You are on page 1of 22

Neural Computing and Applications (2023) 35:20749–20770

https://doi.org/10.1007/s00521-023-08813-5(0123456789().,-volV)(0123456789().
,- volV)

ORIGINAL ARTICLE

An improved weighted mean of vectors algorithm for microgrid


energy management considering demand response
Nehmedo Alamir2,1 • Salah Kamel2 • Mohamed H. Hassan2,3 • Sobhy M. Abdelkader1,4

Received: 20 September 2022 / Accepted: 28 June 2023 / Published online: 25 July 2023
 The Author(s) 2023

Abstract
The integration of demand response programs (DRPs) into the energy management (EM) system of microgrids (MGs)
helps in improving the load characteristics by allowing consumers to interoperate for achieving techno-economic
advantages. In this paper, an improved algorithm is called LINFO is proposed for modifying search ability of the original
weIghted meaN oF vectOrs (INFO) algorithm as well as avoiding its weaknesses like trapping in a local optima. The
improved algorithm’s efficiency is confirmed by comparing its results with those obtained by the original INFO and other
optimization techniques using different standard benchmark test functions. Moreover, this improved algorithm and the
original version are applied for solving the EM problem with the aim of optimizing the operation cost of the MGs in the
presence DRPs. They are used to solve day-ahead EM problem for optimal operation of renewable energy resources, the
optimal generation from a conventional diesel engines (DEs); taking into account the participation of customers in DRP for
minimizing MG operating cost, which includes the cost of DEs fuel and the power transactions cost with the main grid. To
demonstrate the efficacy of the proposed LINFO, simulation results are compared with the results of well-known and newly
developed optimization techniques.

Keywords Microgrid  Optimization  Energy management  Demand response  LINFO

1 Introduction

1.1 Motivation

Increasing concern about global warming and depletion of


fossil fuel sources have emphasized the importance of
& Nehmedo Alamir
Nehmedo.Alamir@ejust.edu.eg electricity generation from Renewable Energy Sources
(RESs) [1–5] and the integration of Energy storage system
Salah Kamel
skamel@aswu.edu.eg with their high efficiency and stable operation [6, 7]. So the
integration of these RESs as sub-systems production or
Mohamed H. Hassan
mohamedhosnymoee@gmail.com Distributed Generation (DG), which is known as Micro-
grids (MGs), into the main grid is essential for solving
Sobhy M. Abdelkader
sobhy.abdelkader@ejust.edu.eg energy-related issues [8]. The advantages of incorporating
RESs into the MG include lower operating costs, higher
1
Electrical Power Engineering, Egypt-Japan University of profitability for investors, reduced level of greenhouse gas
Science and Technology, New Borg El-Arab, emissions, and cheaper power obtained from the grid [9].
Alexandria 21934, Egypt
2
The MG has three distinct advantages: (1) Technical
Department of Electrical Engineering, Faculty of advantages, such as enhancing energy efficiency and
Engineering, Aswan University, Aswan 81542, Egypt
3
increasing power system resilience; (2) Economic benefits,
Ministry of Electricity and Renewable Energy, Cairo, Egypt such as reducing fuel and interruption costs; and (3)
4
Department of Electrical Engineering, Faculty of
Engineering, Mansoura University, Mansoura 35516, Egypt

123
20750 Neural Computing and Applications (2023) 35:20749–20770

techno-economic and environmental advantages [18] pre-


DSM
sents a MG model that aims for maximum reliability. In
Ref. [23], the authors suggested an Ant Colony Opti-
mization (ACO) technique for solving the EM and deter-
Energy DRP
Efficiency mining optimal scheduling for energy sources in the MG in
order to lower the overall generation cost. Optimal
Incen ve scheduling for generation sources to get economic benefits
Price based
based of MG comprised of RESs and Energy storage system
(ESS) is proposed using PSO in [24] and improved PSO in
[25]. In [26] the optimal operation of MG based on Water
Direct load control
Cycle Algorithm (wCA) is solved to minimizing the
operating cost. In [27] a Mixed Integer Distributed Ant
Demand biding Colony Optimization is developed to solve the MG eco-
nomic dispatch problem for cost minimization. According
to the aforementioned works, the main research considered
Emergency DR passive customers’ participation in solving EM in MG.
Implementing solutions to alleviate the network’s demand
Load curtailment side may help further optimize the Microgrid’s EM
problem.
Fig. 1 Categorization of demand side management techniques
In literature, different types of DRPs have been dis-
cussed in order to optimize the MG operation. [28] Used a
Environmental benefits, such as reduced greenhouse gas hybrid augmented weighted e-constraint technique and
emissions [10]. lexicographic optimization to solve optimal power flow in
EM in MG is considered as operation and planning MG considering DR in a combined heat and power (CHP)
stage, which is considered a cost-driven scheduling for MG system. Annual operating cost is minimized in [29] while
sources, to supply load demand with maintaining some maximizing customer satisfaction. In [30], different types
constraints [11]. Moreover, with the advancements in smart of DRP have been introduced and compared for Multi-
grid technology, there has been a rise in the load demand Microgrid using linear programming with Mixed integer
[12], thereby necessitating the implementation of an energy mathematical models. In [31], optimal scheduling for the
management system to ensure a balance between load and MG with RES considering its stochastic nature and con-
energy supply [13]. As a result, there has been a significant ventional generators has been presented. In addition, an IB-
research focus on MG Energy Management to achieve DR program has been used to minimize energy cost and
optimal operation of MGs [14–19]. ensure customers’ benefits using the technique of weighted
Although peak hours are just a few hours per day, sig- sum and the fuzzy satisfying method. An IB-DR for
nificant investment in generation, transmission, and distri- deterministic EM was proposed in [32] using Honey
bution infrastructure is necessary to supply peak demand. Badger Optimizer, and Probabilistic EM using Artificial
As a consequence, the cost of power supply increases. So, Hummingbird Algorithm in [33] A PB-DR is formulated in
Demand Side Management programs (DSMP) are inte- [34] to minimize the cost of power losses in MG and in
grated into EM systems in MGs, which can enhance the [35] using PSO to maximize the customers’ profit. In [36] a
load curve by allowing the interpolation of the customers multi-objective EM problem is solved based on Genetic
[20]. algorithm (GA) using PB-DR to minimize the net present
These DSMP schemes can be categorized into two main cost. In [19] a formulation for an IB-DR to minimize the
groups energy efficiency and Demand Response Programs, generation cost and maximize the MG operator is pro-
with further sub-categories, incentive-based DR (IB-DR) posed. While Ref. [37] solve the economic dispatch prob-
and price-based DR(PB-DR), as shown in Fig. 1 [21]. lem based on IB-DR using Genetic algorithm Ref. [38] a
hybrid technique for EM is implemented based on Pelican
1.2 Related work Optimization Algorithm to solve the optimal operation and
to reduce the demand during the peak periods. In [39] an
In [22], using Particle Swarm Optimization (PSO), a co- IB-DR has been used for bi-level programing; in the upper-
optimization technique for MG planning was developed to level based on unit commitment the marginal prices is
identify the appropriate sizing of DG sources while opti- achieved, and the cost is minimized in the lower-level. A
mizing yearly fuel costs. To maximize solar-based MGs’ multi-objective DR problem using a-constrained domi-
nated differential evolution algorithm for reducing the

