You are on page 1of 21

Journal of Ambient Intelligence and Humanized Computing

https://doi.org/10.1007/s12652-020-02692-7

ORIGINAL RESEARCH

An enhanced moth‑swarm algorithm for efficient energy management


based multi dimensions OPF problem
Bachir Bentouati1 · Aboubakr Khelifi1 · Abdullah M. Shaheen2 · Ragab A. El‑Sehiemy3 

Received: 25 May 2020 / Accepted: 8 November 2020


© Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract
The optimal power flow (OPF) problem is very important issue in operation, planning and energy management of power
systems. OPF analysis aims to find the optimal solution of system nonlinear algebraic equations with satisfying operational
constraints. Economic, environmental and technical objectives are considered for multi-dimensions efficient energy manage-
ment. These objectives involve the reduction of the production costs, reduction of the environmental emissions, improving
the voltage profile, reducing the power losses and enhancing the system stability. This paper presents a new high-efficiency
technology that proposes a multi-objective version of the recently proposed moth swarm algorithm (MSA) i.e. enhanced
MSA (EMSA). The modification is implemented based on quasi-opposition-based learning. In order to verify the efficacy of
proposed EMSA, the simulations are done in the IEEE 30-bus and IEEE 57-bus test systems. The scalability of the proposed
method is proved on the IEEE 118-bus test network. The outcomes are compared with that obtained by MSA and the reported
methods in the literature. From the outcomes obtained, it is strongly confirmed that proposed EMSA performs considerably
better than MSA to address different test objectives with significant improvements of the considered complex power system.

Keywords  Enhanced moth swarm algorithm · Fuel cost · Voltage stability improvement · Real power loss · Multi-objective
optimal power flow

Abbreviations nh Small group of moths walks into


ai , bi , ci Cost coefficients of ith generator random spiral path inside the neighbor-
𝛼i , 𝛽i , 𝛾i , 𝜉i and 𝜆i Emission coefficients of ith unit hood of light source
PGi Real power bus generator nol Small moth group drifts towards the
PD , QD Active and reactive load demands moonlight
Gij Transfer conductance 𝜎jt Normalized form of dispersal degree at
Bij Susceptance between bus i and bus j iteration
VG Voltage levels at generation buses 𝜇t Relative dispersion
TG Transformers tap setting cp Crossover points
QC Shunt VAR compensation t
Bestglobal Best global solution
xjmin , xjmax Upper and lower limit of candidate ngw Gaussian walk
solutions xit+1 Onlooker moths
Xqo Quasi-opposite point

* Ragab A. El‑Sehiemy
elsehiemy@eng.kfs.edu.eg 1 Introduction
1
Electrical Engineering Department, LMSF Laboratory,
The optimal power flow (OPF) problem is one of the basic
Amar Telidji University of Laghouat, 03000 Laghouat,
Algeria problems with given or fixed load power and generator
2 power. Over the past few decades, OPF has been one of the
Electrical Engineering Department, Suez University, Suez,
Egypt most effective tools for analyzing such systems (Attia et al.
3 2018). The main target of OPF is to minimize the cost of
Electrical Engineering Department, Kafrelsheikh University,
Kafr El‑Sheikh, Egypt generating electricity by controlling the optimal setting of

13
Vol.:(0123456789)
B. Bentouati et al.

variables, which are the actual active power of the network computing (QMFOA) (Dabba et al. 2020), coyote optimiza-
and the generator bus voltage. While optimizing the cost tion algorithm (El-Ela et al. 2020). It is necessary to demon-
of power generation, we should meet the constraints of the strate the effectiveness of the algorithm in a small-executed
system on generator capacity, line capacity, bus voltage and time and fast convergence manner. In addition, starting with
current balance (Abaci and Yamacli 2016). The predeter- multiple-goal issues, they require higher computing costs
mined generator power, the complex power flow in the line, and multiple goals at the same time.
and the bus voltage vector determined during optimization The literature shows that these methods have been widely
represent the optimal operating state of the system. The tar- used since the present invention and can solve real-world
get function includes the total fuel cost of the generator set, problems. There are popular algorithms that are developed
the emission rate from generator set, the system power loss for various power system applications as:the real coded
and the voltage safety index, etc. Renewable sources like genetic algorithm in El-Sehiemy et al. (2013), crow search
solar power have added advantage of zero environmental algorithm (Abou El Ela et al. 2017), moth flame optimizer
emission (Bouchekara et al. 2016). Moreover, the evolution (Elsakaan et al. 2018, 2020), enhanced grey wolf optimizer
of modular distributed generators (DGs) calls for smaller (GWO) in Shaheen and El-Sehiemy (2020), non-dominated
space, less construction time and lower investment. Hence, sorting genetic algorithm (NSGA) (Yuan 2006) and multi-
the OPF problem becomes a multi-objective optimization objective particle swarm optimization (MOPSO) (Coello
problem. The main challenge in multi-objective optimiza- and Lechuga 2002; Barakat et al. 2020).
tion is to generate high-quality solutions that generate evenly Most of the recently proposed single-objective algo-
distributed methods. On the other hand, automatic solutions rithms are equipped with operators to solve multi-objective
have become more and more important with the develop- problems. Some of the most recent multi-target algorithms
ment of algorithms, where several conventional methods (Elsakaan et al. 2018; Barakat et al. 2020). Multi-objective
have been applied to solve OPF problems, such as Newton GWO (Mirjalili et al. 2016), multi-objective hybrid particle
method, linear and nonlinear problems (Shafik et al. 2019). swarm and salp optimization algorithm (MPPSO-SSO) (El
In recent years, there are many intelligent optimization Sehiemy et al. 2020), multi objective cuckoo search algo-
methods in the gradient reduction optimization algorithm rithm (MOCS) (Rao et al. 2020), and multi-objective differ-
to solve the problems represented by mathematical formulas ential-based harmony search algorithm (MODH) (Abbasi
by seeking derivatives. The gradual development of com- et al. 2020). The Jaya optimization algorithm was developed
puting resources has promoted the development of the so- for solving two OPF frameworks that cover single and multi-
called meta-heuristic methods. These methods can overcome objective cases in El-Sattar et al. (2019) and El-Fergany and
some of the obstacles that have been presented to solve OPF Hasanien (2020).
problems, such as improving the algorithm through modified In El-Fergany and Hasanien et al. (2018), another devel-
sine–cosine algorithm (MSCA) (Attia et al. 2018), differen- opment of the salp swarm optimizer was suggested for
tial search algorithm (Abaci and Yamacli 2016), improved solving the OPF problem. This work was associated with
colliding bodies optimization (Bouchekara et  al. 2016), a comprised analysis of the voltage stability. In Shaheen
adaptive parallel seeker optimization (Luo and Semlyen et al. (2017a, b), the tree-seed optimization algorithm was
1989), adaptive group search optimization (AGSO) (Dary- applied for finding the optimal solution of OPF problem and
ani et al. 2016), moth swarm algorithm (MSA) (Mohamed concentrating on large-scale power systems. The differential
et al. 2017; Duman 2018), TLBO with using Lévy mutation evolution algorithm was solved the OPF problem in Shaheen
operator (LTLBO) (Ghasemi et al. 2015), fruit fly optimiza- et al. (2016). To enhancing the OPF solution, a forced ini-
tion algorithm (FFOA) (Abou El-Ela et al. 2018). tialized mechanism is applied for the differential evolution
Several efforts were been applied in the area of optimi- optimization algorithm in Medina et al. (2014). The decom-
zation algorithms that enhance the solution quality of large position process that is employed for solving multi-objective
scale problem such as OPF. Sample of recent efforts are: framework in Ghasemi et al. (2014). In Warid et al. (2018),
monarch butterfly optimization (MBO) (Wang et al. 2019), a modified imperialist competitive optimization algorithm
earthworm optimization algorithm (EWO) (Wang et  al. was developed for optimizing the production fuel costs,
2015), elephant herding algorithm (EHO) (Wang et  al. minimizing the resulted emissions, enhancing the voltage
2016), improved moth flame optimization (IMFO) (Wu profile by minimizing the voltage deviation and minimizing
et al. 2019), hybrid algorithm based on water cycle and moth the transmission power losses.
flame optimization(WCMFO) (Khalilpourazari and Khalil- Recently the optimal power flow problem in Biswas et al.
pourazary 2019), Bio-inspired voice evaluation mechanism (2017) is solved with the existence of stochastic renew-
(Połap et al. 2019), Neuro-heuristic voice recognition (Połap able resources as wind and solar powers. Duman (2018)
2016), chaotic salp swarm optimizer (CSSO) (Bentouati presented a modified version of the optimization algo-
et al. 2019), moth flame optimization algorithm and quantum rithm called moth swarm algorithm. It is equipped with an

