You are on page 1of 19

Energy 235 (2021) 121407

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Dynamic economic emission dispatch considering renewable energy


generation: A novel multi-objective optimization approach
Zhi-Feng Liu a, b, Ling-Ling Li a, b, *, Yu-Wei Liu a, b, Jia-Qi Liu a, b, Heng-Yi Li a, b,
Qiang Shen a, b
a
State Key Laboratory of Reliability and Intelligence of Electrical Equipment, Hebei University of Technology, Tianjin, 300130, China
b
Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability of Hebei Province, Hebei University of Technology, Tianjin, 300130, China

a r t i c l e i n f o a b s t r a c t

Article history: This study contributes to construct the mathematical model of hybrid dynamic economic emission
Received 31 January 2021 dispatch (HDEED) considering renewable energy generation and propose a novel solving approach based
Received in revised form on enhanced moth-flame optimization algorithm. Renewable energy power generation technology has
4 June 2021
an important impact on reducing pollutant emissions and promoting sustainable development. There-
Accepted 2 July 2021
Available online 7 July 2021
fore, this study aims to investigate the HDEED problem in consideration of renewable energy generation
and improve the economic and environmental benefits of the power system. First, a moth-flame opti-
mization algorithm based on position disturbance updating strategy (MFO_PDU) was proposed aiming at
Keywords:
Renewable energy
the non-convex, non-linear and high-dimensional characteristics of HDEED problem. Second, the
Multi-objective optimization mathematical model of HDEED on the basis of Wind-Solar-Thermal integrated energy was constructed,
Dynamic economic emission dispatch while taking into account the valve point effect, equality constraints and inequality constraints. Finally,
Moth-flame optimization algorithm three cases including test systems of different scales were formulated and employed to verify the pro-
Sustainable development posed approach, and the compromise solution was determined through membership function. The re-
sults revealed that the fuel cost obtained by the MFO_PDU algorithm was 11.31%, 4.01% and 5.27% smaller
than those of HHO, TSA and MFO algorithms for small-scale test system. Accordingly, the research
outcomes contribute in reducing the fuel cost and pollutant emissions of power generation system, and
further improving the utilization and penetration rate of renewable energy.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction load demand when the slope rate constraints of the generator set
are considered. DEED solves output power scheduling under the
Economic load dispatch (ELD) as a key technique improves the dynamic load demand within 24 h, while minimizing the fuel cost
economic and environmental benefits of the power system under and emissions of the generator set [5,6]. The comparison of static
the condition of equality and inequality constraints [1]. Currently, ELD, DEED and HDEED problems is as follows:
ELD is divided into static ELD, dynamic economic emission dispatch Through the above comparison, it is found that HDEED con-
(DEED) considering both cost and emission objectives, and hybrid siders both renewable energy power generation and thermal po-
DEED (HDEED) considering renewable energy power generation wer generation. In addition, the constraints of the HDEED problem
[2e4]. For the static ELD, the output power of the thermal unit is are more complicated. Considering both the constraints of thermal
allocated at a certain time and under a specific load demand to power units and renewable energy generating units is necessary
minimize the fuel cost, but static ELD cannot adapt to changes in when solving the HDEED problem. The HDEED problem extends
the dispatch scope to a time period compared with the static ELD
problem, and considers the interaction between various time sec-
tions when scheduling. Therefore, the HDEED is a complex multi-
* Corresponding author. State Key Laboratory of Reliability and Intelligence of
objective optimization problem. At present, many experts are
Electrical Equipment, Hebei University of Technology, Tianjin, 300130, China.
E-mail addresses: tjliuzhifeng@126.com (Z.-F. Liu), lilingling@hebut.edu.cn,
committed to studying the HDEED problem of considering wind
lilinglinglaoshi@126.com (L.-L. Li), yuweiresearch@126.com (Y.-W. Liu), power generation, ignoring photovoltaic power generation. How-
201921401082@stu.hebut.edu.cn (J.-Q. Liu), 201931403060@stu.hebut.edu.cn ever, photovoltaic power generation accounts for a high proportion
(H.-Y. Li), 201921401065@stu.hebut.edu.cn (Q. Shen).

https://doi.org/10.1016/j.energy.2021.121407
0360-5442/© 2021 Elsevier Ltd. All rights reserved.
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Nomenclature B Loss coefficient


r Satisfaction degree
Fit Fitness value w Weight coefficient
T 24-h scheduling cycle UR Upper bound of generator output power fluctuation
np Number of solutions in the Pareto solution set (MW/h)
v, t, u Cost coefficients ($/MW2h), ($/MWh), ($/h) DR Lower bound of generator output power fluctuation
d, f Valve point effect coefficient ($/h), (rad/MW) (MW/h)
x, y, z, q, j Emission coefficients (lb/MW2h), (lb/MWh), (lb/h), Pmin Lower constrain of generator output power (MW)
(lb/h), (1/MW) Fmax Upper bound in the Pareto solution set
Pmax Upper constrain of generator output power (MW) Fmin Lower bound in the Pareto solution set
FCOST Cost objective ($) Num Number of generators
Femission Emission objective (lb) n Number of objectives
Numw Number of wind turbines fir Moth position
Numpv Number of photovoltaic units b Spiral coefficient
Pw Output power of wind turbine (MW) Dis Distance between moth and flame
Ppv Output power of photovoltaic generator (MW) Iter Current number of iterations
Pdemand Load demand (MW) Max_iter Maximum number of iterations
Ploss Transmission loss (MW) aa Adaptive parameter
pos Individual position F Disturbance coefficient
D Dimension N Number of moths

Comparison

Type Objective Constraints Time scale of dispatch Dispatch units


Static Single Static (Constraints of thermal power units) Specific time section (At some Thermal power units
ELD point)
DEED Multiple Dynamic (Constraints of thermal power units) One time period (24-h) Thermal power units
HDEED Multiple Dynamic (Constraints of thermal power units and renewable energy One time period (24-h) Thermal power units and renewable energy
units) units

of renewable energy power. To explore the impact of renewable systems, and satisfied results are obtained. However, Ref. [20]
energy power generation on the fuel cost and pollutant emissions lacked to analyze the transmission loss of the scheduling results.
of the power system, this research considers both wind power and Chang et al. [21] had proven a distributed robust optimization al-
photovoltaic power generation in the HDEED problem and pro- gorithm for solving static ELD problem and analyzed the impact of
poses a new approach to solve the HDEED problem. renewable energy generation volatility. The power line blocking
constraint and transmission loss are considered in the solving
1.1. Literature review process of static ELD problem. Additionally, Al-Betar et al. [22]
utilized an improved GWO to solve static ELD problem, but ignored
Currently, two types of methods for solving static ELD have been to address the influence of valve point effect on fuel cost. Alsumait
devised, namely mathematical method and artificial intelligence et al. [23] developed a hybrid algorithm to solve static ELD problem
algorithm. Mathematical methods have focused on iterative and verified the proposed algorithm through different test systems;
method, quadratic programming method and gradient method. however, Ref. [23] ignored to address the high complexity of the
Mathematical methods can achieve better results when there are hybrid algorithm. When solving the ELD problem in the above-
few constraints in static ELD. However, the constraints considered mentioned literatures, only a single objective of fuel cost is
are more complicated as the system scale becomes larger, and the considered. However, in practice, the pollutant emissions cannot be
solution results becomes worse. Artificial intelligence algorithms, ignored. Therefore, the static ELD problem with fuel cost as a single
which are widely used to solve ELD problems, have powerful objective does not meet the actual demand.
solving capability and adapt to multi-objective optimization prob- In order to make the static ELD more suitable for the actual
lems [7]. Various artificial intelligence algorithms have been needs, researchers developed the DEED. Kheshti et al. [24] pointed
developed, such as particle swarm optimization (PSO) [8], equi- that DEED problem considers both fuel cost and pollutant emis-
librium optimization algorithm [9], tunicate swarm algorithm [10], sions, which makes the solving process more challenging. The
etc. dispatch period for DEED is 24 h, and the hourly output of the
To date, scholars have proposed a variety of artificial intelligence generator set is changed to adapt to different load requirements.
algorithms to solve static ELD and achieved good dispatch results, Both fuel cost and pollutant emissions are minimized under the
such as artificial algae algorithm [11], salp swarm algorithm [12], premise of meeting various constraints. Pandit et al. [25] argued
grey wolf optimization algorithm (GWO) [13], sine cosine algorithm that the optimized dispatch results changed due to different con-
[14], differential evolution algorithm (DE) [15], and a series of straints considered. Prior studies have developed a variety of multi-
improved algorithms [16e19], etc. Fu et al. [20] presented an objective optimization algorithms to solve the DEED problem. For
improved bird swarm optimization algorithm for solving static ELD, example, Liang et al. [26] presented a multi-objective bat optimi-
while considering the valve point effect and slope rate constraint. zation algorithm to solve the DEED problem, which considered the
The proposed algorithm is tested with 6-unit and 15-unit test valve point effect, slope rate constraint, and the impact of peak