123
Neural Computing and Applications (2023) 35:20749–20770 20751

customers’ bill and enhance the energy efficiency is pro- second objective is to optimize MG’s operation consider-
posed with integration of DR [40]. ing incentive DR based on the proposed algorithm and
According to the above-mentioned researches, many compare its performance with other optimization meta-
optimization approaches have effectively handled the heuristic and newly developed algorithms.
engineering problems, particularly those involving EM. The main contributions of this paper to the aforemen-
The findings of these studies demonstrated the importance tioned investigation may be stated as follows:
and potential for developing new and improved optimiza-
1. Proposing an improved algorithm called LINFO to
tion approaches for solving specific EM problems; also
avoid the drawbacks of the original INFO algorithm of
According to the No-Free-Lunch (NFL) theorem [41] there
being trapped in local optima. And Validating the
are either no metaheuristic optimization algorithms capable
performance of the proposed algorithm by comparing it
of solving all optimization problems. These two reasons
with newly developed algorithms to validate its
motivate us to propose a new optimization technique,
performance, based on fitness value using different
called Leader-based mutation-selection INFO (LINFO)
test benchmark functions based on different a statistical
based on the weIghted meaN oF vectOrs (INFO) algorithm
terms.
to prevent the possibility to be trapped into local minima.
2. Solving MG energy management problem considering
We chose the algorithms presented here due to their out-
demand response program based on the proposed
standing performance in solving a variety of mathematics
LINFO algorithm. The performance of the proposed
and engineering design challenges.
LINFO algorithm is compared with well-known and
newly developed algorithms in solving the EM
1.3 Contribution
problem.
The first objective of this paper is to propose an improved
version of INFO algorithm to overcome its weakness; the

Main Grid

Central control

Wind System Diesel Generator Residenal customers Solar PV system


Fig. 2 Grid-connected MG system

123
20752 Neural Computing and Applications (2023) 35:20749–20770

1.4 Paper organization Fig. 4 Qualitative metrics of eight benchmark functions including: c
2D image of the functions, search history, average fitness history, and
convergence curve using LINFO technique
The rest of this paper is structured as follows: Sect. 2
presents modeling of the Grid-connected Microgrid with
modeling of DRP. Section 3 discusses the EM optimization 8   9
problem modeling; Sect. 4 focuses on the proposed LINFO   < ct  Pgt  Pg t [ 0 =
algorithms. The obtained simulation results are presented Cg Pgt ¼ 0   Pg t ¼ 0 ð1Þ
: ;
in Sect. 5. Finally, the paper is concluded in Sect. 6. ct  Pgt  Pgt \0

2 Grid-connected MG scheme • Solar power modeling


For a given area, the hourly output power from the solar
2.1 Modeling
PV generator is given as [44]:
• MG model Pst ¼ gPV Ac IPvt ð2Þ

The structure of proposed MG is shown in Fig. 2. MG’s where, gpv is solar PV system’s efficiency, which is a
resources are a PV system, WT system, non-renewable function of the incident solar irradiation ðIPvt Þ(kW h/m2),
sources such as DE and customers with DRP. Integrating and the ambient temperature and Ac is the PV array area
DG (including renewable sources) requires power conver- [44].
sion units; renewable sources need a power electronics • WT modeling
converter to track the maximum power and synchroniza-
tion with the grid [42]. In this study, it is assumed that The hourly power produced by a WT (Pwt ) is entirely
power is transacted between the MG and the main grid. dependent on hourly wind speed at a specific height ðvhubt Þ,
Cost of Power transacted (Pgt ) at any time interval t is air density, swept area and converter efficiency. Using
  power-law equation, wind speed at the required hub height
Cg Pgt and can be obtained as [43]:

Initialize dim, population Calculate Objective Select three randomly


Start Generate initial poputalion
size, max iteration function of each solution solutions (xa, xb, xc)
NO

Calculate ul by Check if rand < pCr Calculate the solutions z1


YES Update σ (Eq. (25))
Eq.(26.1) & rand < 0.5 and z2 by Eq. (24)

NO NO

Calculate ul by YES
Check if rand < pCr
Eq.(26.2) & rand > 0.5 Update ul by
NO
Eq.(27.1) xl+1 = ul
YES YES

Calculate ul by Eq.(26.3) Check if Check if


Check if rand < 0.5
f(ul) < f(xl) l < Np
YES NO NO

Check if rand < 0.5 Update ul by xl+1 = xl YES


Eq.(27.2)

NO

(Leader-based mutation-selection )
Update the position of each individual Check if
YES
using Eq. (28) g < Maxg

Update the best solution found so far Evaluate the objective function value of
End
xbest using Eq. (30) each individual using Eq. (29)

Fig. 3 Flowchart of proposed LINFO algorithm

123
Neural Computing and Applications (2023) 35:20749–20770 20753

123
20754 Neural Computing and Applications (2023) 35:20749–20770

Table 1 Unimodal results for


Function LINFO INFO BOA TSA GWO
benchmark functions
F1
Best 4.01E-55 1.58E-49 3.87E-08 3.79E-08 4.47E-12
Mean 3.22E-53 1.77E-43 4.96E-08 3.92E-07 3.12E-11
Median 2.21E-53 2.92E-45 4.95E-08 1E-07 2.46E-11
Worst 1.11E-52 1.63E-42 6E-08 4.09E-06 8.73E-11
std 3.33E-53 4.65E-43 4.94E-09 9.2E-07 2.31E-11
F2
Best 1.19E-27 6.99E-25 4.26E-06 2.44E-06 1.42E-07
Mean 2.54E-26 5.36E-23 5.71E-06 1.9E-05 2.77E-07
Median 1.96E-26 7.68E-24 5.77E-06 1.86E-05 2.66E-07
Worst 6.85E-26 3.89E-22 7.58E-06 3.68E-05 4.78E-07
std 1.77E-26 1.17E-22 9.92E-07 9.44E-06 9.9E-08
F3
Best 2.65E-52 5.67E-37 3.85E-08 0.027608 0.008462
Mean 5.33E-34 1.51E-31 4.67E-08 1.122677 0.610441
Median 3.51E-50 4.57E-34 4.61E-08 0.772195 0.185412
Worst 1.07E-32 2.52E-30 5.57E-08 3.914695 3.567009
std 2.38E-33 5.63E-31 5.02E-09 1.096313 0.827115
F4
Best 5.37E-28 2.61E-22 8.45E-06 0.67531 0.002608
Mean 5.81E-18 3.55E-19 1.02E-05 3.616654 0.008
Median 5.46E-27 9.66E-21 1.02E-05 3.022253 0.007092
Worst 1.16E-16 3.23E-18 1.15E-05 9.361516 0.016667
std 2.6E-17 8.49E-19 8.51E-07 2.343658 0.003845
F5
Best 24.04967 24.01246 28.89058 27.18973 25.92515
Mean 24.97259 24.79812 28.92369 39.01094 27.18903
Median 25.0525 24.84 28.91978 28.66203 27.09814
Worst 26.96976 25.97111 28.96927 239.7785 28.79035
std 0.811402 0.541049 0.021273 47.26339 0.72182
F6
Best 4.62E-07 8.78E-07 4.311051 2.886997 0.252254
Mean 2.74E-05 9.46E-06 5.211726 3.800719 0.647554
Median 8.08E-06 8E-06 5.06303 3.736935 0.611378
Worst 0.000124 2.95E-05 6.168001 4.850371 1.172757
std 4.07E-05 6.79E-06 0.509499 0.527851 0.280888
F7
Best 0.00024 0.000143 0.000983 0.007604 0.001477
Mean 0.001463 0.002586 0.002696 0.019206 0.004433
Median 0.000991 0.002572 0.002776 0.018479 0.003685
Worst 0.007414 0.008415 0.005116 0.04436 0.01033
std 0.001575 0.001809 0.001104 0.007628 0.002554
Bold is an indication for the best obtained results in the comparison