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

arithmetic for rapid global convergence. The modified algo- the performance of the proposed EMSA is better described
rithm was developed for solving the OPF problem. In this to other evolutionary methods reported in the literature. The
work, the modification is employed by using the arithmetic main contributions of this paper can be itemized as follows:
crossover. To boost the optimization aspect, a modified MSA
was applied to minimize three different objectives. Another 1. This paper proposes an enhanced moth swarm algorithm
framework was developed for the OPF problems in Refs. for solving the optimal power flow problem.
Abdul-hamied et al. (2020) and (Elattar et al. 2020) con- 2. The modification carries out the quasi-opposition-based
sidering hybrid AC/DC grid. In this framework, the impact learning for enhancing the moth swarm optimization
of power electronic devices is studied when the renewable algorithm.
energy resources are penetrated. 3. The proposed EMSA is applied on IEEE 30-bus and
The No-Free Lunch (Wolpert and Macready 1997) theo- IEEE 57-bus test systems. While, the scalability of the
rem for optimization allows researchers to propose new algo- proposed method is proved on the IEEE 118-bus test
rithms or improve one’s because they logically prove that network.
there are no optimization algorithms that solve all optimi- 4. The simulation outcomes are compared with the reported
zation problems. This applies to single-objective and multi- methods in the literature.
objective optimization techniques. A recent work based on 5. The proposed EMSA has better performance compared
the MSA that conceptualizes indigenous flying pattern of with MSA.
moths towards the light source (Duman 2018). Research- 6. Significant improvements are proceed for single and
ers proved that MSA has a high success rate in achieving multi-objective cases.
viable OPF-solution set of good quality to address power
system issues. It was validated in Duman (2018) that by The rest parts of this paper are organized as follows:
allowing a diversified search operation, different test objec- Sect. 2 presents the mathematical models’ representation of
tives achieved their optimum value. the considered OPF problem. Section 3 describes the idea
In spite of various modification approaches, the solu- behind the proposed EMSA to solve different OPF prob-
tions of Rahnamayan et al. (2008) and Wang et al. (2007) lems. Section 4 reports the simulated applications with a
accounted some drawbacks which are poor intensification of detailed discussion on the outcomes while the conclusions
the solutions, poor trade-off between the intensification and are derived in the Sect. 5.
diversification and a delayed convergence profile. Hence, to
realize well intensified solution, proper trade-off between
intensification and diversification and a fast convergence
rate as well, this paper blends opposition-based learning
2 Mathematical OPF models
(OBL) (Rahnamayan et al. 2008) with traditional MSA.
The optimization techniques can be applied for solving the
The basic concept of opposition-based learning (OBL)
OPF problem. The optimal power flow problem is expressed
(Rahnamayan et  al. 2008) was originally introduced by
as non-linear optimization problem of ­nobj objective func-
Tizhoosh. The main idea behind OBL is for finding a better
tions subject to equality and inequality constraints. There-
candidate solution and the simultaneous consideration of
fore, the most common formulation of the OPF problem is
an estimate and its corresponding opposite (i.e., guess and
presented as follows:
opposite guess) which is closer to the global optimum. OBL
has been applied to many EAs, such as differential evolu- {
min F = fi (x, u) i = 1, 2, ....nobj (1)
tion (Rahnamayan et al. 2008), particle swarm optimization
(Wang et al. 2007) etc…
The OBL strategy is carried out to half the population to subject to ∶
minimize time complexity. The population is then updated gn (x, u) = 0 n = 1, 2, ...neq (2)
based on the current solution and the best solution for its hk (x, u) ≤ 0 k = 1, 2, ...ninq
opposing parts. The proposed OBL along with MSA is
applied to solve different OPF problems. The effectiveness Equation (1) has three parts, the first one refers to the
of the proposed EMSA is tested on three standard systems objective function/s to be optimized. It formulated as
called IEEE 30-bus, IEEE 57-bus and the large-scale IEEE single objective by considering one objective to be opti-
118-bus test systems. The proposed procedure is tested mized at certain time. While, multi-objective framework
for five objective problems. These are (1) fuel cost mini- is implemented by considering two or more objectives to
mization, (2) emission minimization, (3) voltage deviation be optimized simultaneously. The other two parts reflect
minimization (4) active power loss minimization and (5) the equality and inequality operation constraints of the
enhancement of voltage stability. It has been observed that considered OPF problem. In this study, several OFs are

13
B. Bentouati et al.

formulated incorporating hence. These OFs represent tech- | NG


∑ |
| Vi ||
nical, economic and environmental issues in the power |
Li = | 1 − Fji | ∀j = 1, 2 … NL (7)
Vj |
system operation. The mathematical representation of |
| i=1 |
these OFs is described as:
where, ­Y1 and ­Y2 are the system YBUS sub-matrices.
• OF#1: The first economic OF aims to optimize the fuel The objective describes the system stability of complete
cost of the committed generation units. This objective system L-index that can be expressed as follows (Abaci
is represented as a polynomial quadratic any generator and Yamacli 2016):
as follows (Attia et al. 2018; Abaci and Yamacli 2016): ( )
L = max Li ∀j = 1, 2 … NL (8)
NG ( )

Fi = ai + bi PGi + ci P2G (3)
i=1
i
The transmission system has several constraints which
can be categorized as:
where, ai  , bi and ci are the cost coefficients of ith
generator. • Equality constraints
• OF#2: The second OF aims to minimize the total power
losses of the transmission network. This objectives
The equality constrains represent the balanced load flow
function is represented as such (Bouchekara et al. 2016;
equations. Equations (9) and (10) represent the active and
Mohamed et al. 2017):
reactive powers inequality constraints that satisfy the load
• The active power losses
requirements and the transmission line losses.
NB NB
∑ ∑ ( ) NB
Fi = Pi = PGi − PDi (4) ∑ [ ]
i=1 i=1
PGi − PDi − Vi Vj Gij cos 𝜃ij + Bij sin 𝜃ij = 0 (9)
j=1
• OF#3: This objective function aims to enhance the volt-
age profile by improving the bus voltage levels at load NB
∑ [ ]
buses. This OF is achieved by minimizing the devia- QGi − QDi − Vi Vj Gij sin 𝜃ij − Bij cos 𝜃ij = 0 (10)
tions between the load buses voltages and the specified j=1

value ( VLi
SP
= 1.0 ) (Bouchekara et al. 2016). However,
where, 𝜃ij = 𝜃i − 𝜃j , NB is the number of buses, PG is
it can be evaluated by Eq. (5):
the active power generation,QG is the reactive power gen-
NL
∑ eration, PD and QD are active and reactive power demand,
| SP |
Fi = |VLi − VLi | (5) respectively, Gij and Bij are elements of admittance matrix
| |
i=1 representing conductance and susceptance between buses
• OF#4: This objective function is concerned with the i and j, respectively.
environmental emission reduction by minimizing the
• Inequality constraints
emission level of pollutants which can be written in
the following form (Bouchekara et al. 2016; Mohamed
et al. 2017): The inequality constrains can be classified as follows:

NG
∑ ( ) ( ) • Generation constraints
Fi = 10−2 𝛼i + 𝛽i PGi + 𝛾i P2G + 𝜉i exp 𝜆i PGi (6)   Following same sequence in Eq. (13), VGi is the gen-
i
i=1
eration bus voltage, NG is the total number of genera-
where, 𝛼i , 𝛽i , 𝛾i , 𝜉i and 𝜆i are the emission coefficients of tors or generator buses.
ith unit.
Pmin ≤ PGi ≤ Pmax (11)
• OF#5: Voltage stability index (L) is an important value G i G i

proposed. This value is calculated for all load buses,


which vary between 0 (no load case) to 1 maximum Qmin ≤ QGi ≤ Qmax (12)
G G
loading point (voltage collapse case).Hence, the mini-
i i

mizing of this value keeps the system far away from


voltage collapse of power system. However, this objec- VGmin ≤ VGi ≤ VGmax (13)
i i
tive function can be formulated as follows:

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

• Security constraints – Prospectors ( nh ): This small group of moths walks into
random spiral path inside the neighborhood of light
Vimin ≤ Vi ≤ Vimax ∀i ∈ NL (14) source.
– Onlookers ( nol ): this small moth group drifts towards
Slimin ≤ Slimax ∀i ∈ nl (15) the moonlight i.e. the fittest global solution, found by
the prospectors.
• Transformer constraints
Different phases of the algorithm are stated below:
(16)
min max
TGi ≤ TGi ≤ TGi ∀i ∈ NT
1. Initialization
• Shunt VAR compensator constraints
  As the flight begins, the initialization of the search
Qmin ≤ Qci ≤ Qmax ∀i ∈ NC (17) agent moths, i.e. the initial candidate solutions are gen-
ci ci
erated for a problem of—dimension and a population
of—number by the d n following equations,
Equation (14) defines the voltage at load buses Vi with ( )
NL being the number of load buses. The constraints xij = rand[0, 1]. xjmax − xjmin + xjmin , ∀i ∈ {1, ....., n}, j ∈ {1, ....., d}
of transmission line loadings (Sli ) must be restricted by (18)
respectively their acceptable
( ) boundaries as in Eq. (15). where andxjmin correspond to the upper and lower
xjmax
Transformer tap settings TGi must be preserved by their
limit of candidate solutions.
permissible boundaries as in Eq. (16). The reactive power
  In this phase, after initialization, types of each moth
of shunt VAR compensator (Qci ) must be controlled by
are determined based on their calculated fitness.
their upper and lower limits in Eq. (17).
2. Reconnaissance
  Throughout the optimization process, there is a ten-
dency of the moths to be concentrated in a region of
search space which seems potentially good. This may
3 Proposed solution methodology impact the convergence characteristic and the diversity
in the solutions. To overcome this situation, the path-
3.1 A systematic outline of moth swarm algorithm
finder moth’s positions are updated with the help of
levy-mutation and adaptive-crossover operation. It is
Moth swarm algorithm (MSA) is an evolutionary tech-
depicted in detail in five steps that are to be followed in
nique, which is proposed by Mohamed et al. (2017). In
the reconnaissance phase.
proposed MSA, in search of more diversified solutions,
3. Proposed diversity index for crossover points
a novel adaptive crossover and mutation scale by means
  This step determines a novel strategy to select cross-
of Lévy flights are employed by the pathfinders in a pure
over points in order to maintain population diversity.
exploration stage. In the meantime, the trade-off between
To obtain the crossover points, the diversity index is
diversification and intensification is maintained by the
implemented in the first step. The normalized form of
prospectors that follow the logarithmic spiral motion.
dispersal degree at t  th iteration is calculated in this step
To achieve an enhanced convergence speed, conversion
as formulated below,
of navigation is done between transverse orientation and
celestial navigation, adaptively. In the proposed MSA, two �
∑ nf
� nf

�2
1 1
assumptions are made to characterize the problem. These nf i=1
t
xi,j − nf
t
xi,j
are the position of light source that resembles possible 𝜎jt =
i=1 (19)
1 ∑ nf
solutions of the optimization problem and the second one nf
xt
i=1 i,j
is luminescence intensity which is similar to the fitness
of the problem. Three moth groups are included in the
proposed MSA as: The relative dispersion is expressed below,
d
– Pathfinder ( nf  ): This moth group finds potential regions 1∑ t
(20)
t
𝜇 = 𝜎
in search space on the basis of First in–Last Out con- d j=1 j
cept. This group guides the main swarm towards the
light source. Lévy flight It is a random process based on α-stable dis-
tribution. This step enables the pathfinders to have journey
through a long-distance with different step sizes.

13
B. Bentouati et al.

( ))
Difference vector levy-mutation In this step, a sub-trial t log t ( t t
𝜀1 ∼ random(size(d)) ⊕ N Bestglobal , xi − Bestglobal
vector is created for nc ∈ cp crossover points. Further, for t
the synthesis of the sub-trail, the vector mutation strategy (25)
may be used. where𝜀1 is a random sample based on Gaussian stochastic
Proposed adaptive crossover operation based on the pop- distribution; Bestglobal
t
corresponds to the best global
ulation diversity In order to attain the complete trail solution, solution;𝜀2 and 𝜀3 are two random numbers with an inter-
the position of each pathfinder solution is updated via an val of [0, 1].
adaptive crossover operation, which incorporates mutated – Associative learning mechanism with immediate mem-
variables of the sub-trail vector into the equivalent variables ory
of the host vector.   Remaining onlooker moths, which are driven by an
Selection strategy This step calculates and compares the immediate memory, apply associative learning mecha-
fitness of the completed trail solution with the same cor- nism to drift towards the moonlight and while doing
responding host solution. The selected fitter solutions are this, they behave real moths in nature. The onlooker
retained for the next generation. moth with this mechanism is updated as follows,
Transverse orientation In this phase, the group of moths (
that has the next best luminescence intensity is marked as | | g)
xit+1 = xit + 0.001 × G|xit − ximin , ximax − xit | + 1 −
the prospector. Through over the iterations, these types of (
|
) 2.g (
| G
)
moths are decreased by the following formula, t
× r1 × Bestp − xi + t
× r2 × Bestglobal − xit
t
G
(( ) ) (26)
nq = round nm − nh × (1 − G) (21)
  The detailed version of MSA is depicted in Mohamed
where G = t
tmax
 ; nh is the prospector moth number, nm is et al. (2017).
total number of moths, t is he current iteration and tmax signi- – Pareto optimal method
fies the highest limit of iteration.   The effective method to solve the multi-objective
Afterward, each prospector moth is positionally updated, optimization problem is the Pareto optimal method,
following a spiral flight path as expressed by following which can lead to the optimal solution set which is gen-
formula, erally known as Pareto optimal solution (Bouchekara
et al. 2016; Warid et al. 2018; Wolpert and Macready
xit+1 = ||xit − xht || × e𝜙 × cos 2𝜋𝜙 × xht (22) 1997). Pareto’s optimality is based on the concept of
advantages and can be determined by considering two
( ) ( ) candidate solutions k1 and k2. k1 say to dominate k2 if
∀h ∈ 1, 2, ..., nh ;i ∈ nh + 1, nh + 1, ..., nq (23)
[ ] the following conditions are met:
where 𝜙 ∈ 1 − t t , 1 is a randomly generated value to { } ( ) ( )
max ∀i ∈ 1, 2, … , xob ∶ Fi k1 ≤ Fi k2 (27)
characterize a spirally oriented path.
{ } ( ) ( )
4. Celestial navigation ∀j ∈ 1, 2, … , xob ∶ Fj k1 ≤ Fj k2 (28)

  According to (27) and (28), there are two possibili-


To enhance the convergence velocity towards the global
ties: one is superior to the other, or neither is dominant.
best, the onlookers are forced by the MSA to have a more
More precisely, when a solution results in a small or
efficient search by means of zooming in on the hot spots of
equal value for the target being considered, it domi-
the prospector. The onlookers are categorized into two parts.
nates in the other method; Solutions that do non-domi-
nate throughout the search space are specified as Pareto
– Gaussian walkGaussian walk
optimal solutions and establish the best front in Pareto.
  The first part of onlooker moths ( ngw ) move through
  The common framework for all population-based
a series of steps Gaussian walk. This type of onlooker
multi-objective algorithms are almost identical. They
moths is updated for the subsequent iteration by the fol-
start the optimization process with multiple candidate
lowing expression,
solutions. In each step of optimization, the non-domi-
[ ]
xit+1 = xit + 𝜀1 + 𝜀2 × Bestglobal
t
{
− 𝜀3 × xit ∀i ∈ 1, 2..., ngw
} nant solution is stored in the repository, and the algo-
rithm attempts to improve it in the next iteration. The
(24)
reason that one algorithm differs from another is that
it uses different methods to enhance the non-dominant

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

( ) ( )
solution. Non-dominant solutions that use random X = x1 , x2 , … xn  . Now, if f Xqo < f (X)(for a minimi-
algorithms should be improved from two perspectives: zation problem), then point can be replaced with Xqo ;
convergence (accuracy) and coverage (distribution) otherwise, the process is continued with X  . Hence, the
(Rahnamayan et al. 2008; Wang et al. 2007). The for- point and its quasi-opposite point are evaluated simulta-
mer refers to the process of improving the accuracy of neously in order to continue with the fitter one.
the non-dominated solutions.
  The main challenge of using random algorithms for 3.2 Proposed EMSA method
multi-objective optimization is that convergence and cov-
erage conflict. If the algorithm focuses only on improving The framework of the EMSA is described to ameliorate
the accuracy of the non-dominant solution, the coverage the performance of the proposed algorithm in this section.
will be poor. Instead, consider only the accuracy of non- The MSA is modified by combining the original MSA with
advantaged solutions that can negatively affect coverage. OBL to present the capability to explore the search domain
– Quasi-opposition-based learning in-depth and quickly achieve the best value. The suggested
  Quasi-opposition-based learning was commenced by algorithm is named EMSA, and the procedure of the combi-
Rahnamayan et al. (2008). It is proved that a quasi-oppo- nation of MSA method and OBL is summarized as follows:
site number is usually closer than an opposite number to Step 1: Initialization the input parameters
the solution.
– Definition of opposite number and quasi-opposite num- • Set the number of pathfinders (population size) and set
ber the max iteration
  If be a real number between [lb, lu] , its opposite num- • Get the function details (lower bound, upper bound and
ber ( xo ) and its quasi-opposite number ( xqo ) are defined variable dimension) and function evaluation
as
Step 2: Oppositional based learning process
x0 = lb + lu − x (29)

