0% found this document useful (0 votes)
16 views15 pages

Research Article

Uploaded by

fatemehof5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
16 views15 pages

Research Article

Uploaded by

fatemehof5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Hindawi

International Transactions on Electrical Energy Systems


Volume 2022, Article ID 7091937, 15 pages
https://doi.org/10.1155/2022/7091937

Research Article
Multi-Objective Optimal Power Flow including Wind and Solar
Generation Uncertainty Using New Hybrid Evolutionary
Algorithm with Efficient Constraint Handling Method

Ravi Kumar Avvari and Vinod Kumar D. M.


Department of Electrical Engineering, National Institute of Technology Warangal, Warangal, Telangana 506004, India

Correspondence should be addressed to Ravi Kumar Avvari; [email protected]

Received 10 February 2022; Revised 4 May 2022; Accepted 9 May 2022; Published 2 July 2022

Academic Editor: mahdiyeh eslami

Copyright © 2022 Ravi Kumar Avvari and Vinod Kumar D. M.. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.

A new hybrid decomposition-based multiobjective evolutionary algorithm is proposed for optimal power flow (OPF) including
wind and solar generation uncertainty. This study recommends a novel constraint-handling method, which adaptively adds the
penalty function and eliminates the parameter dependency on penalty function evaluation. The summation-based sorting and
improved diversified selection methods are utilized to enhance the diversity of multiobjective optimization algorithms. The OPF
problem is modeled as a multiobjective optimization problem with four objectives such as minimizing (i) total fuel cost (TC)
including the cost of renewable energy source (RES), (ii) total emission (TE), (iii) active power loss (APL), and (iv) voltage
magnitude deviation (VMD). The impact of RESs such as wind and solar energy sources on integration is considered in optimal
power flow cost analysis. The costs of RESs are considered in the OPF problem to minimize the overall cost so that the impact of
intermittence and uncertainty of renewable sources is studied in terms of cost and operation wise. The uncertainty of wind and
solar energy sources is described using probability distribution functions (PDFs) such as Weibull and lognormal distributions. The
efficiency of the algorithm is tested on IEEE 30-, IEEE 57-, and IEEE 118-bus systems for all possible conditions of renewable
sources using Monte Carlo simulations.

1. Introduction A significant amount of research has been carried out in


the domain of OPF with the incorporation of RESs in the
In recent times, RES penetration has drastically increased in power system using both deterministic and meta-heuristic
the power system. The penetration of RESs has introduced optimization algorithms. The gradient method is proposed
many challenges to the power system. The intermittent [1] to develop the dynamic OPF to include wind farms
nature of RESs makes the system more complex in terms of without considering the costs of wind power. For solving the
operation and control. The uncertain nature of RESs is OPF model in the presence of a wind plant, the authors [2]
required to be accurately modeled to examine the dynamic used the Newton method and interior-point methods. The
functioning of the power system. Due to its unpredictable uncertain nature of wind power has been estimated and is
nature, protection schemes need to be updated for operating added to the overall cost function. However, deterministic
the power system in a secure region. In a power system, the methods are problem-specific, exhibit poor convergence
main aim is to operate it with optimal cost and simulta- characteristics, and are stuck at local optima points.
neously satisfy the operating and security constraints. The Moreover, these methods are unable to solve real-world
OPF determines the optimal control settings by the satis- optimization issues. To overcome the drawbacks of deter-
fying system and security constraints to economically ministic methods, meta-heuristic methods have been
operate. introduced.
2 International Transactions on Electrical Energy Systems

In [3], the authors used the SHADE algorithm with the prices are agreed by ISO to pay for the scheduled wind and
SF method for arriving at the solution to OPF with RESs. solar energy. Direct prices have not been addressed in most
Similarly, in [4–8] the authors proposed several meta- of the literature.
heuristic optimization methods for solving OPF with RESs. The above literature review reveals the following:
However, these are formulated as single-objective optimi-
zation problems. In the real world, the OPF problem is (i) Most of the authors designed the OPF problem as
multiobjective and the trade-off between multiple objects single-objective optimization. In real time, multiple
gives better optimal conditions for operation. objectives play a key role in the economic viability of
In [9], the authors introduced a modified JAYA al- the power system.
gorithm for solving the MOOPF problem incorporating (ii) The weighted sum-based methods depend on
RESs with four different objectives. In this study, the weights assigned to each objective, and it affects the
authors transformed multiple objectives into a single optimal solution.
objective problem with price and weights. Similarly, in (iii) In most of the literature, the Pareto dominance
[10–12] the authors proposed a weighted sum-based method is used, and in the Pareto dominance
MOOPF problem with various objectives. The weighted method, nondominated sorting (NDS) technique to
sum-based methods are simple in combining multi- select the best solutions is used, which improves the
objectives into a single objective with suitable weights. diversity and convergence. When all the dominant
However, this approach heavily depends on the weights solutions have been removed, the diversity of the
that are assigned to each objective value, and these, in turn, population is lost. NDS selection is complex and
affect the optimal solution. Moreover, the weighted sum- time-consuming.
based methods fail to obtain the best-compromised so-
lution when needed. (iv) The constraints are handled using the penalty factor
In [13], the authors concentrated on the analysis of the method, a specific method that is inefficient, due to
MOOPF solution with RESs using the hybrid DE and SOS parameter dependency.
algorithms, which have been tested under different operating (v) In calculating the uncertainty cost of RESs, only
conditions. Similarly, in [14–16] the authors used the overestimation and underestimation costs are
nondominated sorting (NDS) technique to pick the best considered, while the direct cost is neglected.
solutions for parents in an elitist fashion. When the dom-
inant solutions are removed from the population, the ef- In this study, a new hybrid MOEA based on decom-
fective exploration capability will be lost. Besides this, the position and summation of normalized objectives with an
nondominated sorting selection is challenging and time- improved diversified selection method is used for the
consuming. Moreover, the constraints are handled using the MOOPF problem. An SF strategy is employed to tackle
penalty factor method, which is inefficient. various constraints (i.e., equality and inequality) of the
In OPF, constraints play a key role to obtain feasible MOOPF problem.
optimal solutions. The constraint-handling techniques used The major contributions of the research work include the
in optimization techniques are divided into two categories; following:
(i) generic methods and (ii) specific methods. The generic (1) Proposing a novel MOEA based on decomposition
methods are penalty function-based methods. These are and summation of normalized objectives with im-
simple and mostly used in optimization algorithms as they proved diversified selection for the MOOPF
do not demand additional changes in the algorithm. When a problem.
constraint violation occurs, a penalty is added to its fitness.
However, these methods may not provide satisfactory results (2) Integrating RESs like wind and solar power plants
for all types of constraints. On the other hand, specific with conventional OPF to consider the impact of the
constraint-handling methods can be applied to convex re- uncertain nature of these sources.
gion problems and large variable problems. The cutting (3) Modeling the uncertain nature of wind and solar
plane method and gradient method are the commonly used power plants using PDF and calculating the un-
methods to handle specific constraints [17, 18]. However, the certain cost using Monte Carlo simulations.
drawback of specific methods is that, as the number of (4) Multiobjective OPF (MOOPF) with TC, TE, APL,
variables increases, the computing time also increases. The and VMD as four objectives.
performance of both methods depends on fine-tuning dif-
(5) Utilizing an efficient constraint-handling technique
ferent parameters of constraint handling, which also affects
(CHT) called the superiority of feasible solution (SF)
the fitness value.
to tackle complex constraints in MOOPF problems.
The conventional generators are subjected to different
costs as they run on fuel. RESs such as wind and solar do not The study is structured as follows: Section 2 presents a
require any fuel. Therefore, fuel costs are not considered for wind and solar uncertainty modeling. Section 3 describes the
wind and solar power generation. In the case of wind and problem formulation of MOOPF with RES. Section 4
solar generations owned by anyone other than ISO, direct presents the framework of the proposed algorithm. In
cost needs to be added to the total cost, which is in the form Section 5, simulation case studies are discussed and con-
of maintenance costs and renewal charges [19]. The direct clusions are made in Section 6.
International Transactions on Electrical Energy Systems 3