123
Neural Computing and Applications (2023) 35:20749–20770 20755

Table 2 Multimodal results for


Function LINFO INFO BOA TSA GWO
benchmark functions
F8
Best - 11,639.6 - 9982 - 921.028 - 1394.45 - 1495.31
Mean - 9084.4 - 9002.69 - 766.513 - 1212.82 - 1245.57
Median - 8935.77 - 9158.23 - 778.594 - 1232.52 - 1224.18
Worst - 7529.7 - 7772.56 - 647.792 - 976.635 - 1123.85
std 1106.954 599.3694 61.76107 122.0762 104.0153
F9
Best 0.00 0.00 5.17E-09 156.667 1.062467
Mean 0.00 0.00 0.003376 228.0177 9.801018
Median 0.00 0.00 3.86E-06 228.634 9.824713
Worst 0.00 0.00 0.047754 331.7581 24.96968
std 0.00 0.00 0.010836 46.40919 5.565812
F10
Best 8.88E-16 8.88E-16 1.67E-05 20.81133 20.76487
Mean 8.88E-16 8.88E-16 4.77E-05 20.9608 20.92344
Median 8.88E-16 8.88E-16 4.55E-05 20.99356 20.94465
Worst 8.88E-16 8.88E-16 7.94E-05 21.0961 21.06309
std 0.00 0.00 1.69E-05 0.091505 0.083433
F11
Best 0.00 0.00 3.23E-08 1.3E-09 6.56E-13
Mean 0.00 0.00 4.29E-08 0.007018 0.009891
Median 0.00 0.00 4.22E-08 1.44E-08 4.55E-12
Worst 0.00 0.00 5.81E-08 0.029126 0.055407
std 0.00 0.00 6.29E-09 0.010243 0.015766
F12
Best 5.48E-07 6.44E-08 0.33315 0.374956 0.006066
Mean 0.000319 0.01037 0.565424 2.805889 0.026151
Median 1.38E-06 6.06E-07 0.562862 2.009833 0.023474
Worst 0.006332 0.10368 0.754521 7.656863 0.047176
std 0.001415 0.03191 0.108748 2.128936 0.013414
F13
Best 0.012347 2.35E-05 2.497296 2.372295 0.09955
Mean 0.464452 0.117949 2.894224 3.298085 0.613832
Median 0.372533 0.054429 2.982946 3.22876 0.609981
Worst 1.629995 0.885887 3.109356 4.16073 1.044
std 0.38461 0.202976 0.153028 0.565835 0.280029
Bold is an indication for the best obtained results in the comparison

8
may be converted from the anemometer height as follow- > 0 v  vci and v  vco
>
< v2 2
ing [45]: hubt  vci
  Pwt ¼ :Pnom vci  vhubt  vnom ð4Þ
> 2 2
hhub b : vnom  vci
>
vhubt ¼ vref t ð3Þ Pnom vnom \vhubt  vco
href
where Pnom is the rated power of WT, vnom is the rated
where hhub is hub height; vref t is the hourly wind speed at
wind speed,vco the cut-out wind speed, and vci the cut-in
the reference hub height href . The power law exponent b is
wind speed, and, with Pw and v denoting output power and
usually is in the range between 14 and 17.
wind speed.
The hourly wind power is calculated as [45]:

123
20756 Neural Computing and Applications (2023) 35:20749–20770

Table 3 Composite results for benchmark functions


Function LINFO INFO BOA TSA GWO

F14
Best 0.998004 0.998004 0.998004 0.998004 0.998004
Mean 1.842329 1.540789 1.301281 8.298683 3.892106
Median 0.998004 0.998004 1.024436 10.76318 2.982105
Worst 9.952028 5.928845 2.983027 18.30431 12.67051
std 2.097625 1.51696 0.534994 5.533952 3.727681
F15
Best 0.000307 0.000307 0.000315 0.000308 0.00031
Mean 0.001615 0.004456 0.000487 0.007136 0.003547
Median 0.000308 0.000307 0.000405 0.000505 0.000546
Worst 0.020363 0.020363 0.000917 0.031699 0.020363
std 0.004433 0.008167 0.000173 0.010606 0.007255
F16
Best - 1.03163 - 1.03163 - 1.40747 - 1.03163 - 1.03163
Mean - 1.03163 - 1.03161 - 1.18199 - 1.0253 - 1.03158
Median - 1.03163 - 1.03163 - 1.18517 - 1.03163 - 1.03163
Worst - 1.03163 - 1.03131 - 1.07213 - 0.99999 - 1.03063
std 1.23E-10 7.03E-05 0.088213 0.012981 0.000223
F17
Best 0.397887 0.397887 0.398293 0.39789 0.397888
Mean 0.397887 0.397887 0.409332 0.397927 0.397891
Median 0.397887 0.397887 0.406611 0.397907 0.397891
Worst 0.397887 0.397887 0.461881 0.398082 0.397897
std 6.19E-12 0.00 0.014049 4.53E-05 3.01E-06
F18
Best 3.00 3.00 3.000586 3.000009 3.00
Mean 3.00 3.00 3.092676 8.400078 3.000068
Median 3.00 3.00 3.054728 3.000084 3.000036
Worst 3.00 3.00 3.425476 84.00001 3.000238
std 3.43E-11 1.24E-15 0.108993 18.78799 6.53E-05
F19
Best - 3.86278 - 3.86278 - 0.3005 - 0.30048 - 0.30048
Mean - 3.86278 - 3.86278 - 0.30048 - 0.30048 - 0.30048
Median - 3.86278 - 3.86278 - 0.30048 - 0.30048 - 0.30048
Worst - 3.86278 - 3.86278 - 0.30048 - 0.30048 - 0.30048
std 1.01E-10 2.17E-15 3.74E-06 1.14E-16 1.14E-16
F20
Best - 3.322 - 3.322 - 3.3E-05 - 3.32148 - 3.32198
Mean - 3.26255 - 3.2566 - 1.6E-06 - 3.07223 - 3.22876
Median - 3.26255 - 3.2031 - 1.5E-40 - 3.20118 - 3.26239
Worst - 3.2031 - 3.2031 - 2E-134 - 0.20816 - 2.84039
std 0.060991 0.060685 7.35E-06 0.679321 0.125558
F21
Best - 10.1532 - 10.1532 - 4.61081 - 10.0895 - 10.1502
Mean - 9.77686 - 8.89113 - 4.0759 - 5.89545 - 8.51218
Median - 10.1532 - 10.1532 - 4.12522 - 4.90994 - 10.1413
Worst - 2.63047 - 2.63047 - 3.18003 - 2.58642 - 2.62918
std 1.682085 2.648789 0.379957 2.775111 2.963153