[( ) ] • Create oppositional based population using Eq. (31)


lb + lu • Utilize OBL with 50% solution
xqo = rand , (lb + lu − x) (30)
2
  Similarly, this definition can be extended to higher Step 3: Optimization start
dimensions (Wolpert and Macready 1997) as stated in
the next subsection. • Theoppositional based population of moths with the opti-
– Definition of mal fitness value.
( opposite point
) and quasi-opposite point
 Let X = x1 , x2 ,[… xn ] be a point in n-dimensional • Check if pathfinders go out of the search space and bring
space where xi (∈ lbi , lui and )∈ 1, 2, … , n] . The oppo- it back.
site point Xo = xo1 , xo2 , … xon is completely defined by • Update the position of moth.
its components as in (31). • Obtain the fitness value for each oppositional based pop-
ulation.
xoi = lbi + ubi − xi (31)
( ) Step 4: Stopping condition
  The quasi-opposite point Xqo = xqo1 , xqo2 , … xqon is
completely defined by its components as in (32).
• If the maximum iteration is satisfied, the EMSA stop and
[( ) ]
lbi − lui ( ) obtain the optimal parameters value and the best solution.
xqoi = rand , lbi − lui − xi (32) • Else, go to step 2
2

  By employing the definition of quasi-opposite point, After completing the exploration phase of MSA, the OBL
the quasi-opposition-based optimization is defined in the strategy will be applied to maintain 50% of the domain space
following sub-section.
( ) calculated by MSA. This step enables the primary domain
 Let X = x1 , x2 , … xn be a point in n-dimensional space to quickly reach the best value and repair out-of-range
space where i.e. a candidate solution. Assume f (⋅) is a values. The proposed method is presented in Algorithm 1 as
fitness function which is used to measure the candidate’s shown in Fig. 1.
fitness. According
( to the definition
) of the quasi-opposite
point Xqo = xqo1 , xqo2 , … xqon is the quasi-opposite of

13
B. Bentouati et al.

Algorithm 1: Pseudocode for EMSA Algorithm.


1) Declaration of parameters- search agents’ number, maximum iteration number tmax, number of total pathfinders,
get the function details (lower bound, upper bound and variable dimension) and function evaluation.
2) Initialize each population and calculate the corresponding target value;
3) Create oppositional based population according equation (31) and utilize OBL with 50% solution;
4) Sort the population according to the best to worst target value, and mark the best populations as a pathfinders;
5) set iteration t=1;
6) while t tmax
7) Start for reconnaissance phase
● The oppositional based population of moths with the optimal fitness value.
● Check if pathfinders go out of the search space and bring it back.
● Update the position of moth.
● Obtain the fitness value for each oppositional based population.
10) end for reconnaissance phase;
11) Update xi using equation (24) and (25)
12) set iteration t=t+1;
13) end while

Fig. 1  Pseudo code algorithm of the proposed EMSA

Table 1  Summery of studied Test system Cases # Economic Technical Environmental Technical


cases
Fuel costs (OF1) P losses (OF2) VD (OF3) Emissions (OF4) L-max (OF5)

IEEE-30 1 √ – √ – –
2 √ – – – √
3 √ √ – – –
4 √ – – – √
5 √ – √ √ –
6 √ √ √ – –
7 √ √ – √ –
IEEE-57 8 √ – – – √
9 √ – √ – –
10 √ – – – √
IEEE-118 11 √ – – – –
12 √ √ – – –

4 Discussion on simulation outcomes PC with i7 processor, 8 GB RAM and 2.60 GHz clock speed.
The size of the population for both techniques is 30. For each
To establish the effectiveness and high performance of simulation instance, the number of separate runs for both
the proposed EMSA, 11 instances are simulated, and the techniques is 30. It is clear from Table 1 that for simula-
outcomes are compared with the same obtained by MSA. tion instances 1–7, the considered test network is the IEEE
Moreover, to authenticate the highly efficient functionality, 30-bus and the rest of simulation instances are valid for
each simulation instances are briefed as given in Table 1. IEEE 57-bus network. It is very significant to state that the
The simulations are done using MATLAB 2012b and in a load flow routine of the studied test networks under study is

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

performed using Newton–Raphson technique that is usefully are considered through the cases 1–4 and the optimal val-
built in MATPOWER (Zimmerman and Murillo-Sanchez ues of control variables and their corresponding objectives
2020) with high proven accuracy. are pointed out in Table 2.
Figure 2 represents the Pareto optimal solution acquired
4.1 IEEE 30‑bus test network for optimizing fuel costs and active power losses. The graph
illustrates the capability of both algorithms in the suggested
In this network, 9 buses are appointed for shunt VAR com- EMSA and MSA algorithms to generate a true and evenly
pensators at buses number of {10, 12, 15, 17, 20, 21, 23, distributed Pareto optimal front. It may be noted that the
24, and 29} where the detailed network data are recorded suggested EMSA algorithm can yield pareto non-dominant
in (Zimmerman and Murillo-Sanchez 2020; Shaheen solutions that are superior to MSA.
et al. 2017a, b). The total capacity of generation systems From Table 2, the absolute outperformance of the sug-
equals 435.0 MW (Bouchekara et al. 2016). The allow- gested EMSA is clear for cases 1 and 2. This derivation is
able bounds of ± 10% are prescribed for the voltage mag- achieved since the proposed EMSA acquires the minimum
nitude of buses and the transformer tap regulations whilst for both considered objectives compared to the classical
the VAR output of the shunt compensators is restricted MSA. For case 1, the proposed EMSA produces fuel costs
by 5 MVAr. Both MSA and proposed EMSA are applied of 803.4286 $/h and voltage deviation of 0.1073 in per unit
to the different multiple objectives that are formulated in which they are lower than the produced MSA objectives
Table 1. Firstly, two simultaneous conflicting objectives with fuel costs of 803.874 $/h and voltage deviation of 0.118

Table 2  Output results for two contradictory objectives using MSA and proposed EMSA

13
B. Bentouati et al.

Fig. 2  Pareto’s best front for


case 2

in per unit. For case 2, the obtained fuel costs of 799.3582 The obtained voltage values using the proposed MSA and
$/h using the proposed EMSA is better than the compared EMSA are shown in Fig. 3. It is cleared that the voltages at
MSA objective with 800.0275 $/h. Not only that, but also all system buses are preserved between the accepted bounda-
the proposed EMSA produces smaller value of L ­ max stability ries for voltages. Also, the voltages using the proposed MSA
of 0.1185 compared to 0.1209 for MSA. For cases 3 and 4, and EMSA are greatly improved in contrast to the initial
very competitive objectives are acquired via the proposed condition. Similarly, Fig. 4 elucidates that the related trans-
EMSA and MSA. The results show that the proposed EMSA mitted MW power through the lines are maintained lower
and MSA algorithms are superior to the previous algorithms than their capacity.
in terms of optimization and feasibility of the solution. It is For solving three objectives (cases 5–7), the obtained
important to note that some previous heuristic optimization results using the proposed MSA and EMSA are staggered
algorithms are not feasible due to voltage violations on the in Table  3. Both techniques provide the power system
network load bus and reactive power production constraint operator with different applicable operating points to mini-
discipline violations at the power generation unit. mize the considered objectives. The applicability of the

