
Hence, the evolutionary algorithms that are population‐based algorithms have substantial advantages over individ-
ual‐based algorithms as follows: It does not depend upon either the solved problem or the derivation of the modal
equations, it avoids trapping at local optima, and it is simple algorithm. Numerous new optimization techniques based
on the principle of population have been proposed to solve the complicated optimization problems.19 Many heuristic
methods have been used to solve OPF problem such as genetic algorithm (GA),20 DE,21-24 biogeography‐based optimi-
zation algorithm (BBO),25 evolution strategy26, probability based incremental learning,27 simulated annealing (SA),28,29
gravitational search algorithm (GSA),30 improved colliding bodies optimization (ICBO) algorithm,31 black hole,32 Big‐
Bang big‐crunch,33 multi‐objective forced initialized differential evolution algorithm (MO‐DEA),34 hybrid modified
particle swarm optimization and the shuffle frog leaping algorithms,35 and particle swarm optimization (PSO),36 moth
flame optimization (MFO) algorithm innovated by Seedily Mirjalili,37 ant colony optimization, first proposed by Dorigo
et al,38 artificial bee colony (ABC) algorithm,39 teaching learning based optimization (TLBO),40,41 Lévy mutation
teaching‐learning‐based optimization (LTLBO) algorithm,42 fuzzy harmony search algorithm (FHSA),43Tabu (Taboo)
search,44 group search optimizer,45 and league championship algorithm.46 Proposal of new algorithms is supported
by “No Free Lunch theorem.”47
Artificial intelligence techniques or metaheuristic methods have the merit of obtaining better solutions at low com-
puting time simultaneously. GA simulates the method of evolution, where there is a great chance for the promising
individuals to carry their gents to the elite offspring. This is achieved for the population through successive generations.
Tabu (Taboo) search is a method that can expand its search beyond the limits of search space with systematic
preventing of some solutions from cycling process and stagnation into local optima. The latest solutions have the
possibility to search through attainable points with a reasonable sequence of perturbations. SA is inspired from the
cooling process of materials according to suitable cooling schedule. SA has the ability to avoid stagnation into local
optima and give fast convergence time. PSO technique is inspired from the velocity of a vector to update its position
with respect to each particle in the swarm. Social relationships between the individuals are responsible for assigning
the position of the particles with respect to each other. PSO has the advantages of less computing time and better
solutions when compared with other metaheuristic techniques.48
As the nonlinear optimization algorithms are widely spread due to their high efficiency and better solutions of OPF
problems, the convergence cannot be guaranteed as a result of nonlinearity and nonconvexity of OPF problem.49 The
OPF equations are relaxed to inequality constrains which show the convex region.50,51 If the required conditions to
secure the convex relaxation are not considered, the obtained solutions are not obvious from the physical meaning side.
OPF methods approximate the main nonlinear power flow equations to find linearized equalities. According to the
aimed approximations, the obtained solutions should be nearby to the optimum solutions of the original nonlinear
model. The obtained solutions by linear approximations have the merit of good quality without changing under differ-
ent operating conditions, although the global optimum of the original AC OPF model cannot be secured. In practice,
linear approximation approaches are used by most operators to solve OPF problems due to secure convergence and
TAHER ET AL. 3 of 28

numerical efficiency. However, their drawback is less modeling accuracy.52 Linear approximations of the OPF problem
for the above‐mentioned approaches have been characterized by cold start,53 where the final obtained solutions are not
related to the initial points. Another type is “hot‐start” solutions, where linearization of equations is mathematically
performed by empirical knowledge.53-55
Moth flame optimization is one of the new proposed optimization techniques inspired by the movement of moth
with respect to the moon. Movement of moth towards the moon is called the transverse orientation. When moths are
moved around an artificial light, it will move in a spiral path until falling into the light.37
The aim of this paper is to develop improved moth flame optimization (IMFO) algorithm to solve OPF problems.
However, the main contributions of this paper can be summarized as follows:

• Develop an efficient version of MFO algorithm, called IMFO, to avoid local optima stagnation and enhance the con-
vergence characteristics of the traditional MFO.
• The developed optimization algorithm is mainly based on the concept of MFO with modifying the path of moths in
new spirals around the flame.
• Single and multi‐objective functions are considered to confirm the capability of the proposed optimization technique
to solve different types of OPF problems.
• Comparison among IMFO, traditional MFO, and other well‐known optimization algorithms is established in order
to confirm the efficiency and robustness of the proposed technique.
• Standard IEEE 30‐bus, IEEE 57‐bus, and IEEE 118‐bus test systems are used to validate and prove the efficiency and
robustness of IMFO to solve OPF problems of various power system scale sizes.

The rest of this paper is organized as follows: Section 2 represents the mathematical formulation of OPF. Section 3
explains the proposed optimization (IMFO) algorithm. In Section 4, the simulations carried out and the most relevant
results obtained are reported. Section 5 presents the conclusion of this paper and the expected future work.

2 | OPTIMA L POWER FLOW PR O B L E M FO R M ULA T I O N

Optimal power flow problem can be represented mathematically in the following formula:

Min F ðx; uÞ (1)

Subjected to

gj ðx; uÞ ¼ 0 j ¼ 1; 2; …; m (2)

hj ðx; uÞ ≤ 0 j ¼ 1; 2; …; p (3)

where, F is the objective function to be optimized, x is a vector that represents the dependent variables (state vari-
ables), u is a vector that represents the independent (control) variables, g(x,u) is the equality constraints and h(x,u) is
the inequality constraints, m is the number of equality constraints, and p is the number of inequality constraints.
The state and control variables of OPF can be explained as follows:

2.1 | State variables

The state variables are the dependent variables x that describe a unique state of the power system. The state variables of
OPF can be represented as follows:

x ¼ ½PG1 ; V L1 …V LNPQ ; QG1 …QGNPV ; STL1 …STLNTL  (4)

where PG1 is the slack bus active power, VL is the voltage magnitude of load bus, QG is the generator reactive power
output, STL is the apparent power flow in transmission line, NPQ is the number of load busses, NPV is the number
of generation busses, and NTL is the number of transmission lines.
4 of 28 TAHER ET AL.

2.2 | Control variables

The control variables are the independent variables u that can fulfill the power flow equations. The control variables of
OPF can be represented as follows:

u ¼ ½PG2 …PGNG ; V G1 …V GNG ; QC1 …QCNC ; T 1 …T NT  (5)

where PG is the output active power of generator, VG is the voltage of generation bus, QC is the injected reactive power of
shunt compensator, T is the tap setting of regulating transformer, NG is the number of generators, NC is the number of
shunt compensator units, and NT is the number of transformers.

2.3 | Operating constraints

Optimal power flow constraints are divided into two categories: equality constraints and inequality constraints. The
equality and inequality constraints can be explained as follows:

2.3.1 | Equality constraints

Equality constraints are the load flow equations that balance between active and reactive power flow, as follows:
NB

 
PGi − PDi ¼ jV i j∑j¼1
V j
Gij cosδij þ Bij sinδij (6)

NB

 
QGi − QDi ¼ jVi j∑j¼1
V j
Gij cosδij þ Bij sinδij (7)

where PGi is the generator active power, PDi is the demand active power, QGi is the generator reactive power, and QDi is
the demand reactive power. Gij is the transfer conductance and Bij is the susceptance between bus i and bus j,
respectively.

2.3.2 | Inequality constrains

The inequality constraints are the limits of control variables given as follows:

• Generators active power output

Gi ≤ P Gi ≤ P Gi i ¼ 1; 2; …; NPV
Pmin max
(8)

• Generators bus voltages

Gi ≤ V Gi ≤ V Gi
V min i ¼ 1; 2; …; NPV
max
(9)

• Generators reactive power output

Gi ≤ QGi ≤ QGi
Qmin i ¼ 1; 2; …; NPV
max
(10)

• Transformer tap settings

T min
i ≤ T i ≤ T max
i i ¼ 1; 2; …; NT (11)

• Shunt VAR compensator

Ci ≤ QCi ≤ QCi
Qmin i ¼ 1; 2; …; NC
max
(12)
TAHER ET AL. 5 of 28

Apparent power flow in transmission lines

SLi ≤ Smax
Li i ¼ 1; 2; …; NTL (13)

• Voltage magnitude of load busses

Li ≤ V Li ≤ V Li
V min i ¼ 1; 2; …; NPQ
max
(14)

2.3.3 | Constraints formulation

The inequality constraints are combined into extra terms added to the objective function equation. The added terms are
called the penalty function that emphasize the feasibility of the proposed technique. Penalty function is a quadratic term
can be expressed as follows56:

NPV NPQ NTL

F g ðx; uÞ ¼ F i ðx; uÞ þ K G ðΔPG1 Þ2 þ K Q ∑i¼1 ðΔQGi Þ2 þ K V ∑i¼1 ðΔV Li Þ2 þ K S ∑i¼1 ðΔSLi Þ2 (15)

where,KG,KQ, KV, and KS are the penalty factors, these values are high positive. ΔPG1, ΔQGi, ΔVLi, and ΔSLi are penalty
terms which can be defined as follows:
8  
>
< PG1 − Pmax
G1 PG1 > Pmax
G1
 
ΔPG1 ¼ PG1 − Pmin PG1 < Pmin (16)
>
:
G1 G1
0 Pmin
G1 < PG1 < Pmax
G1

8  
>
< QGi − Qmax
Gi QGi > Qmax
Gi
 
ΔQGi ¼ QGi − Qmin QGi < Qmin
>
:
Gi Gi
0 Qmin max
Gi < QGi < QGi
(17)

8  
>
< V Li − V max
Li V Li > V max
Li
 
ΔV Li ¼ V Li − V min V Li < V min
>
:
Li Li
0 V min
Li < V Li < V max
Li
(18)

(  
SLi − Smax
Li SLi > Smax
Li
ΔSLi ¼ (19)
0 Smin max
Li < SLi < SLi

3 | IMPROVED MOTH FLAME OPTIMIZATION (IMFO)

3.1 | Inspiration
This algorithm is inspired from the movement of moth at night according to the moon light. Moths fly at night accord-
ing to constant angle relative to the moon position. As the distance from earth to the moon is too long, the path can be
considered as a straight line. The movement of moths towards the moon is called transverse orientation. Moths move
around an artificial light in spiral paths.37
In the presence of artificial light, the moths follow the same angle with respect to the moon position, fly in a spiral
path, and dropped into the light.57 The basic flying feature of moths is shown in Figure 1.
The previous behaviour of moth is modeled mathematically in the MFO.
6 of 28 TAHER ET AL.