2. Wind and Solar Power Uncertainty Modeling where lognormal PDF parameters μ � 6 and σ � 0.6. The
standard solar irradiance (Gstd ) � 800 W/m2, and particular
The wind speed at a given geographical area is most likely irradiation point (Rc ) � 120 W/m2.
distributed according to Weibull distributions. Mathemat-
ically, the Weibull PDF is written as follows: 3. Problem Formulation with Renewable
k v (k−1) (− v/c)k
Energy Sources
f(v) � 􏼠 􏼡􏼒 􏼓 (e) , 0 < v < ∞. (1)
c c In this study, a wind generator and solar generator are
located at two different buses in the test system. Since wind
The PDFs for two different shape and scale factors are and solar powers are intermittent, the Monte Carlo simu-
given in [20]. The relationship between wind speed and lations are used to account for uncertainty and to calculate
power generation is as follows: the uncertainty cost. The estimated price for the intermit-
tency of wind and solar power is reflected in three ways:


⎪ 0 for v < vin and v > vout

⎪ direct price, reserve price, and penalty price. Whenever



⎪ power is underestimated, extra unusable power is wasted;

⎨ v − vin however, in practical power system applications, such power
Pw (v) � ⎪ Pwr 􏼠 􏼡 for vin ≤ vw ≤ vr , (2)

⎪ vr − vin can be saved in an energy storage system and thus counted as

⎪ the reserve price. The price of overestimating power that is



⎪ lower than the scheduled power is considered a penalty price

Pwr for vr < vw ≤ vout
in the case of overestimation.
The probability of obtaining a rated and zero power
output is given by the following: 3.1. Direct Price Calculation of Wind and Solar Power Plants.
In contrast to conventional generators, wind and solar
vin k v k
power generators do not require any fuel. When an ISO
fw Pw � 0􏼁 � 1 − exp􏼠−􏼒 􏼓 􏼡 + exp􏼠−􏼒 out 􏼓 􏼡, (3)
c c owns wind/PV facilities, the direct fuel cost may not occur
except if the ISO intends to allocate any compensation for
v k v k setting up or charging it as a renewal cost and repair work
fw Pw � Pwr 􏼁 � exp􏼠−􏼒 r 􏼓 􏼡 + exp􏼠−􏼒 out 􏼓 􏼡. (4)
c c [22]. When private agencies own wind/PV plants, however,
ISO proportionally pays for the agreed-upon scheduled
The probability for the linear part of the wind speed is power.
given by the following: The direct price associated with jth wind plants is as
follows:
(k− 1)
k vr − vin 􏼁 vin Pwr + Pw vr − vin 􏼁
f w Pw 􏼁 � 􏼠 􏼡􏼠 􏼡 Cw,j 􏼐Pws,j 􏼑 � gj Pws,j . (8)
cPwr cPwr
(5) Similarly, the direct price of kth PV plant is as follows:
k
⎝−􏼠 vin Pwr + Pw vr − vin 􏼁 ⎠ Cs,k 􏼐Pss,k 􏼑 � hk Pss,k . (9)
· exp⎛ 􏼡 ⎞.
cPwr

where Weibull PDF parameters k� 2 and c � 10. The wind


speeds vin � 3 m/sec, vout � 25 m/sec, and vr � 16 m/sec. 3.2. Uncertainty Price Calculation of the Wind and Solar Power
Similarly, the power output of a solar energy system is a Plants. If the actual output power of the wind farm is lower
factor of solar irradiance (Gs ) and it likely follows the than the predicted value, to ensure a constant supply of
lognormal distribution [21]. The PDF for the lognormal electricity to the consumers, the operator requires some
distribution is as follows: spinning reserve. It is called the overestimation of power
from unreliable sources. The cost incurred to maintain the
1 − ln Gs − μ􏼁
2 spinning reserve is known as the reserve cost [23].
f G Gs 􏼁 � √��� exp􏼨 􏼩 for Gs > 0. (6) The reserve price of the jth wind plant is as follows:
Gs σ 2π 2σ 2
CRw,j 􏼐Pws,j − Pwav,j 􏼑 � KRw,j 􏼐Pws,j − Pwav,j 􏼑
The PV unit’s solar irradiance to energy generation is
[22]as follows: Pws,j (10)
� KRw,j 􏽚 􏼐Pws,j − pw,j 􏼑fw 􏼐pw,j 􏼑dpw,j .
0



⎪ G2s

⎪ P sr 􏼠 􏼡 for 0 < Gs < Rc In contrast to the overestimation scenario, when the

⎪ Gstd Rc
⎨ actual power output of wind exceeds the predicted output,
Ps G s 􏼁 � ⎪ , (7)

⎪ the surplus power generated by WT cannot be used and is

⎪ G

⎪ s wasted. This is called the underestimation of power from
⎩ Psr 􏼠G 􏼡 for Gs ≥ Rc
std uncertain sources. In this case, ISO must pay a penalty for
excess power.
4 International Transactions on Electrical Energy Systems

The penalty price of the jth wind plant is as follows: The penalty price for a kth PV plant is as follows:
CPw,j 􏼐Pwav,j − Pws,j 􏼑 � KPw,j 􏼐Pwav,j − Pws,j 􏼑 CPs,k 􏼐Psav,k − Pss,k 􏼑 � KPs,k 􏼐Psav,k − Pss,k 􏼑
Pwr,j (11)
� KPs,k ∗ fs 􏼐Psav,k > Pss,k 􏼑 (13)
� KPw,j 􏽚 􏼐pw,j − Pws,j 􏼑fw 􏼐pw,j 􏼑dpw,j .
Pws,j
∗ 􏽨E􏼐Psav,k > Pss,k 􏼑 − Pss,k 􏽩,
In the same way as the wind plant, the PV plant also has
intermittency in power output. The reserve and penalty price where the direct, penalty, and reserve price coefficients of
equations for PV plants are described as follows [24]. wind and PV plants are 1.6, 1.5, and 3, respectively.
Reserve price for kth PV plant is as follows:
CRs,k 􏼐Pss,k − Psav,k 􏼑 � KRs,k 􏼐Pss,k − Psav,k 􏼑
3.3. Objective Functions. The MOOPF problem assumed the
� KRs,k ∗ fs 􏼐Psav,k < Pss,k 􏼑 (12) minimization of four objectives: (i) TC, (ii) TE, (iii) APL,
and (iv) VMD. The objectives can be described as follows:
∗ 􏽨Pss,k − E􏼐Psav,k < Pss,k 􏼑􏽩.

NTG
fTC � 􏽘 ai + bi PTGi + ci P2TGi
i�1
NWG
+ 􏽘 􏽨Cw,j 􏼐Pws,j 􏼑 + CRw,j 􏼐Pws,j − Pwav,j 􏼑 + CPw,j 􏼐Pwav,j − Pws,j 􏼑􏽩 (14)
j�1
NSG
+ 􏽘 􏽨Cs,k 􏼐Pss,k 􏼑 + CRs,k 􏼐Pss,k − Psav,k 􏼑 + CPs,k 􏼐Psav,k − Pss,k 􏼑􏽩,
k�1

NTG
fTE � 􏽘 􏼐αi + βi PTGi + ci P2TGi + ξ i eλi PTGi 􏼑, (15)
i�1

NL
fAPL � 􏽘 􏼐Gk 􏼐V2i + V2j − 2Vi Vj cos θij 􏼑􏼑, (16)
k�1

NPQ 􏼌 􏼌􏼌
􏼌
fVMD � 􏽘 􏼌􏼌􏼌􏼐Vi − Vref 􏼑􏼌􏼌􏼌, (17)
i�1

where Vref � 1.0 p.u., i.e., reference voltage. 3.4.2. Inequality Constraints. Generator constraints

3.4. Constraints Pmin max


(20)
TGi ≤ PTGi ≤ PTGi ; ∀i ∈ NTG ,
3.4.1. Equality Constraints. The overall demand and losses
throughout the system are equal to the total real and reactive Pmin max
WGi ≤ PWGi ≤ PWGi ; ∀i ∈ NWG , (21)
power delivered.
Pmin max
SGi ≤ PSGi ≤ PSGi ; ∀i ∈ NSG , (22)
NB
PGi − PDi − Vi 􏽘 Vj 􏽨Gij cos θij + Bij sin θij 􏽩
j�1 (18) Qmin max
TGi ≤ QTGi ≤ QTGi ; ∀i ∈ NTG , (23)