123
Neural Computing and Applications (2023) 35:20749–20770 20757

Table 3 (continued)
Function LINFO INFO BOA TSA GWO

F22
Best - 10.4029 - 10.4029 - 4.76031 - 10.3637 - 10.4024
Mean - 10.0211 - 9.3053 - 3.74931 - 7.02119 - 10.0134
Median - 10.4029 - 10.4029 - 3.64889 - 9.8942 - 10.3959
Worst - 2.7659 - 2.7659 - 2.93305 - 1.82478 - 2.76526
std 1.707695 2.68678 0.479377 3.57071 1.706042
F23
Best - 10.5364 - 10.5364 - 4.51577 - 10.4599 - 10.5348
Mean - 9.72522 - 9.46058 - 3.38426 - 5.50502 - 9.74305
Median - 10.5364 - 10.5364 - 3.60414 - 2.83596 - 10.5274
Worst - 2.42173 - 2.42173 - 1.95854 - 1.66783 - 2.42135
std 2.496783 2.640833 0.720921 3.728197 2.418464
Bold is an indication for the best obtained results in the comparison

• Modeling of DR maximization MG benefit while meeting operational


restrictions.
The formulation for customer’s cost function ðC ðh; xÞÞ
Thus, the formulation of the EM problem may be
can be obtained mathematically as [46]:
described mathematical as follows:
C ðh; xÞ ¼ k1 x2 þ k2 xð1  hÞ ð5Þ T X
X I X
T  
where k1 , and k2 are cost coefficients, and h is the customer min f1 ð xÞ ¼ min Ci ðPit Þ þ Cg Pgt ð8Þ
t¼1 i¼1 t¼1
willing rate; it ranges between 0 and 1, with the most
willing Costumer having a value of 1 and the least willing While Ci ðPit Þ is the ith conventional generator’s fuel
customer having a value of 0. cost at any time interval t and it is represented by a
Customers will reduce their consumption in response to quadratic model as follows [44]:
DRP only if their benefit Bj  0 and so the benefit for any Ci ðPit Þ ¼ ai p2it þ bi Pit ð9Þ
customer j can be expressed as [43]:
  where ai and bi are fuel cost coefficients.
Bj ðh; y; xÞ ¼ yj  k1 x2j þ k2 xj  k2 xj h ; From (9), it is seen that the first term in this first
ð6Þ
for j ¼ 1; 2; . . .J objective function includes the total operating cost of the
DE units. While the second term represents the cost of
where yj is the incentive that customers receive, xj is the
power transaction between the main grid and MG.
amount of consumption reduction in (KW or MW).
The second objective of EM system is maximizing the
Moreover, MG benefit is calculated as [43]:
MG operator benefit, which is expressed as in (7); the
X
J objective function can be calculated as
B0 ¼ kj x j  y j : ð7Þ
j¼1
T X
X I
max f2 ð xÞ ¼ max k j xj  yj ð10Þ
t¼1 i¼1

So that the mathematical model of the objective function


3 The operating cost function for MG management is given as:
" #
XT X I XT
As previously stated, the Grid-connected MG is composed min w1 Ci ðPit Þ þ Cr ðPrt Þ
of various generating sources, including conventional
"t¼1 i¼1 #t¼1
generators and RESs, as well as load with a DRP. The X T X I
operational cost is composed of two components: the cost þ w2 y j  kj x j ð11Þ
t¼1 i¼1
of DEs generation and the cost of power transactions. So,
the EM system’s objective in MG is to optimally operate where w is the weighting factor and the following equation
MG’s resources by minimizing generation costs and should be satisfied:

123
20758 Neural Computing and Applications (2023) 35:20749–20770

Fig. 5 The convergence characteristics of all optimization algorithms for 23 benchmark functions

w1 þ w2 ¼ 1: ð12Þ X
I X
J
Pit þ Pwt þ Pst þ Prt ¼ Dt  xj;t : ð13Þ
3.1 Operating constraints i¼1 j¼1

• Power balance constraints [43]:

123
Neural Computing and Applications (2023) 35:20749–20770 20759

Fig. 5 continued

The total generated power from MG sources (DEs, WT, • Generation constraints [43]:
and PV), and power transacted should match the load Pimin  Pit  Pimax ð14Þ
demand.
Where Dt and xj;t are the total demand and the jth cus-
tomer power curtailed at time t, respectively.

123
20760 Neural Computing and Applications (2023) 35:20749–20770

Fig. 6 Boxplots for l optimization algorithms for 23 benchmark functions

DRi  Pitþ1  Pit  URi ð15Þ where Pimin and Pimax are the minimum and maximum power
generated from any generator i, respectively; URi and DRi
0  Pst  Pstmax ð16Þ
are the maximum ramp up and ramp down rates for gen-
0  Pwt  Pwtmax ð17Þ erator I;Pstmax and Pwtmax are the forecasted maximum power
from PV and wind generators at any time t.

123
Neural Computing and Applications (2023) 35:20749–20770 20761

Fig. 6 continued

Constraint (14) ensures that any DE generation is within • Power transaction constraints:
the maximum and minimum limits, while constraint (15)
The power transacted power between utility grid and
states that the ramping up and down rate limits should not
MG should not exceed the maximum permissible limit
be violated. Equations (16) and (17) represent the con-
Pgmax as [43]:
straints of the maximum and minimum generation limits of
WT generator and solar PV generation, respectively. Pgmax  Pgt  Pgmax ð18Þ