2
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

shaving and valley filling of electric vehicles. The battery degra- multi-objective optimization into single-objective optimization
dation cost is taken into account in the objective function, and a 6- through weighting technique to obtain minimum fuel cost and
unit system is employed to verify the solving ability of the multi- pollutant emission. In addition, Hu et al. [39] introduced emission
objective bat algorithm. However, Ref. [26] ignored the feature performance standards and lower carbon price limits when solving
that bat algorithm is easy to fall into local extremum in the process the HDEED problem, meanwhile, the waste and storage of wind
of solving the DEED problem. Al bahrani et al. [27] analyzed the energy were analyzed; however, Ref. [39] ignored to take into ac-
influence of load demand management on DEED and utilized an count the non-convex characteristics caused by valve point effect in
orthogonal PSO algorithm to minimize cost and emission, but the fuel cost objective function. Although the above-mentioned
Ref. [27] lacked to further verify the convergence performance of literatures consider renewable energy power generation in the
the orthogonal PSO algorithm. In addition, Ghasemi et al. [28] HDEED problem, most studies only analyze the impact of wind
improved the convergence ability of the PSO algorithm. The power generation on power system dispatch, and lack to address
forbidden operation domain and transmission loss are taken into the impact of photovoltaic power generation on the power system
account in the solving process, and the competitiveness of the dispatch. With the increasing proportion of photovoltaic power
proposed algorithm is proved by two test systems. Although the generation, it is necessary to study the impact of photovoltaic po-
improved PSO obtained satisfied scheduling results, Ref. [28] wer generation on the economic and environmental benefits of
ignored to consider the pollutant emission objective. For the DEED power system.
problem, fuel cost objective and pollutant emission objective are
equally important, so both should be taken into consideration.
1.2. Innovative contributions
Barbosa-Ayala et al. [29] had proven the algebraic technique for
solving DEED and presented numerical polynomial homotopy
Aiming at the gaps in the above-mentioned literatures, this
continuation. The iteration time and solution results are more
study considers both wind power generation and photovoltaic
competitive compared with existing multi-objective algorithms
power generation in the HDEED problem, as well as the contra-
such as nondominated sorting genetic algorithm (GA) II [30],
dictory relationship between the economic benefit and environ-
sequential quadratic programming [31]. However, Ref. [29] ignored
mental benefit in each time period. Meanwhile, a new solution
the impact of the non-convex characteristics caused by the valve
method is proposed on the basis of the characteristics of the HDEED
point effect on scheduling results, resulting in a certain deviation
problem, such as nonlinear, non-convex, high-dimensional, and
between the dispatch results and the actual demands. Although the
difficult to solve. Finally, three cases including small-scale, me-
DEED problem takes into account the economic and environmental
dium-scale and large-scale systems are used to verify the proposed
benefits of the power system, with the continuous progress of
solution method. Additionally, the innovative contributions of this
renewable energy power generation technique, the DEED problem
study are as follows:
including renewable energy power generation has become a
research hotspot.
C A Wind-Solar-Thermal integrated energy dispatching model
In sum, prior studies lacked to tackle the impact of renewable
is constructed, which includes wind power generation,
energy power generation. HDEED problem takes into account the
photovoltaic power generation and thermal power
impact of renewable energy power generation on fuel costs and
generation;
pollutant emissions on the basis of DEED problem. Xiong and Shi
C A novel MFO_PDU algorithm with strong optimization per-
[32] pointed that equality and inequality constraints of HDEED
formance is proposed aiming at the non-convex, non-linear
problem are more complex compared with static ELD and DEED
and high-dimensional characteristics of HDEED problem.
problems, and higher solving ability of the algorithms is required.
MFO_PDU algorithm has stronger convergence ability and
In addition, Suresh et al. [33] and Ma et al. [34] argued that most of
better effect of avoiding local solutions compared with the
the existing HDEED problems consider the impact of wind power
state-of-the-art algorithms in Refs. [40e42] and
generation on fuel cost and pollutant emissions of power system.
Refs. [44,45];
For example, Alham et al. [35] analyzed both demand side man-
C Employing MFO_PDU-based optimization strategy to solve
agement and energy storage to reduce the impact of wind energy
HDEED problem through three cases, the dispatching
randomness, and developed an improved GA algorithm to solve
scheme obtained by the proposed algorithm has been proven
HDEED. Although the improved GA algorithm in Ref. [35] avoids the
to outperform the results obtained by the existing
problem of premature convergence to a certain extent, the
algorithms;
convergence stability of improved GA still needs to be enhanced
C MFO_PDU algorithm effectively reduces the fuel cost and
when solving the HDEED problem. To strengthen the adaptive
pollutant emissions of power generation system, and further
ability of the DE algorithm, Qiao et al. [36] proposed a DE algorithm
improves the utilization efficiency and penetration rate of
with adaptive factor to solve the HDEED problem under grid con-
renewable energy.
nected conditions, and utilized the charging and discharging of
electric vehicles to balance the fluctuation of wind power genera-
tion. However, Ref. [36] ignores the characteristics that DE algo-
rithm is apt to plunge into local extreme values, which limits the 1.3. Organization
potentiality of DE algorithm for solving HDEED problem. Chinna-
durrai and Victoire [37] proved a multi-objective crisscross opti- The remaining part of this paper is organized as follows. Section
mization for solving the DEED considering the uncertainty of wind 2 introduces three challenges in solving the HDEED problem. Sec-
power, but lacked to analyze the impact of wind power forecasting tion 3 shows the process of solving HDEED problem by the
errors on the dispatch results. Consequently, Pourghasem et al. [38] MFO_PDU algorithm. Section 4 adopts three cases to verify the
considered wind power forecast errors and load demand under the proposed algorithm. Section 5 presents the findings, contributions,
condition of meeting the operation constraints, and converted limitations and future research.

3
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

2. Formulation of the HDEED problem dioxide, sulfur oxide). Wind energy and solar energy are clean
energy, and do not emit pollutant gases such as carbon dioxide and
There are three challenges that need to be addressed simulta- sulfur oxide during the process of renewable energy power gen-
neously when solving HDEED problem. First, the multi-objective eration. Therefore, there is no need to consider the emission
HDEED optimization problem needs to be solved, and a competi- characteristics of wind power and photovoltaic power generation
tive solution is obtained to minimize power generation costs and in the emission objective function. The function relationship be-
pollutant emissions. Second, feasible solutions need to satisfy tween pollutant emissions and generator output is independent. In
equality and inequality constraints, such as power balance, slope this study, pollutant emission comprehensive model is used to
rate, and power output constraints. Third, both fuel cost and simulate the relationship between pollutant emission and the
pollutant emissions are in contradiction. Due to the inconsistency output of the thermal power units, that is, quadratic function plus
between the fuel cost and pollutant emission objectives, the opti- exponential function, as depicted follows [27]:
mization of one objective will sacrifice the other when w is a fixed
value, that is, the two optimization objectives conflict and influence X X
T Num  
each other. Reducing fuel costs will increase pollutant emissions,
Femission ¼ 2
xj Pj;i þ yj Pj;i þ zj;i þ qj exp jj Pj;i (3)
i¼1 j¼1
and reducing pollutant emissions will increase fuel costs. Hence,
the solution is not unique when w is different. The objective
where xj (lb/MW2h), yj (lb/MWh), zj (lb/h), qj (lb/h), jj (1/MW)
function F focuses more on the optimization of the fuel cost when
represent the emission coefficients of the jth generation unit.
the value of w is smaller than 0.5, and the objective function F fo-
To date, the research of solving multi-objective problem has
cuses more on reducing pollutant emissions when the value of w is
focus on two aspects, namely priori method and posterior method
greater than 0.5. How to choose a compromise solution to balance
[43]. The priori method is to convert the multi-objective into a
the economic and environmental benefits of the power system
single-objective solution by weight w; the posterior method is to
under different values of w is a key. The following content explains
find the Pareto frontier solution by developing multi-objective
these three challenges.
optimization algorithms, and then the compromise solution is
determined. Compared with the posterior method, the priori
2.1. Objective functions for HDEED method does not need to construct a complex multi-objective
optimization algorithm, and the solution process is simpler.
HDEED is a multi-objective optimization problem, including Therefore, this study selects the priori approach to solve HDEED
cost (FCOST) and emissions (Femission). Based on the statistical analysis problem using w to convert the multi-objective optimization
of the fuel cost of a large number of thermal power units, re- problem to single-objective optimization problem. The final opti-
searchers believe that the relationship between the fuel cost and mization objective F is depicted in Eq. (4).
the output of thermal power units is a quadratic function when the
valve point effect is not considered [45]. When wind energy con- minF ¼ w  FCOST þ ð1  wÞ  Femission (4)
version systems and solar energy conversion systems are owned by
the system operator, renewable energy represented by wind energy Eq. (4) reveals that Femission is the only optimization objective
and solar energy does not consume fuel in the process of power when w is 0, and FCOST is the only optimization objective when w is
generation, so the cost of wind and solar power generation does not 1. In addition, Eq. (4) presents that the two objectives are contra-
need to be considered in the cost function objective. Hence, FCOST is dictory. In the HDEED optimization process, FCOST and Femission
expressed by a quadratic equation as Eq. (1) [46]. cannot be minimized at the same time. The final optimization goal F
is more focused on reducing FCOST when the value of w is in the
X X
T N um  interval [0, 0.5), and the final optimization goal F is more focused on
2
FCOST ¼ vj Pj;i þ tj Pj;i þ uj;i (1) reducing Femission when the value of w is in the range of (0.5, 1].
i¼1 j¼1 Therefore, how to determine the value of w so that the solution of
HDEED minimizes both fuel cost and pollutant emissions is the key
where T is 24 h; Num is the number of generators; vj ($/MW2h), tj to solving the HDEED problem.
($/MWh), uj ($/h) are the cost coefficients of the jth generator; Pj,i
(MW) represents the output power of the jth generator at the ith
hour. 2.2. HDEED constraints
In practice, the opening of the intake valve of the generator set
makes the cost curve of the generator set non-convex, and the valve (1) Power balance constraint
point effect is simulated by introducing a sine function. Hence,
when considering the valve point effect, the relationship between The power balance constraint refers to that the sum of thermal
the fuel cost and the output of the thermal power units is a power, wind power and photovoltaic power generation should
quadratic function plus a sine function, as follows [47]: meet the load demand and transmission loss, as depicted in Eq. (5).