� 0 :: ∀i ∈ NB , Qmin max
(24)
WGi ≤ QWGi ≤ QWGi ; ∀i ∈ NWG ,

NB
Qmin max
SGi ≤ QSGi ≤ QSGi ; ∀i ∈ NSG , (25)
QGi − QDi − Vi 􏽘 Vj 􏽨Gij sin θij − Bij cos θij 􏽩
j�1 (19)
Vmin max
Gi ≤ VGi ≤ VGi ; ∀i ∈ NG . (26)
� 0 :: ∀i ∈ NB .
International Transactions on Electrical Energy Systems 5

By incorporating these three rules into the proposed al-


Transformer constraints gorithm to solve the MOOPF problem, two situations arise,
Tmin max
(27) the first of which is when the population size is lower than the
k ≤ Tk ≤ Tk ; ∀k ∈ NT .
number of feasible solutions, and the second method is to
Shunt VAR compensator constraints ignore nonfeasible solutions. The use of the summation-based
method is to select feasible solutions if the number of feasible
Qmin max
ci ≤ Qci ≤ Qci ; ∀i ∈ NC . (28) solutions is greater than the population size.
Security constraints
4. Proposed Algorithm
Vmin max
Lp ≤ VLp ≤ VLp ; ∀p ∈ NL , (29)
The MOEAs are normally modeled to handle different
Slq ≤ Smax conflicting goals, such as maximizing the spread of solutions
lq ; ∀q ∈ nl . (30)
along the Pareto front (i.e., diversity) and minimizing the
Two equality constraints (equations (18) and (19)) are distance between the solutions along the Pareto front (i.e.,
automatically satisfied when the power flow converges to an convergence) [30]. The trade-off between convergence and
optimal solution. The generator buses’ real power (excluding diversity is important to choose the best solution among the
slack bus), transformer tap ratios, voltage limits, and shunt obtained solutions. Therefore, to attain a balance between
compensator ranges are considered as control variables that exploration and exploitation in this study, a new method is
are self-limiting. The remaining inequality constraints proposed.
require constraint-handling techniques. In this study, a summation of normalized objective
In OPF, generator reactive power capacities are sig- values (SNOVs) with improved diversified selection
nificant. In the case of thermal generators, the ranges are (IDS) is proposed and integrated with the multiobjective
considered as in [25, 26]. In recent years, WTs with evolution algorithm based on the decomposition
complete reactive power capability have become com- (MOEA/D) [31] method to solve the MOOPF problem
mercially viable [27]. Enercon FACTS-WT can deliver with RES. The MOEA/D method decomposes the mul-
reactive power in the range of -0.4p.u.to 0.5p.u. The neg- tiobjective optimization problem into several single
ative sign signifies the generator’s ability to absorb. Rooftop scalar optimization problems and optimizes them all at
solar PV is designed as load buses with zero reactive power. the same time using weight vectors. The weight vectors’
However, because utility-based solar PVs have converters distance is used to create neighborhoods. In every pop-
built-in, full generator modeling is required due to the ulation evolution, information from the neighborhood is
converters’ dynamic behavior [28]. In this study, the re- used to find a solution. The nondominated sorting used in
active power capabilities of solar PV are assessed between MOEA/D is complex and time-consuming. Some useful
−0.4p.u and 0.5p.u. information may be lost if the dominant solutions are
completely discarded. In addition, diversity may be lost
during the search process and lead to local optima. To
3.5. Superiority of Feasible Solution (SF) Method. The most overcome these problems, the summation of normalized
commonly used constraint-handling technique is the pen- objective values with IDS [32] is employed in this study
alty function method. When a constraint violation occurs, its instead of nondominated sorting selection to get uni-
solution is penalized. Owing to its simplicity and ease of formly distributed Pareto front and improved conver-
operation, the outcome of this method is strongly contingent gence characteristics.
on the penalty factor, which is to be chosen using trial and A new constraint-handling strategy called the superiority
error, going to cause the fitness value to deteriorate. This of feasible solution (SF) method is employed to handle the
study deployed a new CHT called the SF technique [29], various constraints (i.e., equality and inequality) of the
which does not require any penalty coefficient. MOOPF problem. The proposed algorithm utilizes the fuzzy
Since MOOPF is a constrained optimization problem, it method to get the best-compromised values. The outcomes
requires a better-constrained handling method. In this of the proposed method are compared with popular
study, the SF technique [29] was employed to solve the methods like MOEA/D [33], NSGA-II [34], and MOPSO
MOOPF problem with RESs. The steps followed when [35] for different cases.
comparing two solutions are as follows: The pseudocode of the proposed method is as follows:

(1) While comparing two nonfeasible solutions, the


solution having the smallest constraint violation is Step 1. Initialization: Generate the initial population (Pt ) of
selected. size N. Using SSA [36], generate uniformly distributed
weights, and the number of weight vectors is defined as
(2) When two feasible solutions are compared, the one follows:
with a better fitness solution is selected.
D+M−1
(3) When a feasible solution is compared to a non- N(D, M) � 􏼠 􏼡 for D > 0. (31)
feasible solution, the feasible solution is selected. M−1
6 International Transactions on Electrical Energy Systems

Step 2. Run the load flow and evaluate the fitness values of Table 1: Various case studies considered in this study.
the selected objective functions and total constraint Case name Test system TC TE APL VMD
violations.
Case-1 ✓ ✓ — —
Case-2 ✓ — ✓ —
Step 3. Using angle criteria [37], locate neighbors with the Case-3 IEEE 30-bus ✓ ✓ ✓ —
smallest angles for each weight vector. The following is an Case-4 ✓ ✓ — ✓
example of the angle criteria: Case-5 ✓ ✓ ✓ ✓
Case-6 ✓ ✓ — —
d2
tan θ � , (32) Case-7 IEEE 57-bus ✓ ✓ ✓ —
d1 Case-8 ✓ ✓ ✓ ✓
Case-9 ✓ — ✓ —
where d1 � (‖wTi wj ‖/‖wj ‖), d2 � ‖wi − d1 (wj /‖wj ‖)‖, Case-10
IEEE 118-bus
✓ — ✓ ✓
i, j � 1, 2, . . . , N, and i ≠ j, θ � angle between d1 and d2 .

Step 4. Evaluate the smaller objective values to form the Step 14. The solutions are dominated by stopping points,
present ideal point. and also the individuals who were not selected will be sent to
the backup set.
Step 5. Evaluate the larger objective values to form the
present nadir point. Step 15. Apply the fuzzy min-max method [38] to get the
best-compromised values.
Step 6. Reproduction: Angle criteria are used to choose N
pairs of mating parents. A set of mating parents is picked
with a probability of δ each weight. 5. Simulation Results
Step 7. To generate the new population (Qt ), use two-point In this study, to tackle the MOOPF problem including wind
crossover and mutation. and photovoltaic uncertainties, the proposed method,
MOEA/D [33], NSGA-II [34], and MOPSO [35] are dem-
Step 8. The new population is formed by combining the onstrated on IEEE 30-, IEEE 57-, and IEEE 118-bus power
original population (Pt ) with the newly generated pop- systems. It is implemented in MATLAB R2016a and runs on
ulation (Qt ). an i3 processor with 4 GB RAM.
In general, more than two objectives are treated as a
Step 9. For each objective and solution, calculate the nor- multiobjective optimization (MOO) problem. While formu-
malized objective values. lating the MOO problem, the objectives are chosen such that
the objectives conflict with each other. The conflict between
Step 10. By adding all of the normalized objective values for objectives depends on the correlation among the objectives.
each solution, obtain the sum of the normalized objective Different objectives will have different degrees of correlation
values [32]. among the combination of objectives. To formulate the
For m � 1 to M, combination of objectives, four different objective functions
Calculate the max and min objectives of the mth objective are considered, which are as follows: (i) TC, (ii) TE, (iii) APL,
and find its range. and (iv) VMD. A total of ten different case studies are con-
Normalize the mth objective values using the expression: sidered on three standard test systems to test the efficiency of
the proposed method for the MOOPF problem. The various
fm (x) − fmin
′ (x) �
fm . (33) case studies considered in this study are given in Table 1.
fmax − fmin Numerous trials with various control parameters were
conducted, and the best findings obtained are summarized
End.
in this study. The parameters chosen for each method are
For i � 1 to N.
listed in Table 2.
Add up all normalized objectives to get a unique value.
End.
5.1. Modified IEEE 30-Bus System. The IEEE 30-bus power
Step 11. Calculate the Euclidean space between all of the system has 6 thermal generators placed at buses 1, 2, 5, 8, 11,
solutions and the reference point. and 13 (# 1 generator as a slack generator), 41 lines. In this
study, 4 off-nominal transformers are considered between
Step 12. Set a stopping point for the individual with the lines 6–10, 6–9, 4–12, and 27–28, and 9 shunt VAR com-
shortest path to the original point. pensators are placed at the buses. The whole real and reactive
power demand on the system is 238.40 MW and
Step 13. Divide the objective range into 100 bins, and scan 126.20MVAR, respectively. In addition to the above thermal
all bins till you reach the stopping point. The solution having generators, one wind generator and one solar generator are
the least summation value will be picked to enter into the added to buses 22 and 25, respectively. Detailed information
preferential set for each scanned bin. about the test system is provided in [39, 40].
International Transactions on Electrical Energy Systems 7

