See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/351807120

Single and Multi-Objective Optimization Algorithms

Article · July 2019

DOI: 10.21608/ijaes.2019.169954

CITATIONS READS

3 36

3 authors, including:

Ahmed Ewais Ashraf Mohamed Hemeida

Aswan University Aswan University
10 PUBLICATIONS 21 CITATIONS 125 PUBLICATIONS 1,469 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Energy-conversion View project

Micro-Grids for rural areas in Egypt View project

All content following this page was uploaded by Ahmed Ewais on 08 October 2022.

The user has requested enhancement of the downloaded file.

 International Journal of Applied Energy Systems, Vol. 1, No. 2, July 2019

Single and Multi-Objective Optimization Algorithms

R. M. Kotb, A. M. Ewais, A. M. Hemeida
Electrical Engineering Department
Faculty of energy Engineering, Aswan University
Aswan, Egypt

Abstract— Hybrid and multi optimization techniques are such as: Multi-objective Grey Wolf Optimizer [14], Multi-
used extensively for solving optimal power flow problems. In objective Bee Algorithm [15], Multi-objective Particle
this paper, particle swarm optimization (PSO), is incorporated Swarm Optimization [16, 17], Multi-objective Bat Algorithm
with grey wolf optimization, (GWO) to form hybrid algorithm [18], Non-dominated Sorting Genetic Algorithm [19-21].
called HPSOGWO and using the multi-objective optimization
of this algorithm, which called MO-HPSOGWO and Hybrid algorithms are a combination between two or
comparing them. The HPSOGWO and MO-HP SOGWO are more algorithms such as hybrid Particle Swarm Optimization
implemented to enhance the optimal power flow solution of with Gravitational Search Algorithm (PSOGSA) [22,23], and
ieee-30 bus system. Five objective functions (OPF) optimizing Particle Swarm Optimization with Dragonfly Algorithm
separately by HPSOGWO and simultaneously in a single run (PSODA) [24], and Particle Swarm Optimization with
by MO-HPSOGWO. The Matlab software is used to solve the Firefly Algorithm (PSOFA) [25], and Particle Swarm
system. Optimization with Multi Verse Optimizer (PSOMVO) [26].
Keywords— Hybrid and multi optimization techniques, Multi
optimization techniques, Single objective functions (OPF). Regarding to the No-Free Lunch (NFL) theorem, there
is no optimization technique for solving all optimization
I. INTRODUCTION problems [27] making researchers are able to formulate new
Optimization techniques are the best techniques for algorithms or improve it.
optimal results for any problem in any field. Traditional
II. PROBLEM FORMULATION
optimization techniques are less used due to its key
disadvantage: local optimal solution [1, 2]. The modern A. Single objective OPF Problem Formulation
optimization techniques are mostly used. It is adjusted by a The mathematical formulation of the OPF problem is
set of random candidate solutions for a given problem in presented as a non-linearly constrained optimization
order to progress them over an adjusted number of steps. problem:
Optimization of real world problems requires handling
various difficulties such as: multi-objectives [3], 𝑢 = [𝑄𝐶𝑇 𝑇𝐶 𝑇 𝑉𝐶𝑇 𝑃𝐶𝑇 ]
uncertainties [4], constraints [5], false global solutions [6], Where:
local solutions [7], despite of the advantages of these u = the control variables
techniques. 𝑄𝐶 = reactive power supplied by all shunt reactors
A multi-objective optimization problem consists of a TC = magnitudes of transformer load tap changer
several objective functions more than one function [8]. There 𝑉𝐺 = voltage magnitude at generator buses
is a several non-dominated solutions for a multi-objective 𝑃𝐺 = active power generated at generator buses
problem because of the problems nature [9]. On the other 𝑥 = [𝑉𝐿𝑇 𝜃 𝑇 𝑃𝑆𝐺 𝑄𝐺𝑇 ]
hand, a single objective problem is featured by only one Where:
global (best) solution. x = the state variables
There are two basic methods to solve multi-objective 𝑉𝐿𝑇 = voltage magnitude at load buses
optimization problem: a posteriori versus a priori [3, 10]. For 𝜃 = voltage angles of all buses excluding the slack
a priori method, we converted a multi-objective optimization bus
problem to a single objective by a set of weights. In this 𝑃𝑆𝐺 = active power generated at the slack bus
method, an algorithm must be run multiple times to defined 𝑄𝐺 = reactive power generated at all generator units
the Pareto optimal set, which is considered as the main 𝑁𝐿 = load buses number
disadvantages of this method. And some special Pareto 𝑁𝐺 = generator buses number.
optimal fronts cannot be defined with this method [11-13]. Optimization problem as OPF problem is presented as
At A posterior method, multi-objective formulation of a maximizing or minimizing objective function to be subjected
multi-objective optimization problem is maintained to to a set of equality and inequality constraints.
determine the Pareto optimal set by one run. Moreover, any B. Multi-Objective OPF problem Formulation
type of Pareto front can be defined with this method. The
main disadvantage of this method is that it has higher Multi-Objective optimization problem consists of several
computational cost and managing multiple objectives at the objective functions optimized simultaneously [28, 29]. The
same time. There are other popular optimization methods Multi-Objective OPF problem is presented as:

77
Faculty of Energy Engineering - Aswan University - Aswan - Egypt

Minimize f(x) where power at each load bus in a system and the Ploss must be
𝑓(𝑥) = [𝑓1 (𝑥)𝑓2 (𝑥)𝑓3 (𝑥)𝑓4 (𝑥)𝑓5 (𝑥)] more than zero.
𝑁 𝑁𝐺 𝑁𝑑

Subject to: 𝐹3 = 𝑃𝐿 = ∑ 𝑃𝑖 = ∑ 𝑃𝐺𝑖 − ∑ 𝑃𝑑𝑖 (3)

𝑥∈𝑋 𝑖=1 𝑖=1 𝑖=1
Where X is a feasible region: where
𝑋 = [𝑥 ∶ 𝑥 ∈ 𝑅𝑛 , 𝑔𝑖 (𝑥) ≤ 0, 𝑥𝑗 ≥ 0 ∀𝑖 , 𝑗] 𝑃𝑖 real power in each bus:
where 𝑃𝑑𝑖 the demand real power
R = set of real numbers 𝑁𝑑 is the load buses number in a network.
𝑔𝑖 (𝑥)= set of constraints 𝑁𝐺 is the generator buses number.
x = set of decision variables.
III. PROBLEM OBJECTIVES D. Reactive Power Transmission Loss Minimization
A. Fuel Cost Minimization Reactive Power Transmission Loss Minimization lead to
voltage stability margin (VSM) increasing and enhancing
The economic distribution of a load defined among the and guarantee good transportation real power from sources to
different generators of a system, the variable operating costs
sinks in a network and the Qloss can be positive or negative
must be presented as active power generated at each
value.
generator in a system. The fuel cost is the essential cost in a 𝑁
thermal or nuclear unit. Then the fuel cost must be presented
as active power generated at each generator in a system. 𝐹4 = 𝑄𝑙𝑜𝑠𝑠 = ∑ 𝑄𝑖 = Σ𝑄𝐺𝑖 − Σ𝑄𝑑𝑖 (4)
Other costs, such as the operation and maintenance costs, can 𝑖=1
also be presented as the power output. Fixed costs, such as E. Reactive Power Reserve Margin Maximization
the capital cost, depreciation ..., are not containing in the fuel
cost. Reactive Power Reserve Margin Maximization leads to
minimize reactive power losses and to improve voltage
A quadratic function of active power generated by each unit
stability and voltage stability under increased load condition
in a system approximates the fuel cost curve as:
or system disturbances. The fast reactive sources are
𝑁𝐺 2
𝐹1 = ∑𝑖=1 (𝑎𝑖 + 𝑏𝑖 𝑃𝐺𝑖 + 𝑐𝑖 𝑃𝐺𝑖 )(𝑅𝑠./ℎ) (1) generators, synchronous condensers and FACTs.
𝑁
𝐺 𝑄𝑖2
𝑝5 = ∑𝑖=1 [ ] (5)
where: 𝑄𝑖 𝑚𝑎𝑥
𝑃𝐺𝑖 is the active power generated at an ith generator in a
system IV. PROBLEM CONSTRAINTS
𝑁𝐺 is the generators number in a system A. Equality constraints
𝑎𝑖 , 𝑏𝑖 , 𝑐𝑖 fuel cost coefficients of an ith generator in a Equality constraints condition can be presented as:
system. 𝑁𝐺

∑ 𝑃𝐺𝑖 − 𝑃𝐷 − 𝑃𝑙𝑜𝑠𝑠 = 0 (6)

