Electric Power Systems Research 221 (2023) 109370

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Optimal energy management strategy for a renewable based microgrid with

electric vehicles and demand response program
Tao Hai a, b, c, Jincheng Zhou a, d, e, **, Alireza Rezvani f, g, *, Binh Nguyen Le f, g, Hitoshi Oikawa h
a
School of Computer and Information, Qiannan Normal University for Nationalities, Duyun, Guizhou 558000, China
b
Key Laboratory of Advanced Manufacturing Technology of the Ministry of Education, Guizhou University, Guizhou 550025, China
c
State Key Laboratory of Public Big Data, Guizhou University, Guiyang, Guizhou 550025, China
d
Key Laboratory of Complex Systems and Intelligent Optimization of Guizhou Province, Duyun 558000, China
e
Key Laboratory of Complex Systems and Intelligent Optimization of Qiannan, Duyun 558000, China
f
Institute of Research and Development, Duy Tan University, Da Nang, Viet Nam
g
School of Engineering and Technology, Duy Tan University, Da Nang, Viet Nam
h
Solar Energy and Power Electronic Co, Ltd, Tokyo, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: This paper suggests an optimal management for a microgrid containing renewable sources and electrical vehicles
Electrical vehicle (EVs) with responsive loads for minimizing operation costs and emissions. The role of EVs is coping with peak
ISFLA algorithm load conditions while responsive loads can deal with uncertainties of renewable-based sources. A two-stage
Microgrid
model is implemented for clearing the energy and reserve markets. The costs of generation and reserve power
Optimal management
Renewable energy
are minimized first, and then the costs of variations in the management owing to the changes in wind turbine and
photovoltaic (PV) are optimized. To optimize the objective function, an improved shuffled frog leaping algorithm
(ISFLA) is used. Simulation results are provided for different case studies in a microgrid and the outcomes
confirm the superiority of the recommended algorithm in comparison with other conventional approaches. The
use of EVs and responsive loads provides advantages such as minimizing the operational cost, emissions and the
changing behavior of the PV and wind turbines.

reservation in microgrids is necessary. To solve the mentioned chal­

1. Introduction lenges, it is essential for microgrids to use different controllable re­
sources and various storage systems that the optimal operation with the
Reducing the operating costs of distribution networks, including aim of minimizing costs in these systems is a critical challenge. In
distributed generations (DGs) and renewable sources is one of the most addition to minimizing costs, achieving the minimum amount of envi­
attractive topics in recent years. The use of microgrid is one of the ronmental pollution is also considered as one of the important issues in
effective solutions to decrease the operating costs in distribution net­ microgrids [3,4]. Therefore, it is necessary to use new technologies to
works, with the help of which the challenge of excessive load growth minimize costs and environmental pollution at the same time in
and the problem of environmental pollution can be solved [1]. Micro­ microgrids [5,6].
grids have the ability to operate in two modes, connected to the distri­ The emergence of electric vehicles (EVs), both PEVs and plug-in
bution network and in island mode; however, in normal conditions, the hybrid electric vehicles (PHEVs) can reduce environmental pollution
microgrid is worked in the grid-tied mode. The advantages of microgrids produced by old-style transportation systems, and their use in micro­
include dropping environmental pollution caused by traditional power grids can solve the challenge of providing electricity in peak load con­
production, technical benefits such as decreasing losses and enhancing ditions. The use of EVs causes fewer fossil resources to be used and these
the voltage profile, as well as economic benefits due to the use of resources are preserved as a reserve for critical situations. These ad­
renewable energy resources (RES) with minimal operating costs [2]. vantages have caused the growth of EVs around the world. The use of
However, one of the challenges of using RES such as photovoltaic (PV) EVs in microgrids for exchanging active and reactive power between EVs
wind turbine (WT) is their random behavior, and therefore the need for and microgrids, introduced them as effective tools for the power system

* Corresponding author at: Institute of Research and Development, Duy Tan University, Da Nang, Vietnam.
** Co-corresponding author.
E-mail addresses: guideaaa@126.com (J. Zhou), alirezarezvani@duytan.edu.vn (A. Rezvani).