FIGURE 1 Basic moth flying characteristics

3.2 | MFO algorithm

Moth flame optimization algorithm formulates the spiral movement of moth towards the flame (light). In MFO, the can-
didate solutions are the moths, and the control variables are the position of the moths in the search space. The set of
moths M is represented in the following matrix:
2 3
m1;1 m1;2 ⋯m1;d
6 m2;1 m2;2 ⋯m2;d 7
6 7
M¼6 7 (20)
4 ⋮ ⋱ ⋮ 5
mn;1 mn;2 ⋯mn;d

where n is the number of moths and d is the number of variables.

The fitness values sorting can be given by the following array:
2 3
OM 1
6 OM 2 7
6 7
OM ¼ 6 7: (21)
4 ⋮ 5
OM n

The fitness value is the return value of each moth, where all moths are passed through the fitness function. The out-
put of the fitness function is identical to its fitness value in OM array.
A basic matrix of MFO is represented by flames. Flames matrix can be described as follows:
2 3
F 1;1 F 1;2 ⋯F 1;d
6 F 2;1 F 2;2 ⋯F 2;d 7
6 7
F¼6 7: (22)
4 ⋮ ⋱ ⋮ 5
F n;1 F n;2 ⋯F n;d

Dimension of moth's matrix is equal to the dimension of flames matrix. The fitness values of flames can be sorted in
the following array:
2 3
OF 1
6 OF 2 7
6 7
OF ¼ 6 7: (23)
4 ⋮ 5
OF n

It is worth mentioning that both of moths and flames are solutions. It is possible to recognize the difference between
them when analyzing the way to treat and update moths and flame positions during running process. Moths are the
TAHER ET AL. 7 of 28

search agents seek for best position, while flames are the flags or the best position of moths. The mathematical equation
represents the movement of moths with respect to flame position, which can be formulated as follows:
 
Mi ¼ S Mi; Fj : (24)

where,

• Mi represents the moth at the order i.

• Fj represents the flame at the order j.
• S represents the spiral path function.

In MFO algorithm, the movement of moths proceeds as logarithmic spiral function to update the position of each
moth with respect to flame; any spiral movement should satisfy the following conditions:

• Spiral movement starts from moth.

• Spiral movement ending at flame position.
• Domain of spiral movement is restricted by search space.

MFO logarithmic spiral function can be given by the following equation:

S ðMi; FjÞ ¼ Di:ebt :cosð2πt Þ þ Fj (25)

where:

• Di represents the distance of moth at the order i with respect to flame at the order j.
• b is a constant that assigns the shape of spiral algorithm.
• t is a random number that lies in between [−1, 1].

The distance D can be calculated from the following equation:

Di ¼ ∣F j − M i ∣: (26)

Equation 25 indicates the spiral path of the flying moths which update their position according to the flame position.
Parameter t defines the closeness of moth to flame, where at (t = −1) the moth is at the closest position with respect to
flame, while at (t = 1) the moth is at the farthest position with respect to flame in the hyper ellipse of moths around the
flame in all directions within the search space. The logarithmic spiral path of moths around flame and the positions are
indicated at different values of t, as shown in Figure 2. Seeking to find better solutions is enhanced by considering the
best solutions as the flames. Moths update their position with respect to matrix F according to Equation 22 when the
optimization process is proceeding, where n is the latest best solution. Exploitation is more fulfilled by considering it as
random [r, 1], where r is the convergence constant that decreases linearly in the range [−1: −2], so that moths exploit
their identical flames precisely in rates proportional to iterations number. The number of flames is decreasing with the
progress of iterations running as shown in Figure 3 and emphasized in the following equation:
 
N −1
Flame no ¼ N − l * (27)
T

where,

∘ l represents the current number of iteration.

∘ N represents the maximum number of flames.
∘ T represents the maximum number of iterations.

The progressive decrease in the number of flames over the course of iterations balances between exploration and
exploitation within the search space.
8 of 28 TAHER ET AL.

FIGURE 2 Logarithmic spiral space around a flame and its position with respect to t

FIGURE 3 Number of flame is decreased adaptively over the course of iterations

3.3 | IMFO algorithm

Aforementioned moths movement behaviour is transverse orientation. Equation 24 represents the movement of moths
in hyper sphere space. IMFO technique is characterized by logarithmic spiral movement of moths around the flames in
the search space as indicated in Equation 25. The spiral movement of moths is considered the basic fundamental of
IMFO that shows the extent of the updated moth locations around the flames. Spiral equation is characterized by the
possibility of moths to move around flames and not necessary in the space between them; hence, the exploration and
exploitation of the search space are fulfilled. In Soliman et al,58 the authors proposed to modify Equation 25. This mod-
ification reveals new spirals for the expected paths of the moths around the flame.

3.3.1 | Theory of hyperbolic spirals

Euclidean model of hyperbolic geometry is chosen to clarify the concept of hyperbolic spiral. Hyperbolic geometry cannot
be included within Euclidean space with smooth distance. The Poincare's circle is suitable model due to two reasons:
firstly it completely lies inside the bounding circle in the Euclidean plane, Secondly, it maintains the correct angles
between directions within small areas though distorting distances. Hyperbolic distances are performed as small Euclidean
distances. The bounding circle is considered as a unit circle lies at the origin of the Euclidean plane. This circle is consid-
ered as the origin of the hyperbolic plane. Let rH represent the hyperbolic distance from the origin, Euclidean distance
TAHER ET AL. 9 of 28

sinh x
from the origin rE is represented by rE, = tanh (rH ), where tanh is the hyperbolic function; tanh x ¼ . It is worth
cosh x
mentioning that most of the Euclidean spirals are mathematically represented in the polar form r = f (θ). Hyperbolic spiral
can be represented in the form rH = f (θ); then, it can be formulated as rH = tanh (f (θ)). Archimedean spiral is considered
as the most widely used model of Euclidean spirals which represented by r = aθ.59

3.3.2 | Principle of IMFO

To improve MFO, it is proposed that moths move according to Arithmetic spiral as follows:
IMFO moves in hyperbolic spiral in transcendental plane curve. The transcendental functions sine and cosine were
tabulated from physical measurements. The most familiar transcendental functions are the logarithm, the exponential
(with any nontrivial base), the trigonometric, the hyperbolic functions, and the inverses of all of these functions. To
explain the idea of the hyperbolic spiral, it is worth mentioning that the concept of Archimedean spiral starts in the
origin and makes a curve with three rounds in the shape of spiral as shown in Figure 4.
The hyperbolic spiral is the reverse of Archimedean spiral, where the origin is the end point in hyperbolic spiral
(r = a/θ starting from infinity). Simultaneously, the origin is the starting point in Archimedean spiral as shown in
Figures 5 and 6.

FIGURE 4 Basic Archimedean spiral

FIGURE 5 The reciprocal spiral r = a/θ

10 of 28 TAHER ET AL.

FIGURE 6 The hyperbolic version of the reciprocal spiral

Spiral branches should be separated by equidistance between each other. In other words, the distances of intersec-
tion points along a line through the origin are the same. More generally, a power spiral is governed by the following
equation:
1
r ¼ aθn : (28)

Assuming a ˃ 0, according to varying the values of n, different spirals can be obtained as shown in Figure 7, where θ
is the polar angle, and n is a constant that assigns the extent of spiral wrapping.
Archimedean spiral is the special case at a = 1 which can be formulated as follows:

r ¼ aθ: (29)

To formulate the second proposed spiral path in the original MFO, the Archimedes'spiral function is substituted in
Equation 25 to obtain the new Equation 30 as follows:

S ðMi; FjÞ ¼ Di:t :cosð2πt Þ þ Fj: (30)

The solution process of OPF using IMFO algorithm is shown in Figure 8.