B. Emission Minimization 𝑖=1

The function of emission can be aggregated of all types of Where 𝑃𝐷 and 𝑃𝑙𝑜𝑠𝑠 are demand power and power losses,
emission considered, such as NOx, SO2, thermal emission, respectively.
etc., As shown in this equation, the amount of emissions is 𝑁𝐺
presented as a function of active power generated at each
∑ 𝑄𝐺𝑖 − 𝑄𝐷 − 𝑄𝑙𝑜𝑠𝑠 = 0 (7)
generator in a system, which is the sum of quadratic and
𝑖=1
exponential functions:
𝑁𝐺
2
𝐹2 = ∑[10−2 ∗ (𝛼𝑖 + 𝛽𝑖 𝑃𝐺𝑖 + 𝛾𝑖 𝑃𝐺𝑖 ) B. Inequality constraints
𝑖=1 -Constraints of generation capacity
+ 𝜀𝑖 exp(𝜆𝑖 𝑃𝐺𝑖 )](𝑡/ℎ) (2) The generator outputs and bus voltage is restricted by min
where: and max limits as:
𝑚𝑖𝑛 𝑚𝑎𝑥
𝛼𝑖 , 𝛽𝑖 , 𝛾𝑖 , 𝜀𝑖 and 𝜆𝑖 are the emission characteristics co- 𝑃𝐺𝑖 ≤ 𝑃𝐺𝑖 ≤ 𝑃𝐺𝑖 (8)
𝑚𝑖𝑛 𝑚𝑎𝑥
efficients of the ith generator. 𝑄𝐺𝑖 ≤ 𝑄𝐺𝑖 ≤ 𝑄𝐺𝑖 (9)
C. Total Real Power Loss Minimization 𝑄𝑖𝑚𝑖𝑛 ≤ 𝑄𝑖 ≤ 𝑄𝑖𝑚𝑎𝑥 (10)
𝑣𝑖𝑚𝑖𝑛 ≤ 𝑣𝑖 ≤ 𝑣𝑖𝑚𝑎𝑥 (11)
The term PL represents the total I2R loss in the transmission
- Constraints of line flow
lines and transformers of the system. From equation (3) total
This constraints can be presented as:
active power loss equal the sum of generated active power at 𝑚𝑎𝑥
each generator in a system subtract the sum of an active |𝑃𝐿𝑓,𝑘 | ≤ 𝑃𝐿𝑓,𝑘 𝑘 = 1,2, … … , 𝐿 (12)