Fig. 3  Bus-voltage profiles for cases 1, 3 and 4 for MSA and EMSA

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Fig. 4  Transmitted MW power for cases 1–4 for MSA and EMSA

acquired operating points using the proposed MSA and regulations whilst the VAR output of the shunt compen-
EMSA is clear as their buses voltages and transmission sators is restricted by 30 MVAr. Both MSA and proposed
power are within their permissible bounds as displayed in EMSA are applied to the different multiple objectives that
Figs. 5 and 6, respectively. It is cleared that the voltages are formulated in Table 1. Table 4 depicts their obtained
at all system buses are preserved between the accepted control variables and their corresponding objectives. This
boundaries for voltages. Figure 7 shows the reactive power detailed comparison, for all cases, in Table 4 points out that
variation at different bus for cases 5–7 using the MSA and the numerical results of the OPF problem using the pro-
EMSA. posed EMSA are very significant to that achieved by the
Figure 8 represents the Pareto optimal solution that is standard MSA. This viability is obtained since the proposed
obtained to optimize the three objective functions “fuel EMSA acquires the minimum for both considered objec-
cost, emissions and voltage profile”. Figure 9 shows Pareto tives. For case 1, the proposed EMSA produces fuel costs
optimal solution is presented for three objectives called fuel of 41,666.2449 $/h which are lower than the produced MSA
cost minimization, power loss minimization and the voltage fuel costs of 41,672.2992 $/h. For case 2, the obtained fuel
profile enhancement. From these figures, well distributed costs and voltage deviation of 41,712.5095 $/h and 0.671
solutions are obtained for three objective cases that are opti- in per unit, respectively using the proposed EMSA which
mized simultaneously. they are lower than the produced MSA objectives with fuel
costs of 41,724.5194 $/h and voltage deviation of 0.7174 in
4.2 IEEE 57‑bus test network per unit. Similar results are achieved in case 3; the obtained
fuel costs of 41,667.5279 $/h using the proposed EMSA is
In the IEEE 57-bus test network, 7 committed generators are better than the compared MSA objective with 41,675.7775
interconnected by 80 transmission lines with 17 installed tap $/h. Not only that, but also the proposed EMSA produces
changing transformers. Also, 3 buses are appointed for VAR smaller value of ­Lmax stability of 0.2756 compared to 0.2794
compensators at buses {15, 25, 53} where the detailed net- for MSA. Nevertheless, the feasibility of the completed oper-
work data are recorded in (Zimmerman and Murillo-Sanchez ating points using the proposed MSA and EMSA is clear
2020). The allowable bounds of ± 10% are prescribed for as their buses voltages are within their permissible bounds
the voltage magnitude of buses and the transformer tap as displayed in Fig. 10. The reactive power outputs of the

13
B. Bentouati et al.

Table 3  Output results for three contradictory objectives using MSA and proposed EMSA
Case 5 Case 6 Case 7
Control variables
EMSA MSA EMSA MSA EMSA MSA
PG1 (MW) 160.6648 160.1998 106.7438 100.1751 100.9424 98.2925
PG2 (MW) 51.7890 51.4499 53.3298 65.3112 58.7184 58.5700
PG5 (MW) 21.0332 21.2662 39.6176 33.0547 37.0949 38.6194
PG8 (MW) 26.7234 26.4903 35.0000 34.1145 35.0000 34.9460
PG11 (MW) 15.5933 15.0396 30.0000 24.1860 29.4863 29.9888
PG13 (MW) 16.5662 17.8552 23.6837 31.5229 26.4594 27.2263
V1(p.u) 1.0443 1.0260 1.0399 1.0698 1.1000 1.1000
V2(p.u) 1.0314 1.0116 1.0294 1.0597 1.0914 1.0935
V5(p.u) 1.0155 1.0038 1.0070 1.0316 1.0701 1.0719
V8(p.u) 1.0112 0.9964 1.0091 1.0418 1.0794 1.0804
V11(p.u) 0.9994 1.0861 1.0134 1.0301 1.0995 1.1000
V13(p.u) 0.9735 1.0619 1.0662 1.0284 1.1000 1.0990
Qc10(Mvar) 3.1317 2.4107 1.8181 1.0187 5.0000 3.2904
Qc12(Mvar) 3.9154 3.0612 0.0394 2.4531 0.7299 1.7110
Qc15(Mvar) 4.7637 0.8020 0.4262 4.8742 5.0000 1.3570
Qc17(Mvar) 0.0045 0.2766 4.1123 1.7789 0.4006 3.7957
Qc20(Mvar) 0.8546 1.6496 4.1324 1.8024 4.9844 0.0754
Qc21(Mvar) 4.9546 0.8176 0.6605 4.6161 0.2051 0.0635
Qc23(Mvar) 5.0000 0.0017 0.6038 2.8001 2.7539 0.9964
Qc24(Mvar) 5.0000 4.6733 5.0000 3.2916 5.0000 1.7819
Qc29(Mvar) 1.0938 1.3063 0.0383 4.8094 1.7426 1.2332
T6–9 1.0013 1.0543 0.9974 1.0724 1.0054 1.0254
T6–10 0.9020 0.9049 0.9313 0.9250 0.9246 0.9537
T4–12 0.9160 1.0450 1.0546 1.0463 1.0032 1.0387
T28–27 0.9535 0.9514 0.9489 1.0072 0.9695 0.9833
Fuel cost ($/h) 806.6047 806.7882 858.5696 855.4377 857.3284 863.6009
VD 0.1191 0.1694 0.1879 0.3273 1.8599 1.3998
Lmax 0.1473 0.1470 0.1458 0.1468 0.1274 0.1320
Emission (ton/h) 0.3238 0.3224 0.2331 0.2300 0.2283 0.2255
Ploss 8.9699 8.9010 4.9852 4.9644 4.3014 4.2430

reactive power resources are preserved below their bending SSA (El-Fergany and Hasanien 2020), Tree seed algorithm
limits as presented in Fig. 11. The obtained voltages using TSA (El-Fergany and Hasanien 2018), differential evolution
the proposed MSA and EMSA are greatly improved in con- (DE) (Shaheen et al. 2017a, b), MODE (Shaheen et al. 2016)
trast to the initial condition. and MSA (Mohamed et al. 2017).
Figure 12 shows the Pareto optimal solution achieved The comparison shows the advantages of the proposed
to optimize fuel costs and voltage stability for case 8. The EMSA. The best and feasibility of the solution have been
diagram illustrates the ability of the two algorithms in the demonstrated by several previous algorithms, consequently
proposed EMSA and MSA algorithms to generate a true and demonstrating their global effectiveness and aspiring. Addi-
evenly distributed Pareto optimal front. It may be noted that tionally, the proposed algorithm is superior to MSA in pro-
the proposed EMSA algorithm can produce a pareto non- ducing better compromise solutions. It is important to note
dominant solution that is superior to MSA. Furthermore, the that some previous heuristic optimization algorithms are not
proposed MSA and EMSA are compared with other reported feasible due to voltage violations on the network load bus
algorithms for case 8 in Table 5. This comparison demon- and reactive power production constraint discipline viola-
strates the high effectiveness of the proposed EMSA in mini- tions at the power generation unit.
mizing the fuel costs compared to salp swarm algorithm

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Fig. 5  Bus-voltage profiles for cases 5–7