4 | S I M U L A TI O N A N D RE S U L T S

IEEE 30‐bus, IEEE 57‐bus, and IEEE 118‐bus test systems are used to validate the proposed IMFO to solve OPF prob-
lems. Fifteen case studies are implemented as outlined in Table 1. The maximum number of iterations is set to 500. The
population size is selected as 50 for IEEE 30‐bus and 57‐bus systems and 100 for IEEE 118‐bus system, respectively. The
obtained results by IMFO are compared with those obtained by MFO,37 GA,20 PSO,38 and TLBO.40,41 The software

FIGURE 7 Archimedean spiral for t = 0 to 6π

TAHER ET AL. 11 of 28

FIGURE 8 IMFO algorithm flowchart

program is written in the commercial MATLAB 7.7 (R2009b) using selected programs from MATPOWER on a PC with
an Intel Core 7 CPU 2.4 GHz 6.0 GB RAM. The penalty factors KG,KQ, KV, and KS indicate the violation of inequality
constraints. The values of penalty factors in this paper equal to zero for all cases, where there is no violation for all
inequality constraints of dependent variables which consist of real power generation output, load bus voltage magni-
tude, reactive power generation output, and line loading.

4.1 | IEEE 30‐bus system

IEEE 30‐bus system is considered the first test system used to check the performance of IMFO. The system base
apparent power is 100 MVA, the total system demand active power is 283.4 MW, and the total system demand reac-
tive power is 126.2 MVAR. System data including bus data, line data, and generator data are given in Alsac and
Stott.8 The number of search agents is set to 50. The control variable limits for the various case studies are given in
Table 1.60 The characteristics of 30‐bus system are indicated in Appendix A.1. The cost and emission coefficients are
given in Appendix A.2 and Appendix A.3, respectively. The results obtained by IMFO, MFO, GA, PSO, and TLBO
are achieved under the same equality and inequality constraints. All obtained results are tabulated in Appendix A.4,
Table 2, and Table 3.
12 of 28 TAHER ET AL.

TABLE 1 Different cases presented in this paper

Case Name Objective Function Test System

Case 1 Quadratic fuel cost IEEE 30

Case 2 Piecewise quadratic fuel cost
Case 3 Quadratic fuel cost with valve‐point loading
Case 4 Emission cost
Case 5 Active power loss
Case 6 Quadratic fuel cost considering active power loss
Case 7 Quadratic fuel cost considering voltage profile
Case 8 Quadratic fuel cost considering voltage stability
Case 9 Quadratic fuel cost considering emission
Case 10 Quadratic fuel cost considering power loss, voltage
profile, and voltage stability
Case 11 Quadratic fuel cost IEEE 57
Case 12 Quadratic fuel cost considering voltage profile
Case 13 Quadratic fuel cost considering voltage stability
Case 14 Quadratic fuel cost considering emission
Case 15 Quadratic fuel cost IEEE 118

TABLE 2 Comparison of OPF solution obtained by IMFO, MFO, GA, PSO, and TLBO for cases (1‐5) of IEEE30‐bus test system

Case No Objective Functions IMFO MFO GA PSO TLBO

Case 1 Fuel cost ($/h) 800.3848 800.6206 800.4346 800.4075 800.4104

Emission (ton/h) 0.3658 0.3663 0.3683 0.3660 0.3663
PLoss (MW) 8.9990 9.0790 9.037 9.006 9.008
QLoss (MW) −10.638 −8.693 −11.221 −11.924 −10.326
VD (p.u.) 0.9035 0.6343 0.9155 0.9129 0.9068
Lmax 0.1274 0.1310 0.1275 0.1276 0.1269
Case 2 Fuel cost ($/h) 645.8958 646.0591 646.7964 647.0333 646.4812
Emission (ton/h) 0.2836 0.2833 0.2834 0.2829 0.2835
PLoss (MW) 6.6100 6.625 6.8325 6.819 6.7539
QLoss (MW) −17.828 −18.485 −18.625 −20.901 −18.333
VD (p.u.) 0.9115 0.6497 0.9152 0.8218 0.9028
Lmax 0.1274 0.1324 0.1270 0.1290 0.1282
Case 3 Fuel cost ($/h) 832.1023 832.3871 832.4400 832.1095 832.0944
Emission (ton/h) 0.4382 0.4385 0.4369 0.4373 0.4378
PLoss (MW) 10.712 10.800 10.6479 10.652 10.673
QLoss (MW) −4.717 −2.923 −6.131 −6.535 −6.948
VD (p.u.) 0.8318 0.5417 0.7282 0.8627 0.8637
Lmax 0.1281 0.1320 0.1289 0.1290 0.1280
Case 4 Fuel cost ($/h) 944.2572 944.3942 944.4295 944.4481 944.3278
Emission (ton/h) 0.20480 0.20483 0.20483 0.20483 0.20482
PLoss (MW) 3.2601 3.254 3.266 3.263 3.221
QLoss (MW) −30.983 −29.161 −28.683 −30.898 −30.000
VD (p.u.) 0.8881 0.9011 0.8551 0.8815 0.8988
Lmax 0.1276 0.1287 0.1280 0.1271 0.1271
Case 5 Fuel cost ($/h) 967.5900 967.6938 967.6486 967.6066 967.5841
Emission (ton/h) 0.2073 0.2073 0.2073 0.2073 0.2073
PLoss (MW) 3.0905 3.139 3.118 3.103 3.088
QLoss (MW) −32.582 −33.041 −31.490 −32.117 −30.316
VD (p.u.) 0.9192 0.9080 0.8325 0.9202 0.9029
Lmax 0.1274 0.1269 0.1293 0.1277 0.1274
TAHER ET AL. 13 of 28

TABLE 3 Comparison of OPF solution obtained by IMFO, MFO, GA, PSO, and TLBO for cases(6‐10) of IEEE 30‐bus test system.

Case No Objective Functions IMFO MFO GA PSO TLBO

Case 6 Fuel cost ($/h) 858.7455 858.9658 859.0075 858.8508 859.0236

Emission (ton/h) 0.2291 0.2289 0.2288 0.2290 0.2289
PLoss (MW) 4.582 4.556 4.5501 4.537 4.531
QLoss (MW) −25.144 −26.216 −27.497 −28.226 −27.161
VD (p.u.) 0.9559 0.9549 0.9339 0.9320 0.9317
Lmax 0.1274 0.1258 0.1272 0.1272 0.1271
Case 7 Fuel cost ($/h) 803.5715 803.5173 803.2347 803.4736 803.5675
Emission (ton/h) 0.3636 0.3672 0.3644 0.3639 0.3636
PLoss (MW) 9.814 9.870 9.747 9.821 9.813
QLoss (MW) −2.302 −2.730 −3.465 −2.899 −3.110
VD (p.u.) 0.0954 0.1007 0.1018 0.0978 0.0939
Lmax 0.1366 0.1369 0.1374 0.1371 0.1369
Case 8 Fuel cost ($/h) 800.4762 800.9415 800.4385 800.5815 800.4738
Emission (ton/h) 0.3662 0.3700 0.3662 0.3662 0.3658
PLoss (MW) 9.016 9.222 9.013 9.056 9.012
QLoss (MW) −12.952 −4.768 −10.543 −11.098 −13.075
VD (p.u.) 0.9086 0.7789 0.9272 0.8792 0.9432
Lmax 0.1255 0.1266 0.1254 0.1280 0.1247
Case 9 Fuel cost ($/h) 836.4809 836.5734 836.5425 836.4357 836.4141
Emission (ton/h) 0.2423 0.2422 0.2422 0.2423 0.2423
PLoss (MW) 5.374 5.400 5.396 5.365 5.363
QLoss (MW) −23.011 −20.117 −25.181 −25.595 −25.067
VD (p.u.) 0.9206 0.9375 0.8697 0.9296 0.9314
Lmax 0.1280 0.1279 0.1273 0.1268 0.12741
Case 10 Fuel cost ($/h) 829.8682 830.3227 830.5804 830.1222 830.1058
Emission (ton/h) 0.2535 0.2530 0.2524 0.2530 0.2529
PLoss (MW) 5.6756 5.638 5.577 5.591 5.590
QLoss (MW) −22.122 −18.487 −24.116 −22.830 −22.238
VD (p.u.) 0.2778 0.3142 0.3084 0.2972 0.2988
Lmax 0.1358 0.1357 0.1358 0.1355 0.1361

Case 1: OPF for fuel cost minimization

This is the basic case of this paper. In this case, the objective function is minimization of total fuel cost for the system
generated power. The objective quadratic function is formulated as follows:

   
NG $
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ Ci þ Penalty (31)
h

where, ai, bi, and ci are the cost coefficients of i‐th generator. The obtained control variables by IMFO are tabulated in
column 4 in Appendix A.4. The obtained fuel cost by IMFO is 800.3848 $/h, which is the lowest value compared with
those obtained by the other algorithms as given in Table 2. Whereas the fuel cost for MFO, GA, PSO, and TLBO are
800.6206, 800.4346, 800.4075, and 800.4104 $/h, respectively. The convergence characteristics of IMFO, traditional
MFO, and the other reported optimization techniques are shown in Figure 9.

Case 2: OPF for piecewise quadratic fuel cost minimization

In the normal operating conditions of thermal power plants, different types of fuel sources such as natural gas and
coal are consumed according to the operating state and the required power level to the reduce operating cost. Piece
wise quadratic equation is used instead of normal quadratic equation. Generators 1 and 2 considered as a set of
constraints as follows:
14 of 28 TAHER ET AL.