78
 International Journal of Applied Energy Systems, Vol. 1, No. 2, July 2019

𝑚𝑎𝑥
where, 𝑃𝐿𝑓,𝑘 is the active power flow of line k; 𝑃𝐿𝑓,𝑘 is the to social hierarchy of wolves, considering the fittest solution
active power flow high limit of line k and L is the as the alpha (α). The second and third best solutions are
transmission lines number. presented as beta (β) and delta (δ), respectively. The
residual of the candidate solutions are supposed to be omega
V. THE PROPOSED ALGORITHMS (ω).
The mathematical model for each optimization techniques Three basic principles of GWO algorithm, namely hunting,
is explained in this section. chasing, and tracking for prey, encircling prey, and
attacking prey which are considered as the behavior of grey
A. Particle Swarm Optimization (PSO) wolves and using for designing GWO. The encircling
The PSO algorithm was firstly invented by Kennedy and behavior can be presented as:
Eberhart in 1995 [30,31] and it is based on the imitation of ⃗ = |𝑐 . 𝑋𝑝 (𝑡) − 𝑋(𝑡)|
𝐷 (16)
the social behavior of fish, birds and insects and its
𝑋(𝑡 + 1) = 𝑋𝑝 (𝑡) − 𝐴. 𝐷 ⃗ (17)
movement with communication as bird flocking and fish
schooling. The word particle indicates, for example, a bee in where, t indicates the current iteration, D,A, and C indicate
a colony or a bird in a swarm. Each individual or particle in coefficient vectors, Xp is the prey position vector, and X
a swarm be in a organized way by its own intelligence and denotes the grey wolf position vector. The vectors A and C
the collective or group intelligence of the swarm. When one are determined as :
particle finds a good route to food, other particles in a 𝐴 = 2𝑎 . 𝑟1 − 𝑎 (18)
swarm will also be able to follow the good path instantly 𝐶 = 2. 𝑟2 (19)
even if their site is remote from the swarm. This simulating the hunting behavior of grey wolves, assuming
optimization methods based on swarm intelligence are that the alpha (α), beta (β) , and delta (δ) have better
called behaviorally inspired techniques as opposed to the knowledge about the probable site of prey. The hunting
genetic algorithms, which are called evolution-based behavior can be presented as :
procedures. It is a population-based technique (a population ⃗ 𝛼 = |𝐶1 . 𝑋𝛼 − 𝑋|, 𝐷
𝐷 ⃗ 𝛽 = |𝐶2 . 𝑋𝛽 − 𝑋|, 𝐷⃗𝛿
of particles) and used for optimizing optimization problems.
= |𝐶3 . 𝑋𝛿 − 𝑋| (20)
Each particle is supposed to have two characteristics (a
position and a velocity). Each particle be around in the 𝑋1 = 𝑋𝛼 − 𝐴1 . (𝐷 ⃗ 𝛼) (21)
search space and can be the best position when evaluated the ⃗
𝑋2 = 𝑋𝛽 − 𝐴2 . (𝐷𝛽 ) (22)
value of objective function. The particles can be updated a 𝑋3 = 𝑋𝛿 − 𝐴3 . (𝐷 ⃗ 𝛿 ) (23)
good positions and their velocities based on equations (14)
and (15). This approach is learned from swarms behavior to 𝑋1 + 𝑋2 + 𝑋3
𝑋(𝑡 + 1) = (24)
optimize global optimization functions solution and every 3
individual in the swarm is called a particle [32]. These At |A| < 1, the wolves are forced to attack the prey, where A
mathematical equations are: is random value. Searching for prey is the exploration ability
𝜔𝑚𝑎𝑥 − 𝜔𝑚𝑖𝑛 and attacking the prey is the exploitation ability. At |A| > 1
𝜔 = 𝜔𝑚𝑎𝑥 − 𝑘 ∗ (13) the wolves are enforced to splay from the prey.
𝑀𝑎𝑥𝑖𝑡𝑒
𝑘+1 𝑘 𝑘 𝑘
𝑉𝑖𝑗 = 𝜔 ∗ 𝑉𝑖𝑗 + 𝑐1 ∗ 𝑟1 ∗ (𝑃𝑏𝑒𝑠𝑡𝑖,𝑗 − 𝑋𝑖,𝑗 ) + 𝑐1 ∗ 𝑟1 ∗ C. A Newly Hybrid Algorithm
(𝐺𝑏𝑒𝑠𝑡𝑗𝑘 − 𝑋𝑖,𝑗 𝑘
) (14) There are a lot of hybridization techniques for heuristic
𝑘+1 𝑘 𝑘+1
𝑋𝑖,𝑗 = 𝑋𝑖,𝑗 + 𝑉𝑖,𝑗 (15) techniques. According to Talbi [34,35], which can be
Where, population size is indicated as N and dimension D is hybridized two techniques or more for hybridization
presented as X = [X1, X2,..., XN]T , where T indicates the techniques. HPSOGWO is a combination of PSO and
transpose operator. Each particle is presented as Xi (i = 1, 2, GWO. HPSOGWO combines the best strength of both PSO
... , N) is presented as 𝑋𝑖 = [𝑋𝑖,1 , 𝑋𝑖,2 , … , 𝑋𝑖,𝐷 ]. Also, the in exploration and in exploitation stage across the targeted
initial velocity of the population is indicated as V =[V1, optimum solution by replacing the best Value of PSO with
V2,..., VN]T . Thus, the velocity of each particle in a grey wolf position value of GWO. In HPSOGWO, first
population Xi (i = 1, 2, .. ,N) is presented as Vi = three agents position is updated in the search space by the
[𝑉𝑖,1 , 𝑉𝑖,2 , … . , 𝑉𝑖,𝐷 ]. The index i mutates from 1 to N equations (25-27) with addition inertia constant (β) to
whereas the index j mutates from 1 to D. control the exploration and exploitation of the grey wolf in
the search space. The modified equations are presented as:
B. Grey Wolf Optimizer (GWO) ⃗ 𝛼 = |𝐶1 . 𝑋𝛼 − 𝜔 ∗ 𝑋| (25)
𝐷
The GWO algorithm mimes the leadership hierarchy and
hunting technique of grey wolves in nature submitted by ⃗ 𝛽 = |𝐶2 . 𝑋𝛽 − 𝜔 ∗ 𝑋|
𝐷 (26)
Mirjalili et al. [33]. Grey wolves are considered to be at the
top of food series and they are living in a collection. Four
species of grey wolves such as alpha (α), beta (β), delta (δ), ⃗ 𝛿 = |𝐶3 . 𝑋𝛿 − 𝜔 ∗ 𝑋|
𝐷 (27)
and omega (ω) are simulating the leadership hierarchy and
as basic parameters of GWO. As designing GWO according