X X
T Num     X
Num Xw
Num Xpv
Num
2  min  Pj;i þ w
Pj;i þ
pv
Pj;i ¼ Pi demand þ Piloss (5)
FCOST ¼ vj Pj;i þ tj Pj;i þ uj;i þ dj sin fj Pj;i  Pj;i 
i¼1 j¼1 j¼1 j¼1 j¼1

(2)
where Numw represents the number of wind turbines; Numpv de-
w (MW) indicates the
notes the number of photovoltaic units; Pj;i
where dj ($/h) and fj (rad/MW) represent the valve point effect
pv
min (MW) is the minimum
coefficient of the jth generator set; Pj;i output power of the jth wind turbine at time period i; Pj;i (MW) is
output power of the jth generator set at the ith hour. the output power of the jth photovoltaic generator at time period i;
Femission refers to pollutant emissions (Nitrogen oxide, carbon Pi demand (MW) represents the load demand at time period i; Piloss

4
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

(MW) denotes the transmission loss at time period i. 3. Solving HDEED problem based on MFO_PDU algorithm
The transmission loss is calculated through the loss factor B:
3.1. MFO_PDU
X Num
Num X X
Num
Piloss ¼ Pt;j Bt;j Pj;i þ Pt;i Bt0 þ B00 (6) MFO algorithm, which simulates the dynamic behavior of moths
t¼1 j¼1 t¼1 when they fly to the fire, is a new bionic algorithm proposed by
Mirjalili in 2015, and has been applied in various fields, such as
where Bt,j, Bt0, B00 are loss coefficients. parameters optimization, economic dispatch, data mining and so
on [44].
(2) Generator set output power constrains In MFO algorithm, the moth position is assumed to be the so-
lution of the problem to be addressed, and the optimal solution is
Pjmin  Pj  Pjmax (7) approximated by position updating during the iterative process.
The moth position set is represented as follows:
where Pjmin (MW) and Pjmax (MW) are the lower and upper con- 2 3
pos1;1 pos1;2 / pos1;D
strains of generator output power. 6 pos2;1
6 pos2;2 / pos2;D 7
7
POS ¼ 4 5 (11)
« «
(3) Slope rate constraint
posN;1 posN;2 / posN;D
The output power is limited to a certain range through the slope
where posi,j is the position of the ith individual in the j-dimensional
rate constraint to prevent the generator set from being damaged by
space; D denotes the dimension of the problem to be solved; and N
high-temperature steam.
is the number of moths.
 The moth position vector is substituted into the fitness function
Pj;i  Pj;i1  URj Pj;i > Pj;i1
(8) (objective function) in solving practical problems. The fitness
Pj;i þ DRj  Pj;i1 Pj;i < Pj;i1
values of moths are as follows:
where URj (MW/h) and DRj (MW/h) are the upper and lower h i
Fit pos ¼ Fit1pos Fit2pos / pos
FitN (12)
bounds of generator output power fluctuations.

where Fiti is the fitness value of the ith moth.


Flames in the MFO algorithm guide the movement of moths, and
2.3. Compromise selection each moth corresponds to a flame. The flame positions are as
follows:
In the process of optimizing F, the value of w is from 0 to 1 with
2 3
an interval of 0.04, then 26 solutions are obtained and saved in the fir1;1 fir1;2 / fir1;D
Pareto solution set. The membership function is employed to 6 fir2;1 fir2;2 / fir2;D 7
FIR ¼ 6
4
7
5 (13)
determine the best compromise solution. First, the satisfaction of « «
each objective in the Pareto solution set is calculated: firN;1 firN;2 / firN;D
8
>
> 1 Fi;j  Fimin where firi,j represents the position of the ith moth in the jth
>
>
>
> dimension.
< F max  F
i;j Similarly, the fitness value corresponding to each flame is ob-
ri;j ¼ i
Fimin < Fi;j < Fimax (9)
>
> Fimax  Fimin tained by substituting the position of the flame into the fitness
>
>
>
> function. The fitness value set corresponding to the flame is as
:
0 Fi;j  Fimax follows:
h i
where ri,j denotes the satisfaction of the ith objective in the jth so- Fit fir ¼ Fit1fir Fit2fir / fir
FitN (14)
lution set. Fimax and Fimin represent the upper and lower bounds of
the ith objective in the Pareto solution set. Both moths and flames are search agents, and the positions are
Second, the satisfaction of each solution in the Pareto solution feasible solutions to the problem to be solved. However, the posi-
set is calculated: tion update strategies of moths and flames are different. MFO al-
gorithm includes three parts: initialization, position update and
P
n optimal judgment in the solution process, as follows:
ri;j
rj ¼ i¼1
(10) MFO ¼ ð Initialization Position End Þ (15)
P
np Pn
ri;j Initialization is used to initialize the positions and fitness values
j¼1 i¼1
of the moths:
where np is the number of solutions in the Pareto solution set; n
Initialization/fPOS; Fit pos g (16)
(n ¼ 2) is the number of objectives.
Finally, the solution corresponding to the maximum ri,j value is The position of the moth in the search space is updated by
chosen as the compromise solution. Position:

5
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Position : POS/POS (17)


End determines whether the algorithm meets the termination
conditions, if the conditions are met, return true, if not, return false:

End/ftrue; falseg (18)


Moths fly toward the flame with a spiral motion, and cannot
exceed the search space during the flight. The position is updated
according to the position of the flame, as depicted follows:

POSi ¼ Disi ,ebt ,cosð2ptÞ þ firj (19)

where b is the spiral coefficient, which is used to define the spiral


degree; firj denotes the position of the jth flame; Disi represents the
distance between the ith moth and the jth flame:
 
Disi ¼ posi  firj  (20)
Fig. 1. Adaptive parameter aa
t defines how close the moth is to the flame.
The moth is the farthest away from the flame when t ¼ 1, and
the moth is the closest to the flame when t ¼ 1. The mathematical parameter aa, as depicted follows:
model of t is as follows: 
3 p Iter
aa ¼  1  cos pþ , (25)
t ¼ ða  1Þ*rand þ 1 (21) 2 2 Max iter

where rand represents a random number in the interval [0,1]; a aa changes as presented in Fig. 1 during the algorithm optimization
linearly decreases from 1 to 2, and its mathematical model is as process.
follows: The aa value changes quickly at the beginning of the iteration,
which strengthens the global development capability; aa value
1 changes slowly at the end of the iteration, which enhances the local
a ¼  1 þ Iter, (22)
Max iter search capability. aa presents a dynamic change as the number of
iterations increases, which makes the MFO algorithm have a dy-
where Iter denotes the current number of iterations; and Max_iter namic adaptability.
represents the maximum number of iterations. Moths in MFO search around the flame, which causes the al-
Although the MFO algorithm has higher solution accuracy than gorithm to fall into local extremes easily. Hence, a location distur-
traditional optimization algorithms, there are still limitations need bance update strategy is proposed to address this limitation. When
to be improved. In light of the limitations of traditional MFO al- the number of iterations reaches the disturbance frequency P, the
gorithm, this research proposes a MFO algorithm based on the position of each moth is updated again according to the average
position disturbance update strategy. First, the initialization value of the set of moth positions. The mathematical model of the
method of the MFO algorithm is improved. A higher-quality position disturbance update strategy is as follows:
initialization solution makes the algorithm converge to the
 
optimal value faster, reduce the calculation cost of the algorithm posIterþ1 ¼ posIter þ posmean  posIter ,F (26)
i i i
and avoid local extreme values. To this end, this study uses the
randomness and ergodicity of the chaotic map to initialize the
where posmean denotes the average position; and F(F ¼ rand 
population, as follows:
0.4 þ 0.5) is the disturbance coefficient.
posi ¼ posmax þ Mi ðposmax  posmin Þ (23)

where posmax represents the maximum value of the search space; 3.2. Performance test of the MFO_PDU algorithm in the priori
posmin denotes the minimum value of the search space; Mi is the method
chaotic mutation scale, and mathematical model is as follows:
Previous studies usually use benchmark functions to test the
Mi ¼ aMi ð1  Mi Þ (24) convergence performance of the single-objective optimization al-
gorithms in the priori method, such as Refs. [41,42]. This study uses
where Mi is a random number in the interval [0,1]; and the model is a priori approach to solve the HDEED problem, so in this section, a
in a state of chaos when a ¼ 4. set of benchmark functions are applied to test the convergence
The parameter t in MFO, which is determined by the parameter performance of the MFO_PDU algorithm. The benchmark functions
a, affects the distance between the moth and the flame. a linearly include two forms: multimodal and unimodal. Unimodal bench-
decreases from 1 to 2, but the balance between local develop- mark functions are employed to test the convergence ability of the
ment and global search is not considered during the algorithm algorithms, and multimodal functions are used to test the ability of
solving process, which limits the convergence speed and accuracy the algorithms to jump out of local extremes. Therefore, three
of the MFO algorithm. Therefore, this study proposes an adaptive unimodal and two multimodal functions in this study were

6
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Table 1
Standard test functions [43,44].

f Range D Optima Attribute


PD 2 [-100, 100] 10 0 Unimodal
f1 ðuÞ ¼ i¼1 ui
PD QD [-10, 10] 10 0 Unimodal
f2 ðuÞ ¼ i¼1 i j þ
ju i¼1 jui j
PD Pi 2 [-100, 100] 10 0 Unimodal
f3 ðuÞ ¼ i¼1 ð j¼1 uj Þ
PD
f4 ðuÞ ¼ i¼1 ½u2i  10 cosð2pui Þ þ 10D [-5.12, 5.12] 10 0 Multimodal