FIGURE 9 Convergence characteristics of different optimization techniques (case 1)

8
>
> ai1 P2Gi þ bi1 PGi þ Ci1 Pmin max
Gi < P Gi < PGi
>
>
>
> ai2 P2Gi þ bi2 PGi þ Ci2 Pmin max
>
> Gi1 < P Gi < PGi2
>
>
>
> :
>
>
<
F mf cost ¼ : (32)
>
>
>
> :
>
>
>
>
>
> :
>
>
>
>
:
ai2 P2Gi þ bi2 PGi þ C iK Pmin max
GiK−1 < P Gi < PGi

where, ai1, ai2, and aik are the cost coefficient of the i‐ th generator and the k‐th fuel source. The cost coefficients can be
found from APPENDIX A.2.
Fuel equation cost of the other units is still unchanged as the objective function. The piece wise quadratic equation
can be formulated as follows61:
   
2 NG $
f ¼ ∑i¼1 aik P2Gi þ bik PGi þ Cik þ ∑i¼3 ai P2Gi þ bi PGi þ Ci Þ þ Penalty : (33)
h

The obtained control variables by IMFO are tabulated in column 5 in Appendix A.4. The obtained fuel cost by IMFO
is 645.8958 $/h which is the minimum value compared with the other algorithms as given in Table 2. Whereas the fuel
cost for MFO, GA, PSO, and TLBO are 646.0591, 646.7964, 647.0333, and 646.4812 $/h, respectively. The convergence
characteristics of IMFO, traditional MFO, and the other reported optimization techniques prove the prevalence of IMFO
to attain the OPF solution with fast convergence as shown in Figure 10.

Case 3: OPF for minimization of quadratic fuel cost with valve‐point loadings

The fuel cost of generators 1 and 2 is considered separately in this case by the same procedure of Case 2, but the
valve point loading is considered instead of considering piecewise quadratic functions as a result of using multi‐fuel
option. Steam turbine generators with multi valve are subjected to vary fuel cost due to ripple like effect of the generator
heat rate curve.62 The mathematical expression of cost function with valve point loading can be expressed by adding an
absolute value of rectifying sinusoidal term to the basic quadratic cost function.34 The fuel cost function of this case can
be formulated as follows63:

FIGURE 10 Convergence characteristics of different optimization techniques (case 2)

TAHER ET AL. 15 of 28

    
NG
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ C i þjdi sin ei Pmin
Gi − P Gi j þ Penalty (34)

where, di and ei are the cost coefficients of valve‐point loading effect.

The obtained control variables are tabulated in column 6 in Appendix A.4. It is noted that, the fuel cost is increased
from 800.3848 $/h for the basic fuel cost of Case 1 to 832.1023 $/h due to valve‐point loading effect. In this case, TLBO
gives the best solution of 832.0944 $/h, where, it is slightly lower than the solution obtained by IMFO. Whereas the
fuel cost obtained by MFO, GA, and PSO are 832.3871, 832.4400, and 832.1095 $/h, respectively, as indicated in
Table 2. The convergence characteristics of IMFO, traditional MFO, and the other reported techniques are indicated in
Figure 11.

Case 4: OPF for emission cost minimization

The aim of this case is to reduce the environment pollution due to released emission gases from thermal power
plants. The main gases to be reduced are carbon dioxide (CO2), NOx, and sulfur oxide (SOX). These gases are a family
of chemical compounds released from fossil fuel combustion. The objective function for emission gases can be formu-
lated as follows64:

Emission ¼ ∑i¼1 γi P2Gi þ βi PGi þ αi þ ζi eðλi PGi Þ þ Penalty ðton=hÞ

NG
(35)

where, γi, βi, αi, ζi, and λi are the emission coefficient of the i‐ th generator. Values of emission coefficients can be found
in Appendix A.2. The obtained value of emission gases by IMFO is 0.20480 ton/h, which is the minimum value com-
pared with these obtained by the other algorithms. The obtained control variables for this case are tabulated in column
7 in Appendix A.4. The results obtained by IMFO, MFO, GA, PSO, and TLBO algorithms for this case are indicated in
Table 2. The emission gases obtained by MFO, GA, and PSO have the same value of 0.20483 ton/h, while the emission
gases obtained by TLBO are 0.20482 ton/h. The convergence characteristics of IMFO, traditional MFO, and the other
reported optimization techniques are shown in Figure 12.

Case 5: OPF for power loss minimization

FIGURE 11 Convergence characteristics of different optimization techniques (case 3)

FIGURE 12 Convergence characteristics of different optimization techniques (case 4)

16 of 28 TAHER ET AL.

The purpose of this case is to reduce the active power loss from the power source to the customer through the trans-
mission lines. The objective function represents minimization of active power loss through transmission lines, which
can be formulated as follows:

nl nl   
f ¼ ∑i¼1 ∑j¼1 Gij V 2i þ V 2i − 2Vi Vi cos δi − δj þ Penalty ðMW Þ: (36)

The obtained control variables for this case are tabulated in column 8 in Appendix A.4. The results obtained by
IMFO, MFO, GA, PSO, and TLBO algorithms for this case are indicated in Table 2. In this case, TLBO gives
3.088 MW, where it is slightly lower than the solution obtained by IMFO which is 3.0905 MW, whereas the power loss
obtained by MFO, GA, and PSO are 3.139, 3.118, 3.103, and 3.088 MW, respectively. The convergence characteristics of
IMFO, traditional MFO, and the other techniques are indicated in Figure 13.

Case 6: OPF for minimization of fuel cost and active power losses

In this case, minimization of both of basic quadratic fuel cost and the active power losses are considered as a multi‐
objective function. The required objective function can be formulated as follows:
  nl nl   
NG
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ Ci Þ þ λP ∑ ∑ Gij V 2i þ V 2i − 2V i V i cos δi − δj þ Penalty
i¼1 j¼1

(37)

where, λP is a weighting factor. This weighting factor assign the importance of power losses compared with fuel cost.
The weighting factor λP value is chosen as 40. The obtained control variables for this case are tabulated in column 9
in Appendix A.4. The obtained OPF solutions by IMFO, MFO, GA, PSO, and TLBO algorithms are indicated in
Table 3. The obtained fuel cost by IMFO is 858.7455$/h which is the minimum value compared with the other algo-
rithms, whereas the fuel cost by MFO, GA, PSO, and TLBO are 858.9658, 859.0075, 858.8508, and 859.0236 $/h, respec-
tively. The power loss obtained by IMFO is slightly higher than the solutions obtained by the other reported techniques,
where the power loss obtained by IMFO, GA, PSO and TLBO are 4.582, 4.556, 4.5501, 4.537, and 4.531 MW, respectively.

Case 7: OPF for minimization of fuel cost and voltage deviation

In this case, minimization both of basic quadratic fuel cost and VD is considered as a multi‐objective function, to
improve the obtained poor voltage profile in Case 1, where fuel cost was optimized as objective function.65 VD minimi-
zation can be achieved by reducing the load bus voltage deviation close to 1 p.u. The required objective function for
minimizing both of fuel cost and VD can be formulated as follows66:
 
NG NG
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ Ci þ λ VD ∑i¼1 ∣V Li − 1∣ þ Penalty (38)

where, λVD is a weighting factor. This weighting factor assigns the importance of VD regarding to fuel cost. The
weighting factor λVD value is chosen as 100. The obtained control variables for this case are tabulated in column 10
in Appendix A.4. The obtained OPF solutions by IMFO, MFO, GA, PSO, and TLBO algorithms are given in Table 3.

FIGURE 13 Convergence characteristics of different optimization techniques (case 5)

TAHER ET AL. 17 of 28

Comparison of VD values reveals that it is optimized compared with Case 1; it is reduced from 0.9035 to 0.0954 p.u.
The improvement is shown in Figure 14, where the voltage profile of the PQ busses is greatly enhanced via the evolu-
tion of objective function through iterations.

Case 8: OPF for minimization of fuel cost with voltage stability enhancement

In this case, OPF is obtained by considering minimization of basic quadratic fuel cost and enhancement of voltage
stability. The voltage stability is to keep the voltage at all the busses at the permissible operating limits. Voltage insta-
bility may lead to voltage collapse which cause undesirable low voltage profile. Voltage stability indicator is known as L‐
index which is very important objective function even at the planning or the operation stage of power systems. The
function of L‐index depends upon the extent of voltage collapse at j‐th bus. L‐index can be represented as follows67:

NG Vi
Lj ¼ ∣1 − ∑i¼1 F Ji ∣∀ j ¼ 1; 2; …; NL (39)
VJ

Fji ¼ − ½Y 1 −1 ½Y 2  (40)

where,
Y1 and Y2 represent the system YBUS sub‐matrices.
The objective function represents L‐index of the whole system can be represented as follows:

L ¼ maxðLJ Þ∀ j ¼ 1; 2; …; NL: (41)

The purpose of this case is to achieve minimization of basic quadratic fuel cost and to avoid voltage collapse. The
objective function of this case can be formulated as follows:
 
NG
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ C i þ λ Lmax maxðLJ Þ þ Penalty (42)

where,
λL is a weighting factor. This weighting factor assigns the importance of voltage stability compared with fuel cost.
The weighting factor λL value is chosen to be 100. The obtained control variables for this case are tabulated in column
11 in Appendix A.4. The obtained OPF solutions by IMFO are compared with those obtained by MFO, GA, PSO, and
TLBO algorithms as given in Table 3. The fuel cost by IMFO, MFO, GA, PSO, and TLBO are 800.4762, 800.9415,
800.4385, 800.5815, and 800.4738 $/h, respectively. The voltage stability obtained by IMFO, GA, PSO, and TLBO are
0.1255, 0.1266, 0.1254, 0.1280, and 0.1247 p.u., respectively.