123
20762 Neural Computing and Applications (2023) 35:20749–20770

4 Solution method
• Demand response constraints [43]: In this section, the weIghted meaN oF vectOrs (INFO)
The demand management formulation in (6) is extended algorithm is briefly presented then the process of the pro-
for the whole day; so customers’ benefit can be expressed posed leader INFO (LINFO) technique is described.
as in (19) to ensure the incentive is higher than the cost that
the customer should pay. 4.1 Original INFO algorithm
XT  
yj;t  k1 x2j;t þ k2 xj:t  k2 xj;t h  0 ð19Þ The INFO algorithm was introduced in 2022 [47]. For
t¼1 updating the vectors’ location, INFO algorithm use four
main processes: initialization stage, updating rule, vector
X
T X
T
yj;t  ðk1 x2j;t þ k2 xj;t  k2 xj;t hÞ  yj1;t  ðk1 x2j1;t combining, and a local search.
t¼1 t¼1
• Initialization stage
þ k2 xj1;t  k2 xj1;t hÞ
For j ¼ 2; 3; . . .; J The INFO algorithm is composed of n vectors popula-
ð20Þ tion in search space with diminution D. The initial popu-
lation is generated at random as following:
X
J
xj;t  CMj ð21Þ Xn ¼ Xmin þ rand ð0; 1Þ:ðXmax  Xmin Þ ð23Þ
t¼1
where Xn denotes the nth vector, Xmin ; Xmax are the bounds
X
T X
J
of the solution domain in each problem and rand ð0; 1Þ
yj;t  UB ð22Þ
t¼1 j¼1
return a random number in an interval of [0, 1].

where UB and CMj are daily MG budget limit and cus- • updating rule stage
tomer j daily power curtailment limit. This stage enhances the diversity of the population
Constraint in (20) ensures that the bigger the customer’s throughout the search process. This operation creates new
curtailment, the larger the incentive they will get. The total vectors using the weighted mean of existing vector. The
MG budget limit constraint is described in Eq. (21) to main formulation of the updating rule is presented as:
ensure the daily budget is lower than the limit. Equa-

ð24Þ

tion (22) is to ensure the total curtailment of any customer j where z1gl and z2gl are the new vectors in the gth genera-
is within the permissible limits. tion; and r is the scaling rate of a vector, and it is defined
as:

123
Neural Computing and Applications (2023) 35:20749–20770 20763

r ¼ 2a  rand  a xrnd ¼ /  xavg þ ð1  /Þ  ð/  xbt þ ð1  /Þ  xbs Þ


 
g ð25Þ ð27:3Þ
a ¼ 2  exp 4:
Maxg ðx a þ xb þ xc Þ
xavg ¼ ð27:4Þ
3

• vector combining stage where / is a random number in between 0 and 1; and xrnd
denotes a new solution that randomly combines the three
In order to enhance the diversity of the population in the solutions (xavg , xbt , and xbs ); Which increase proposed
INFO algorithm, using the z1gl and z2gl that calculated in the algorithm’s nature of randomness, and so better search
previous stage (new vectors) in a combination with vector ability in the search space. t1 and t2 are two random
xgl based on the following equations as: numbers given by the following equations:

2  rand if p [ 0:5
t1 ¼ ð27:5Þ
1 otherwise
Where l have a value of 0:05  randn.
rand if p\0:5
• local search stage t2 ¼ ð27:6Þ
1 otherwise
The search ability of this stage prevents the algorithm to where p is a random number in the range of (0, 1).
drop into local optima. Considering the local operator using
the global position (xgbest ), a new vector can be generated
around this global position as follows:

In which

123
20764 Neural Computing and Applications (2023) 35:20749–20770

Table 4 DE units data and customer data (Test system 1)


i,j DE Customer
2
ai ($/MW h) bi ($/KW h) Pi,min (KW) Pi,max (KW) DRi (KW/h) URi (KW/h) h K1,j K2,j CMj (KW)

1 0.06 0.5 0 4 3 3 0 1.079 1.32 30


2 0.03 0.25 0 6 5 5 0.45 1.378 1.36 35
3 0.04 0.3 0 9 8 8 0.9 1.847 1.64 40

Table 5 ki,t values and total initial hourly demand (test system 1) Table 6 Comparison of the EM problem for test system 1
Time (h) ki,t ($) Total demand (KW) Total operating cost ($)

t=1 1.57 31.83 Technique Worst Best Mean SD


t=2 1.4 31.4 PSO [52] 807.35 477.4 658.49 90.47
t=3 2.2 31.17 JAYA [53] 790.67 647.49 711.32 41.57
t=4 3.76 31 HBA [54] 672.81 460.39 545.31 40.11
t=5 4.5 31.17 INFO [47] 747.1355 509.374 615.9763 66.60
t=6 4.7 32.1 LINFO 356.18 532.38 441.24 38.85
t=7 5.04 32.97
t=8 5.35 34.1
t=9 6.7 37.53
t = 10 6.16 38.33
t = 11 6.38 40.03
t = 12 6.82 41.17
t = 13 7.3 39.67
t = 14 7.8 41.7
t = 15 8.5 42.1
t = 16 7.1 41.67
t = 17 6.8 40.7
t = 18 6.3 40.07
t = 19 5.8 38.63
t = 20 4.2 36.4
t = 21 3.8 34.1
t = 22 3.01 32.8
t = 23 2.53 32.5
t = 24 1.42 32
Fig. 8 Convergence characteristic curve of test system 1

4.2 Proposed leader INFO algorithm

The development of the Leader-based mutation-selection


[48] to solve the possibility of the optimal value may drop
into a local minima. This modification depends on the best
vector’s position xtbest , the following best vector’s position
(second-best) xtbest1 and the third-best vector position
vector xtbest2 based on the value of the objective function
on the new vector’s location xi ðnewÞ between the number
of population. Then the new mutation vector’s position
vector xi ðmutÞ is given by:

Fig. 7 Hourly solar and wind power for test system 1

123
Neural Computing and Applications (2023) 35:20749–20770 20765

Table 7 Total amount of energy saved and the total amount of


incentive received (Test system1)
j Saving (kWh) Incentive ($)

1 27.69 54.67
2 34.4 67.42
3 38.99 75.18
Total 101.13 197.2724

 
t
xi ðmutÞ ¼ xi ðnewÞ þ 2  1 
 Max it  
 ð2  rand  1Þ2  xtbest  xtbest1
 þ xtbest2
þ ð2  rand  1Þ xtbest  xi ðnewÞ
ð28Þ
Fig. 9 Hourly DE units generated power and grid power for test
system 1 Then, the next location is updated using the following
equation:
xi ðmutÞ f ðxi ðmutÞÞ\f ðxi ðnewÞÞ
x i ð t þ 1Þ ¼ ð29Þ
xi ðnewÞ f ðxi ðmutÞÞ  f ðxi ðnewÞÞ
Finally, the best solution is updated as follows:
xi ðmutÞ f ðxi ðmutÞÞ\f ðxbest Þ
xbest ¼ ð30Þ
xi ðnewÞ f ðxi ðnewÞÞ\f ðxbest Þ
The flowchart of the proposed LINFO technique is
displayed in Fig. 3. The place of Leader-based mutation-
selection in the proposed algorithm is presented in this
figure. This modification leads to enhance the exploration
of the proposed LINFO algorithm based on the simulta-
Fig. 10 Load demand and final load with DR neous crossover and mutation using the three best leaders.