1 XD 2 QD u [-600, 600] 10 0 Multimodal
f5 ðuÞ ¼ u  i¼1 cos piffi þ 1
4000 i¼1 i i

Table 2 enhanced. The Avg value of the MFO algorithm for f3 is 611.11, which
The values of parameters. is much greater than the 1.40E-60 of the MFO_PDU algorithm; the
Algorithm Parameters setting Avg values of the MFO_PDU and MFO algorithms for f4 are 1.39 and
24.85. The superiority of the proposed algorithm in the priori
Values Max_iter pop
method was verified through standard test functions.
PSO wmax ¼ 0.9, wmin ¼ 0.1, c1 ¼ c2 ¼ 1.5 1000 30
SSA v0 ¼ 0
DA s1 ¼ 0.1, s2 ¼ 0.1, s3 ¼ 0.7, s4 ¼ 1, s5 ¼ 1 3.3. HDEED solving process on the basis of MFO_PDU
CSA FL ¼ 2; AP ¼ 0.1
MFO b¼1
MFO_PDU b ¼ 1, PF ¼ 25 The specific process of using the MFO_PDU algorithm to solve
the HDEED is depicted in Fig. 2.
Fig. 2 illustrates that the HDEED problem solving process mainly
includes three parts: HDEED modeling, MFO_PDU algorithm opti-
employed to verify the MFO_PDU-based optimization strategy. The
mization, and compromise solution selection. The specific process
optimal solution, variable dimension, and variable range of the test
is as follows:
functions are listed in Table 1.
PSO [40], Salp Swarm Algorithm (SSA) [41], Dragonfly Algorithm
(1) Construct the mathematical model of HDEED problem;
(DA) algorithm [42], and Crow Search Algorithm (CSA) [45] were
(2) Set population parameters, such as population size, distur-
selected as comparison algorithms to verify and analyze the
bance frequency, number of iterations, etc.;
convergence characteristics of the MFO_PDU algorithm. The
(3) Initialize the position and fitness value of the moth;
parameter setting values of each algorithm are revealed in Table 2.
(4) Determine whether the disturbance frequency PF is reached;
Each algorithm has the same number of iterations and popula-
if reach PF, update the position of the moth according to Eq.
tion size to ensure the objectivity of the test results. wmax and
(26), otherwise update the position of the moth by Eq. (19),
wmin are the upper and lower bounds of the PSO algorithm; c1 and
and calculate the fitness value, as depicted in Eq. (27):
c2 are the cognitive coefficients. FL and AP are the flight length and
perception probability of CSA algorithm. v0 is the initial speed of fitness ¼ minðw  FCOST þ ð1  wÞ  Femission Þ (27)
slap in the SSA algorithm. s1-s5 are the weight coefficients of the
DA algorithm. b specifies the spiral degree of the MFO and
MFO_PDU algorithms, and PF is the disturbance frequency. (5) Check the slope rate constraint and power boundary
For each test function, the algorithm repeats the calculation 30 constraint according to Eqs. (7) and (8);
times and counts the average (Avg) and standard deviation (std) of (6) Update the flame position;
the convergence results, as presented in Table 3. (7) Determine whether the termination iteration is reached, if
Table 3 revealed that compared to the other five algorithms, the not, return to Step 3;
MFO_PDU algorithm obtained competitive results for both unim- (8) Output the optimal solution and optimal position;
odal and multimodal functions. Compared with the MFO algorithm, (9) Select the compromise solution by Eqs. (9) and (10).
the convergence ability of the MFO_PDU algorithm is significantly

4. Results and discussion

Table 3
Statistical results of test experiments.
In this section, three cases (including test systems with different
scales) were employed to verify the feasibility of the proposed al-
f PSO SSA DA gorithm. Case1 is a small-scale system with five thermal units;
Avg std Avg std Avg std Case2 presents a medium-scale system with 10 thermal units; and
f1 0.17 0.47 6.35E-10 2.59E-10 10.44 11.62 Case3 shows a large-scale system with 15 thermal units. The pro-
f2 0.21 0.49 3.47E-04 1.90E-03 1.14 0.61 posed algorithm was used to dispatch thermal units within 24 h for
f3 18.96 29.04 1.81E-09 8.61E-10 79.15 86.91 small-scale, medium-scale and large-scale systems, while consid-
f4 17.55 8.47 19.07 9.18 30.88 11.11 ering the valve point effect, power balance constraint, output po-
f5 0.26 0.16 0.21 0.11 0.83 0.32
wer constraint, and slope rate constraint. The unit coefficients for
f MFO MFO_PDU CSA small-scale, medium-scale and large-scale systems are listed in
Avg std Avg std Avg std Appendix A1, A2 and A3 respectively. The data of wind output
f1 1.74E-30 3.95E-30 1.73E-63 6.39E-63 4.62E-08 6.34E-08 power and PHOTOVOLTAIC output power used in this study is
f2 0.33 1.82 8.56E-35 3.97E-34 3.06E-02 0.11 referenced in Ref. [48], and the load demand data in the three cases
f3 611.11 1.88Eþ03 1.40E-60 3.85E-60 5.96E-04 1.00E-03 is referenced in Ref. [49]. The 24-h output power and load demand
f4 24.85 14.65 1.39 2.74 8.98 3.84 are depicted in Fig. 3.
f5 0.17 8.20E-02 3.12E-02 5.22E-02 0.17 9.57E-02
The wind power as presented in Fig. 3 is not limited by the time
7
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 2. HDEED problem solving process.

period because the wind speed is not affected by the time period. Table 5.
However, the illumination is affected by the time period, so there is After considering renewable energy power generation, the
no photovoltaic output power within 0e7 h and 18e24 h. emissions obtained by the HHO and TSA algorithms were reduced
by 15.60% and 17.30%. Additionally, the emissions obtained by the
MFO and MFO_PDU algorithms were reduced by 19.40% and
4.1. Case1
15.59%. Meanwhile, the fuel cost was reduced. The fuel cost and
pollutant emission obtained by the MFO_PDU algorithm were the
The proposed algorithm was validated in two situations: Situ-
smallest. For Situation2, the cost obtained by the MFO_PDU algo-
ation1 does not consider renewable energy power generation, and
rithm was 11.31%, 4.01%, and 5.27% smaller than those of the HHO,
Situation2 considers renewable energy power generation (2 wind
TSA and MFO algorithms. In addition, the pollutant emission ob-
units and 1 photovoltaic unit). The solution set for Situation2 in
tained by the MFO_PDU algorithm was 14.67%, 8.72%, and 3.08%
Case1 obtained by the MFO and MFO_PDU algorithms is depicted in
lower than those of the HHO, TSA and MFO algorithms,
Fig. 4.
respectively.
The compromise solution of MFO_PDU and MFO algorithm as
Fig. 5 presents the power output, load demand, transmission
revealed in Fig. 4 was obtained by calculating the membership
loss and power balance errors of each unit in each period obtained
function, which were (4.31  104, 1.57  104) and (4.55  104,
by the HHO, TSA, MFO and MFO_PDU algorithms for Situation2 in
1.62  104). The solution set and compromise solution of the
Case1.
MFO_PDU algorithm were more competitive compared with the
Each stacked histogram in Fig. 5(a) represents the sum of the
MFO algorithm. The compromise solution obtained by the
power generation of 5-unit system at that moment, and the bars
MFO_PDU algorithm gives the dispatch results of the five thermal
with different colors represent the power output of different units.
units in each period and the corresponding transmission loss, as
Fig. 5(b) depicts the power balance errors of each algorithm in
listed in Table 4.
different time periods. By comparing the power output of each unit,
The existing algorithms (SA [50], PSO [51], PPSO [52], PS [50])
the dispatch scheme given by MFO_PDU was more competitive,
and HHO, TSA, MFO algorithms were selected as comparison al-
and the output power of each unit in each period was more
gorithms for Situation1; HHO, TSA, MFO algorithms were selected
balanced. Meanwhile, the dispatch results of the MFO and
as comparison algorithms for Situation2. For two situations in
MFO_PDU algorithms basically satisfied the power balance
Case1, the compromise solution of each algorithm is presented in

8
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Table 4
Dispatch results based on MFO_PDU algorithm for Situation2 in Case1.

Hour 5-unit generation and transmission loss

P1 P2 P3 P4 P5 Loss

1 74.99 92.50 112.65 42.28 50.00 2.95


2 75.00 89.40 112.67 83.51 50.00 3.49
3 46.75 80.63 112.66 124.90 88.71 4.18
4 40.49 97.77 112.67 97.779 138.71 4.85
5 70.49 98.13 112.67 124.90 124.50 5.72
6 75.00 98.53 112.67 158.52 112.49 6.35
7 75.00 98.54 144.38 128.74 139.75 6.93
8 75.00 104.59 125.99 178.74 139.75 7.94
9 69.41 98.44 115.66 203.52 139.36 8.08
10 75.00 107.36 132.99 209.83 141.19 9.10
11 75.00 117.33 160.70 159.83 139.80 8.61
12 75.00 122.73 143.99 209.83 139.76 9.80
13 75.00 97.26 129.19 207.42 139.74 8.61
14 55.79 98.53 112.67 208.45 139.75 7.84
15 62.59 98.54 116.74 158.45 139.75 6.75
16 75.00 91.97 112.67 108.45 122.07 5.28
17 58.87 84.96 111.96 124.70 129.99 5.26
18 75.00 98.53 113.17 127.76 139.75 6.24
19 73.84 98.53 147.70 177.76 139.75 8.21
20 75.00 98.30 129.86 209.80 139.76 8.73
21 75.00 98.53 169.86 166.31 139.75 8.48
22 75.00 116.48 129.86 125.07 139.76 6.99
23 71.80 95.03 112.67 124.90 89.76 4.98
24 49.98 91.46 112.67 124.90 50.00 3.83

Table 5
Running results of each algorithm for Case1.