Case 9: OPF for minimization of fuel cost with emission

In this case, OPF is obtained by considering minimization of basic quadratic fuel cost and emission. The purpose of
this case is to reduce basic quadratic fuel cost in the same time with minimizing the released emission gases as
explained in case 4 by reducing the pollutants to the environment mainly from NOx, and SOX. The objective function
of this case can be formulated as follows:

FIGURE 14 Voltage profile comparison for IEEE 30‐bus test system

18 of 28 TAHER ET AL.

 
NG
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ C i þ λ i ðEmissionÞ þ Penalty (43)

where,
λi is a weighting factor; this weighting factor assigns the importance of emission gases compared with fuel cost. The
weighting factor λi value is chosen to be 1000. The obtained control variables for this case are tabulated in column 12 in
Appendix A.4. The obtained OPF solutions by IMFO are compared with those obtained by MFO, GA, PSO, and TLBO
algorithms and indicated in Table 3. The fuel cost by IMFO, MFO, GA, PSO, and TLBO are 836.4809, 836.5734,
836.5425, 836.4357, and 836.4141 $/h, respectively. The emission gases obtained by IMFO, GA, PSO, and TLBO are
0.2423, 0.2422, 0.2422, 0.2423, and 0.2423 p.u., respectively.

Case 10: OPF for minimization of cost, emission, voltage deviation, and losses

In this case, OPF solutions are obtained by considering minimization of four single objective functions which are
basic quadratic fuel, emission, active power losses, and voltage deviation in the same time. The multi‐objective function
for this case can be formulated as follows:
     
f ¼ ∑i¼1 ai P2Gi þ bi PGi þ Ci þ λE ∑i¼1 γi P2Gi þ βi PGi þ αi þ ζi eðλi PGi Þ þ λ VD ∑i¼1 jV Li − 1j
NG NG NG

nl nl    (44)
þ λ P λ P ∑ ∑ Gij V 2i þ V 2i − 2Vi Vi cos δi − δj þ Penalty
i¼1 j¼1

where;
λE, λVD, and λP are weighting factors chosen for balance between the single objectives as 19, 21, and 22, respectively.
The obtained control variables for this case are tabulated in column 13 in Appendix A.4. The obtained OPF solutions by
IMFO along with those obtained by MFO, GA, PSO, and TLBO algorithms are indicated in Table 3.

4.2 | IEEE 57‐bus system

Improved moth flame optimization algorithm is applied upon the IEEE 57‐bus system which is a comparatively large‐
scale system to clarify its leverage. The system base apparent power is 100 MVA, the total system active power is
1250.8 MW, and the total system reactive power is 336.4 MVAR. System data including bus data, line data, and gener-
ator data are given in Zimmerman and Murillo‐Sánchez.68 The number of population size is set to 50. The maximum
number of iterations is set to 500. The characteristics of IEEE 57 bus system are indicated in Appendix A.5. The cost
and emission coefficients are given in Appendix A.6. The optimal control parameters limits and solutions for the chosen
four case studies are tabulated in Table 4.
The obtained results by IMFO are compared with solutions by other algorithms; MFO, GA, PSO, and TLBO algo-
rithms as indicated in Table 5.

Case 11: Minimization of basic fuel cost

This case aims to minimize the basic quadratic fuel cost same as case 1 using Equation 31. The optimal
control variable settings obtained by IMFO are tabulated in Table 6. The obtained optimal fuel cost by IMFO is
41 667.1497 $/h which is the best solution compared with the adduced algorithms in the literature review as
indicated in Table 6.

Case 12: Minimization of fuel cost and voltage deviation

This case aims to minimize both of the basic quadratic fuel cost and VD same as Case 7 using multi‐objective func-
tion stated in Equation 38; λVD is selected as 100. The obtained results given in Table 5 indicate that the voltage devi-
ation is reduced from 1.3944 p.u. in Case 11 to 0.71824 p.u. in Case 12. However, the fuel cost is increased from
41 667.1497 $/h in Case 11 to 41 692.7178 $/h in Case 12. The impact of IMFO to improve the voltage profile when
it is considered in the evolution of objective function is shown in Figure 15.
TAHER ET AL. 19 of 28

TABLE 4 OPF solution values obtained by IMFO for IEEE 57‐bus test system

Variable Case 11 Case 12 Case 13 Case 14

PG1 143.601 143.452 143.025 142.771

PG2 86.593 88.626 86.136 99.996
PG3 45.508 45.124 45.310 48.001
PG6 74.668 75.915 74.164 99.999
PG8 461.009 463.451 462.607 422.345
PG9 92.001 87.310 92.360 98.065
PG12 362.260 362.460 362.247 353.980
VG1 1.0513 1.01920 1.034 1.0496
VG2 1.0524 1.0216 1.037 1.0496
VG3 1.0396 1.0163 1.029 1.0486
VG6 1.037 1.0297 1.042 1.0616
VG8 1.0548 1.0495 1.049 1.0713
VG9 1.0375 1.0275 1.028 1.0473
VG12 1.0314 1.0106 1.019 1.042
T19 (4_18) 1.0104 0.9289 0.954 1.0997
T20 (4_18) 0.97867 1.0202 1.0325 0.9358
T31 (21_20) 1.034 0.9989 1.0096 1.0549
T35 (24_25) 1.0885 0.9777 0.9358 1.0019
T36 (24_25) 0.9000 0.9996 1.0999 1.1
T37 (24_26) 1.0175 1.0109 1.0167 1.0288
T41 (7_29) 0.97644 0.9985 0.9802 0.99667
T46 (34_32) 0.95816 0.9405 0.9599 0.96543
T54 (11_41) 0.9000 0.9000 0.9097 0.99221
T58 (15_45) 0.96229 0.9589 0.9597 0.97282
T59 (14_46) 0.95526 0.9595 0.9569 0.95833
T65 (10_51) 0.96279 0.9759 0.9555 0.97368
T66 (13_49) 0.92438 0.9274 0.9185 0.93235
T71 (11_43) 0.9615 0.9659 0.9467 0.95766
T73 (40_56) 0.98557 1.0263 0.9935 1.0236
T76 (39_57) 0.95505 0.9000 0.9687 0.99983
T80 (9_55) 0.98059 1.0113 0.9769 0.99772
QC18 9.4129 7.9773 20.0000 0.054559
QC25 12.971 16.1944 16.9581 20.000
QC53 13.409 16.1125 13.6869 13 468
Fuel cost ($/h) 41 667.1497 41 692.7178 41 673.6204 41 723.6915
VD 1.3944 0.71824 1.17625 1.5788
Lmax 0.2403 0.24259 0.23525 0.2283
Ploss (MW) 15.130 15.7552 15.299 14.355
Qloss (MVar) −54.213 −46.605 −50.806 −58.547
Emission (ton/h) 1.3592 1.3681 1.36413 1.2203
20 of 28 TAHER ET AL.

TABLE 5 Comparison of OPF solution obtained by IMFO, MFO, GA, PSO, and TLBO for IEEE57‐bus test system

Case No Objective Functions IMFO MFO GA PSO TLBO

Case 11 Fuel cost ($/h) 41 667.1497 41 679.3749 41 676.4786 41 673.6339 41 670.7809

Emission (ton/h) 1.3592 1.3703 1.3415 1.3617 1.3592
PLoss (MW) 15.1297 15.2996 15.0115 15.2806 15.2707
QLoss (MW) −54.213 −53.856 −53.680 −53.014 −51.990
VD (p.u.) 1.3944 1.3559 1.1513 1.2895 1.4081
L‐index 0.2403 0.2491 0.2380 0.2546 0.2379
Case 12 Fuel cost ($/h) 41 692.7178 41 719.8471 41 700.4162 41 684.4009 41 694.7778
Emission (ton/h) 1.3681 1.3515 1.3717 1.3623 1.3594
PLoss (MW) 15.7552 16.327 15.453 15.5632 15.7426
QLoss (MW) −46.605 −40.187 −51.456 −48.967 −45.794
VD (p.u.) 0.71824 0.75514 0.80517 0.76240 0.7120
L‐index 0.24259 0.2396 0.24144 0.24065 0.2409
Case 13 Fuel cost ($/h) 41 673.6204 41 688.6522 41 670.0872 41 670.1755 41 685.353
Emission (ton/h) 1.36413 1.3360 1.3423 1.3614 1.3537
PLoss (MW) 15.299 15.475 15.165 15.205 15.487
QLoss (MW) −50.806 −50.854 −53.074 −53.238 −49.752
VD (p.u.) 1.17625 1.3988 1.2473 1.4323 1.3453
L‐index 0.23525 0.2395 0.2413 0.2420 0.24787
Case 14 Fuel cost ($/h) 41 723.6915 41 751.6148 41 732.5892 41 727.8988 41 724.8730
Emission (ton/h) 1.2203 1.2198 1.2194 1.2188 1.2224
PLoss (MW) 14.355 15.0139 14.348 14.651 14.425
QLoss (MW) −58.547 −51.992 −58.917 −56.599 −58.928
VD (p.u.) 1.5788 1.8155 1.3381 1.3474 1.5022
Lmax 0.22833 0.22489 0.2452 0.2436 0.23288

TABLE 6 Comparison between IMFO and some other algorithms for case 11

Algorithm Fuel cost ($/h)

IMFO 41 667.1497
42
LTLBO 41 679.5451
34
MO‐DEA 41 683.0000
30
GSA 41 695.8717
39
ABC 41 693.9589

FIGURE 15 Voltage profile comparison for IEEE 57‐bus test system

Case 13: Minimization of fuel cost with voltage stability enhancement

This case aims to minimize both of the basic quadratic fuel cost and the voltage stability simultaneously same as
Case 8 using multi‐objective function stated in Equation 38, λL is selected as 100. The obtained results given in
Table 5, where L‐index is reduced from 0.2403 p.u. in Case 11 to 0.23525 p.u. in Case 13. However, the fuel cost is
increased from 41667.1497 $/h in Case 11 to 41 673.6204 $/h in Case 13.
TAHER ET AL. 21 of 28

Case 14: Minimization of fuel cost with emission

The aim of this case is to reduce both of basic quadratic fuel cost and emission gases same as Case 9. Equation 42 is
used, λE is selected as 1000. The value of emission is reduced from 1.3592 ton/h. in Case 11 to 1.2203 ton/h in Case 14.
However, the fuel cost is increased from 41 667.1497 $/h in Case 11 to 41 723.6915 $/h in Case 14.