Fig. 6  Transmitted MW power for cases 5–7 for MSA and EMSA

13
B. Bentouati et al.

Fig. 7  Reactive power resources sin cases 5–7

Fig. 8  Pareto’s best front for


case 5

Fig. 9  Pareto’s best front for


case 7

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Table 4  Output results for two objectives using MSA and EMSA for IEEE 57 bus network
Control variable Case 8 Case 9 Case 10
EMSA MSA EMSA MSA EMSA MSA
PG1 (MW) 142.7824 145.3854 145.9230 146.6199 144.3343 144.3599
PG2 (MW) 92.0760 86.4249 82.5509 76.6906 91.2683 95.4542
PG3 (MW) 44.2990 48.3503 46.7154 43.5575 45.7665 44.8424
PG6 (MW) 70.2550 64.9552 70.1644 77.6419 71.8160 80.4393
PG8 (MW) 460.8321 467.8370 467.2501 472.6534 459.3115 469.6406
PG9 (MW) 99.4497 90.4110 89.0891 81.6286 90.5763 65.7548
PG12 (MW) 355.9357 362.2878 364.9793 368.1983 362.3171 365.3562
V1 (p.u) 1.0716 1.0729 1.0213 1.0146 1.0695 1.0719
V2 (p.u) 1.0753 1.0758 1.0267 1.0186 1.0720 1.0750
V3 (p.u) 1.0597 1.0643 1.0147 1.0089 1.0595 1.0627
V6 (p.u) 1.0625 1.0648 1.0255 1.0168 1.0649 1.0629
V8 (p.u) 1.0779 1.0821 1.0477 1.0415 1.0789 1.0767
V9 (p.u) 1.0713 1.0733 1.0290 1.0277 1.0714 1.0696
V12 (p.u) 1.0584 1.0601 1.0102 1.0050 1.0574 1.0588
Qc18 (Mvar) 11.9310 0.5662 2.4658 3.5493 10.4149 5.7283
Qc25 (Mvar) 17.4150 9.4949 17.1494 12.7784 11.7961 13.8933
Qc53 (Mvar) 12.2484 6.3154 17.3539 2.7186 11.6739 8.5595
T4–18 1.0964 1.0195 1.0391 0.9727 0.9992 0.9466
T4–18 1.0048 1.0221 0.9305 0.9306 1.0012 1.0232
T21–20 1.0662 0.9995 0.9799 1.0151 1.0234 0.9994
T24–25 1.0146 0.9339 0.9670 0.9845 1.0556 0.9272
T24–25 1.0838 0.9984 1.0544 1.0241 0.9383 1.0868
T24–26 1.0375 1.0257 0.9970 1.0144 1.0369 1.0244
T7–29 1.0037 1.0008 1.0108 0.9741 1.0031 0.9959
T34–32 0.9942 0.9741 0.9332 0.9288 0.9860 0.9849
T11–41 0.9158 0.9009 0.9009 0.9028 0.9669 0.9686
T15–45 0.9902 0.9923 0.9471 0.9593 0.9869 0.9888
T14–46 0.9776 0.9881 0.9525 0.9557 0.9730 0.9820
T10–51 0.9916 1.0048 0.9837 0.9816 0.9901 0.9915
T13–49 0.9380 0.9645 0.9243 0.9198 0.9410 0.9540
T11–43 1.0268 1.0393 0.9878 0.9448 1.0516 1.0015
T40–56 0.9784 0.9785 0.9566 0.9982 1.0517 0.9818
T39–57 1.0172 0.9867 0.9297 0.9605 1.0060 0.9614
T9–55 1.0241 1.0296 1.0185 0.9916 1.0352 1.0124
Fuel cost ($/h) 41666.2449 41672.2992 41712.5095 41724.5194 41667.5279 41675.7775
VD 1.6710 1.3506 0.6710 0.7174 1.6305 1.6022
Lmax 0.2840 0.2831 0.2944 0.2944 0.2756 0.2794
Emission (ton/h) 1.3482 1.3832 1.3872 1.4145 1.3542 1.3990
Ploss 14.8301 14.8519 15.8723 16.1901 14.5902 15.0476

4.3 IEEE 118‑bus test network is added for SOF in Table 7. The obtained fuel costs of
135,262.57$/h using the proposed EMSA is better than the
To prove the scalability of the proposed EMSA, the third compared MSA objective with 135,310.84 $/h. Also, the
system, IEEE 118-bus test system, is used. In this network, EMSA has the better performance compared with other opti-
54 committed generators are interconnected by 200 trans- mization algorithms like as DE (Bouchekara et al. 2016),
mission lines where the detailed network data are recorded CBO (Bouchekara et al. 2016), ICBO (Bouchekara et al.
in Elattar et al. (2020). The total generation of this network 2016), ECBO (Bouchekara et al. 2016), BBO (Bouchekara
equals 9966.2 MW and its main characteristics are given et al. 2016) and ABC (Bouchekara et al. 2016) algorithms
(Bouchekara et al. 2016). that were reported in Table 7 as 158,920.81, 135,297.22,
The proposed EMSA is developed for single and multi- 136,353.31, 135,582.15, 135,640.65 and 135,759.015 $/h.
objective cases, cases 11 and 12, in Tables 6 and 7. Table 6 Table 8 depicts their obtained control variables and their
depicts their obtained control variables and their correspond- corresponding objectives for case 12, the costs of produc-
ing objectives for case 11, the costs of production are mini- tion and power loess minimization are optimized simulta-
mized as the primary objective function. An assessment of neously. An assessment of the proposed EMSA with the
the proposed EMSA with the MSA method in the literature MSA method in the literature is added for MOF in Table 9.

13
B. Bentouati et al.

Fig. 10  Bus-voltage profiles for cases 8–10

Fig. 11  Reactive power varia-


tions for cases 8–10

Figure 13 shows the Pareto’s Best Front for case 12. Well 5 Conclusions
distributed solutions are obtained. For case 12, the proposed
EMSA produces fuel costs of 142,898,8$/h and power losses This paper specifically confirms the superior performance
of 42,5378 in MW which they are lower than the produced of proposed EMSA over MSA in view of a small as well as
MSA objectives with fuel costs of 142,878,6$/h and power a large test network. The effectiveness of proposed EMSA in
losses of 40,04,927 in MW. minimizing various challenging objectives of two different
test networks which includes different simulation instances.
From the different convergence curves plotted in differ-
ent figures, it is confirmed that proposed EMSA has fast

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Fig. 12  Pareto’s best front for


case 9

Table 5  Comparison study for Algorithms Fuel cost ($/h) Algorithms Fuel cost ($/h)
case 8
Proposed EMSA 41,666.2449 DE (2017) (Shaheen et al. 2017a, b) 41,682
MSA 41,672.2992 MODE (2016) (Shaheen et al. 2016) 41,683
SSA (2020) (El-Fer- 41,672.3 MSA (2016) (Mohamed et al. 2017) 41,673.72
gany and Hasanien.
2020)
TSA (2018) (El-Fer- 41,685.07
gany and Hasanien.
2018)

Bold values refer to the objective functions

convergence ability than MSA. Also, from the bus voltage Even the proposed solution meth has good solution
profiles shown in this paper, it is strongly confirmed that capability for all studied cases with acceptable merits. But
proposed EMSA is highly suitable for constrained operation. it needs more search space concentration and parametric
The voltage for all studied cases as well as the transmitted study to fine choice of the algorithm control variables. In
powers among transmission systems are preserved within the future work, it will be intended to enhance the explora-
the operating boundaries. Not only that, from the statisti- tion performance to reach more acceptable levels at lower
cal analysis done in this paper, it is also intensely verified computation time. To extend this work, two viewpoints
and authenticated that in terms of statistical robustness and can be considered. The first one considers the improve-
dependability, proposed EMSA proves its highly sophisti- ments in the problem formulation side while the other
cated performance. The proposed EMSA leads to enhanced one enhances the solution quality of the proposed opti-
operation characteristics in different operation dimensions mizer. In the first side, the future work that is considered
technical and economical at acceptable emissions levels. as extension of the current study will be cover solving the

13
B. Bentouati et al.

Table 6  Output results of EMSA for IEEE 118 bus network (case 11)
Variable Value Variable Value Variable Value Variable Value Variable Value

PG1 39.80624 PG65 174.5769 VG1 1.043904 VG65 1.048995 Qc5 10.99141
PG4 32.03278 PG66 230.7042 VG4 1.057815 VG66 1.068083 Qc34 6.748959
PG6 45.89895 PG70 36.60536 VG6 1.049087 VG69 1.048305 Qc37 0.182296
PG8 57.84287 PG72 30.57019 VG8 1.023427 VG70 1.030294 Qc44 6.069339
PG10 310.8617 PG73 35.90984 VG10 0.982941 VG72 1.083423 Qc45 7.798877
PG12 64.82168 PG74 31.14104 VG12 1.045192 VG73 1.026041 Qc46 20.94922
PG15 42.58369 PG76 38.18126 VG15 1.031873 VG74 1.010062 Qc48 7.669938
PG18 43.06132 PG77 65.68069 VG18 1.025931 VG76 1.048971 Qc74 17.54638
PG19 30.32197 PG80 325.926 VG19 1.037818 VG77 1.052246 Qc79 21.89991
PG24 31.07873 PG85 30.52296 VG24 1.063867 VG80 1.056575 Qc82 2.559621
PG25 115.4123 PG87 31.2 VG25 1.068494 VG85 1.057604 Qc83 5.676478
PG26 185.4341 PG89 390.1434 VG26 1.084135 VG87 1.01186 Qc105 20.60822
PG27 39.64801 PG90 40.5811 VG27 1.069689 VG89 1.080443 Qc107 11.71019
PG31 32.13977 PG91 32.93069 VG31 1.045221 VG90 1.037809 Qc110 1.248113
PG32 31.88483 PG92 31.52139 VG32 1.052909 VG91 1.030789 T8 0.99985
PG34 45.5834 PG99 41.05055 VG34 1.001533 VG92 1.066463 T32 1.083448
PG36 48.02503 PG100 166.9348 VG36 0.991404 VG99 1.075618 T36 0.974909
PG40 34.24176 PG103 46.22992 VG40 1.008061 VG100 1.076266 T51 1.009621
PG42 39.72246 PG104 32.09886 VG42 0.993858 VG103 1.055975 T93 0.914551
PG46 40.01474 PG105 43.89417 VG46 1.07415 VG104 1.033493 T95 0.948674
PG49 178.9462 PG107 49.43733 VG49 1.062534 VG105 1.029837 T102 0.969954
PG54 57.44336 PG110 40.63367 VG54 1.06105 VG107 1.054453 T107 1.013391
PG55 30.74573 PG111 41.37919 VG55 1.054407 VG110 1.027384 T127 1.005025
PG56 34.89669 PG112 31.70861 VG56 1.060626 VG111 1.016988 Fuel cost ($/h) 13,5262.57
PG59 113.8223 PG113 76.40707 VG59 1.058186 VG112 1.02457
PG61 86.9607 PG116 50.18298 VG61 1.027913 VG113 1.03786 loss (MW) 53.447
PG62 34,73304 VG62 1,039259 VG116 1,059685