C and E Situation1

SA PSO PPSO PS HHO TSA MFO MFO_PDU

C (104) ($) 4.86 5.08 4.83 4.79 5.22 4.83 4.79 4.72
E (104) (lb) 2.11 2.01 2.36 1.89 2.18 2.08 2.01 1.86

C and E Situation2

HHO TSA MFO MFO_PDU

C (104) ($) 4.86 4.49 4.55 4.31


E (104) (lb) 1.84 1.72 1.62 1.57

condition, that is, the sum of the output of 5 units per hour was
equal to the sum of the load demand and the transmission loss.
However, the power balance errors obtained by the HHO and TSA
Fig. 3. Renewable energy power curves and load demand.
algorithms were larger, and the maximum error exceeded 4 MW.
Fig. 6 presents the cost and emissions obtained by the HHO, TSA,
MFO, and MFO_PDU algorithms for Case1.
The cost of each period obtained by the HHO algorithm was
significantly higher than that of the TSA, MFO, and MFO_PDU al-
gorithms for Situation1. In most periods, the cost values obtained
by the MFO_PDU algorithm were smaller than the HHO, TSA and
MFO algorithms. The emission values obtained by the HHO algo-
rithm at 11 o'clock was the largest, reaching 1567.03 lb. The emis-
sion values of the MFO_PDU algorithm were significantly lower
than the other three algorithms from 10:00 to 15:00. In addition,
MFO_PDU obtained the lowest cost value in most periods when
considering renewable energy power generation. The emission
value of the HHO algorithm at 12 o'clock reached 1254.11. Analysis
results for Case1 demonstrated that the dispatch scheme obtained
by the MFO_PDU algorithm was the most satisfactory.
Fig. 7 presents the hourly convergence curve of TSA, MFO and
MFO_PDU algorithms during the process of searching the
compromise solution.
Fig. 7(a) revealed that the iterative curve obtained by the
Fig. 4. Solution set for Situation 2 in Case1. MFO_PDU algorithm was lower than those of the TSA and MFO
algorithms, indicating that the convergence accuracy of MFO_PDU

9
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 6. Generation cost and emission of each unit per hour for Case1.

was higher. Meanwhile, the MFO_PDU algorithm showed a faster


convergence speed as depicted in Fig. 7(b).

4.2. Case2

For Case2, Situation1 consists of 10 thermal units, and Situa-


tion2 includes 10 thermal units, 2 wind units and 2 photovoltaic
units. Compared with Case1, Case2 has increased the number of
thermal power generation units and photovoltaic power genera-
tion units, which requires the solving algorithm to have stronger
optimization capability. When considering renewable energy po-
wer generation, the solution set and compromise solution obtained
by the MFO and MFO_PDU algorithms are depicted in Fig. 8.
Fig. 8 presented the minimum costs and emissions obtained by
the MFO and MFO_PDU algorithms under different values of w. The
Fig. 5. Power balance verification. solution set distribution obtained by the MFO_PDU algorithm was
more uniform compared with the MFO algorithm, and the frontier
solution distribution was smoother. The solution set of the
MFO_PDU algorithm was under the MFO algorithm solution set,
indicating that the costs and emissions obtained by the MFO_PDU
10
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 7. Iterative curves for Case1.

algorithm were more competitive. For Situation 2 in Case2, Table 6


lists the optimal dispatch results of the MFO_PDU algorithm
including the power output and transmission loss of 10 units in
each period.
For Case2, when renewable energy generation was not consid-
ered, the results obtained by existing models were used as a
comparison, including PSO [53], PSOCS [49], MONND [54], and
NSGA II [30] algorithms. In addition, the compromise solutions of
HHO, TSA, and MFO algorithms were selected as a comparison
when considering renewable energy power generation. In both
situations, the compromise solution of each algorithm is listed in
Table 7.
For Case2, the MFO_PDU algorithm also obtained competitive
results. For Situation1, the cost and emission obtained by the
MFO_PDU algorithm were the smallest, which were 2.52  106$
and 2.97  105lb. Similar to the running results obtained in Case1,
the costs and emissions were reduced after considering the
renewable energy generation. After considering the energy gener-
Fig. 8. Solution set for Situaion2 in Case2.
ation, the cost and emission of the HHO algorithm were reduced by
6.34% and 9.74%; the cost and emission of the TSA algorithm were
reduced by 6.20% and 14.80%; the cost and emission of the MFO

11
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Table 6
Dispatch results of MFO_PDU algorithm for Situation2 in Case 2.

Hour 10 units generation and transmission loss

P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 Loss

1 150.00 135.00 73.00 101.06 122.85 118.44 93.08 85.31 80.00 55.00 19.64
2 150.00 135.00 83.11 116.68 122.86 122.44 123.08 115.31 80.00 55.00 22.54
3 150.00 135.00 137.17 120.41 172.73 160.00 129.59 120.00 80.00 55.00 28.51
4 150.00 135.00 177.62 161.53 222.59 160.00 130.00 120.00 80.00 55.00 35.64
5 161.89 215.00 197.33 193.50 172.59 160.00 130.00 120.00 80.00 55.00 40.01
6 192.53 222.23 191.16 242.85 222.59 160.00 130.00 120.00 80.00 55.00 49.02
7 220.82 222.29 238.22 237.00 243.00 160.00 130.00 120.00 80.00 55.00 54.01
8 233.27 294.53 290.02 187.00 243.00 160.00 130.00 120.00 80.00 55.00 59.40
9 277.39 305.66 297.39 237.00 243.00 160.00 130.00 120.00 80.00 55.00 71.07
10 303.24 316.89 333.36 287.00 243.00 160.00 130.00 120.00 80.00 55.00 79.54
11 320.11 335.93 340.00 300.00 243.00 160.00 130.00 120.00 80.00 55.00 87.87
12 356.16 366.34 340.00 300.00 243.00 160.00 130.00 120.00 80.00 55.00 92.47
13 345.37 286.34 340.00 300.00 243.00 160.00 130.00 120.00 80.00 55.00 84.43
14 265.37 309.53 319.12 250.00 193.00 160.00 130.00 120.00 80.00 55.00 70.89
15 221.87 229.53 239.12 239.75 243.00 160.00 130.00 120.00 80.00 55.00 60.15
16 163.76 216.87 194.64 189.75 193.00 160.00 130.00 120.00 80.00 55.00 44.68
17 150.00 173.79 185.19 180.83 227.37 160.00 130.00 120.00 80.00 55.00 40.30
18 176.62 222.26 195.46 230.83 243.00 160.00 130.00 120.00 80.00 55.00 49.31
19 228.40 253.45 275.46 263.43 243.00 160.00 130.00 120.00 80.00 54.99 59.37
20 288.63 307.92 297.42 300.00 243.00 160.00 130.00 120.00 80.00 55.00 74.91
21 302.77 309.53 303.03 250.00 243.00 160.00 130.00 120.00 80.00 55.00 71.05
22 222.77 229.53 223.03 200.00 229.63 160.00 130.00 120.00 80.00 55.00 49.50
23 150.00 149.53 143.03 157.18 179.63 160.00 130.00 120.00 80.00 55.00 32.07
24 150.00 135 105.50 113.89 172.77 124.92 113.01 119.99 80.00 55.00 25.34

Table 7 Each stacked histogram in Fig. 9(a) represents the sum of power
Running results of each algorithm for Case2. generation of 10 units at that moment, and the bars with different
C and E Situation1
colors represent the power output of each unit. Fig. 9(b) shows the
power balance errors of four algorithms. Fig. 9(a) presented that the
PSO PSOCS MONND NSGA II HHO TSA MFO MFO_PDU
output power of 10 units obtained by the MFO_PDU algorithm in
6
C (10 ) ($) 2.60 2.53 2.56 2.52 2.84 2.58 2.54 2.52 each hour was more balanced. For the HHO algorithm, in the 10e12
E (105) (lb) 3.11 2.98 2.97 3.10 3.90 3.58 3.07 2.97
period, the sum of the power generation of 10 units did not meet
C and E Situation2 the load demand and transmission loss. The power balance error
HHO TSA MFO MFO_PDU curve of the HHO algorithm in Fig. 9(b) fluctuated greatly, and the
6 maximum error exceeded 200 MW. In addition, the power balance
C (10 ) ($) 2.66 2.42 2.43 2.38
E (105) (lb) 3.52 3.05 2.72 2.64 error curve of the TSA algorithm fluctuated greatly in the 10e13
period. The power balance errors of the MFO and MFO_PDU algo-
rithms were the smallest, indicating that the sum of the hourly
power generation of 10 units met the sum of the load demand and
algorithm were reduced by 4.33% and 11.40%; the cost and emission
the transmission loss.
of the MFO_PDU algorithm were reduced by 5.55% and 11.11%
The cost and emission of 10 units in each period are presented in
respectively.
Fig. 10 under the two situations in Case2.
The analysis results revealed that for two situations in Case2, the
The costs and emissions of 10 units were reduced accordingly
MFO_PDU algorithm achieved satisfactory costs and emissions. The
when considering renewable energy power generation as depicted
cost of NSGAII was 2.52  106$ for Situation1, but the emissions
in Fig. 10. For Situation2, the highest cost obtained by the MFO_PDU
were 4.19% higher than MFO_PDU. The emission obtained by the
algorithm was 1.43  105$, which was 7.74%, 5.29%, and 2.05%
MONND algorithm was 2.97  105lb for Situation1, but the cost was
smaller than that of the HHO, TSA, and MFO algorithms respec-
1.58% higher than MFO_PDU. For Situation1, the emission obtained
tively. The maximum emission obtained by the MFO_PDU algo-
by the MFO_PDU algorithm was reduced by 23.84%, 17.03%, and
rithm was 2.13  104lb, which was 22.82%, 18.70%, and 5.75%
3.25% than those obtained by the HHO, TSA, and MFO algorithms.
smaller than that of the HHO, TSA, and MFO algorithms, respec-
Additionally, the cost obtained by the MFO_PDU algorithm was
tively. For Case2, the dispatch scheme obtained by MFO_PDU out-
reduced by 11.26%, 2.32%, and 0.78% compared with costs obtained
performs those obtained by other algorithms.
by the HHO, TSA, and MFO algorithms. Meanwhile, the MFO_PDU
Fig. 11 depicts the iterative curves of the TSA, MFO, and
algorithm obtained the most competitive cost and emissions for
MFO_PDU algorithms in each period.
Situation2. For fuel cost, MFO_PDU algorithm was 10.52%, 1.65% and
Fig. 11 (a) and (b) presented the global and local graphs of the
2.05% smaller than HHO, TSA and MFO algorithm respectively. For
iterative curves. The convergence speed of MFO and MFO_PDU al-
emissions, MFO_PDU algorithm was 25%, 13.44% and 2.94% smaller
gorithms was faster than the TSA algorithm, and the convergence
than HHO, TSA and MFO algorithm respectively.
values obtained by MFO_PDU algorithm were more competitive in
For Situation2, the power output and power balance errors of
most periods.
10-unit in each period obtained by each algorithm are depicted in
Fig. 9.