4.3 | IEEE 118‐bus test system

The considered large‐scale system in this paper is IEEE 118‐bus test system to check the IMFO algorithm validation and
performance at large system sizes. The system base apparent power is 100 MVA, the total system demand active power
is 4242 MW, and the total system demand reactive power is 1439 MVAR. System data including bus data, line data, and

TABLE 7 OPF solution values obtained by IMFO for IEEE118‐bus test system

Variables Value Variables Value Variables Value Variables Value Variables Value

PG1 68.781 PG62 0.008 PG116 0.273 VG61 1.0155 VG112 0.9688
PG4 35.368 PG65 362.857 VG1 0.9611 VG62 1.0162 VG113 0.9730
PG6 0.0 PG66 365.456 VG4 1.0222 VG65 0.9930 VG116 0.9503
PG8 29.323 PG70 3.807 VG6 0.9796 VG66 1.0605 T8 0.9741
PG10 185.000 PG72 26.505 VG8 0.9886 VG69 1.0391 T32 0.9787
PG12 82.210 PG73 30.519 VG10 0.9855 VG70 0.9924 T36 1.0130
PG15 0.002 PG74 0.005 VG12 0.9706 VG72 1.0350 T51 0.9420
PG18 13.460 PG76 0.181 VG15 0.9532 VG73 0.9500 T93 0.9805
PG19 99.995 PG77 51.552 VG18 0.9500 VG74 0.9688 T95 0.9464
PG24 26.700 PG80 458.925 VG19 0.9561 VG76 0.9546 T102 0.9531
PG25 216.479 PG85 0.000 VG24 1.0738 VG77 0.9955 T107 0.9400
PG26 99.994 PG87 3.981 VG25 1.0623 VG80 1.0130 T127 0.9984
PG27 52.467 PG89 468.224 VG26 1.0483 VG85 0.9882 Qc5 10.000
PG31 9.148 PG90 0.801 VG27 1.0059 VG87 0.9954 Qc34 8.59
PG32 21.522 PG91 0.631 VG31 0.9776 VG89 1.0117 Qc37 3.89
PG34 14.475 PG92 0.768 VG32 0.9977 VG90 0.9813 Qc44 9.20
PG36 43.080 PG99 1.780 VG34 0.9793 VG91 0.9500 Qc45 9.78
PG40 72.481 PG100 229.190 VG36 0.9750 VG92 0.9862 Qc46 6.33
PG42 38.895 PG103 37.884 VG40 0.9680 VG99 0.9500 Qc48 2.89
PG46 11.715 PG104 0.000 VG42 1.0090 VG100 0.9865 Qc74 4.14
PG49 148.000 PG105 14.501 VG46 1.0222 VG103 0.9679 Qc79 3.81
PG54 49.437 PG107 50.901 VG49 1.0331 VG104 0.9535 Qc82 10.000
PG55 39.665 PG110 0.357 VG54 0.9687 VG105 0.9500 Qc83 9.93
PG56 49.077 PG111 31.577 VG55 0.9621 VG107 0.9500 Qc105 9.29
PG59 159.232 PG112 26.804 VG56 0.9683 VG110 0.9636 Qc107 8.61
PG61 99.996 PG113 36.472 VG59 1.0024 VG111 0.9744 Qc110 10.000
Fuel cost ($/h) 131.8200
Loss (MW) 77.6522
QLoss (MW) ‐910.020
VD (p.u.) 1.5944
Lmax 0.0807
PG69 (MW) 407.192
22 of 28 TAHER ET AL.

TABLE 8 Comparison between IMFO and some other reported algorithms for case 15

Algorithm Fuel cost ($/h)

IMFO 131.8200
43
FHSA 132 138.30
HSA43 132 319.60
31a
ICBO 135 121.5704
39
ABC 135 304.3584
25
BBO 135 263.7289
a
Limit of shunt VAR compensators is 5 MVAR.

FIGURE 16 Voltage profile comparison

for multi‐objective function cases of IEEE
30‐bus test system

generator data are given in Zimmerman and Murillo‐Sánchez.68 The characteristics of IEEE 118‐bus system are
indicated in Appendix A.7.

Case 15: Minimization of fuel cost

In this case, the basic quadratic fuel cost is considered to be minimized same as cases 1 and 11 for IEEE 30‐bus and
IEEE57‐bus test systems. The objective function of this case is given in Equation 31. The OPF solutions of control
variables are given in Table 7. From this table, the value of minimum fuel cost is 131.8200 $/h, which is the best solution
compared with the adduced algorithms in the literature review as indicated in Table 8.

4.4 | Checking the feasibility

Feasibility of the papers focusing on the optimization topic is confirmed when the optimal values of the control
variables and state variables are restricted with both of equality and inequality constraints. In this paper, the
precision of the obtained OPF solutions by IMFO is checked, where all values of the required control variables have
been achieved without violation of the constraints. The validity of obtained state variables is checked by penalty
functions as given in (15: 19). Obvious example of the feasibility of IMFO is the voltage profile comparison for
the multi‐objective cases of IEEE30‐bus system. The improvement is shown in Figure 16, where the voltage profile
of the PQ busses is close to unity when the voltage deviation is considered in the evolution of multi‐objective
functions in cases 7 and 10.

5 | CONCLUSIONS

In this paper, an IMFO algorithm has been proposed in order to solve the OPF problems. The IMFO algorithm based on
the concept of MFO with some modifications related to the path of moths in new spirals around the flame. Various
objective functions have been tested on three standard IEEE test systems. The cost functions have been restricted with
both of equality and inequality constraints to show the effectiveness and accuracy of OPF solutions of the developed
TAHER ET AL. 23 of 28

algorithm. IMFO has been implemented to solve single and multi‐objective functions in case of small, medium, and
large‐scale systems. The results reveal that IMFO algorithm is capable of finding precise and better OPF solutions
compared with the other well‐known techniques: traditional MFO, GA, PSO, and TLBO. A comparison among the
convergence characteristics of IMFO technique and the other techniques proves the prevalence of IMFO to attain the
OPF solution with fast convergence. In the future research, the IMFO technique can be applied to solve other optimi-
zation problems in power systems such as optimal allocation of shunt capacitive banks, FACTS devices, and distributed
generation to achieve single and multi‐objective functions.

ORCID
Francisco Jurado http://orcid.org/0000-0001-8122-7415

R EF E RE N C E S
1. Mukherjee A, Mukherjee V. Solution of optimal power flow using chaotic krill herd algorithm. Chaos, Solitons Fractals. 2015;78:10‐21.
2. Hazra J, Sinha A. A multi‐objective optimal power flow using particle swarm optimization. Int Trans Electr Energy Syst.
2011;21:1028‐1045.
3. Ghasemi M, Ghavidel S, Ghanbarian MM, Gharibzadeh M, Vahed AA. Multi‐objective optimal power flow considering the cost, emission,
voltage deviation and power losses using multi‐objective modified imperialist competitive algorithm. Energy. 2014;78:276‐289.
4. Lee K, Park Y, Ortiz J. A united approach to optimal real and reactive power dispatch. IEEE Trans Power Syst. 1985;PAS‐104(5):1147‐1153.
5. Attia A‐F, El Sehiemy RA, Hasanien HM. Optimal power flow solution in power systems using a novel sine‐cosine algorithm. Int J Electr
Power Energy Syst. 2018;99:331‐343.
6. Attia A‐F, Al‐Turki YA, Abusorrah AM. Optimal power flow using adapted genetic algorithm with adjusting population size. Electr Power
Compon Syst. 2012;40(11):1285‐1299.
7. Vaisakh K, Srinivas L. Genetic evolving ant direction HDE for OPF with non‐smooth cost functions and statistical analysis. Expert Syst
Appl. 2011;38(3):2046‐2062.
8. Alsac O, Stott B. Optimal load flow with steady‐state security. IEEE Trans Power Syst. 1974;PAS‐93(3):745‐751.
9. Shoults RR, Sun D. Optimal power flow based upon PQ decomposition. IEEE Trans Power Syst. 1982;PAS‐101(2):397‐405.
10. Bottero M, Galiana F, Fahmideh‐Vojdani A. Economic dispatch using the reduced hessian. IEEE Trans Power Syst. 1982;
PAS‐101(10):3679‐3688.
11. Momoh JA. A generalized quadratic‐based model for optimal power flow, in Systems, Man and Cybernetics, 1989. Conference Proceedings.,
IEEE International Conference on, 1989: 261‐271.
12. Momoh JA, Adapa R, El‐Hawary M. A review of selected optimal power flow literature to 1993. I. Nonlinear and quadratic programming
approaches. IEEE Trans Power Syst. 1999;14(1):96‐104.
13. Stott B, Hobson E. Power system security control calculations using linear programming, part I. IEEE Trans Power Syst. 1978;
PAS‐97(5):1713‐1720.
14. Stott B, Hobson E. Power system security control calculations using linear programming, part II. IEEE Trans Power Syst. 1978;
PAS‐97(5):1721‐1731.
15. BoussaïD I, Lepagnot J, Siarry P. A survey on optimization metaheuristics. Inform Sci. 2013;237:82‐117.
16. Frank S, Steponavice I, Rebennack S. Optimal power flow: a bibliographic survey I. Energy Syst. 2012;3(3):221‐258.
17. Kirkpatrick S. Optimization by simulated annealing: quantitative studies. J Stat Phys. 1984;34(5‐6):975‐986.
18. Wang G‐G, Deb S, Coelho LdS. Elephant herding optimization, in Computational and Business Intelligence (ISCBI), 2015 3rd International
Symposium on, 2015: 1‐5.
19. Blum C, Puchinger J, Raidl GR, Roli A. Hybrid metaheuristics in combinatorial optimization: a survey. Appl Soft Comput.
2011;11(6):4135‐4151.
20. Holland JH, Reitman JS. Cognitive systems based on adaptive algorithms. Acm Sigart Bull. 1977;(63):49‐49.
21. Storn R, Price K. Differential evolution—a simple and efficient heuristic for global optimization over continuous spaces. J Glob Optim.
1997;11(4):341‐359.
22. Wang Y, Li H‐X, Huang T, Li L. Differential evolution based on covariance matrix learning and bimodal distribution parameter setting.
Appl Soft Comput. 2014;18:232‐247.
23. Wang Y, Cai Z, Zhang Q. Differential evolution with composite trial vector generation strategies and control parameters. IEEE Trans Evol
Comput. 2011;15(1):55‐66.
24 of 28 TAHER ET AL.