Table 7  Comparison study for case 11


Objectives MSA Proposed ICBO CBO ECBO DE (Bouchek- ABC BBO
EMSA (Bouchekara (Bouchekara (Bouchekara ara et al. 2016) (Bouchekara (Bouchekara
et al. 2016) et al. 2016) et al. 2016) et al. 2016) et al. 2016)

Fuel cost, 135,310.84 135,262.57 136,353.31 135,297.22 135,582.15 158,920.81 135,759.015 135,640.65
($/h)

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Table 8  Output results for two objectives EMSA for IEEE 118 bus network (case 12)
Variable Value Variable Value Variable Value Variable Value Variable Value

PG1 37.03309 PG65 192.9355 VG1 1.082261 VG65 1.092747 Qc5 12.76325
PG4 53.21267 PG66 234.8078 VG4 1.084874 VG66 1.082724 Qc34 10.61849
PG6 37.6815 PG70 46.14463 VG6 1.086741 VG69 1.084618 Qc37 0.636957
PG8 49.56929 PG72 52.87319 VG8 1.067515 VG70 1.079044 Qc44 10.48639
PG10 220.6996 PG73 44.23424 VG10 1.072732 VG72 1.082352 Qc45 13.5909
PG12 66.14814 PG74 57.11745 VG12 1.077271 VG73 1.081884 Qc46 6.352021
PG15 37.15012 PG76 70.72791 VG15 1.080415 VG74 1.07946 Qc48 9.505847
PG18 47.144 PG77 53.12932 VG18 1.083175 VG76 1.084461 Qc74 1.577404
PG19 45.97165 PG80 330.0685 VG19 1.075181 VG77 1.092229 Qc79 5.727518
PG24 47.44588 PG85 43.67001 VG24 1.093157 VG80 1.063089 Qc82 5.043068
PG25 111.6945 PG87 38.60344 VG25 1.066948 VG85 1.075635 Qc83 0.419336
PG26 159.308 PG89 250.7708 VG26 1.075449 VG87 1.079903 Qc105 11.09776
PG27 52.12908 PG90 37.21778 VG27 1.082963 VG89 1.074208 Qc107 6.72475
PG31 37.16082 PG91 46.66734 VG31 1.074551 VG90 1.084929 Qc110 1.497216
PG32 58.72713 PG92 36.16116 VG32 1.085321 VG91 1.084137 T8 1.031751
PG34 53.24964 PG99 50.73784 VG34 1.079471 VG92 1.078493 T32 1.06339
PG36 63.74147 PG100 154.3336 VG36 1.069669 VG99 1.080064 T36 1.039911
PG40 65.0576 PG103 48.24028 VG40 1.081862 VG100 1.086274 T51 1.02342
PG42 50.36039 PG104 45.50073 VG42 1.086592 VG103 1.071546 T93 1.018808
PG46 41.67633 PG105 46.68615 VG46 1.086421 VG104 1.063657 T95 1.029814
PG49 171.8697 PG107 41.41623 VG49 1.082832 VG105 1.076544 T102 1.038015
PG54 77.12713 PG110 52.30023 VG54 1.077964 VG107 1.078594 T107 1.024116
PG55 38.06695 PG111 50.56897 VG55 1.071153 VG110 1.073905 T127 1.047655
PG56 57.27306 PG112 40.99431 VG56 1.078319 VG111 1.093063 Fuel cost ($/h) 142,878.6
PG59 109.4155 PG113 39.7787 VG59 1.07658 VG112 1.08274
PG61 124.733 PG116 52.72987 VG61 1.089757 VG113 1.090847 Loss (MW) 40.04927
PG62 44.48982 VG62 1.081067 VG116 1.085528

Table 9  Comparison study for Objectives MSA Proposed EMSA


case 12 between the MSA and
the proposed EMSA Fuel cost, ($/h) 142,898.8 142,878.6
Power loss, MW 42.5378 40.04927