Fig. 11 Power curtailed and


incentive paid to customers for
test system 1

123
20766 Neural Computing and Applications (2023) 35:20749–20770

Table 8 Ten conventional


I ai ($/MW2h) bi ($/MWh) Pi,min (MW) Pi,max (MW) DRi (MW/h) URi (MW/h)
generators units data for test
system 2 1 0.00043 21.6 30 370 200 200
2 0.00063 21.05 35 360 200 200
3 0.000394 20.81 33 240 150 150
4 0.0007 23.9 30 200 100 100
5 0.00079 21.62 33 143 80 80
6 0.00056 17.87 37 60 30 30
7 0.00211 16.51 20 30 30 30
8 0.0048 23.23 27 120 60 60
9 0.10908 19.58 20 80 40 40
10 0.00951 22.54 25 55 30 30

Table 9 Customer type,


J h K1,j K2,j
obtained from four recent techniques including, gray wolf
coefficients of the customer cost optimizer (GWO) [49], tunicate swarm algorithm (TSA)
function, and daily customer 1 0 1.847 11.64 180
curtailment limitations (test
[50], butterfly optimization algorithm (BOA) [51], and the
system 2) 2 0.14 1.378 11.63 230 conventional INFO algorithm [47]. Figure 4 displays the
3 0.26 1.079 11.32 310 Qualitative metrics on F1, F4, F5, F8, F9, F12, F15, and
4 0.37 0.9124 11.5 390 F18: 2D views of the functions, convergence curve, aver-
5 0.55 0.8794 11.21 440 age fitness history, and search history.
6 0.84 1.378 11.63 530 Statistical analysis of the proposed LINFO technique
7 1 1.5231 11.5 600 and the other four optimization algorithms is presented in
Tables 1, 2, and 3 when applied for three types of bench-
mark functions (unimodal, multimodal, and composite,
5 Simulation results respectively). The optimal values achieved by the LINFO,
INFO, BOA, TSA, and GWO techniques are displayed in
5.1 The performance of LINFO optimizer bold. It can be noted that the LINFO technique achieves the
optimum solution for the majority of the tested benchmark
The proposed LINFO technique’s capability and profi- functions. Figure 5 illustrates the convergence curves of all
ciency are tested on the numerous benchmark functions, by techniques for those functions. Whereas Fig. 6 depicts the
the statistical measurement including the best values, worst boxplots for these techniques. It is clear from Figs. 5 and 6
values, median values, mean values, and standard deviation that the proposed LINFO technique has achieved a
(STD) for optimal solutions reached by the proposed stable point for all functions and that it is boxplots are
LINFO algorithm and the other recent and well-known stable and very narrow for the majority of functions when
optimization algorithms. The results reached with applying compared to those of the other algorithms.
the proposed LINFO technique are compared to those

Fig. 12 Initial hourly load demand for test system 2

123
Neural Computing and Applications (2023) 35:20749–20770 20767

In this test system, the MG is comprised of three DE


units, one WT unit, one PV generator, and three residential
customers with DR. Table 4 includes the cost coefficients
of the DE units, three customers data, and the daily energy
interruptibility limits for each customer. The values of
power interabtabilty at each hour for customers are
assumed to be the same as listed in Table 5. Figure 7 shows
the hourly output power of WT and PV generators [43]. It
is assumed that the power transacted cost between the MG
and the main grid has symmetrical values; MG operator’s
daily budget is $500 for this test system.
All simulations were conducted using MATLAB 2021b
on a 2.9-GHz i7 PC with 8-GB of RAM. To address the
EM problem and integrate the DRP for the purpose of cost
minimization and benefit maximization, 20 separate runs
were carried out using PSO [52], JAYA [53], HBA [54],
Fig. 13 Hourly solar and wind power for test system 2 original INFO [47] and the proposed INFO algorithms. The
results obtained from these runs were compared, and the
Table 10 Power interabtabilty
(ki,t) values for test system 2
Time (h) ki,t ($) findings are presented in Table 6. Notably, it is evident that
the LINFO technique outperforms the other techniques in
t=1 27.61
terms of cost minimization.
t=2 29.41
The convergence characteristics curve of the objective
t=3 28.24
function is shown is Fig. 8, which demonstrate the effec-
t=4 26.69
tiveness and robustness of LINFO algorithm.
t=5 29.01
Using LINFO, the hourly generated power from the
t=6 33.96
three DEs and the power transacted with the grid are shown
t=7 83.97
in Fig. 9; it can be noted that during the period of renew-
t=8 81.10
able sources generation, the power is mainly sold to the
t=9 110.60 main grid from the MG with total power transaction of -
t = 10 74.12 18.69 kW (the power sold to main grid is higher than that
t = 11 78.95 bought from it). The initial load and load with DR are
t = 12 66.85 shown in Fig. 10; the power curtailed from all customers is
t = 13 47.98 shown Fig. 11, as well as the incentive they get with total
t = 14 66.82 power curtailment of 102 KW. Total power curtailment for
t = 15 48.50 each customer during the day is shown Table 7.
t = 16 49.21
t = 17 66.65
• MG test system 2
t = 18 61.49 For scalability validation of the proposed algorithm, a
t = 19 56.19 second test system with larger MG is simulated; this test
t = 20 57.92 system the MG is comprising of 10 DE units, ten WT
t = 21 49.16 generators, ten PV generator and seven customers with DR.
t = 22 54.00 Tables 8, and 9 list the cost coefficients of the DE units and
t = 23 34.37 seven customers’ data and the daily energy interruptibility
t = 24 30.30 limits for each customer, respectively. The initial load for
this test system is shown in Fig. 12; Fig. 13 shows the
hourly values of WT and PV generators [43]. The values of
5.2 Real-world application power interabtabilty are shown in Table 10.
For this test system, Fig. 14 shows the hourly generated
To validate the effectiveness of the proposed approach in power from the ten DEs and the power transacted with the
solving EM problem, 2 MG test systems were simulated. utility grid Fig. 14; the initial load and load with DR are
The grid- connected MG system is as illustrated in Fig. 2, shown in Fig. 15, as it is shown the DRP help in decreasing
• MG test system 1 the load demand mainly in the period where the renewable