24. Wang Y, Cai Z, Zhang Q. Enhancing the search ability of differential evolution through orthogonal crossover. Inform Sci.
2012;185(1):153‐177.
25. Simon D. Biogeography‐based optimization. IEEE Trans Evol Comput. 2008;12(6):702‐713.
26. Rechenberg I. Evolutionsstrategien. In: Simulationsmethoden in der Medizin und Biologie. Hannover: Springer; 1978:83‐114.
27. Baluja S. Population‐based incremental learning. A method for integrating genetic search based function optimization and competitive
learning, Carnegie‐Mellon Univ Pittsburgh Pa Dept Of Computer Science 1994.
28. Kirkpatrick S, Gelatt CD, Vecchi MP. Optimization by simulated annealing. Science. 1983;220(4598):671‐680.
29. Černý V. Thermodynamical approach to the traveling salesman problem: an efficient simulation algorithm. J Optim Theory Appl.
1985;45(1):41‐51.
30. Rashedi E, Nezamabadi‐Pour H, Saryazdi S. GSA: a gravitational search algorithm. Inform Sci. 2009;179(13):2232‐2248.
31. Bouchekara H, Chaib A, Abido MA, El‐Sehiemy RA. Optimal power flow using an improved colliding bodies optimization algorithm.
Appl Soft Comput. 2016;42:119‐131.
32. Hatamlou A. Black hole: a new heuristic optimization approach for data clustering. Inform Sci. 2013;222:175‐184.
33. Erol OK, Eksin I. A new optimization method: big bang–big crunch. Adv Eng Softw. 2006;37(2):106‐111.
34. Niknam T, Narimani MR, Azizipanah‐Abarghooee R. A new hybrid algorithm for optimal power flow considering prohibited zones and
valve point effect. Energ Conver Manage. 2012;58:197‐206.
35. Ghasemi M, Ghavidel S, Rahmani S, Roosta A, Falah H. A novel hybrid algorithm of imperialist competitive algorithm and teaching
learning algorithm for optimal power flow problem with non‐smooth cost functions. Eng Appl Artif Intel. 2014;29:54‐69.
36. Eberhart R, Kennedy J. A new optimizer using particle swarm theory, in Micro Machine and Human Science, 1995. MHS'95., Proceedings
of the Sixth International Symposium on, 1995: 39‐43.
37. Mirjalili S. Moth‐flame optimization algorithm: a novel nature‐inspired heuristic paradigm. Knowl‐Based Syst. 2015;89:228‐249.
38. Dorigo M, Birattari M. Ant Colony Optimization Encyclopedia of Machine Learning. New York: Springer; 2010:36‐39.
39. Karaboga D, Basturk B. A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm. J
Glob Optim. 2007;39(3):459‐471.
40. Rao RV, Savsani VJ, Vakharia D. Teaching–learning‐based optimization: an optimization method for continuous non‐linear large scale
problems. Inform Sci. 2012;183(1):1‐15.
41. Rao RV, Savsani VJ, Vakharia D. Teaching–learning‐based optimization: a novel method for constrained mechanical design optimization
problems. Comput Aided Des. 2011;43(3):303‐315.
42. Ghasemi M, Ghavidel S, Gitizadeh M, Akbari E. An improved teaching–learning‐based optimization algorithm using Lévy mutation strat-
egy for non‐smooth optimal power flow. Int J Electr Power Energy Syst. 2015;65:375‐384.
43. Pandiarajan K, Babulal C. Fuzzy harmony search algorithm based optimal power flow for power system security enhancement. Int J
Electr Power Energy Syst. 2016;78:72‐79.
44. Glover F. Tabu search and adaptive memory programming—advances, applications and challenges. In: Interfaces in Computer Science
and Operations Research. Boston: Springer; 1997:1‐75.
45. He S, Wu QH, Saunders J. Group search optimizer: an optimization algorithm inspired by animal searching behavior. IEEE Trans Evol
Comput. 2009;13(5):973‐990.
46. Kashan AH. League championship algorithm (LCA): an algorithm for global optimization inspired by sport championships. Appl Soft
Comput. 2014;16:171‐200.
47. Wolpert DH, Macready WG. No free lunch theorems for optimization. IEEE Trans Evol Comput. 1997;1(1):67‐82.
48. Yang Z, Zhong H, Xia Q, Kang C. Fundamental review of the OPF problem: challenges, solutions, and state‐of‐the‐art algorithms. J
Energy Eng. 2017;144:04017075.
49. Yang Z, Zhong H, Xia Q, Kang C. Solving OPF using linear approximations: fundamental analysis and numerical demonstration. IET
Gener Transm Distrib. 2017;11(17):4115‐4125.
50. Madani R, Sojoudi S, Lavaei J. Convex relaxation for optimal power flow problem: mesh networks. IEEE Trans Power Syst.
2015;30(1):199‐211.
51. Sojoudi S, Lavaei J. Physics of power networks makes hard optimization problems easy to solve, in Power and Energy Society General
Meeting, 2012 IEEE, 2012: 1‐8.
52. Wang H, Murillo‐Sanchez CE, Zimmerman RD, Thomas RJ. On computational issues of market‐based optimal power flow. IEEE Trans
Power Syst. 2007;22(3):1185‐1193.
53. Stott B, Jardim J, Alsaç O. DC power flow revisited. IEEE Trans Power Syst. 2009;24(3):1290‐1300.
54. Lu S, Zhou N, Kumar NP, Samaan N, Chakrabarti BB. Improved DC power flow method based on empirical knowledge of the system, in
Transmission and Distribution Conference and Exposition, 2010 IEEE PES, 2010: 1‐6.
TAHER ET AL. 25 of 28

55. Qi Y, Shi D, Tylavsky D. Impact of assumptions on DC power flow model accuracy, in North American Power Symposium (NAPS), 2012,
2012: 1‐6.
56. Mahdad B, Srairi K. Blackout risk prevention in a smart grid based flexible optimal strategy using Grey Wolf‐pattern search algorithms.
Energ Conver Manage. 2015;98:411‐429.
57. Frank KD, Rich C, Longcore T. Effects of artificial night lighting on moths, Ecological consequences of artificial night lighting, 2006:
305‐344
58. Soliman GM, Khorshid MM, Abou‐El‐Enien TH. Modified moth‐flame optimization algorithms for terrorism prediction.
59. Dunham D. Hyperbolic spirals and spiral patterns, in Meeting Alhambra, ISAMA‐BRIDGES Conference Proceedings, Granada, Spain, 2003:
521‐528.
60. Yao X, Liu Y, Lin G. Evolutionary programming made faster. IEEE Trans Evol Comput. 1999;3:82‐102.
61. El Ela AA, Abido M, Spea S. Optimal power flow using differential evolution algorithm. Electr Pow Syst Res. 2010;80(7):878‐885.
62. Hardiansyah H. A modified particle swarm optimization technique for economic load dispatch with valve‐point effect. Int J Intell Syst
Appl. 2013;5(7):32‐41.
63. Alsumait J, Sykulski J, Al‐Othman A. A hybrid GA–PS–SQP method to solve power system valve‐point economic dispatch problems. Appl
Energy. 2010;87(5):1773‐1781.
64. Abido MA. Environmental/economic power dispatch using multiobjective evolutionary algorithms. IEEE Trans Power Syst.
2003;18(4):1529‐1537.
65. Yuryevich J, Wong KP. Evolutionary programming based optimal power flow algorithm. IEEE Trans Power Syst. 1999;14(4):1245‐1250.
66. He S, Wen J, Prempain E, Wu Q, Fitch J, Mann S. An improved particle swarm optimization for optimal power flow, in Power System
Technology, 2004. PowerCon 2004.2004 International Conference on, 2004: 1633‐1637.
67. Kessel P, Glavitsch H. Estimating the voltage stability of a power system. IEEE Trans Power Delivery. 1986;1(3):346‐354.
68. Zimmerman RD, Murillo‐Sánchez CE. Matpower 6.0 User's Manual, ed: Dec, 2016.