Fig. 13  Pareto’s best front for


case 12

13
B. Bentouati et al.

optimal power flow problem under emergency events like Duman S (2018) A modified moth swarm algorithm based on an arith-
line and generation outages. Added to that, considering the metic crossover for constrained optimization and optimal power
flow problems. IEEE Access. https​://doi.org/10.1109/ACCES​
high penetration of renewable energy resources with mod- S.2018.28495​99
eling of power electronics components. In the viewpoint El Sehiemy RA, Selim F, Bentouati B, Abido MA (2020) A novel
of optimization area, developing new hybrid version of multi-objective hybrid particle swarm and salp optimization algo-
moth-swarm optimization algorithm with other optimiza- rithm for technical–economical–environmental operation in power
systems. Energy. https​://doi.org/10.1016/j.energ​y.2019.11681​7
tion methods like particle swarm and Jaya optimization Elattar EE, Shaheen AM, Elsayed AM, El-Sehiemy RA (2020) Opti-
methods to enhance the solution quality. mal power flow with emerged technologies of voltage source con-
verter stations in meshed power systems. IEEE Access 8:166963–
166979. https​://doi.org/10.1109/ACCES​S.2020.30229​19
El-Ela AAA, El-Sehiemy RA, Shaheen AM, Ellien AR (2020) Optimal
References allocation of distributed generation units correlated with fault cur-
rent limiter sites in distribution systems. IEEE Syst J. https​://doi.
Abaci K, Yamacli V (2016) Differential search algorithm for solving org/10.1109/JSYST​.2020.30090​28
multi-objective optimal power flow problem. Int J Electr Power El-Fergany AA, Hasanien HM (2018) Tree-seed algorithm for solving
Energy Syst. https​://doi.org/10.1016/j.ijepe​s.2015.12.021 optimal power flow problem in large-scale power systems incor-
Abbasi M, Abbasi E, Mohammadi-Ivatloo B (2020) Single and multi- porating validations and comparisons. Appl Soft Comput J. https​
objective optimal power flow using a new differential-based har- ://doi.org/10.1016/j.asoc.2017.12.026
mony search algorithm. J Ambient Intell Human Comput. https​ El-Fergany AA, Hasanien HM (2020) Salp swarm optimizer to solve
://doi.org/10.1007/s1265​2-020-02089​-6 optimal power flow comprising voltage stability analysis. Neural
Abdul-hamied DT, Shaheen AM, Salem WA, Gabr WI, El-Sehiemy Comput Appl. https​://doi.org/10.1007/s0052​1-019-04029​-8
RA (2020) Equilibrium optimizer based multi dimensions Elsakaan AA, El-Sehiemy RA-A, Kaddah SS, Elsaid MI (2018) Eco-
operation of hybrid AC/DC grids. Alex Eng J. https​: //doi. nomic power dispatch with emission constraint and valve point
org/10.1016/j.aej.2020.08.043 loading effect using moth flame optimization algorithm. Adv Eng
Abou El Ela AA et al (2017) Application of the crow search algo- Forum 28:139–149. https​://doi.org/10.4028/www.scien​tific​.net/
rithm for economic environmental dispatch. In: 2017 Nine- aef.28.139
teenth International Middle East Power Systems Conference Elsakaan AA, El-Sehiemy RA, Kaddah SS, Elsaid MI (2020) Opti-
(MEPCON), Cairo, pp 78–83. https​://doi.org/10.1109/MEPCO​ mal economic–emission power scheduling of RERs in MGs
N.2017.83011​66 with uncertainty. IET Gener Transm Distrib 14(1):37–52. https​
Abou El-Ela AA, El-Sehiemy RAA, Mouwafi MT, Salman DAF (2018) ://doi.org/10.1049/iet-gtd.2019.0739
Multi-objective fruit fly optimization algorithm for OPF solution El-Sattar SA, Kamel S, El Sehiemy RA, Jurado F, Yu J (2019) Sin-
in power system. In: 2018 Twentieth International Middle East gle- and multi-objective optimal power flow frameworks using
Power Systems Conference (MEPCON). IEEE, pp 254–259 Jaya optimization technique. Neural Comput Appl 31(12):8787–
Attia AF, El Sehiemy RA, Hasanien HM (2018) Optimal power flow 8806. https​://doi.org/10.1007/s0052​1-019-04194​-w
solution in power systems using a novel Sine-Cosine algorithm. El-Sehiemy RA, El-Hosseini MA, Hassanien AE (2013) Multiobjec-
Int J Electr Power Energy Syst. https​://doi.org/10.1016/j.ijepe​ tive real-coded genetic algorithm for economic/environmental
s.2018.01.024 dispatch problem. Stud Inform Control. https​://doi.org/10.24846​
Barakat AF, El-Sehiemy RA-A, Elsaid M (2020) Close accord on /v22i2​y2013​01
particle swarm optimization variants for solving non-linear opti- Ghasemi M, Ghavidel S, Ghanbarian MM et al (2014) Multi-objec-
mal reactive power dispatch problem. Int J Eng Res Afr 46:88– tive optimal power flow considering the cost, emission, volt-
105. https​://doi.org/10.4028/www.scien​tific​.net/jera.46.88 age deviation and power losses using multi-objective modi-
Bentouati B, Javaid MS, Bouchekara HREH, El-Fergany AA (2019) fied imperialist competitive algorithm. Energy. https​: //doi.
Optimizing performance attributes of electric power systems org/10.1016/j.energ​y.2014.10.007
using chaotic salp swarm optimizer. Int J Manag Sci Eng Ghasemi M, Ghavidel S, Gitizadeh M, Akbari E (2015) An improved
Manag. https​://doi.org/10.1080/17509​653.2019.16771​97 teaching-learning-based optimization algorithm using Lévy muta-
Biswas PP, Suganthan PN, Amaratunga GAJ (2017) Optimal power tion strategy for non-smooth optimal power flow. Int J Electr
flow solutions incorporating stochastic wind and solar power. Power Energy Syst. https​://doi.org/10.1016/j.ijepe​s.2014.10.027
Energy Convers Manag. https ​ : //doi.org/10.1016/j.encon​ Khalilpourazari S, Khalilpourazary S (2019) An efficient hybrid algo-
man.2017.06.071 rithm based on water cycle and moth-flame optimization algo-
Bouchekara HREH, Chaib AE, Abido MA, El-Sehiemy RA (2016) rithms for solving numerical and constrained engineering opti-
Optimal power flow using an Improved colliding bodies optimi- mization problems. Soft Comput. https​://doi.org/10.1007/s0050​
zation algorithm. Appl Soft Comput J. https​://doi.org/10.1016/j. 0-017-2894-y
asoc.2016.01.041 Luo GX, Semlyen A (1989) Hydrothermal optimal power flow based
Coello CAC, Lechuga MS (2002) MOPSO: a proposal for multiple on a combined linear and nonlinear programming methodology.
objective particle swarm optimization. In: Proceedings of the con- IEEE Trans Power Syst. https​://doi.org/10.1109/59.19382​6
gress on evolutionary computation, Honolulu, HI, pp 1051–1056 Medina MA, Das S, CoelloCoello CA, Ramírez JM (2014) Decompo-
Dabba A, Tari A, Meftali S (2020) Hybridization of Moth flame opti- sition-based modern metaheuristic algorithms for multi-objective
mization algorithm and quantum computing for gene selection optimal power flow—a comparative study. Eng Appl Artif Intell.
in microarray data. J Ambient Intell Human Comput. https​://doi. https​://doi.org/10.1016/j.engap​pai.2014.01.016
org/10.1007/s1265​2-020-02434​-9 Mirjalili S, Saremi S, Mirjalili SM, Coelho LDS (2016) Multi-
Daryani N, Hagh MT, Teimourzadeh S (2016) Adaptive group objective grey wolf optimizer: a novel algorithm for multi-cri-
search optimization algorithm for multi-objective optimal power terion optimization. Expert Syst Appl. https​://doi.org/10.1016/j.
flow problem. Appl Soft Comput J. https​://doi.org/10.1016/j. eswa.2015.10.039
asoc.2015.10.057

13
An enhanced moth‑swarm algorithm for efficient energy management based multi dimensions OPF…

Mohamed AAA, Mohamed YS, El-Gaafary AAM, Hemeida AM Wang H, Liu Y, Zeng S, Li C (2007) Opposition-based particleswarm
(2017) Optimal power flow using moth swarm algorithm. Electr algorithm with Cauchy mutation. In: Proceedings of the IEEE
Power Syst Res. https​://doi.org/10.1016/j.epsr.2016.09.025 Congress on Evolutionary Computation, pp 4750–4756
Połap D (2016) Neuro-heuristic voice recognition. In: Proceedings of Wang GG, Deb S, Coelho LDS (2015) Earthworm optimization algo-
the 2016 federated conference on computer science and informa- rithm: a bio-inspired metaheuristic algorithm for global optimiza-
tion systems, pp 487–490 tion problems. Int J Bioinspired Comput. https​://doi.org/10.1504/
Połap D, Woźniak M, Damaševičius R, Maskeliūnas R (2019) Bio- ijbic​.2015.10004​283
inspired voice evaluation mechanism. Appl Soft Comput 80:342– Wang GG, Deb S, Gao XZ, Dos Santos CL (2016) A new metaheuristi-
357. https​://doi.org/10.1016/j.asoc.2019.04.006 coptimisation algorithm motivated by elephant herding behaviour.
Rahnamayan S, Tizhoosh HR, Salama MM (2008) Opposition- Int J Bioinspired Comput 8(6):394–409
based differential evolution. Stud Comput Intell. https​://doi. Wang GG, Deb S, Cui Z (2019) Monarch butterfly optimization. Neural
org/10.1007/978-3-540-68830​-3_6 Comput Appl. https​://doi.org/10.1007/s0052​1-015-1923-y
Rao NT, Sankar MM, Rao SP, Rao BS (2020) Comparative study of Warid W, Hizam H, Mariun N, Abdul Wahab NI (2018) A novel
Pareto optimal multi objective cuckoo search algorithm and multi quasi-oppositional modified Jaya algorithm for multi-objective
objective particle swarm optimization for power loss minimization optimal power flow solution. Appl Soft Comput J. https​://doi.
incorporating UPFC. J Ambient Intell Human Comput. https:​ //doi. org/10.1016/j.asoc.2018.01.039
org/10.1007/s1265​2-020-02142​-4 Wolpert DH, Macready WG (1997) No free lunch theorems
Shafik MB, Chen H, Rashed GI, El-Sehiemy RA (2019) Adaptive multi for optimization. IEEE Trans Evol Comput. https ​ : //doi.
objective parallel seeker optimization algorithm for incorporating org/10.1109/4235.58589​3
TCSC devices into optimal power flow framework. IEEE Access Wu Z, Shen D, Shang M, Qi S (2019) Parameter Identification of
7:36934–36947. https​://doi.org/10.1109/acces​s.2019.29052​66 single-phase inverter based on improved moth flame opti-
Shaheen AM, El-Sehiemy RA (2020) Optimal co-ordinated alloca- mization algorithm. Electr Power Compon Syst. https​://doi.
tion of distributed generation units/capacitor banks/voltage regu- org/10.1080/15325​008.2019.16079​22
lators by EGWA. IEEE Syst J. https​://doi.org/10.1109/JSYST​ Yuan G (2006) Research and application of non-dominated sorting
.2020.29866​47 genetic algorithm (NSGA). Zhejiang University, Hangzhou
Shaheen AM, El-Sehiemy RA, Farrag SM (2016) Solving multi-objec- Zimmerman RD, Murillo-Sanchez CE (2020) Matpower [Software].
tive optimal power flow problem via forced initialised differen- https​://matpo​wer.org
tial evolution algorithm. IET Gener Transm Distrib. https​://doi.
org/10.1049/iet-gtd.2015.0892 Publisher’s Note Springer Nature remains neutral with regard to
Shaheen AM, El-Sehiemy RA, Farrag SM (2017a) A reactive power jurisdictional claims in published maps and institutional affiliations.
planning procedure considering iterative identification of VAR
candidate buses. Neural Comput Appl 31(3):653–674. https:​ //doi.
org/10.1007/s0052​1-017-3098-1
Shaheen AM, Farrag SM, El-Sehiemy RA (2017b) MOPF solution
methodology. IET Gener Transm Distrib. https:​ //doi.org/10.1049/
iet-gtd.2016.1379

13

You might also like