123
20768 Neural Computing and Applications (2023) 35:20749–20770

Fig. 14 Hourly DE units


generated power and grid power
for test system 2

6 Conclusion

In this paper, a LINFO algorithm has been proposed in


order to solve the EM problems in MGs with DRP. The
superiority of LINFO over the standard INFO algorithm, as
well as other optimization techniques (e.g., BOA, TSA, and
GWO) in terms of convergence curve, average fitness
history, and search history is demonstrated using different
bench functions. Also, a statistical analysis of the proposed
LINFO technique and four optimization algorithms are
compared for three unimodal, multimodal, and composite
benchmarks, which demonstrate that LINFO has achieved
the optimal solution for most of test functions. The pro-
Fig. 15 Load demand and final load with DR test system 2 posed LINFO is used to operate the MG with minimum
DEs generation cost and minimum cost of power transac-
Table 11 Total amount of energy saved and the total amount of
tion. Different optimization approaches were used to solve
incentive received (Test system 2) the energy management problem in MG over a period of
24 h in order to get optimal operation of MG sources. A
j Saving (kWh) Incentive ($)
comparison among the convergence characteristics of
1 179.9986 8273.561 proposed LINFO algorithm and the original INFO algo-
2 229.8189 10,836.18 rithm prove the adequacy of the proposed LINFO in
3 309.2934 17,421.04 solving the EM problem with fast convergence. LINFO
4 389.5757 19,529.12 results are compared with well-known and new developed
5 439.0237 20,331.28 optimization techniques as PSO, JAYA, HBA, and INFO;
6 529.9821 27,380.45 the results demonstrated the robustness of the proposed
7 598.8072 28,910.92 algorithm to solve the EM problem in MG. to test system
Total 2676.5 132,682.6 were simulated; in test system 101 KWh reduction in
energy consumption, and 2677 MWh in the second test
system. The results of two test systems for the EM problem
under consideration have been described.
generation is low, Total power curtailment for each cus- In addition, the proposed LINFO algorithm can be used
tomer during the day is shown Table. 11. to solve an probabilistic EM problem. Ref. [13] proposed a
cloud based infrastructure for handling the data of smart
buildings in Nano- Microgrid. The dependent on the main
grid is reduced in the peak hours. The proposed LINFO
algorithm with the demand response, which is an ancillary

123
Neural Computing and Applications (2023) 35:20749–20770 20769

service in the smart grid, can be employed to solve the EM 5. Nchofoung TN, Fotio HK, Miamo CW (2023) Green taxation and
in the cloud based nano-grid. renewable energy technologies adoption: a global evidence.
Renew Energy Focus 44:334–343
6. Ansari S, Ayob A, Hossain Lipu MS, Hussain A, Saad MHM
Acknowledgements The icons used through this paper was developed
(2022) Remaining useful life prediction for lithium-ion battery
by Freepik, AmethystDesign, Arkinasi, and Smashicons from www.
storage system: a comprehensive review of methods, key factors,
flaticon.com.
issues and future outlook. Energy Rep 8:12153–12185
7. Zhang J, Jiang Y, Li X, Huo M, Luo H, Yin S (2022) An adaptive
Author contributions NA contributed to conceptualization, method-
remaining useful life prediction approach for single battery with
ology, software; SK contributed to conceptualization, methodology,
unlabeled small sample data and parameter uncertainty. Reliab
software; MHH contributed to conceptualization, software, writing—
Eng Syst Saf 222:108357
original draft preparation; SMA contributed to visualization, software,
8. Shivam, Dahiya R (2018) Stability analysis of islanded DC
writing—original draft preparation.
microgrid for the proposed distributed control strategy with
constant power loads. Comput Electr Eng 70:151–162
Funding Open access funding provided by The Science, Technology
9. Hirsch A, Parag Y, Guerrero J (2018) Microgrids: a review of
& Innovation Funding Authority (STDF) in cooperation with The
technologies, key drivers, and outstanding issues. Renew Sustain
Egyptian Knowledge Bank (EKB). Open access funding provided by
Energy Rev 90:402–411
The Science, Technology & Innovation Funding Authority (STDF) in
10. Phani Raghav L, Seshu Kumar R, Koteswara Raju D, Singh AR
cooperation with The Egyptian Knowledge Bank (EKB).
(2022) Analytic hierarchy process (AHP)—swarm intelligence
based flexible demand response management of grid-connected
Data availability Data sharing are not applicable to this article as no
microgrid. Appl Energy 306:118058
datasets were generated or analyses during the current study.
11. Li Y, Zhao T, Wang P, Gooi HB, Wu L, Liu Y et al (2018)
Optimal operation of multimicrogrids via cooperative energy and
Declarations reserve scheduling. IEEE Trans Ind Inf 14:3459–3468
12. Masera M, Bompard EF, Profumo F, Hadjsaid N (2018) Smart
Conflict of interest The authors declare that there is no conflict of (electricity) grids for smart cities: assessing roles and societal
interest regarding the publication of this manuscript. impacts. Proc IEEE 106:613–625
13. Kumar N, Vasilakos AV, Rodrigues JJPC (2017) A multi-tenant
Ethical approval This article does not contain any studies with human cloud-based DC nano grid for self-sustained smart buildings in
participants or animals performed by any of the authors. smart cities. IEEE Commun Mag 55:14–21
14. Wang Z, Chen B, Wang J, Kim J (2016) Decentralized energy
Informed consent Not applicable. management system for networked microgrids in grid-connected
and islanded modes. IEEE Trans Smart Grid 7:1097–1105
Open Access This article is licensed under a Creative Commons 15. Nikmehr N, Najafi-Ravadanegh S (2015) Optimal operation of
Attribution 4.0 International License, which permits use, sharing, distributed generations in micro-grids under uncertainties in load
adaptation, distribution and reproduction in any medium or format, as and renewable power generation using heuristic algorithm. IET
long as you give appropriate credit to the original author(s) and the Renew Power Gener 9:982–990
source, provide a link to the Creative Commons licence, and indicate 16. Chen T, Cao Y, Qing X, Zhang J, Sun Y, Amaratunga GAJ
if changes were made. The images or other third party material in this (2022) Multi-energy microgrid robust energy management with a
article are included in the article’s Creative Commons licence, unless novel decision-making strategy. Energy 239:121840
indicated otherwise in a credit line to the material. If material is not 17. Bukar AL, Tan CW, Said DM, Dobi AM, Ayop R, Alsharif A
included in the article’s Creative Commons licence and your intended (2022) Energy management strategy and capacity planning of an
use is not permitted by statutory regulation or exceeds the permitted autonomous microgrid: Performance comparison of metaheuristic
use, you will need to obtain permission directly from the copyright optimization searching techniques. Renew Energy Focus
holder. To view a copy of this licence, visit http://creativecommons. 40:48–66
org/licenses/by/4.0/. 18. Tostado-Véliz M, Kamel S, Hasanien HM, Turky RA, Jurado F
(2022) Uncertainty-aware day-ahead scheduling of microgrids
considering response fatigue: An IGDT approach. Appl Energy
310:118611
References 19. Harsh P, Das D (2021) Energy management in microgrid using
incentive-based demand response and reconfigured network
1. Malekpour AR, Pahwa A (2017) Stochastic networked microgrid considering uncertainties in renewable energy sources. Sustain
energy management with correlated wind generators. IEEE Trans Energy Technol Assess 46:101225
Power Syst 32:3681–3693 20. Palensky P, Dietrich D (2011) Demand side management:
2. Alamir N, Ismeil MA, Orabi M (2017) New MPPT technique demand response, intelligent energy systems, and smart loads.
using phase-shift modulation for LLC resonant micro-inverter. IEEE Trans Ind Inf 7:381–388
In: 2017 nineteenth international Middle East power systems 21. Jordehi AR (2019) Optimisation of demand response in electric
conference (MEPCON), pp 1465–1470 power systems, a review. Renew Sustain Energy Rev
3. Steffen B (2020) Estimating the cost of capital for renewable 103:308–319
energy projects. Energy Econ 88:104783 22. Yuan C, Illindala MS, Khalsa AS (2017) Co-optimization
4. Khasanov M, Kamel S, Rahmann C, Hasanien HM, Al-Durra A scheme for distributed energy resource planning in community
(2021) Optimal distributed generation and battery energy storage microgrids. IEEE Trans Sustain Energy 8:1351–1360
units integration in distribution systems considering power gen- 23. Marzband M, Yousefnejad E, Sumper A, Domı́nguez-Garcı́a JL
eration uncertainty. IET Gener Transm Distrib 15:3400–3422 (2016) Real time experimental implementation of optimum
energy management system in standalone microgrid by using