Table 2: Control parameters used in different methods.


S. No. Method Control parameters
Population size (N) � 100, number of divisions made along with every object (D) � 12, neighborhood size
1. Proposed method
(T) � 20, crossover probability (Pc) � 1.0, mutation probability (Pm) � 0.05, and number of iterations � 100.
Population size (N) � 100, number of divisions made along with every object (D) � 12, neighborhood size
2. MOEA/D [33]
(T) � 20, crossover probability (Pc) � 1.0, mutation probability (Pm) � 0.05, and number of iterations � 100.
Population size (N) � 100, crossover probability (Pc) � 0.8, No. of iterations � 100, and mutation probability
3. NSGA-II [34]
(Pm) � 0.01.
4. MOPSO [35] Population size (N) � 100, C1 � C2 � 2, W � 0.5, and number of iterations � 100.

Table 3: IEEE 30-bus system: best-compromised values obtained by the proposed method for Case-1 to Case-5.
Limits
S. no. Control variables Control variables at bus/line Case-1 Case-2 Case-3 Case-4 Case-5
Min Max
1. 2 20 80 49.3631 43.5939 49.7998 49.9396 49.9289
2. 5 15 50 26.7537 30.5688 36.3853 25.7456 41.2671
3. 8 10 35 27.5324 27.0078 30.5428 24.1533 29.4956
4. Power (MW) 11 10 30 18.8816 20.4392 25.7521 22.0017 21.0325
5. 13 12 40 23.9297 21.8609 27.6149 23.8890 27.9709
6. 22 0 50 31.2931 33.0888 29.7074 32.1679 32.2785
7. 25 0 50 35.4439 29.3993 31.1763 33.7683 28.2922
8. 1 0.95 1.1 1.0427 1.0496 1.0476 1.0301 1.0212
9. 2 0.95 1.1 1.0354 1.0392 1.0387 1.0245 1.0147
10. 5 0.95 1.1 1.0008 1.0131 1.0174 1.0154 1.0011
11. 8 0.95 1.1 1.0113 1.0285 1.0346 0.9869 0.9972
Voltage (p.u)
12. 11 0.95 1.1 1.0126 1.0260 1.0020 1.0104 1.0070
13. 13 0.95 1.1 1.0297 1.0171 1.0148 1.0165 1.0098
14. 22 0.95 1.1 1.0303 1.0256 1.0174 1.0030 1.0067
15. 25 0.95 1.1 1.0386 1.0422 1.0267 1.0190 1.0212
16. 11 0.9 1.1 1.0144 1.0257 0.9974 1.0143 1.0220
17. 12 0.9 1.1 1.0319 1.0350 1.0238 1.0051 0.9908
Tap ratio
18. 15 0.9 1.1 1.0044 0.9896 0.9943 0.9677 0.9719
19. 36 0.9 1.1 0.9880 0.9955 1.0249 0.9660 0.9721
20. 10 0 5 3.1425 1.9772 2.0710 3.0865 2.7684
21. 12 0. 5 1.9108 2.5250 3.7790 1.9405 1.6628
22. 15 0 5 1.8903 2.5177 2.1385 2.9160 3.9393
23. 17 0 5 2.4423 2.9065 3.1406 2.9999 2.5418
24. Shunt VAR compensator (MVAR) 20 0 5 2.2654 2.7249 2.4215 3.6789 2.9067
25. 21 0 5 2.3629 1.3956 2.2982 2.1017 1.7645
26. 23 0 5 2.9082 2.7610 2.8149 1.9246 1.6307
27. 24 0 5 2.3035 2.9588 2.2973 1.9558 3.3332
28. 29 0 5 2.7370 2.6593 2.4329 2.9911 2.3346
1. TC ($/h) - - - 794.0907 798.6845 838.0936 799.7882 851.9069
2 TE (ton/h) - - - 0.2166 - 0.2049 0.2172 0.2057
3. APL (MW) - - - - 3.9899 3.2506 - 3.1972
4. VMD (p.u.) - - - - - - 0.0902 0.1038

5.1.1. Case-1: Simultaneously Minimize TC and TE. In this 5.1.2. Case-2: Simultaneously Minimize TC and APL. In this
case, TC and TE are the objectives considered to simulta- case, TC and APL are the objectives considered to simul-
neously minimize. The optimal decision variables obtained taneously minimize. The optimal decision variables obtained
by the suggested method are included in Table 3. The best- by the suggested method are included in Table 3. The best-
compromised values that could be found using the proposed compromised values that could be found using the proposed
algorithm have a TC of 794.0907$/h and a TE of 0.2166ton/ algorithm have a TC of 798.6845$/h and an APL of
h, which is the lowest value compared with MOEA/D [33], 3.9899 MW, which is the lowest value compared with
NSGA-II [34], and MOPSO [35] as shown in Table 4. The MOEA/D [33], NSGA-II [34], and MOPSO [35] as shown in
Pareto optimal fronts of all the methods are depicted in Table 4. The Pareto optimal fronts of all the methods are
Figure 1. depicted in Figure 2.
8 International Transactions on Electrical Energy Systems

Table 4: IEEE 30-bus system: comparison of the proposed method with MOEA/D [33], NSGA-II [34], and MOPSO [35] for Case-1 to Case-
5.
Case name Objective functions Proposed method MOEA/D [33] NSGA-II [34] MOPSO [35]
TC ($/h) 794.0907 794.2012 794.4894 794.1736
Case-1
TE (ton/h) 0.2166 0.2170 0.2171 0.2203
TC ($/h) 798.6845 798.9000 805.4298 816.1819
Case-2
APL (MW) 3.9899 4.0790 4.0319 4.2422
TC ($/h) 838.0936 840.0000 848.6240 870.9164
Case-3 TE (ton/h) 0.2048 0.2043 0.2062 0.2198
APL (MW) 3.2506 3.7170 3.3119 4.1321
TC ($/h) 799.7880 801.7412 800.9397 831.3916
Case-4 TE (ton/h) 0.2172 0.2164 0.2173 0.2386
VMD (p.u.) 0.0902 0.1271 0.1229 0.1434
TC ($/h) 851.9069 855.4589 858.7833 862.8927
TE (ton/h) 0.2057 0.2101 0.2137 0.2540
Case-5
APL (MW) 3.1972 3.1997 3.2025 3.4352
VMD (p.u.) 0.1038 0.1912 0.2635 0.4925

0.26
7
0.25
6
0.24
APL (MW)

5
TE (ton/h)

0.23
4
0.22
3

0.21 2
0.3
0.2 0.25
TE 0.2 950
(to 900
n/h 850
0.19 ) 0.15 800
750
760 780 800 820 840 860 880 TC ($/h)
TC ($/h)
Proposed Method NSGA-II
Proposed Method NSGA-II
MOEA/D MOPSO
MOEA/D MOPSO
Figure 3: Case-3: IEEE 30-bus system Pareto optimal fronts.
Figure 1: Case-1: IEEE 30-bus system Pareto optimal fronts.