79
Faculty of Energy Engineering - Aswan University - Aswan - Egypt

Where, ω denotes as inertia weight For combining PSO and optimal solutions obtained in each iteration are stored in an
GWO techniques, the velocity and updated equation are archive (Repository) and this archive is updated in each
presented as: iteration to make the domination points deleted. All the
𝑘+1 𝑘 𝑘
𝑉𝑖,𝑗 = 𝜔 ∗ 𝑉𝑖,𝑗 + 𝑐1 ∗ 𝑟1 ∗ (𝑋1 − 𝑋𝑖,𝑗 ) + 𝑐2 ∗ 𝑟2 Pareto optimal solutions in the archive are equally good.
𝑘
∗ (𝑋2 − 𝑋𝑖,𝑗 ) + 𝑐3 ∗ 𝑟3
𝑘
∗ (𝑋3 − 𝑋𝑖,𝑗 ) (28) VII. D. A MULTI-OBJECTIVE HYBRID ALGORITHM
(MO-HPSOGWO(
𝑘+1 𝑘 𝑘+1
𝑋𝑖,𝑗 = 𝑋𝑖,𝑗 + 𝑉𝑖,𝑗 (29)
In order to implement multi-objective optimization by
HPSOGWO we combine two new components. The
D. The basic steps of HPSOGWO components are similar to MOPSO [17,18] and MOGWO
STEP 1: Create an initial population (agents) or (Grey [21].The first one is the repository (archive), which is
wolves). responsible for storing non-dominated Pareto optimal
STEP 2: Initialize a,A,C and ω equations( 13,18,19). STEP solutions obtained so far and there is a maximum number of
3: Fitness evaluation of each agents. solutions for the repository. The second component is a
STEP 4: Calculate the position of Grey Wolf.𝑋𝛼 , 𝑋𝛽 , 𝑋𝛿 leader selection designing that assists to select alpha, beta
and delta solutions as the leader of the hunting process from
equations(25-27) and (21-23).
the repository.
STEP 5: Updating velocity and position equations(28,29).
The MO-HPSOGWO algorithm inherits all the
STEP 6: Repeat STEP (2-5) until the stop criteria is
characteristics of HPSOGWO, which means that there are
reached.
the same exploration and exploitation abilities in two
STEP 7: Stop.
algorithms. The basic difference that MO-HPSOGWO
design about the repository (archive), which the solution is a
set of non-dominated solutions not three best solutions as
VI. MULTI-OBJECTIVE OPTIMIZATION
HPSOGWO algorithm.
Multi-Objective optimization problem consists of several
objective functions to be optimized simultaneously. In these
problems, objective functions are in conflicting with each
other. For example, in Optimal Power Flow problem (OPF),
by minimizing generation fuel cost and consequently active
and reactive power losses are maximizing. Main concepts
related to Multi-Objective Optimization are [36]:
A. Domination
In Multi-Objective optimization the domination is used for
comparing the solutions as in Fig 1. If all the X1 solution
are not worse than all solution X2 in all objectives, or if all
the X1 solutions are equal to X2 but only in one case or one
dimension X1 is better than X2, then it can be said that X1
will dominate X2 and X2 must be deleted from solutions list.
The mathematical expression of the domination part is
presented as : Fig.1. Pareto optimum [37]
𝑥1 𝑑𝑜𝑚𝑖𝑛𝑎𝑡𝑒 𝑥2 𝑖𝑓:
𝑓(𝑥1 ) ≤ 𝑓(𝑥2 )
𝑓(𝑥1 ) < 𝑓(𝑥2 )
B. Pareto front
A solution of Multi-Objective optimization problem is a
curve not a point but a set of points, which every point in
this curve will non-dominate each other. Because if find a
point in which one of the target functions is minimum, there
is another target function which is not minimum at this
point. This curve called Pareto front curve as in Fig 2.

C. Finding best local guide

A solution of Multi Objective optimization problem is a set
of Pareto optimal solutions not the best point. Pareto Fig.2. Concept of pareto optimality [38]