12
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 10. Generation cost and emission of each unit per hour for Case2.

4.3. Case3

For Case3, Situation1 consists of 15 thermal units, and Situa-


tion2 includes 15 thermal units, 3 wind units and 2 photovoltaic
units. The calculation amount of Case3 is increased compared with
Case1 and Case2. When considering renewable energy power
generation, the solution set and compromise solution obtained by
the MFO and MFO_PDU algorithms are depicted in Fig. 12.
Fig. 12 revealed the minimum costs and emissions obtained by
the MFO and MFO_PDU algorithms when considering renewable
energy power generation with different values of w. The solution
set obtained by the MFO_PDU algorithm was evenly distributed in
the frontier solution compared with the solution set obtained by
the MFO algorithm, and the frontier solution of the MFO_PDU al-
Fig. 9. Generator units' output and power balance verification for Case2.
gorithm was smoother. The compromise solutions of MFO and

13
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 11. Iterative curves for Case2.

Table 8
Compromise solution of each algorithm for Case 3.

C and E Situation1

PSO MONNDE HHO TSA MFO MFO_PDU


5
C (10 ) ($) 7.06 7.11 7.39 7.04 6.90 6.89
E (105) (lb) 3.07 2.73 6.64 4.33 3.18 3.02

C and E Situation2

HHO TSA MFO MFO_PDU


5
C (10 ) ($) 7.04 6.74 6.56 6.55
E (105) (lb) 3.98 3.45 2.78 2.57

MFO_PDU algorithms were (6.56  105$, 2.78  105lb) and


(6.55  105$, 2.57  105lb). The cost values obtained by the two
algorithms were similar, but the emission values were significantly
different. The emission obtained by MFO_PDU was 7.55% smaller
than the MFO algorithm. For Case 3, the dispatch results obtained
Fig. 12. Solution set for Situation 2 in Case3. by the MFO_PDU algorithm are listed in Table A4 (in Appendix A).
For Case3, the existing methods (PSO [53], MONNDE [54], HHO,

14
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 14. Generation cost and emission of each unit per hour for Case3.

solutions obtained by the MFO_PDU algorithm were optimal. For


Situation1, the minimum cost obtained by the MFO_PDU algorithm
was 6.89  105$, which was 6.76% and 2.13% smaller than that of
HHO and TSA algorithms. The smallest emission obtained by the
MFO_PDU algorithm was 3.02  105lb, which was 54.51%, 30.25%
and 5.03% less than that of HHO, TSA, and MFO algorithms. The
minimum cost and emission obtained by the MFO_PDU algorithm
for Situation2 were 6.55  105$ and 2.57  105lb. The emission of
the MFO_PDU algorithm was 35.42%, 25.50% and 7.55% smaller than
that of the HHO, TSA, and MFO algorithms.
When renewable energy generation is considered, the power
Fig. 13. Power balance verification for Case3.
output and power balance errors of 15 units in each period are
presented in Fig. 13.
TSA, and MFO) were selected as comparison algorithms to verify Each stacked histogram in Fig. 13(a) represents the sum of the
the proposed MFO_PDU algorithm. In both situations, the power generation of 15 units at that moment, and the different
compromise solution of each algorithm is listed in Table 8. colored bars represent the power output of each unit. The power
Table 8 indicated that in two situations, the compromise balance errors of the four algorithms at different time periods are

15
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Fig. 15. Iterative curves for Case3.

shown in Fig. 13(b). The power balance errors of the MFO and 1.43  104lb and 4.70  104lb for Situation1 and Situation2.
MFO_PDU algorithms were the smallest, and the power balance The iterative curves of the TSA, MFO, and MFO_PDU algorithms
errors of the TSA algorithm fluctuated in a relatively small period of in each hour are depicted in Fig. 15.
time. However, the power balance errors of the HHO algorithm The iterative curves of the TSA algorithm in Fig. 1 (a) were at the
fluctuated greatly during the dispatch period, and the maximum top level, and the iteration curves of the MFO and MFO_PDU al-
error exceeded 30 MW. gorithm were at the bottom level, indicating that the convergence
For both situations in Case3, the cost and emission of 15 units in values of the MFO and MFO_PDU algorithms were better than the
each time period are presented in Fig. 14. TSA algorithm.
Fig. 14 (a) and (b) depicted the cost and emission of 15 units in For the three cases, the MFO_PDU algorithm proposed in this
each time period for both situations in Case3. The cost and emission study gives competitive dispatching schemes compared with the
obtained by the MFO_PDU algorithm were optimal in most periods. existing approaches, indicating that the proposed algorithm is
However, the emission curves of HHO and TSA algorithms fluctu- suitable for solving HDEED problem.
ated greatly. The highest emission of the HHO algorithm reached

16
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

5. Conclusions competitive dispatch plans for three cases; and (3) Research out-
comes contribute in improving the environmental and economic
This study fulfills the gaps from solving the HDEED problem benefits of the power system, and providing a new idea for the
considering both wind power and photovoltaic power generation formulation of dynamic economic emission dispatch strategies for
using a novel MFO_PDU-based optimization strategy. The economic the Wind-Solar-Thermal integrated energy system.
dispatch of the power system after renewable energy generation is Although the algorithm proposed obtains satisfactory dispatch
connected to the grid is a new research topic. Hence, this study results, this study also has certain limitations. First, the general-
proposes the MFO_PDU-based approach algorithm to solve the ization ability and convergence stability of the MFO_PDU-based
economic emission dispatch problem of the integrated energy optimization strategy need to be further improved. Second, the
system. In addition, three cases including small-scale, medium- HDEED problem does not fully consider the impact of the volatility
scale and large-scale systems are employed to verify the feasibility of wind and photovoltaic power. We will focus on addressing these
and effectiveness of the proposed algorithm in two situations. The limitations in future research.
findings include the following parts:
Credit author statement
(1) The convergence performance of the traditional MFO is
improved by introducing chaotic mutation initialization and Zhi-Feng Liu: Conceptualization, Methodology, Original writing.
position disturbance update strategies into the MFO_PDU Ling-Ling Li: Conceptualization, Methodology, Original writing. Yu-
algorithm; Wei Liu: Methodology, Original writing. Jia-Qi Liu: Methodology,
(2) The MFO_PDU algorithm is tested by unimodal and multi- Original writing. Heng-Yi Li: Methodology, Original writing. Qiang
modal benchmark functions, and has been proven to Shen: Methodology, Original writing.
outperform the existing algorithms;
(3) The HDEED in consideration of both wind and photovoltaic Declaration of competing interest
power generation is constructed, and the MFO_PDU-based
approach formulates the most competitive dispatch plan; The authors declare that they have no known competing
(4) The fuel cost and pollutant emissions are significantly financial interests or personal relationships that could have
reduced when considering renewable energy power gener- appeared to influence the work reported in this paper.
ation. For large-scale test system, the minimum cost and
emission obtained by the MFO_PDU algorithm are
Acknowledgements
6.55  105$ and 2.57  105lb, respectively. The emission of
the MFO_PDU algorithm is 35.42%, 25.50% and 7.55% smaller
This study was supported by key project of Tianjin Natural Sci-
than that of the HHO, TSA, and MFO algorithms.
ence Foundation [Project No.19JCZDJC32100], Natural Science
Foundation of Hebei Province of China [Project No. E2018202282]
The innovative contributions of this work are as follows: (1) The
and the Hebei Provincial Innovation Foundation for Postgraduate
mathematical model of HDEED problem on the basis of Wind-
[Project No. CXZZBS2020028].
Solar-Thermal integrated energy is constructed, taking into ac-
count the valve point effect, equality constraints and inequality
constraints; (2) A novel MFO_PDU-based optimization approach is Appendix A
proposed to solve HDEED problem and formulates the most

Table A1
Generator coefficients for small-scale system

Unit v t u d f X y z q j UR DR Pmin Pmax

1 8.0E-3 2.0 25 100 0.042 1.8E-2 0.805 80 0.6550 2.846E-2 30 30 10 75


2 3.0E-3 1.8 60 140 0.040 1.5E-2 0.555 50 0.5773 2.446E-2 30 30 20 125
3 1.2E-3 2.1 100 160 0.038 1.05E-2 1.355 60 0.4968 2.270E-2 40 40 30 175
4 1.0E-3 2.0 120 180 0.037 8.0E-3 0.600 45 0.4860 1.948E-2 50 50 40 250
5 1.5E-3 1.8 40 200 0.035 1.2E-2 0.555 30 0.5035 2.075E-2 50 50 50 300