How to cite this article: Taher MA, Kamel S, Jurado F, Ebeed M. An improved moth‐flame optimization
algorithm for solving optimal power flow problem. Int Trans Electr Energ Syst. 2019;29:e2743. https://doi.org/
10.1002/etep.2743

A P P EN D I X A

A.1 | Characteristics of IEEE 30‐bus test system

Characteristics Value details Details

8
Busses 30
8
Branches 41
Generators 6 Busses: 1, 2, 5, 8, 11, and 13
Load voltage limits 24 [0.95:1.05]
Shunts VAR compensation 9 Busses: 10, 12, 15, 17, 20, 21, 23, 24, and 29.
Transformers 4 Branches: 11, 12, 15, and 36
Control variables 24 ‐
26 of 28 TAHER ET AL.

A.2 | Cost and emission coefficients for the IEEE 30‐bus test system

.Generator Bus a b c d e a b ? ω μ

G1 1 0 2 0.00375 18 0.037 4.091 −5.554 6.49 2.00E − 04 2.857

G2 2 0 1.75 0.0175 16 0.038 2.543 −6.047 5.638 5.00E − 04 3.333
G3 5 0 1 0.0625 14 0.04 4.258 −5.094 4.586 1.00E − 06 8
G4 8 0 3.25 0.00834 12 0.045 5.326 −3.55 3.38 2.00E − 03 2
G5 11 0 3 0.025 13 0.042 4.258 −5.094 4.586 1.00E − 06 8
G6 13 0 3 0.025 13.5 0.041 6.131 −5.555 5.151 1.00E − 05 60667

A.3 | Cost coefficients for generators 1 and 2 of IEEE 30‐bus test system for multi‐ fuel sources

Generator Bus a b c Pmin Pmax a b c Pmin Pmax

G1 1 55 0.7 0.005 50 140 82.5 1.05 0.0075 140 200

G2 2 40 0.3 .01 20 55 80 0.6 0.02 55 80

A.4 | OPF solution values obtained by IMFO for IEEE 30‐bus test system

Variable MIN Max Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10

PG1 (MW) 50 200 176.969 140 198.821 64.176 51.491 102.803 176.249 177.127 112.968 122.620
PG2 (MW) 20 80 48.732 55 45.152 67.484 80 55.505 48.891 48.562 58.968 52.509
PG5 (MW) 15 50 21.387 24.124 18.110 50 50 37.989 21.666 21.466 27.823 31.220
PG8 (MW) 10 35 21.422 35 10.016 35 35 35 22.262 21.389 35 35
PG11 (MW) 10 30 11.872 18.388 10 30 30 29.999 12.146 11.872 27.348 26.752
PG13 (MW) 12 40 12.005 17.489 12.001 40 40 26.686 12 12 26.666 20.969
V1 (PU) 0.95 1.1 1.0824 1.0718 1.0817 1.0548 1.0607 1.0717 1.0371 1.082 1.0678 1.0694
V2 (PU) 0.95 1.1 1.0727 1.0974 1.0988 1.0483 1.0565 1.0249 1.0209 1.0616 1.0565 1.0974
V5 (PU) 0.95 1.1 1.0323 1.0288 1.0246 1.0262 1.0364 1.0458 1.0095 1.0297 1.0298 1.0812
V8 (PU) 0.95 1.1 1.0357 1.0283 1.0343 1.0343 1.0428 1.0438 1.0004 1.0331 1.0398 1.0006
V11 (PU) 0.95 1.1 1.0671 1.0517 1.0402 1.056 1.0564 1.1 1.0357 1.0655 1.0543 1.0229
V13 (PU) 0.95 1.1 1.049 1.0646 1.0675 1.0587 1.057 0.98982 0.99407 1.0502 1.0659 1.0175
T11 (PU) 0.90 1.1 1.0356 1.0326 1.0348 1.0035 1.009 1.0414 1.0526 0.99335 1.0258 1.0631
T12 (PU) 0.90 1.1 0.9123 0.9 0.90833 0.92711 0.94261 0.93552 0.90005 0.99232 0.9 0.94762
T15 (PU) 0.90 1.1 0.97436 1.0046 1.0007 0.99019 0.99796 0.96467 0.95639 0.98237 1.003 1.0199
T36 (PU) 0.90 1.1 0.97237 0.97068 0.97571 0.96136 0.97235 0.97431 0.96715 0.97153 0.96899 0.99971
Qc10 0.00 5 1.373 1.510 4.9799 0.801 2.292 0.183 5 5 1.1931 1.064
(MVAR)
Qc12 0.00 5 2.494 0.000 1.4004 1.048 2.329 2.105 3.158 4.993 0 0.758
(MVAR)
Qc15 0.00 5 4.132 4.806 4.7462 3.851 4.946 4.001 5 5 4.3283 4.998
(MVAR)
TAHER ET AL. 27 of 28

(Continued)

Variable MIN Max Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10
Qc17 0.00 5 4.874 5.000 4.2106 5 5 5 5 5 2.554 5
(MVAR)
Qc20 0.00 5 3.859 4.975 3.5526 4.947 4.608 3.298 5 5 4.253 4.9999
(MVAR)
Qc21 0.00 5 5.000 5.000 5.000 4.972 5 4.998 5 4.991 4.9999 3.808
(MVAR)
Qc23 0.00 5 3.285 3.946 3.2469 3.874 2.592 3.435 5 3.335 2.9772 4.998
(MVAR)
Qc24 0.00 5 5.000 5.000 5 5 5 5 5 5 4.9998 2.450
(MVAR)
Qc29 0.00 5 2.311 1.995 2.2283 1.672 1.783 2.015 2.223 3.090 1.8605 0.10226
(MVAR)
Fuel cost 800.3848 645.8958 832.1023 944.6964 967.5900 858.7455 803.5715 800.4762 836.4809 829.8682
($/h)
Emission 0.3658 0.2836 0.4382 0.20480 0.2073 0.2291 0.3636 0.3662 0.2423 0.2535
(ton/h)
PLoss 8.999 6.6100 10.7122 3.33 3.0905 4.5815 9.8137 9.0161 5.3738 5.676
(MW)
QLoss ‐10.638 ‐17.828 ‐4.717 ‐28.943 ‐32.582 ‐25.144 ‐2.302 ‐12.952 ‐23.011 ‐22.122
(MW)
VD (p.u.) 0.9035 0.9115 0.8318 0.8061 0.9192 0.9559 0.0954 0.9086 0.9206 0.2778
Lmax 0.1274 0.1274 0.1281 0.1288 0.1274 0.1274 0.1366 0.1255 0.1280 0.1358

A.5 | Characteristics of IEEE 57‐bus test system

Characteristics Value details Details

60
Busses 57
60
Branches 88
Generators 7 Busses: 1, 2, 5, 8, 11, and 13
Load voltage limits 50 [0.94:1.06]
Shunts VAR compensation 9 Branches: 18, 25, and 53.
Transformers 17 Branches: 19, 20, 31, 35, 36, 37, 41, 46, 54, 58, 59, 65, 66, 71, 73, 76, and 80.
Control variables 33 ‐
28 of 28 TAHER ET AL.

A.6 | Cost and emission coefficients for IEEE 57‐bus test system

Generator Bus a b c d e a b ? ω μ

G1 1 0 2 0.00375 18 0.037 4.091 −5.554 6.49 2.00E − 04 2.86E − 01

G2 2 0 1.75 0.0175 16 0.038 2.543 −6.047 5.638 5.00E − 04 3.33 E − 01
G3 3 0 3 0.025 13.5 0.041 6.131 −5.555 5.151 1.00E − 05 6.67 E − 01
G4 6 0 2 0.00375 18 0.037 3.491 −5.754 6.39 3.00E − 04 2.66 E − 01
G5 8 0 1 0.0625 14 0.04 4.258 −5.094 4.586 1.00E − 06 8.00 E − 01
G6 9 0 1.75 0.0195 15 0.039 2.754 −5.847 5.238 4.00E − 04 2.88 E − 01
G7 12 0 3.25 0.00834 12 0.045 5.326 −3.555 3.38 2.00E − 03 2.00 E − 01

A.7 | Characteristics of IEEE 118‐bus test system

Characteristics Value details Details

60
Busses 118
60
Branches 186
Generators 54 Busses: 1, 4, 6, 8, 10, 12, 15, 18, 19, 24, 25, 26, 27, 31, 32, 34, 36, 40, 42, 46,
49, 54, 55, 56, 59, 61, 62, 65, 66, 69, 70, 72, 73, 74, 76, 77, 80, 85, 87, 89, 90,
91, 92, 99, 100, 103, 104, 105, 107, 110, 111, 112, 113, and 116
Load voltage limits 50 [0.94:1.06]
Shunts VAR compensation 9 Busses: 5, 34, 37, 44, 45, 46, 48, 74, 79, 82, 83, 105, 107, and 110
Transformers 17 Branches: 8, 32, 36, 51, 93, 95, 102, 107, and 127.
Control variables 130 ‐