123
20770 Neural Computing and Applications (2023) 35:20749–20770

multi-layer ant colony optimization. Int J Electr Power Energy 39. Mahboubi-Moghaddam E, Nayeripour M, Aghaei J, Khodaei A,
Syst 75:265–274 Waffenschmidt E (2018) Interactive robust model for energy
24. Mohammadi M, Hosseinian SH, Gharehpetian GB (2012) Opti- service providers integrating demand response programs in
mization of hybrid solar energy sources/wind turbine systems wholesale markets. IEEE Trans Smart Grid 9:2681–2690
integrated to utility grids as microgrid (MG) under pool/bilateral/ 40. Lu Q, Zeng W, Guo Q, Lü S (2022) Optimal operation scheduling
hybrid electricity market using PSO. Sol Energy 86:112–125 of household energy hub: a multi-objective optimization model
25. Wu K, Zhou H (2014) A multi-agent-based energy-coordination considering integrated demand response. Energy Rep
control system for grid-connected large-scale wind–photovoltaic 8:15173–15188
energy storage power-generation units. Sol Energy 107:245–259 41. Wolpert DH, Macready WG (1997) No free lunch theorems for
26. Yang X, Long J, Liu P, Zhang X, Liu X (2018) Optimal optimization. IEEE Trans Evol Comput 1:67–82
scheduling of microgrid with distributed power based on water 42. Kim H-J, Kim M-K (2019) Multi-objective based optimal energy
cycle algorithm. Energies 11:2381 management of grid-connected microgrid considering advanced
27. Suresh V, Janik P, Jasinski M, Guerrero JM, Leonowicz Z (2023) demand response. Energies 12:4142
Microgrid energy management using metaheuristic optimization 43. Nwulu NI, Xia X (2017) Optimal dispatch for a microgrid
algorithms. Appl Soft Comput 134:109981 incorporating renewables and demand response. Renew Energy
28. Aghaei J, Alizadeh M-I (2013) Multi-objective self-scheduling of 101:16–28
CHP (combined heat and power)-based microgrids considering 44. Tazvinga H, Xia X, Zhang J (2013) Minimum cost solution of
demand response programs and ESSs (energy storage systems). photovoltaic–diesel–battery hybrid power systems for remote
Energy 55:1044–1054 consumers. Sol Energy 96:292–299
29. Chen J, Zhang W, Li J, Zhang W, Liu Y, Zhao B et al (2018) 45. Tazvinga H, Zhu B, Xia X (2014) Energy dispatch strategy for a
Optimal sizing for grid-tied microgrids with consideration of joint photovoltaic–wind–diesel–battery hybrid power system. Sol
optimization of planning and operation. IEEE Trans Sustain Energy 108:412–420
Energy 9:237–248 46. Fahrioglu M, Alvarado FL (2000) Designing incentive compati-
30. Nguyen A-D, Bui V-H, Hussain A, Nguyen D-H, Kim H-M ble contracts for effective demand management. IEEE Trans
(2018) Impact of demand response programs on optimal opera- Power Syst 15:1255–1260
tion of multi-microgrid system. Energies 11:1452 47. Ahmadianfar I, Heidari AA, Noshadian S, Chen H, Gandomi AH
31. Khalili T, Nojavan S, Zare K (2019) Optimal performance of (2022) INFO: An efficient optimization algorithm based on
microgrid in the presence of demand response exchange: a weighted mean of vectors. Expert Syst Appl 195:116516
stochastic multi-objective model. Comput Electr Eng 74:429–450 48. Naik MK, Panda R, Wunnava A, Jena B, Abraham A (2021) A
32. Alamir N, Kamel S, Megahed TF, Hori M, Abdelkader SM leader Harris Hawks optimization for 2-D Masi entropy-based
(2022) Energy management of microgrid considering demand multilevel image thresholding. Multimed Tools Applications
response using honey badger optimizer. Renew Energy Power 80:35543–35583
Qual J 20:12–17 49. Mirjalili S, Mirjalili SM, Lewis A (2014) Grey Wolf Optimizer.
33. Alamir N, Kamel S, Megahed TF, Hori M, Abdelkader SM Adv Eng Softw 69:46–61
(2022) Developing an artificial hummingbird algorithm for 50. Kaur S, Awasthi LK, Sangal AL, Dhiman G (2020) Tunicate
probabilistic energy management of microgrids considering Swarm Algorithm: a new bio-inspired based metaheuristic para-
demand response. Front Energy Res 10:905788 digm for global optimization. Eng Appl Artif Intell 90:103541
34. Soroudi A, Siano P, Keane A (2016) Optimal DR and ESS 51. Arora S, Singh S (2019) Butterfly optimization algorithm: a novel
scheduling for distribution losses payments minimization under approach for global optimization. Soft Comput 23:715–734
electricity price uncertainty. IEEE Trans Smart Grid 7:261–272 52. Moghaddam AA, Seifi A, Niknam T, Alizadeh Pahlavani MR
35. Shehzad Hassan MA, Chen M, Lin H, Ahmed MH, Khan MZ, (2011) Multi-objective operation management of a renewable
Chughtai GR (2019) Optimization modeling for dynamic price MG (micro-grid) with back-up micro-turbine/fuel cell/battery
based demand response in microgrids. J Clean Prod 222:231–241 hybrid power source. Energy 36:6490–6507
36. Gamil MM, Senjyu T, Takahashi H, Hemeida AM, Krishna N, 53. Warid W, Hizam H, Mariun N, Abdul-Wahab NI (2016) Optimal
Lotfy ME (2021) Optimal multi-objective sizing of a residential power flow using the Jaya algorithm. Energies 9:678
microgrid in Egypt with different ToU demand response per- 54. Hashim FA, Houssein EH, Hussain K, Mabrouk MS, Al-Atabany
centages. Sustain Cities Soc 75:103293 W (2021) Honey Badger Algorithm: new metaheuristic algorithm
37. Sasaki Y, Ueoka M, Uesugi Y, Yorino N, Zoka Y, Bedawy A for solving optimization problems. Math Comput Simul
et al (2022) A robust economic load dispatch in community 192:84–110
microgrid considering incentive-based demand response. IFAC-
PapersOnLine 55:389–394 Publisher’s Note Springer Nature remains neutral with regard to
38. Alamir N, Kamel S, Megahed TF, Hori M, Abdelkader SM jurisdictional claims in published maps and institutional affiliations.
(2023) Developing hybrid demand response technique for energy
management in microgrid based on pelican optimization algo-
rithm. Electr Power Syst Res 214:108905

123

You might also like