5.5

5
0.8
4.5
APL (MW)

0.6
VMD (p.u)

4
0.4
3.5
0.2
3
0 950
2.5 0.28 900
0.26
0.24 850
h)
2 TE (t
0.22
0.2
800 ($/
760 780 800 820 840 860 880 900 920 on/h
) 0.18 750 TC
TC ($/h)

Proposed Method NSGA-II Proposed Method NSGA-II


MOEA/D MOPSO MOEA/D MOPSO
Figure 2: Case-2: IEEE 30-bus system Pareto optimal fronts. Figure 4: Case-4: IEEE 30-bus system Pareto optimal fronts.
International Transactions on Electrical Energy Systems 9

Table 5: IEEE 57-bus system: best-compromised values obtained by the proposed method for Case-6 to Case-8.
Limits
S. no. Control variables Control variables at bus/line Case-6 Case-7 Case-8
Min Max
1. 2 0 100 98.1298 67.1046 74.8586
2. 3 0 140 69.4063 55.0118 64.1958
3. 6 0 100 70.9842 98.4255 52.0615
4. 8 0 550 329.0458 306.7587 315.7318
Power (MW)
5. 9 0 100 72.7441 99.3728 98.8024
6. 12 0 410 315.2646 341.6928 378.7730
7. 45 0 80 79.6017 79.9551 79.8243
8. 46 0 80 79.8919 79.9311 79.5757
9. 1 0.95 1.1 1.0481 1.0296 1.0391
10. 2 0.95 1.1 1.0371 1.0246 1.0333
11. 3 0.95 1.1 1.0340 1.0227 1.0229
12. 6 0.95 1.1 1.0275 1.0185 1.0209
13. Voltage (p.u.) 8 0.95 1.1 1.0295 1.0162 1.0318
14. 9 0.95 1.1 1.0169 1.0099 1.0160
15. 12 0.95 1.1 1.0369 1.0268 1.0217
16. 45 0.95 1.1 1.0471 1.0498 1.0514
17. 46 0.95 1.1 1.0209 1.0372 1.0175
18. 19 0.9 1.1 1.0154 1.0139 1.0056
19. 20 0.9 1.1 0.9945 1.0497 1.0367
20. 31 0.9 1.1 1.0183 1.0260 0.9955
21. 35 0.9 1.1 0.9938 1.0263 0.9876
22. 36 0.9 1.1 0.9601 0.9982 0.9821
23. 37 0.9 1.1 0.9943 1.0176 1.0321
24. 41 0.9 1.1 1.0225 0.9911 1.0155
25. 46 0.9 1.1 0.9889 0.9757 0.9456
26. Tap ratio 54 0.9 1.1 0.9999 0.9233 0.9049
27. 58 0.9 1.1 0.9814 0.9802 0.9613
28. 59 0.9 1.1 1.0108 0.9877 1.0070
29. 65 0.9 1.1 0.9914 0.9841 0.9967
30. 66 0.9 1.1 0.9748 0.9484 0.9140
31. 71 0.9 1.1 0.9703 0.9756 0.9547
32. 73 0.9 1.1 1.0158 0.9829 1.0058
33. 76 0.9 1.1 0.9691 0.9769 0.9649
34. 80 0.9 1.1 0.9908 0.9872 1.0199
35. 18 0 20 9.1150 11.4035 11.0379
36. Shunt VAR compensator (MVAR) 25 0 20 9.8438 10.4059 8.2934
37. 53 0 20 11.2830 7.1925 7.7894
1. TC ($/h) - - - 36195.21 36096.69 36207.21
2. TE (ton/h) - - - 1.0182 1.0238 1.1383
3. APL (MW) - - - - 10.3303 9.9732
4. VMD (p.u.) - - - - - 0.6848

5.1.3. Case-3: Simultaneously Minimize TC, TE, and APL. decision variables obtained by the suggested method are
In this case, TC, TE, and APL are the objectives con- included in Table 3. The best-compromised values that
sidered to simultaneously minimize. The optimal decision could be found using the proposed algorithm have a TC
variables obtained by the suggested method are included of 799.7880$/h, a TE of 0.2172ton/h, and a VMD of
in Table 3. The best-compromised values that could be 0.0902p.u., which is the lowest value compared with
found using the proposed algorithm have a TC of MOEA/D [33], NSGA-II [34], and MOPSO [35] as shown
838.0936 $/h, a TE of 0.2049ton/h, and an APL of in Table 4. The Pareto optimal fronts of all the methods
3.2506 MW, which is the lowest value compared with are depicted in Figure 4.
MOEA/D [33], NSGA-II [34], and MOPSO [35] as shown
in Table 4. The Pareto optimal fronts of all the methods are
depicted in Figure 3. 5.1.5. Case-5: Simultaneously Minimize TC, TE, APL, and
VMD. In this case, TC, TE, APL, and VMD are the ob-
jectives considered to be simultaneously minimized. The
5.1.4. Case-4: Simultaneously Minimize TC, TE, and VMD. optimal decision variables obtained by the suggested method
In this case, TC, TE, and VMD are the objectives con- are included in Table 3. The best-compromised values that
sidered to simultaneously minimize. The optimal could be found using the proposed algorithm have a TC of
10 International Transactions on Electrical Energy Systems

Table 6: IEEE 57-bus system: comparison of the proposed method with MOEA/D [33], NSGA-II [34], and MOPSO [35] for Case-6 to Case-
8.
Case name Objective functions Proposed method MOEA/D [33] NSGA-II [34] MOPSO [35]
TC ($/h) 36195.21 36198.87 36399.10 36733.34
Case-6
TE (ton/h) 1.0182 1.0271 1.0912 1.1145
TC ($/h) 36096.69 36990.02 36363.70 39208.74
Case-7 TE (ton/h) 1.0238 1.0782 1.1288 1.0890
APL (MW) 10.3303 10.7016 10.7953 11.0434
TC ($/h) 36207.21 36317.56 36479.38 37321.91
TE (ton/h) 1.0916 1.1256 1.1382 1.2049
Case-8
APL (MW) 9.9732 11.2487 11.3923 14.5232
VMD (p.u.) 0.6848 0.6954 0.8907 0.8323

1.25

1.2
18
1.15
16
1.1
TE (ton/h)

APL (MW)
1.05 14

1 12
0.95
10 4.5 4
0.9 0
4 ×1
)
8 /h
0.85 1.5 ($
1.4 1.3 1.2 1.1 3.5 TC
1 0.9
0.8 TE (ton/h)
3.55 3.63 3.65 3.7 3.75
TC ($/h) ×104 Proposed Method NSGA-II
Proposed Method NSGA-II MOEA/D MOPSO
MOEA/D MOPSO Figure 6: Case-7: IEEE 57-bus system Pareto optimal fronts.
Figure 5: Case-6: IEEE 57-bus system Pareto optimal fronts.

best compromise solution that could be found using the


851.9069$/h, a TE of 0.2057ton/h, APL of 3.1972 MW, and proposed algorithm has a TC of 36195.21$/h and a TE of
VMD of 0.1038p.u., which is the lowest value compared 1.0182ton/h, which is the lowest value compared with
with MOEA/D [33], NSGA-II [34], and MOPSO [35] as MOEA/D [33], NSGA-II [34], and MOPSO [35] as shown in
shown in Table 4. Table 6. The Pareto optimal fronts of all the methods are
depicted in Figure 5.