80
 International Journal of Applied Energy Systems, Vol. 1, No. 2, July 2019

VIII. PSEUDO CODE OF THE MO-HPSOGWO ALGORITHM From TABLE III : in single optimization process there is
Create an initial population Xi(i = 1, 2, ...,n) one best (global) solution for each function without attention
Initialize a, A,C and ω equations(13,18,19) to the value of the other four functions. For example, when
Fitness evaluation of each agents minimizing of generation fuel cost this leads to active and
Find the non-dominated points and initialized the reactive power losses in a system increasing and vice versa.
repository with them
𝑋𝛼 = SelectLeader(rep) From TABLE IV : there are several solutions (non-
Exclude α from the repository tentatively to avoid selecting dominated) for five functions (OPF) as in Fig.8, which are
the same leader optimized simultaneously. And these five functions are
𝑋𝛽 = SelectLeader(rep) conflicting objectives that means when decision optimal
solution need to trade off between them. So the decision
Exclude β from the repository tentatively to avoid selecting
maker (DM) for selecting a compromise solution based on
the same leader
maximum limits as in TABLE II and minimum limits as in
𝑋𝛿 = SelectLeader(rep)
TABLE III.
Add back alpha and beta to the repository
t=1;
TABLE I. Load flow analysis of 30 bus system by using NR method
f o r (t=1: Max iterations)
Bus No. V(p.u.) Delta P(MW) Q(MVAR)
f o r each agents
Update the position of the current search agent by 1 1.050 0.0 353.099 -14.98
equations(25-27) and (21-23) 2 1.038 -3.705 54.28 18.011
Update the velocity and position by equations (28,29) end 3 1.011 -10.514 -145.380 10.748
for 4 1.019 -8.316 -15.700 12.855
Update a, A, C and ω 5 1.091 -8.667 24.280 22.753
Fitness evaluation of each agents 6 1.091 -10.383 24.000 20.940
Find the non-dominated points 7 1.006 -9.554 -45.600 -10.900
Update the repository 8 1.016 -6.857 -15.200 -1.600
If the repository is complete 9 1.048 -9.932 0.000 0.000
Run the grid mechanism to delete one of the current 10 1.031 -11.672 -11.600 -2.000
repository points 11 1.023 -5.708 -4.800 -1.200
Add the new point to the repository 12 1.065 -11.212 -22.400 -7.500
end if 13 1.015 -7.994 0.000 -0.000
𝑋𝛼 = SelectLeader(rep) 14 1.047 -12.088 -12.400 -1.600
Exclude α from the repository tentatively to avoid selecting 15 1.039 -12.108 -16.400 -2.500
the same leader 16 1.043 -11.666 -7.000 -1.800
𝑋𝛽 = SelectLeader(rep) 17 1.030 -11.876 -18.000 -5.800
Exclude β from the repository tentatively to avoid selecting 18 1.024 -12.654 -6.400 -0.900
the same leader 19 1.019 -12.787 -19.000 -3.400
𝑋𝛿 = SelectLeader(rep) 20 1.021 -12.565 -4.400 -0.700
Add back alpha and beta to the repository 21 1.019 -12.132 -35.000 -11.200
t = t+1; 22 1.020 -12.119 -0.000 0.000
return rep 23 1.023 -12.434 -6.400 -1.600
IX. RESULTS AND DISCUSSION 24 1.009 -12.527 -17.400 -6.700
25 1.010 -12.469 -0.000 -0.000
shown in Fig.9 The total active power demands is 283.4
26 0.993 -12.894 -7.000 -2.300
MW and Total reactive power demands is 126.2 MVAR.
27 1.020 -12.163 0.000 0.000
Five objective functions (OPF) are individually optimized as
28 1.012 -8.478 0.000 0.000
a single objective function in optimization process by using
29 1.000 -13.401 -4.800 -0.900
HPSOGWO and are optimized simultaneously by using
MO-HPSOGWO, which are : 30 0.989 -14.290 -21.200 -1.900
F1 Fuel Cost Minimization
TABLE II. IEEE 30-bus system individual objective functions before
F2 Emission Minimization applying optimization technique
F3 Total Active Power Loss Minimization Objective function Objective value
F4 Reactive Power Transmission Loss Minimization
Over all Generation fuel costs 4205.1 ($/h)
F5 Reactive Power Reserve Margin Maximization.
Emission index 2.4681 (t/h)
Active power transmission loss 19.579 (MW)
Reactive power transmission loss 5.827 (MVar)
Reactive power reserve margin 1.1034

81
Faculty of Energy Engineering - Aswan University - Aswan - Egypt

TABLE III. Single objective function values by HPSOGWO. TABLE V. The standard values used
Function Best solution Parameters quantity
F1 (min of fuel cost) 739.838 population size 100
F2 (min of emission) 2.0486E-04 Repository size (rep) 100
F3(min of active power loss) 5.28 No. of iterations 200
F4(min of reactive power loss) -16.02
F5(max of reactive power reserve) 1.17E-16
TABLE IV. Multi-objective functions (OPF) values (non-dominated
solutions) by MO-HPSOGWO
NO. GFC EI Ploss Qloss RPRM
(Rs/h) (t/h) (MW ) (MVar) (p.u)
1 778.892 0.0904 0.901 3.501 0.806
2 1628.2 1.1583 11.581 -7.867 0.213
3 805.713 0.1356 1.354 2.016 0.931
4 785.456 0.0487 0.484 10.316 1.086
5 976.827 0.3706 3.704 -0.181 0.711 Fig.3. Minimization of generation fuel cost by HPSOGWO
6 787.768 0.0011 0.008 8.575 1.377
7 779.825 0.0134 0.132 10.625 1.053
8 1067.4 0.4988 4.986 -2.586 0.223
9 823.879 2.48E-04 2.36E-04 14.316 1.376
10 799.699 0.013 0.129 12.327 1.156
11 805.841 0.132 1.317 4.186 0.584
12 1054.9 0.4445 4.442 -0.419 0.663
13 858.655 0.1986 1.983 3.199 0.695
14 803.943 2.91E-04 4.59E-04 12.252 1.249
15 772.455 0.0506 0.504 5.517 1.143
16 938.129 0.3297 3.295 0.830 0.630
17 986.147 0.3724 3.722 0.239 0.483 Fig.4. Minimization of emission by HPSOGWO
18 812.773 0.0038 0.035 10.605 1.248
19 1885.6 1.449 14.494 -10.689 0.142
20 806.641 3.05E-04 4.44E-04 11.896 1.573
21 794.102 0.0168 0.165 10.329 0.990
22 1360.5 0.8531 8.528 -6.064 0.312
23 888.950 0.2575 2.573 0.417 0.720
24 1332.4 0.8148 8.145 -5.114 0.423
25 765.056 0.0227 0.224 5.935 1.398
26 854.473 0.1956 1.954 3.593 0.668
27 858.135 0.2078 2.076 2.245 0.558
28 761.863 0.0634 0.632 5.078 0.857
29 783.092 0.0165 0.163 10.118 1.122
Fig.5. Minimization of active power loss by HPSOGWO
30 819.717 4.15E-04 0.002 11.945 1.461
31 786.184 0.0061 0.059 9.857 1.398
32 781.588 0.0067 0.064 8.211 1.675
33 804.136 0.0856 0.854 7.028 0.846
34 1477.9 0.9816 9.814 -6.846 0.309
35 1992.6 1.5829 15.826 -12.984 0.258
36 772.595 0.0493 0.491 6.269 1.109
37 1008.3 0.4093 4.090 0.227 0.703
38 774.865 0.0343 0.340 6.896 1.326
39 778.702 0.0255 0.253 9.663 1.128
40 773.374 0.0544 0.542 6.308 0.843 Fig.6. minimization of reactive power loss by HPSOGWO