https://doi.org/10.1016/j.epsr.2023.109370
Received 1 December 2022; Received in revised form 21 February 2023; Accepted 26 March 2023
Available online 15 April 2023
0378-7796/© 2023 Elsevier B.V. All rights reserved.
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Nomenclature Ptw wind turbine power generation at time t

Ctw energy cost suggested by wind power unit
ECt expected cost at time t Prmn mth scenario of probability of renewable sources
t time Ctgrid the suggested price of utility
Ctsi start-up cost t
rGi reserve supplied by the jth aggregator
Ctsimn start-up cost
mn
NL number of load
T total hours for scheduling t
rEV mn the suggested reserve by unit j
Ng the number of distributed generations Ptwsmn output power spillage of WT in the m,n scenario
Uit ON/OFF status of the generation units Ctws WT cost in the m,n scenario
PtGi the real power produced by the units PtPVSmn output power of PV in the m,n scenario
CtGi energy price suggested by the unit i
CtPVS PV power spillage cost
RtGi.u lower bound of spinning reserve of unit i
PGi,min (t) minimum power of DGs
RtGi.d the hourly spinning reserve down supplied by the ith unit
PGrid,min (t) minimum power of main-grid
λtGi.u the hourly spinning reserve up cost supplied by the ith unit PGi,max (t) maximum power of DGs
λtGd.u the hourly spinning reserve down cost supplied by the ith PGrid,max (t) maximum power of main-grid
unit PtPV output of PV
NEV Total EVs
CtPV hourly cost of solar power generation
PtEVj real power of EV j
Ptgrid hourly power transaction with the electrical grid
CtEVj the energy cost presented by aggregator j

[7]. The use of Vehicle-to-Grid (V2G) technology can improve the power smart charging strategy is suggested in [19]. The optimal management
quality of the microgrid through services such as reactive power, har­ of a microgrid with the aim of minimizing the cost considering PEV
monics and unbalanced compensation [8]. aggregators is investigated in [20]. In another study in [21], the optimal
scheduling of microgrid with aim of minimizing the operating costs with
1.1. Literature review DERs and V2G is recommended. For coping with the discharge problem
of PEV, a new method is proposed in [22], considering the different
The operation of microgrids is one of the topics that has received circumstances of each individual vehicle while attaining the maximum
numerous considerations in recent years. This issue has been researched profit. To maximize the profit of a microgrid including multiple PHEV
so far by various objective functions, different optimization methods, battery exchange stations, a model based on deterministic integer pro­
and different DERs. Optimal scheduling of various DERs in microgrids gramming is recommended in [23]. An optimal management for V2G
with different objective functions is a very complex optimization prob­ energy and ancillary services is proposed in [24], to maximize the profit
lem that requires great accuracy. Most of the previous research are of the microgrid with low EV charging costs.
published to reduce the costs or maximize the profit of the microgrid. Moreover, in some papers minimizing the power load deviation is
The bi-level strategy is used for the optimal energy and spinning investigated. In order to optimize the power demand curve in a micro­
reserve scheduling in microgrid in [9]. In Ref [10], the optimal sched­ grid, an effective management procedure for EVs charging and dis­
uling of microgrids with several DERs is suggested and demand response charging is recommended in [25]. Two-stage optimal management for
(DR) programs are applied to minimize the operation costs. Unit charging structure in EVs is proposed in [26]. The main aim of the paper
Commitment (UC) is addressed for a microgrid with several DERs in in the first level is to minimize operational costs. The power load devi­
[11]. All the above-mentioned works are based on single-objective and ation is also minimized in the second stage. In [27], optimal scheduling
the emission is not considered in their works. Consequently, it is for V2G operation is investigated with aim of minimizing the overall
required to consider emission minimization in the optimal scheduling of load variance considering uncertainties.
microgrid at the same time when minimizing the cost to cope with the According to Kabatepe and Türkay [28], the single objective models
difficulties related to greenhouse gas emissions [12]. are not very effective for evaluating the economic and environmental
For addressing the uncertainty of renewable sources, the Weibull and effects of EVs. The paper has investigated the effect of EV charging and
Beta PDFs are applied in [13]. The differential evolution algorithm is market penetration scenarios to improve the performance of EVs. The
also used for the nonlinear optimization approach. In [14], chaos clonal use of DR is another effective method to cope with peak load conditions
evolutionary programming is applied to address the optimization and reduce operational costs. It is done by reduction of load consump­
problem. The optimal scheduling of s distribution systems including tion in peak loads conditions [29]. In Refs. [30,31], an analytical review
wind turbines (WT) is investigated [15]. The confidence interval is written on DR programs. This technology can be also considered as
concept is utilized to model the uncertainties of the WT. The aim is power plant reserves to provide the reserve power for the power system.
minimizing the operating cost and emission using the Modified Bacterial
Foraging Optimization algorithm. 1.2. Contribution and organization
A two-stage stochastic approach by the roulette wheel mechanism
for uncertainty modeling of loads, output power of DERs and market The optimal performance of a MG with EVs and DR is recommended
price in a microgrid is proposed [16]. The Adaptive Modified Firefly in this study. The EVs are used to deal with peak load conditions while
Algorithm (AMFA) is used for the optimization of the problem formu­ the DR program is used as a reserve to deal with WT and PV un­
lation. A scenario-based approach is used for modeling uncertainties in certainties. The proposed model is very effective to minimize the cost
[17]. In [18], the uncertainties are addressed using a Point Estimate and emissions. A two-stage method is used, so that in the first stage, the
Method (PEM). The optimal scheduling is achieved using the AMFA. amount of power generation and reserve power costs are determined.
The EVs in the energy management of microgrid is also an interesting Then, optimal management of generation units is done because of var­
topic that needs more investigation. For controlling the charging de­ iations in WT and PV systems. The major contributions of this paper
mand of PEVs in a microgrid consisting of several DERs and storages, a include:

2
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 1. Proposed pricing scheme of DR program.

1 Optimal scheduling of EVs and responsive loads for peak shaving and that they are plugged into the charging point. Normally, PHEVs have to
coping with the uncertainties of renewable-based power units. travel one in the morning when leaving the home and one in the evening
2 A two-stage modeling is suggested to propose the expected operation when returning home, but the short travels are omitted for optimal
cost of MG considering uncertainties. scheduling. When PHEVs arrive at home, they start charging at 6:00 PM,
therefore, the charging time can be mathematically presented based on
The rest of this paper is: the modeling of microgrid components is the probability density function (PDF) around 6:00 p.m. Then, they can
given in Section 2. the objective function is given Section 3. The algo­ act as storage and discharge their stored power to the grid to cope with
rithm is introduced in Section 4. Simulations are presented in Section 5. peak load conditions and minimize operation costs. The energy and
Conclusions are given in Section 6. reserve prices are assumed to be $0.83 and $0.07, respectively, in this
study.
2. MG modeling
2.2.1. Modeling and constraints of EVs
The mathematical modeling of the microgrids elements are given in The stored energy at the end of period t can be calculated through
this part. Eq. (3). The daily travel profile of the EV is taken into account in vari­
ableETrip (V, t). Moreover, it is needed to use a satisfactory forecast
2.1. DRP modeling approach to predict the daily travel profile. A satisfactory communica­
tion system between all players is also required for having the infor­
The optimal operation of microgrid with the aim of minimizing costs mation of daily travel profile [31]. The data of charge/discharge in
is a very important issue and it is needed to use new technology such as period t and the energy remaining from the earlier period are also
DR programs. The microgrid operator is responsible for dispatching the needed.
resources in the microgrid in order to schedule the next day with the aim Estored(V,t) = Estored(V,t− 1) − ETrip(V,t) + Δt
of reducing costs while predicting PV and WT. One of the basic chal­
1
lenges in microgrids is to address peak load conditions. In this regard, ×ηc(V) × Pch(V,t) − × PDch(V,t) ∀t ∈ {1, ..., T}; ∀V ∈ {1, ..., Nv }; Δt = 1;
ηd(V)
the use of consumption management programs can be suggested as a
good solution to overcome difficult peak load conditions. In the pro­ t = 1→Estored(V,t− 1) = EInitial(V)
posed method, to carry out the DR program, it is supposed to the amount (1)
of load as well as the price of consumption is determined by an aggre­
gator which is responsible for implementing the DR program, and based where:
on this strategy, the DR policy for the next day is determined [12]. Fig. 1
shows the suggested DR program. In addition, msL and πsL in the Fig are Δt: Elementary period
load reduction and the price, respectively. The mathematical modeling ηc(V) : Efficiency in the charging process.
of DR can be given as: ηd(V) : Efficiency in the discharging process.
EInitial(V,t) : Energy available in the battery of the EV when starting the
2.2. Implementation of EVs day.
Estored(V,t) : Energy available in the battery of the EV at the end of the
PHEV is a high-tech hybrid EV with rechargeable batteries that can day.
be fully recharged with the help of the main grid. Also, the design of this ETrip(V,t) : The energy depleted on a trip for the EV at time slot t.
high-tech equipment is such that it can use fossil fuels for long distances
because the amount of battery charge used in these EVs is limited. It is clear that batteries must be charged based on their capacities as
Charger type, battery state of charge (SOC), battery capacity, number of given in Eq. (2). For grantee of the minimum amount of energy for each
PHEVs, charging type, charging time and length of charging are EV, Eq. (3) can be used. It can be considered a reserve energy for
important parameters for modeling PHEVs [26,27]. unpredicted travel in each period t, and to escape the fast degradation of
One of the basic challenges of PHEVs is the uncertainty in their EVs‫ ׳‬batteries [28].
charging demand, which makes the issue of their penetration in
Estored(V,t) ≤ EBatCap(V) ; ∀t ∈ {1, ..., T}; ∀V ∈ {1, ..., NV } (2)
microgrids very complicated. In this regard, three charging patterns are
taken into account for using them in microgrids including (1) uncoor­
Estored(V,t) ≥ EminCharg(V,t) ; ∀t ∈ {1, ..., T}; ∀V ∈ {1, ..., NV } (3)
dinated charging, (2) coordinated charging and (3) smart charging.
In the first pattern, PHEVs are connected to the station at any time