5.2. Modified IEEE 57-Bus System. To show the scalability of


the proposed algorithm, the IEEE 57-bus system is used for 5.2.2. Case-7: Simultaneously Minimize TC, TE, and APL.
solving the MOOPF problem. It contains 7 thermal gen- In this case, TC, TE, and APL are the objectives that need
erators placed at buses 1, 2, 3, 6, 8, 9, and 12 (# 1 generator as simultaneous minimizing. The optimal decision variables
a slack generator), 80 lines. In this study, 15 off-nominal obtained by the suggested method are included in Table 5.
transformers are considered along with 3 shunt VAR The best-compromised values that could be found using the
compensators. The entire real and reactive power demand proposed algorithm have a TC of 36096.69$/h, a TE of
on the system is 1250.80 MW and 336.40MVAR, respec- 1.0238ton/h, and an APL of 10.3303 MW, which is the
tively. In addition to the above thermal generators, one wind lowest value compared with MOEA/D [33], NSGA-II [34],
generator and one solar unit are added at buses 45 and 46, and MOPSO [35] as shown in Table 6. The Pareto optimal
respectively. Detailed information about the test system is fronts of all the methods are depicted in Figure 6.
provided in [39, 40].
5.2.3. Case-8: Simultaneously Minimize TC, TE, APL, and
5.2.1. Case-6: Simultaneously Minimize TC and TE. In this VMD. In this case, TC, TE, APL, and VMD are the ob-
case, TC and TE are the objectives that need to be simul- jectives that need to be simultaneously minimized. The
taneously minimized. The optimal decision variables ob- optimal decision variables obtained by the suggested method
tained by the suggested method are included in Table 5. The are included in Table 5. The best-compromised values that
International Transactions on Electrical Energy Systems 11

Table 7: IEEE 118-bus system: best-compromised values obtained by the proposed method for Case-9 to Case-10.
Limits Bus/line Limits
Variables Bus/line Case-9 Case-10 Variables Case-9 Case-10
min max min max
1 0 100 31.118 52.020 31 0.95 1.1 1.0253 1.0169
4 0 100 21.400 41.520 32 0.95 1.1 1.0195 1.0030
6 0 100 42.530 40.834 34 0.95 1.1 0.9909 1.0094
8 0 100 28.556 26.448 36 0.95 1.1 1.0241 1.0067
10 0 550 273.35 245.999 40 0.95 1.1 1.0096 1.0118
12 0 185 86.663 95.235 42 0.95 1.1 1.0305 1.0113
15 0 100 43.268 37.779 46 0.95 1.1 1.0150 1.0221
18 0 100 99.997 45.659 49 0.95 1.1 1.0118 1.0055
19 0 100 35.781 55.328 54 0.95 1.1 1.0359 1.0023
24 0 100 64.585 35.747 55 0.95 1.1 1.0708 1.0239
25 0 320 182.60 90.928 56 0.95 1.1 1.0661 1.0256
26 0 414 0.000 162.842 59 0.95 1.1 1.0619 1.0271
27 0 100 24.090 46.656 61 0.95 1.1 1.1000 1.0175
31 0 107 22.783 25.781 62 0.95 1.1 1.0982 0.9985
32 0 100 62.386 37.325 65 0.95 1.1 1.0875 1.0049
34 0 100 43.187 40.202 66 0.95 1.1 1.0434 1.0099
36 0 100 100.00 54.841 69 0.95 1.1 1.0436 1.0280
40 0 100 88.192 64.417 70 0.95 1.1 1.0184 1.0096
42 0 100 83.016 49.460 72 0.95 1.1 1.0066 1.0062
46 0 119 19.417 44.338 73 0.95 1.1 1.0092 1.0190
49 0 304 138.66 140.370 74 0.95 1.1 1.0119 1.0264
54 0 148 59.984 98.543 76 0.95 1.1 1.0045 1.0089
55 0 100 74.764 52.457 77 0.95 1.1 1.0367 1.0170
56 0 100 59.627 46.8259 80 0.95 1.1 1.0229 1.0183
59 0 255 117.12 115.881 85 0.95 1.1 0.9985 1.0125
61 0 260 121.66 121.893 87 0.95 1.1 0.9617 1.0290
62 0 100 35.517 44.630 89 0.95 1.1 1.0269 1.0266
Power (MW) 65 0 491 214.33 213.346 90 0.95 1.1 1.0321 1.0253
66 0 492 205.83 187.419 91 0.95 1.1 1.0209 1.0221
70 0 100 12.582 54.0066 92 0.95 1.1 1.0237 1.0056
72 0 100 12.141 40.751 99 0.95 1.1 1.0297 1.0222
73 0 100 55.579 50.455 100 0.95 1.1 1.0412 1.0206
74 0 100 14.137 42.021 103 0.95 1.1 1.0269 1.0326
76 0 100 75.678 37.592 104 0.95 1.1 1.0484 1.0319
77 0 100 82.194 42.539 105 0.95 1.1 1.0531 1.0203
80 0 577 256.72 270.903 107 0.95 1.1 1.0380 1.0324
85 0 100 42.579 42.382 110 0.95 1.1 1.0685 1.0243
87 0 104 0.000 19.159 111 0.95 1.1 1.0854 1.0296
89 0 707 257.13 216.783 112 0.95 1.1 1.0619 1.0349
90 0 100 97.811 36.604 113 0.95 1.1 1.0228 1.0236
91 0 100 8.436 52.187 116 0.95 1.1 1.0519 1.0073
92 0 100 45.760 43.249 64 0.95 1.1 1.0160 1.0149
99 0 100 23.885 40.178 65 0.95 1.1 1.0453 1.0229
100 0 352 113.62 150.771 8 0.9 1.1 0.9897 1.0013
103 0 140 42.612 50.598 32 0.9 1.1 1.0611 1.0167
104 0 100 11.805 45.019 36 0.9 1.1 0.9508 0.9920
105 0 100 100.00 63.106 51 0.9 1.1 1.0001 0.9742
107 0 100 19.672 39.046 Tap ratio 93 0.9 1.1 0.9995 1.0096
110 0 100 56.657 53.624 95 0.9 1.1 0.9000 1.0180
111 0 136 22.867 43.765 102 0.9 1.1 1.0099 1.0270
112 0 100 40.659 39.005 107 0.9 1.1 0.9262 0.9814
113 0 100 16.067 51.229 127 0.9 1.1 0.9770 0.9995
116 0 100 39.564 44.049 34 0 25 6.1168 14.2885
64 0 100 99.998 74.484 44 0 25 14.0212 11.8200
65 0 100 99.998 61.143 45 0 25 24.0656 12.7371
1 0.95 1.1 1.0554 1.0093 46 0 25 9.5124 15.7212
4 0.95 1.1 0.9500 1.0106 48 0 25 5.8489 14.8892
6 0.95 1.1 0.9753 1.0256 74 0 25 19.2222 10.8066
8 0.95 1.1 0.9585 1.0102 Shunt VAR compensator (MVAR) 79 0 25 0.0000 15.8930
10 0.95 1.1 1.0381 1.0205 82 0 25 24.7411 13.8225
12 0.95 1.1 1.0502 1.0141 83 0 25 13.9104 11.6974
Voltage (p.u.) 15 0.95 1.1 0.9529 1.0094 105 0 25 24.9944 13.5413
18 0.95 1.1 0.9960 1.0220 107 0 25 19.0262 15.0743
19 0.95 1.1 1.0077 1.0321 110 0 25 12.1782 11.1250
24 0.95 1.1 0.9956 1.0184
25 0.95 1.1 1.0117 1.0251 TC ($/h) - - - 132958.66 135774.93
26 0.95 1.1 1.0397 1.0227 APL (MW) - - - 31.2916 39.6333
27 0.95 1.1 1.0466 1.0121 VMD (p.u) - - - - 0.4299
12 International Transactions on Electrical Energy Systems

Table 8: IEEE 118-bus systems: comparison of the proposed method with MOEA/D [33], NSGA-II [34], and MOPSO [35] for Case-9 to
Case-10.
Case name Objective functions Proposed method MOEA/D [33] NSGA-II [34] MOPSO [35]
TC ($/h) 132958.66 133249.84 133837.90 134673.5
Case-9
APL (MW) 31.2916 31.8104 31.8664 35.3868
TC ($/h) 135774.93 135801.21 135912.8 136459.9
Case-10 APL (MW) 39.6333 42.0412 45.6904 48.3446
VMD (p.u.) 0.4299 0.4523 0.5074 0.5878

could be found using the proposed algorithm have a TC of 40


36207.21$/h, a TE of 1.0916ton/h, APL of 9.9732 MW, and
VMD of 0.6848p.u., which is the lowest value compared 38
with MOEA/D [33], NSGA-II [34], and MOPSO [35] as
shown in Table 6.
36

APL (MW)
5.3. Modified IEEE 118-Bus System. To show the scalability of 34
the proposed algorithm for a large-scale test system in
solving the MOOPF problem, the IEEE 118-bus system is 32
considered. It contains 54 thermal generators (# 69 gener-
ator as a slack generator) and 186 lines. In this study, 9 off-
30
nominal transformers and 12 shunt VAR compensators are
considered. The sum of real and reactive power demand on
the system is 4242.00 MW and 1439.00MVAR, respectively. 28
1.31 1.32 1.33 1.34 1.35 1.36 1.37
In addition to the above thermal generators, one wind TC ($/h) ×105
generator and one solar unit are added to buses 63 and 64,
respectively. Detailed information about the test system is Proposed Method NSGA-II
provided in [39, 40]. MOEA/D MOPSO
Figure 7: Case-9: IEEE 118-bus system Pareto optimal fronts.