Table A2
Generator coefficients for medium-scale system

Unit v t u d F x y z q j UR DR Pmin Pmax

1 0.1524 38.5397 786.7988 450 0.041 0.0312 2.4444 103.3908 0.5035 0.0207 80 80 150 470
2 0.1058 46.1591 451.3251 600 0.036 0.0312 2.4444 103.3908 0.5035 0.0207 80 80 135 470
3 0.0280 40.3965 1049.9977 320 0.028 0.0509 4.0695 300.3910 0.4968 0.0202 80 80 73 340
4 0.0354 38.3055 1243.5311 260 0.052 0.0509 4.0695 300.3910 0.4968 0.0202 50 50 60 300
5 0.0211 36.3278 1658.5696 280 0.063 0.0344 3.8132 320.0006 0.4972 0.0200 50 50 73 243
6 0.0179 38.2704 1356.6592 310 0.048 0.0344 38.132 320.0006 0.4972 0.0200 50 50 57 160
7 0.0121 36.5104 1450.7045 300 0.086 0.0465 3.9023 330.0056 0.5163 0.0214 30 30 20 130
8 0.0121 36.5104 1450.7045 340 0.082 0.0465 3.9023 330.0056 0.5163 0.0214 30 30 47 120
9 0.1090 39.5804 1455.6056 270 0.098 0.0465 3.9524 350.0056 0.5475 0.0234 30 30 20 80
10 0.1295 40.5407 1469.4026 380 0.094 0.0470 3.9864 360.0012 0.8475 0.0234 30 30 10 55

17
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

Table A3
Generator coefficients for large-scale system

Unit v t u d f x y z q j UR DR Pmin Pmax

1 2.99E-4 10.1 671 100 0.042 0.0180 0.8050 80 0.6550 0.02846 80 120 150 455
2 1.83E-4 10.2 574 140 0.040 0.0150 0.5550 50 0.0180 0.02446 80 120 150 455
3 1.126E-3 8.8 374 160 0.038 0.0105 1.3550 60 0.0150 0.02270 130 130 20 130
4 1.126E-3 8.8 374 180 0.037 0.0080 0.6000 45 0.0105 0.01948 130 130 20 130
5 2.05E-4 104 461 200 0.035 0.0120 0.5550 30 0.0080 0.02075 80 120 150 80
6 3.01E-4 101 630 450 0.041 0.0312 2.4444 103.3908 0.0120 0.02070 80 120 135 80
7 3.64E-4 9.8 548 600 0.036 0.0312 2.4444 103.3908 0.0312 0.02070 80 120 135 80
8 3.38E-4 11.2 227 320 0.028 0.0509 4.0695 300.3910 0.0509 0.02020 65 100 60 65
9 8.07E-4 11.2 173 260 0.052 0.0509 4.0695 300.3910 0.0509 0.02020 60 100 25 60
10 1.203E-3 10.7 175 280 0.063 0.0344 3.8132 320.0006 0.0344 0.02000 60 100 25 60
11 3.586E-3 10.2 186 310 0.048 0.0344 3.8132 320.0006 0.0344 0.02000 80 80 20 80
12 5.513E-3 9.9 230 300 0.086 0.0465 3.9023 330.0056 0.0465 0.02140 80 80 20 80
13 3.71E-4 13.1 225 340 0.082 0.0465 3.9023 330.0056 0.0465 0.02140 80 80 25 80
14 1.929E-3 12.1 309 270 0.098 0.0465 3.9524 350.0056 0.0465 0.02340 55 55 15 55
15 4.447E-3 12.4 323 380 0.094 0.0470 3.9864 360.0012 0.0470 0.02340 55 55 15 55

Table A4
Generation dispatching of 15 units.

Hour Generation dispatching of the first nine units for Case3 (MW)

P1 P2 P3 P4 P5 P6 P7 P8

1 150.00 150.00 62.89 130.00 150.00 135.00 135.00 60.00 References


2 150.00 150.00 91.27 99.75 150.00 135.00 135.00 60.00
3 175.45 228.53 103.01 130.00 150.00 135.00 222.26 60.00
[1] Xiang Y, Wu G, Shen XD, Ma YH, Gou J, Xu WT, Liu JY. Low-carbon economic
4 225.112 303.92 113.73 20.00 238.32 211.31 222.90 60.00
dispatch of electricity-gas systems. Energy 2021;226:1e12.
5 150.00 232.75 129.99 130.00 318.12 211.62 309.53 60.00 [2] Zhang H, Liang S, Ou M, Wei M. An asynchronous distributed gradient algo-
6 244.95 152.75 130.00 130.00 238.12 289.35 229.53 60.00 rithm for economic dispatch over stochastic networks. Int J Electr Power
7 262.33 272.75 126.79 130.00 158.12 288.24 309.53 60.00 Energy Syst 2021:124.
8 224.61 280.17 130.00 106.21 262.21 288.28 309.50 60.00 [3] Shen X, Zou DX, Duan N, Zhang Q. An efficient fitness-based differential
9 246.46 298.57 130.00 130.00 239.76 364.87 229.50 60.00 evolution algorithm and a constraint handling technique for dynamic eco-
10 248.17 305.78 129.99 130.00 335.46 364.82 309.83 159.95 nomic emission dispatch. Energy 2019;186:28.
11 211.63 314.66 129.99 130.00 330.29 364.34 396.41 94.95 [4] Li XZ, Wang WQ, Wang HY, Wu JH, Fan XC, Xu QD. Dynamic environmental
12 288.41 336.94 130.00 130.00 329.89 418.28 316.41 153.86 economic dispatch of hybrid renewable energy systems based on tradable
13 256.78 304.53 130.00 117.17 318.96 364.72 396.31 88.86 green certificates. Energy 2020;193:775e92.
14 233.38 224.53 130.00 130.00 329.51 284.72 316.31 60.00 [5] Hlalele T, Zhang JF, Naidoo RM, Bansal RC. Multi-objective economic dispatch
with residential demand response programme under renewable obligation.
15 241.86 307.07 130.00 130.00 249.51 204.72 236.31 60.00
Energy 2021;218:14.
16 161.86 227.07 61.57 104.90 169.51 288.24 222.26 60.00
[6] Kim J, Kim K-KK. Dynamic programming for scalable just-in-time economic
17 150.00 231.34 130.00 107.40 239.75 208.24 222.26 60.00
dispatch with non-convex constraints and anytime participation. Int J Electr
18 234.03 304.63 130.00 111.33 159.75 213.96 309.50 160.00 Power Energy Syst 2020:123.
19 245.93 304.05 130.00 119.43 241.18 288.25 229.50 95.00 [7] Zheng L, Zhou X, Qiu Q, Yang L. Day-ahead optimal dispatch of an integrated
20 247.54 294.62 130.00 130.00 287.76 364.81 309.51 155.67 energy system considering time-frequency characteristics of renewable en-
21 250.87 305.39 130.00 130.00 306.88 364.86 309.51 90.67 ergy source output. Energy 2020;209.
22 170.87 289.36 130.00 107.51 226.88 284.86 229.51 60.00 [8] Coello CAC, Pulido GT, Lechuga MS. Handling multiple objectives with particle
23 150.00 229.90 20.00 105.24 150.00 204.86 222.26 60.00 swarm optimization. IEEE Trans Evol Comput 2004;8(3):256e79.
24 216.25 228.53 45.33 108.43 239.71 135.00 142.26 60.00 [9] Faramarzi A, Heidarinejad M, Stephens B, Mirjalili S. Equilibrium optimizer: a
novel optimization algorithm. Knowl Base Syst 2020;191:1e21.
Hour Generation dispatching of 9e15 units and transmission loss (MW) [10] Kaur S, Awasthi LK, Sangal AL, Dhiman G. Tunicate Swarm Algorithm: a new
P9 P10 P11 P12 P13 P14 P15 Loss bio-inspired based metaheuristic paradigm for global optimization. Eng Appl
Artif Intell 2020;90:1e29.
1 25.00 25.00 20.00 56.53 25.00 47.05 15.00 76.25 [11] Kumar M, Dhillon JS. Hybrid artificial algae algorithm for economic load
2 85.41 25.00 80.00 80.00 25.00 15.00 15.00 84.30 dispatch. Appl Soft Comput 2018;71:89e109.
3 25.41 124.73 20.00 56.53 25.00 15.00 55.00 149.20 [12] Alkoffash MS, Awadallah MA, Alweshah M, Abu Zitar R, Assaleh K, Al-
4 25.00 64.73 20.00 20.00 63.17 47.05 15.00 140.39 Betar MAA. Non-convex economic load dispatch using hybrid salp swarm
5 25.00 74.86 80.00 20.00 25.00 15.00 15.00 182.41 algorithm. Arabian J Sci Eng 2021:20.
6 124.99 25.00 80.00 79.99 63.30 55.00 15.00 174.70 [13] Pradhan M, Roy PK, Pal T. Grey wolf optimization applied to economic load
dispatch problems. Int J Electr Power Energy Syst 2016;83:325e34.
7 145.83 74.86 80.00 20.00 63.31 55.00 15.00 222.56
[14] El-Sehiemy RA, Rizk-Allah RM, Attia AF. Assessment of hurricane versus sine-
8 85.83 159.97 80.00 56.68 63.42 47.01 48.40 268.34
cosine optimization algorithms for economic/ecological emissions load
9 145.82 124.73 80.00 80.00 85.00 55.00 48.42 279.50 dispatch problem. Int Trans Electr Energy Syst 2019;29(2):1e14.
10 85.83 124.73 80.00 80.00 85.00 47.13 55.00 367.96 [15] Parouha RP, Das KN. Economic load dispatch using memory based differential
11 145.00 124.71 80.00 80.00 85.00 47.14 48.27 364.27 evolution. Int J Bio-Inspired Comput 2018;11(3):159e70.
12 145.85 74.86 80.00 80.00 85.00 55.00 55.00 385.73 [16] Zare M, Narimani MR, Malekpour M, Azizipanah-Abarghooee R, Terzija V.
13 145.69 74.86 80.00 80.00 63.31 47.04 48.46 320.92 Reserve constrained dynamic economic dispatch in multi-area power sys-
14 145.83 124.73 80.00 80.00 85.00 55.00 55.00 317.30 tems: an improved fireworks algorithm. Int J Electr Power Energy Syst
15 85.83 74.86 80.00 80.00 85.00 55.00 55.00 222.53 2021;126:1e14.
16 144.75 74.86 80.00 56.53 85.00 47.05 15.00 173.88 [17] Zhao XG, Zhang ZQ, Xie YM, Meng J. Economic-environmental dispatch of
17 84.75 74.86 80.00 80.00 25.00 15.00 48.42 173.12 microgrid based on improved quantum particle swarm optimization. Energy
18 84.32 25.00 80.00 57.26 25.00 47.04 15.02 217.86 2020;195:1e15.
19 85.41 125.00 80.00. 80.00 85.00 55.00 55.00 263.73 [18] Hassan MH, Houssein EH, Mahdy MA, Kamel S. An improved Manta ray
foraging optimizer for cost-effective emission dispatch problems. Eng Appl
20 25.41 124.57 79.99 80.00 85.00 55.00 48.38 304.33
Artif Intell 2021;100:1e20.
21 125.41 74.87 80.00 80.00 85.00 15.00 15.00 274.00
[19] Chopra N, Brar YS, Dhillon JS. An improved particle swarm optimization using
22 85.41 74.86 80.00 80.00 63.31 55.00 55.00 203.30
simplex-based deterministic approach for economic-emission power dispatch
23 85.41 25.00 80.00 80.00 85.00 47.05 15.00 110.46 problem. Electr Eng 2021:19.
24 25.41 25.00 20.00 80.00 25.00 15.00 15.00 100.67 [20] Fu C, Zhang SQ, Chao KH. Energy management of a power system for