3
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

EBatCap(V) : Capacity of BES (W h) 3.1. Emission function

EminCharg(V,t) : Minimum stored energy (W h).
Some technical limitations can be used for EVs discharge process as In this part, the emission function is defined based on two stages
given in Eqs. (4), (5) [28]. because of uncertainties of WT and PV units. At the first stage, the
pollution of units is determined. After that, the pollution regarding the
PDch(v,t) ≤ PDchLimit(V,t) × XCh(V,t) ; ∀t ∈ {1, ..., T}; ∀V ∈ {1, ..., NV }; Xch(V,t)
changes of arrangement of the units initiated by variations in the per­
∈ 0, 1 formances of WT and PV powers is determined.
(4)
∑
T
C(Pi ) = EEt
1 t=1
× PDch(v,t) × Δt ≤ EStpred(V,t− 1) − Etrip(V,t) ; ∀t ∈ {1, ..., T}; ∀V { N }
ηd(v) ∑
T ∑ g T ∑
∑ Nm

∈ {1, ..., NV }; (5) = Uit PtGi EGi + Ptgrid Egrid

t
+
t=1 i=1 t=1 m=1
{ }
where ∑
Nn Ng
∑
× Prmn t
rGi E + Ptgridmn Egrid
t
(9)
PDchLimit(V,t) : Maximum active power discharge. Eqs. (6) and (7) are n=1 i=1
mn Gi

also limited by [28]:

Pch(v,t) ≤ PchLimit(V,t) × XDch(V,t) ; ∀t ∈ {1, ..., T}; ∀V ∈ {1, ..., NV }; XDch(V,t) 3.2. The proposed objective function

∈ 0, 1 Two mentioned objective functions can be defined as [31]:

(6)
∑
T ∑
T
MinOF = ω1 F(Pi )+ω2 Qr,i C(Pi ) = ω1 Costt + ω2 Qr,i Emissiont
ηd(v) × PCh(v,t) × Δt ≤ EBatcap(V) − EStored(V,t− 1) − Etrip(V,t) ∀t ∈ {1, ..., T}; ∀V t=1 t=1

∈ {1, ..., NV }; (10)

(7) It is noteworthy that OF indicates the objective function, Qr,i denotes
the ith unit’s price penalty factor [33] while the weighting factors of
where
objective functions are depicted byω1 and ω2 , respectively. The co­
PchLimit(V,t) : EVs Maximum real power charge in period t (W).
efficients are defined between ω1 =1,ω2 =0 and ω1 =0,ω2 =1. Moreover,
the weight of cost and emission are ω1 =ω2 =0.5.
2.3. Renewable energies modeling
3.3. Stage 1
For modeling the uncertainties of the WT and PV system, the PDF is
classified into various limit set [32]. Next, the scenario tree is created •Power balance
and employed as an input parameter for stochastic modeling. The pre­
dicted power, and the upper and lower limit are chosen based on the Ng
∑
PtGi + Ptw + PtPV ± PtEV ± Ptgrid = PtL (11)
probabilities of 50%.
i=1