5.3.1. Case-9: Simultaneously Minimize TC and APL. In this


case, TC and APL are the objectives that need to be si-
multaneously minimized. The optimal decision variables
obtained by the suggested method are included in Table 7. 0.7
The best compromise solution that could be found using the
proposed algorithm has a TC of 132958.66$/h and an APL 0.65
of 31.2916 MW, which is the lowest value compared with 0.6
VMD (p.u)

MOEA/D [33], NSGA-II [34], and MOPSO [35] as shown in


Table 8. The Pareto optimal fronts of all the methods are 0.55
depicted in Figure 7. 0.5

0.45
5.3.2. Case-10: Simultaneously Minimize TC, APL, and 0.4
VMD. In this case, TC, APL, and VMD are the objectives 45
1.4
that need simultaneous minimizing. The optimal decision 40 1.38
APL 1.36 5
variables obtained by the suggested method are included in (MW 35 1.32
1.34 ×10
)
) TC ($/h
Table 7. The best compromise solution that could be ob-
tained using the proposed algorithm has a total cost of Proposed Method NSGA-II
135774.93$/h, APL of 39.6333 MW, and VMD of MOEA/D MOPSO
0.4299p.u., which is the lowest value compared to MOEA/D
Figure 8: Case-10: IEEE 118-bus system Pareto optimal fronts.
[33], NSGA-II [34], and MOPSO [35] as shown in Table 8.
The Pareto optimal fronts of all the methods are depicted in
Figure 8. proposed method, MOEA/D, NSGA-II, and MOPSO for all
the cases are given in Table 9. The computational times of the
proposed method are significantly faster than those of other
5.4. Computational Time. In this study, the MOOPF studied methods for all cases. Hence, the proposed method
problem was executed on a 2.00 GHz, i3 processor, with a outperformed the other methods in terms of solution quality
4 GB RAM computer. The computational (CPU) times of the and computing time.
International Transactions on Electrical Energy Systems 13

Table 9: Comparison of CPU time (sec).


Case name Proposed method MOEA/D [33] NSGA-II [34] MOPSO [35]
Case-1 845.19 1333.57 1615.12 1014.31
Case-2 853.92 1413.40 1460.24 1029.56
IEEE 30-bus system Case-3 856.94 1342.81 1700.27 1198.45
Case-4 915.49 1351.51 1500.08 1201.34
Case-5 1009.63 1372.80 1750.88 1279.83
Case-6 1105.04 1672.55 1967.01 1401.21
IEEE 57-bus system Case-7 1146.21 1709.16 2046.42 1456.89
Case-8 1190.14 1887.68 2054.09 1523.47
Case-9 1535.33 2771.16 2946.30 1881.56
IEEE 118-bus system
Case-10 1705.10 2833.67 3219.41 2015.64

6. Conclusions Pwav,j , Psav,k : The actual power output of jth


This study proposes a solution to the MOOPF problem with windmill and kth PV plant, respectively
a combination of thermal, wind, and PV systems using KRw,j , KPw,j : Reserve and penalty price coefficients
MOEA based on decomposition and summation of nor- of jth windmill, respectively
malized objectives with an improved diversified selection KRs,k , KPs,k : Reserve and penalty price coefficients
method. The method also deals with tackling various con- of kth PV plant, respectively
straints in the MOOPF problem using the superiority of the ai , bi , ci : ith Generator cost coefficients
feasible solution (SF) technique. The fuel costs of thermal NTG , NWG , and Number of thermal, wind, and solar
generators and uncertainty prices associated with wind and NSG : power plants, respectively
PV energy systems are minimized along with carbon αi , β i , c i , ξ i , λ i : ith generator emission coefficients
emission, active power losses, and voltage magnitude de- Gij , Bij : Conductance and susceptance between
viation. Monte Carlo simulations were used to assess the buses i and j
uncertainty of wind and solar power. Apart from the con- NB , NTG , NC , NPQ , Number of buses, thermal generators,
ventional cost minimization, this study selects factors to and NT : shunt VAR compensators, PQ buses,
account for the uncertain price of available wind and solar and transformers, respectively
power. It depicts the OPF formulation along with factors Pmin
Gi , PGi :
max
Min-max limits on ith generator real
affecting wind and PV power’s intermittency. To show the power
efficacy of the suggested method, simulations were per- Qmin
Gi , QGi :
max
Min-max limits on ith generator
formed on the same test systems as with MOEA/D, NSGA- reactive power
II, and MOPSO algorithms. The results show the superiority S, Smax : Apparent power flow and its maximum
of the proposed method compared to other methods. Hence, limit, respectively
the proposed method can be effectively used in operation Tmin
k , Tk :
max
Min-max limits of kth transformer tap
and control when wind and solar power generation are positions
included in the power system. Vmin
i , V max
i : Min-max limits of ith bus voltages
θij : Voltage angle between buses i and j
PG i , Q G i : Real and reactive power injection at ith
Abbreviations bus
PDi , QDi : Real and reactive power demand at ith
k, c: Shape and scale factors, respectively bus
v: Wind speed (m/sec) w: Weight vector
vin , vout , vr : Cut-in, cut-out, and rated wind speeds, nl : Number of lines
respectively M: Number of objectives
Gs : Solar irradiance fm′: Normalized mth objective
μ: Mean MOEA: Multiobjective evolutionary algorithm
σ: Standard deviation MOOPF: Multiobjective optimal power flow
Pwr , Psr : Rated power of wind and solar plants, PV: Photovoltaic
respectively WT: Wind turbine
Pws , Pss : Scheduled power of wind and solar ISO: Independent system operator
plants, respectively NSGA-II: Nondominated sorting genetic
g j , hk : Direct price coefficients of jth a algorithm-II
windmill and kth solar plant, MOPSO: Multiobjective particle swarm
respectively optimization.
14 International Transactions on Electrical Energy Systems