18
Z.-F. Liu, L.-L. Li, Y.-W. Liu et al. Energy 235 (2021) 121407

economic load dispatch using the artificial intelligent algorithm. Electronics considering wind uncertainty using non-dominated sorting crisscross opti-
2020;9(1):1e19. mization. IEEE Access 2020;8:94678e96.
[21] Chang X, Xu Y, Sun H, Khan I. A distributed robust optimization approach for [38] Pourghasem P, Sohrabi F, Abapour M, Mohammadi-Ivatloo B. Stochastic
the economic dispatch of flexible resources. Int J Electr Power Energy Syst multi-objective dynamic dispatch of renewable and CHP-based islanded
2021:124. microgrids. Elec Power Syst Res 2019;173:193e201.
[22] Al-Betar MA, Awadallah MA, Krishan MM. A non-convex economic load [39] Hu FT, Hughes KJ, Ingham DB, Ma L, Pourkashanian M. Dynamic economic and
dispatch problem with valve loading effect using a hybrid grey wolf opti- emission dispatch model considering wind power under Energy Market Re-
mizer. Neural Comput Appl 2020;32(16):12127e54. form: a case study. Int J Electr Power Energy Syst 2019;110:184e96.
[23] Alsumait JS, Sykulski JK, Al-Othman AK. A hybrid GA-PS-SQP method to solve [40] Ratnaweera A, Halgamuge SK, Watson HC. Self-organizing hierarchical par-
power system valve-point economic dispatch problems. Appl Energy ticle swarm optimizer with time-varying acceleration coefficients. IEEE Trans
2010;87(5):1773e81. Evol Comput 2004;8(3):240e55.
[24] Kheshti M, Ding L, Ma SC, Zhao B. Double weighted particle swarm optimi- [41] Mirjalili S, Gandomi AH, Mirjalili SZ, Saremi S, Faris H, Mirjalili SM. Salp
zation to non-convex wind penetrated emission/economic dispatch and swarm algorithm: a bio-inspired optimizer for engineering design problems.
multiple fuel option systems. Renew Energy 2018;125:1021e37. Adv Eng Software 2017;114:163e91.
[25] Pandit N, Tripathi A, Tapaswi S, Pandit M. An improved bacterial foraging [42] Mirjalili S. Dragonfly algorithm: a new meta-heuristic optimization technique
algorithm for combined static/dynamic environmental economic dispatch. for solving single-objective, discrete, and multi-objective problems. Neural
Appl Soft Comput 2012;12(11):3500e13. Comput Appl 2016;27(4):1053e73.
[26] Liang HJ, Liu YG, Li FZ, Shen YJ. Dynamic economic/emission dispatch [43] Mirjalili S, Saremi S, Mirjalili SM, Coelho LD. Multi-objective grey wolf opti-
including PEVs for peak shaving and valley filling. IEEE Trans Ind Electron mizer: a novel algorithm for multi-criterion optimization. Expert Syst Appl
2019;66(4):2880e90. 2016;47:106e19.
[27] Al-Bahrani LT, Horan B, Seyedmahmoudian M, Stojcevski A. Dynamic eco- [44] Mirjalili S. Moth-flame optimization algorithm: a novel nature-inspired
nomic emission dispatch with load demand management for the load demand heuristic paradigm. Knowl Base Syst 2015;89:228e49.
of electric vehicles during crest shaving and valley filling in smart cities [45] Askarzadeh A. A novel metaheuristic method for solving constrained engi-
environment. Energy 2020;195:1e13. neering optimization problems: Crow search algorithm. Comput Struct
[28] Ghasemi M, Akbari E, Zand M, Hadipour M, Ghavidel S, Li L. An efficient 2016;169:1e12.
modified HPSO-TVAC-Based dynamic economic dispatch of generating units. [46] Chen MR, Zeng GQ, Lu KD. Constrained multi-objective population extremal
Elec Power Compon Syst 2019;47(19e20):1826e40. optimization based economic-emission dispatch incorporating renewable
[29] Barbosa-Ayala OI, Montanez-Barrera JA, Damian-Ascencio CE, Saldana- energy resources. Renew Energy 2019;143:277e94.
Robles A, Arturo Alfaro-Ayala J, Alfredo Padilla-Medina J, Cano-Andrade S. [47] Li LL, Liu ZF, Tseng ML, Zheng SJ, Lim MK. Improved tunicate swarm algo-
Solution to the economic emission dispatch problem using numerical poly- rithm: solving the dynamic economic emission dispatch problems. Appl Soft
nomial homotopy continuation. Energies 2020;13(17). Comput 2021:108.
[30] Basu M. Dynamic economic emission dispatch using nondominated sorting [48] Mohy-ud-din G. Hybrid dynamic economic emission dispatch of thermal,
genetic algorithm-II. Int J Electr Power Energy Syst 2008;30(2):140e9. wind, and photovoltaic power using the hybrid backtracking search algorithm
[31] Coelho LS, Mariani VC. Combining of chaotic differential evolution and with sequential quadratic programming. J Renew Sustain Energy 2017;9(1):
quadratic programming for economic dispatch optimization with valve-point 1e29.
effect. IEEE Trans Power Syst 2006;21(2):989e96. [49] Qian SQ, Wu HH, Xu GF. An improved particle swarm optimization with clone
[32] Xiong GJ, Shi DY. Hybrid biogeography-based optimization with brain storm selection principle for dynamic economic emission dispatch. Soft Comput
optimization for non-convex dynamic economic dispatch with valve-point 2020;24(20):15249e71.
effects. Energy 2018;157:424e35. [50] Alsumait JS, Qasem M, Sykulski JK, Al-Othman AK. An improved pattern
[33] Suresh V, Sreejith S, Sudabattula SK, Kamboj VK. Demand response-integrated search based algorithm to solve the dynamic economic dispatch problem with
economic dispatch incorporating renewable energy sources using ameliorated valve-point effect. Energy Convers Manag 2010;51(10):2062e7.
dragonfly algorithm. Electr Eng 2019;101(2):421e42. [51] Basu M. Particle swarm optimization based goal-attainment method for dy-
[34] Zou YW, Zhao JF, Ding DK, Miao FD, Sobhani B. Solving dynamic economic and namic economic emission dispatch. Elec Power Compon Syst 2006;34(9):
emission dispatch in power system integrated electric vehicle and wind 1015e25.
turbine using multi-objective virus colony search algorithm. Sustain Cities Soc [52] Ghasemi M, Akbari E, Rahimnejad A, Razavi SE, Ghavidel S, Li L. Phasor particle
2021;67:1e16. swarm optimization: a simple and efficient variant of PSO. Soft Comput
[35] Alham MH, Elshahed M, Ibrahim DK, El Zahab EEA. A dynamic economic 2019;23(19):9701e18.
emission dispatch considering wind power uncertainty incorporating energy [53] Mason K, Duggan J, Howley E. Multi-objective dynamic economic emission
storage system and demand side management. Renew Energy 2016;96: dispatch using particle swarm optimization variants. Neurocomputing
800e11. 2017;270(12):188e97.
[36] Qiao BH, Liu J. Multi-objective dynamic economic emission dispatch based on [54] Mason K, Duggan J, Howley E. A multi-objective neural network trained with
electric vehicles and wind power integrated system using differential evolu- differential evolution for dynamic economic emission dispatch. Int J Electr
tion algorithm. Renew Energy 2020;154:316e36. Power Energy Syst 2018;100:201e21.
[37] Chinnadurrai CL, Victoire TAA. Dynamic economic emission dispatch

19

You might also like