82
 International Journal of Applied Energy Systems, Vol. 1, No. 2, July 2019

X. CONCLUSION
Most of the real world problems in many fields science,
engineering, economics and logistics are multi-objectives
optimization problems, making conflicting objectives. In
this paper using two algorithms HPSOGWO and MO-
HPSOGWO testing through IEEE 30-bus system and
optimizing five objective function (OPF). From results MO-
HPSOGWO is more realistic and efficient than HPSOGWO
because single objective function (HPSOGWO) has one
global solution without attention to the value of the other
Fig.7. maximization of reactive power reserve margin by HPSOGWO
four functions as in TABLE III but multi-objective
functions (MO-HPSOGWO) has a set of non-dominated
solutions and the compromise solution selecting based on
decision maker (DM) as in TABLE IV and these five
functions can be optimized concurrently. So without using
multi-objective optimization in (OPF) only one aspect of the
power system has been optimized.
XI. REFERENCES

[1] C. T. Kelley, ‘‘Detection and remediation of stagnation in the Nelder

Mead algorithm using a sufficient decrease condition, ” (1999).
[2] T. P. Vogl, J. Mangis, A. Rigler, W. Zink, D. Alkon, “ Accelerating
the convergence of the back-propagation method,”. (1988).
[3] R. T. Marler, J. S. Arora, “Survey of multiobjective optimization
methods for engineering,”. (2004)
[4] H. G. Beyer, B. Sendhoff, “Robust optimization a comprehensive
survey,” (2007).
[5] C. Coello, “Theoretical and numerical constraint-handling techniques
used with evolutionary algorithms: a survey of the state of the art,”.
(2002)
[6] K. Deb, D. E. Goldberg, “Analyzing deception in trap functions,”
Fig.8. Pareto optimal for five function (OPF) non-dominated solutions by (1993)
HPSOGWO [7] Knowles JD, Watson RA, Corne DW, ‘‘Reducing local optima in
single objective problems by multi-objective,” In: Evolutionary multi-
criterion optimization, pp.269283, (2001).
[8] Deb K, ‘‘Multi-objective optimization using evolutionary algorithms,
” Vol.16, Wiley,(2001).
[9] Coello CAC, Lamont GB and David A. Van Veldhuisen,
‘‘Evolutionary algorithms for solving multi-objective problems, ”
Springer.
[10] Branke J, Deb K, Dierolf H, Osswald M, ‘‘Finding knees in multi-
objective optimization, ” (2004).
[11] Das I, Dennis JE, ‘‘Normal boundary intersection: A new method for
generating the Pareto surface in nonlinear multicriteria optimization
problems, ” (1998).
[12] Kim IY, De Weck, ‘‘Adaptive weighted-sum method for bi-objective
optimization: Pareto front generation, ” (2005).
[13] Messac, A., and Mattson, C. A., ‘‘Generating Well-Distibuted Sets of
Pareto Points for Engineering Design using Physical Programming, ”
Optimization and Engineering, Vol. 3, Issue 4, December 2002, pp.
431-450.
[14] Mirjalili S, Saremi S, Mirjalili SM, Coelho L. d. s., ‘‘Multi objective
grey wolf optimizer: A novel algorithm for multi-criterion
optimization, ” (2016).
[15] Akbari R, Hedayatzadeh R, Ziarati K, Hassanizadeh B, ‘‘A multi-
objective artificial bee colony algorithm, ” In Swarn and Evolutionary
Computation, February 2012..
[16] Coello CAC, Lechuga MS, ‘‘MOPSO: A proposal for multiple
objective particle swarm optimization, ” (2002).
Fig.9. Single-line diagram of IEEE-30 bus test system [17] Coello CAC, Pulido GT, Lechuga MS,‘‘ Handling multiple objectives
with particle swarm optimization,” IEEE Transactions on
Evolutionary Computation,June 2004.