Data Availability systems via differential evolutionary particle swarm optimi-


zation,” Int. Trans. Elect. Energy Syst., vol. 30, no. 4, p. 12270,
The data supporting these findings are from previously 2020.
reported studies and datasets, which have been cited. [15] S. Duman, J. Li, L. Wu, and U. Guvenc, “Optimal power flow
with stochastic wind power and FACTS devices: a modified
Conflicts of Interest hybrid PSOGSA with chaotic maps approach,” Neural
Computing & Applications, vol. 32, pp. 8463–8492, 2019.
The authors declare that they have no conflicts of interest. [16] S. R. Salkuti, V. Sandeep, B. C. babu, and C.-M. Jung, “Multi-
objective based optimal generation scheduling considering
wind and solar energy systems,” International Journal of
References Emerging Electric Power Systems, vol. 19, no. 5, Article ID
[1] C. Haiyan, C. Jinfu, and D. Xian Zhong, “Multi-stage dynamic 20180006, 2018.
optimal power flow in wind power integrated system,” in [17] K. Deb, Optimization for Engineering Design: Algorithms and
Proceedings of the IEEE/PES Transmission and Distribution Examples, Prentice-Hall, New Delhi, India, 1995.
Conference and Exhibition: Asia and Pacific, pp. 1–5, Dalian, [18] G. V. Reklaitis, A. Ravindran, and K. M. Ragsdell, Engineering
China, June 2005. Optimization Methods and applications, Wiley, New York,
[2] A. Domingos and M. P. M. Pedro, Optimal Power Flow In- NY, USA, 1983.
cluding Wind Generation, Instituto Superior Tecnico, Tech- [19] C. Chun-Lung, T. Y. Lee, and R. M. Jan, “Optimal wind-
nical University of Lisbon, Lisbon, Portugal, 2012. thermal coordination dispatch in isolated power systems with
[3] P. P. Biswas, P. N. Suganthan, and G. A. J. Amaratunga, large integration of wind capacity,” Energy Conversion and
“Optimal power flow solutions incorporating stochastic wind Management, vol. 47, pp. 3456–3472, 2006.
and solar power,” Energy Conversion and Management, [20] J. Hetzer, D. C. Yu, and K. Bhattarai, “An economic dispatch
vol. 148, pp. 1194–1207, 2017. model incorporating wind power,” IEEE Transactions on
[4] M. H. Sulaiman, Z. Mustaffa, A. J. Mohamad, M. M. Saari, and Energy Conversion, vol. 23, no. 2, pp. 603–611, 2008.
M. R. Mohamed, “Optimal power flow with stochastic solar [21] T. P. Chang, “Investigation on frequency distribution of
power using barnacles mating optimizer,” Int. Trans. Electr. global radiation using different probability density functions,”
Energy Syst., vol. 66, pp. 88–93, 2021. Int. J. of Appl. Science and Engg., vol. 8, no. 2, pp. 99–107, 2010.
[5] C. Mishra, S. P. Singh, and J. Rokadia, “Optimal power flow in [22] S. S. Reddy, P. R. Bijwe, and A. R. Abhyankar, “Real-time
the presence of wind power using modified cuckoo search,” economic dispatch considering renewable power generation
IET Generation, Transmission & Distribution, vol. 9, no. 7, variability and uncertainty over scheduling period,” IEEE
pp. 615–626, 2015. Systems Journal, vol. 9, no. 4, pp. 1440–1451, 2015.
[6] K. Teeparthi and D. M. Vinod Kumar, “Security-constrained [23] A. Panda and M. Tripathy, “Security constrained optimal
optimal power flow with wind and thermal power generation power flow solution of wind-thermal generation system using
using fuzzy adaptive artificial physics optimization algo- bacterial foraging algorithm,” Energy, vol. 93, pp. 816–827,
rithm,” Neural Computing & Applications, vol. 29, no. 3, 2015.
pp. 855–871, 2016. [24] L. Shi, C. Wang, L. Yao, Y. Ni, and M. Bazargan, “Optimal
[7] H. T. J. Ranjit Roy, “Optimal power flow solution of power power flow solution incorporating wind power,” IEEE Systems
system incorporating stochastic wind power using Gbest Journal, vol. 6, no. 2, pp. 233–241, 2012.
guided artificial bee colony algorithm,” International Journal [25] O. Aslac and B. Stott, “Optimal load flow with steady-state
of Electrical Power & Energy Systems, vol. 64, pp. 562–578,
security,” IEEE Transactions on Power Apparatus and Systems,
2015.
vol. 3, pp. 745–751, 1974.
[8] S. Duman, J. Li, and L. Wu, “AC optimal power flow with
[26] R. D. Zimmerman, C. E. Murillo-Sanchez, and R. J. Thomas,
thermal-wind-solar-tidal systems using the symbiotic or-
“MATPOWER,” http://www.pserc.cornell.edu/matpower.
ganisms search algorithm,” IET Renewable Power Generation,
[27] T. Ackermann, Wind Power in Power System, John Wiley &
vol. 15, pp. 278–296, 2021.
Sons, New York, NY, USA, Chapter 43, 2012.
[9] E. E. Elattar and S. K. Elsayed, “Modified JAYA algorithm for
[28] S. Eftekharnejad, V. Vittal, G. T. Heydt, B. Keel, and J. Loehr,
optimal power flow incorporating renewable energy sources
considering the cost, emission, power loss and voltage profile “Impact of increased penetration of photovoltaic generation
improvement,” Energy, vol. 178, pp. 598–609, 2019. on power systems,” IEEE Transactions on Power Systems,
[10] Z. Ullah, S. Wang, J. Radosavljevic, and J. Lai, “A solution to vol. 28, no. 2, pp. 893–901, 2013.
the optimal power flow problem considering WT and PV [29] K. Deb, “An efficient constraint handling method for genetic
generation,” IEEE Access, vol. 7, pp. 46763–46772, 2019. algorithms,” Computer Methods in Applied Mechanics and
[11] P. K. Roy and C. Paul, “Optimal power flow using krill herd Engineering, vol. 186, pp. 311–338, 2000.
algorithm,” Int. Trans. Electr. Energy Syst., vol. 25, 2014. [30] K. Li, K. Deb, Q. Zhang, and S. Kwong, “An evolutionary
[12] M. A. Taher, S. Kamel, F. Jurado, and E. Mohamed, “An many-objective optimization algorithm based on dominance
improved moth-flame optimization algorithm for solving and decomposition,” IEEE Transactions on Evolutionary
optimal power flow problem,” Int. Trans. Electr. Energy Syst., Computation, vol. 19, no. 5, pp. 694–716, 2015.
vol. 29, 2018. [31] Q. Zhang and H. Li, “MOEA/D: a multi objective evolutionary
[13] A. Saha, A. Bhattacharya, P. Das, and A. K. Chakraborty, “A algorithm based on decomposition,” IEEE Transactions on
novel approach towards uncertainty modeling in multi-ob- Evolutionary Computation, vol. 11, no. 6, pp. 712–731, 2007.
jective optimal power flow with renewable integration,” Int. [32] B. Y. Qu and P. N. Suganthan, “Multi-objective differential
Trans. Elect. Energy Syst., vol. 29, no. 12, p. 12136, 2019. evolution based on the summation of normalized objectives
[14] S. Duman, S. Rivera, J. Li, and L. Wu, “Optimal power flow of and improved selection method,” in Proceedings of the IEEE
power systems with controllable wind-photovoltaic energy Symposium, pp. 1–8, Anchorage, AL, USA, May 2011.
International Transactions on Electrical Energy Systems 15

[33] J. Zhang, Q. Tang, P. Li, D. Deng, and Y. Chen, “A modified


MOEA/D approach to the solution of multi-objective optimal
power flow problem,” Applied Soft Computing, vol. 47,
pp. 494–514, 2016.
[34] K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and
elitist multi objective genetic algorithm: NSGA-II,” IEEE
Transactions on Evolutionary Computation, vol. 6, no. 2,
pp. 182–197, 2002.
[35] C. Coello, G. T. Pulido, and M. S. Lechuga, “Handling
multiple objectives with particle swarm optimization,” IEEE
Transactions on Evolutionary Computation, vol. 8, no. 3,
pp. 256–279, 2004.
[36] I. Das and J. E. Dennis, “Normal –boundary intersection: a
new method for generating Pareto optimal points in multi-
criteria optimization problems,” SIAM Journal on Optimi-
zation, vol. 8, no. 3, pp. 631–657, 1998.
[37] Y. Zhang and Y. Li, “A many-objective evolutionary algorithm
based on decomposition and local dominance,” 2018.
[38] M. A. Abido, “Environmental/economic power dispatch using
multi-objective evolutionary algorithms: a comparative
study,” IEEE Transactions on Power Systems, vol. 1, no. 4,
pp. 1529–1537, 2003.
[39] A. Shabanpour-Haghighi, A. R. Seifi, and T. Niknam, “A
modified teaching-learning based Optimization for multi-
objective optimal power Flow problem,” Energy Conversion
and Management, vol. 77, pp. 597–607, 2014.
[40] R. D. Zimmerman, C. E. Murillo-Sanchez, and R. J. Thomas,
“MATPOWER: steady-state operations, planning, and anal-
ysis tools for power systems research and education,” IEEE
Transactions on Power Systems, vol. 26, pp. 9–12, 2011.

You might also like