83
Faculty of Energy Engineering - Aswan University - Aswan - Egypt

[18] Xin-She Yang, ‘‘Bat algorithm for multi-objective optimization,” [32] Peter J.Angeline, ‘‘Evolutionary optimization versus particle swarm
,September 2011,Intenational Journal of Bio-Inspired Computation. optimization:Philosophy and performance differences,” In:
[19] Deb K, Agrawal S, Pratap A, Meyarivan T, ‘‘A fast elitist non- Proceedings of 7th Annual Conference Evolutionary Programming,
dominated sorting genetic algorithm for multi-objective optimization: San Diego, pp. 601610,March,1998.
NSGA-II, ” (2000). [33] S. Mirjalili, S. M. Mirjalili, and A. Lewis, ‘‘Grey wolf optimizer,
[20] Deb K, Goel T, ‘‘Controlled elitist nondominated sorting genetic ”Advances in Engineering Software, vol. 69, pp. 4661, (2014).
algorithms for better convergence, ” (2001). [34] Narinder Singh and S. B. Singh, ‘‘Hybrid Algorithm of Particle
[21] Deb K, Pratap A, Agarwal S, Meyarivan T, ‘‘A fast and elitist Swarm Optimization and Grey Wolf Optimizer for Improving
multiobjective genetic algorithm: NSGA-II, ” (2002). Convergence Performance, ” Department of Mathematics, Punjabi
University, Patiala, Punjab 147002, India, (2017).
[22] S. Mirjalili and S. Z. M. Hashim, ‘‘A new hybrid PSOGSA algorithm
for function optimization, ” in Proceedings of the International [35] Talbi, ‘‘A taxonomy of hybrid metaheuristics, ” Journal of Heuristics,
Conference on Computer and Information Application (IC- CIA ’10), vol. 8, no. 5, pp. 541564, (2002).
pp. 374377, Tianjin, China, November (2010). [36] Seyed Saeed Hosseini, Sajad Ahmad Hamidi, Motahar Mansuri and
[23] Hardiansyah, ‘‘A Novel Hybrid PSO-GSA Method for Non-convex Ali Ghoddosian, ‘‘Multi Objective Particle Swarm Optimization
Economic Dispatch Problems, ” Published Online November (2013) (MOPSO) for Size and Shape Optimization of 2D Truss Structures,
in MECS DOI: 10.5815/ijieeb.2013.05.01. ”pp.9-14, (2015).
[24] Trivedi I.N., Jangir P., Kumar A., Jangir N., Bh- esdadiya R.H., [37] Carlos A. Santos Silva,‘‘Multi-objective Optimization, ” MIT
Totlani R., ‘‘A Novel Hybrid PSO-DA Algorithm for Global Portugal.
Numerical Optimization, ” (2018). [38] Ankit Yadav and Dr.Sanjay K.Jain, ‘‘MULTIOBJECTIVE
[25] S. Arunachalam, T. AgnesBhomila, and M. Ramesh Babu, ‘‘Hybrid OPTIMAL POWER FLOW, ” July 2010,Thapar university,Patiala.
Particle Swarm Optimization Algorithm and Firefly Algorithm Based
Combined Economic and Emission Dispatch Including Valve Point
Effect, ” in July (2015).
[26] Pradeep Jangir, Siddharth A.Parmar, Indrajit N.Trivedi, and R.H.
Bhesdadiya, ‘‘A novel hybrid Particle Swarm Optimizer with multi
verse optimizer for global numerical optimization and Optimal
Reactive Power Dispatch problem, ” in April (2017).
[27] Wolpert DH, Macready WG, ‘‘No Free Lunch theorems for
optimization, ” April,1997, IEEE Transactions on Evolutionary
Computation
[28] S.A.H. Soliman, A.H. Mantawy, ‘‘Modern Optimization Techniques
With Applications in Electric Power Systems, ” Energy System
Springer, NewYork, (2012).
[29] Wadhwa, C.L., Jain, N.K, ‘‘Multiple objective optimal load flow: A
new perspective, ” (1990).
[30] J. Kennedy and R. Eberhart, ‘‘Particle swarm optimization, ” in Proc.
IEEE Int. Conf. Neural Networks (ICNN’95), Perth, Australia, IV:
19421948, (1995).
[31] Y. Shi and R. Eberhart, ‘‘A modified particle swarm optimizer,
”Proceedings of IEEE International Conference on Evolutionary
Computation, Anchorage, Alaska, 69-73, (1998).

84

View publication stats