3. Cost-function where PL (t) is the load.

•Power generation bounds
The objective function of this paper is given as a linear cost function
to minimize the operation cost by F(Pi )according to Eq. (8). This PtGi,min uit ≤ PtGi ≤ PGi,max uit (12)
objective can be attained by the appropriate arrangement of DERs in the
microgrid. The EVs play a important role in minimizing the cost of PGrid,min uit ≤ PtGi ≤ PGrid,max uit (13)
microgrid through proper charging and discharging patterns. The
operator of the microgrid must pay for EVs when discharging. On the Ptw,LB ≤ Ptw ≤ Ptw,UB (14)
other hand, EVs should pay when charging. For implementing the V2G
strategy, the remuneration tariffs need to be used for the EVs batteries PtPV,LB ≤ PtPV ≤ PtPV,UB (15)
injecting energy to the network, while considering batteries’ degrada­
tion cost [32]. where Ptw,LB and Ptw,UB are bounds of WT at time t, correspondingly.
{ N PtPV,LB and PtPV,UB signify bounds of PV at time t. uit is a binary variable.
Ng Ng
∑ ∑ ∑ ∑ ∑ •Spinning reserve limit:
T T g

F(Pi ) = ECt = Csit + Uit PtGi CGi

t
+ Uit RtGi.u λtGi.u
( ) ( )
t=1 t=1 i=1 i=1 i=1
0 ≤ RtGi.u ≤ PGi,max − PtGi uit ∀i, t0 ≤ RtGi.d ≤ PtGi − PGi,min uit ∀i, t (16)
Ng
∑ ∑ ∑
+ Uit RtGi.d λtGi.d + rCDRtL + ECDRtL • DR costs can be given as:
i=1 rd e
L∈SDRP
L∈Sru
DRP
orSDRP
} ∑
NSL
∑
NEV rCDRtL = ΔsLt π sLt ZsLt ∀t, L ∈ SDRP
ru rd
orSDRP (17)
+ PtEVj CEV
t
+ Ptw Cwt + PtPV CPV
t
+ Ptgrid Cgrid
t
s=1
(8)
j
j=1
{ Ng Ng ∑
NSL
∑
T ∑
Nm ∑
Nn ∑ ∑
+ Prmn Csit mn + t
rGi t
CGi ECDRtL = ΔsLt πesLt ZsLt ∀t, L ∈ SDRP
e
(18)
mn
t=1 m=1 n=1 i=1 i=1 s=1

∑ ∑
Nl ∑
NEV It is noted that the capacity and energy cost of the demand response L
± rEDRtLmn + VOLLLt shedLt mn + t
rEVjmn
t
CEV j at time slot t is denoted by πsLt and πesLt , respectively.
•Generating unit start-up cost
ru orSrd
L∈SDRP L=1 j=1
DRP
}
+Ptwsmn Cws
t
+ PtPVsmn CPVs
t
+ Ptgridmn Cgrid
t ( t )
Csi ≥ ηSU
t
it Ui − Uit− 1 ∀i, t (19)

4
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Csit ≥ 0∀i, t (20) CCsit mn ≥ 0∀i, t, k (36)

where ηSU
it is the cost of starting of unit i in period t. where CCtsimn in period t and scenario mn shows the actual start-up cost.

3.4. Stage 2 4. SFLA algorithm

• The balance of power can be defined as: One of the strong meta-heuristic algorithms is the SFLA, which is
Ng modeled based on the search of frogs in order to find food [34,35]. In
∑ ∑
NL
(
PtGimn + Ptwmn + PtPVmn ± PtEVmn ± PtGridmn + shedLt mn − LCLt mn ) =PL (t) this algorithm, frogs are members of the population of the algorithm,
i=1 L=1 and local and global methods are used to generate different populations
(21) and improve the performance of the objective function. In the structure
of the algorithm, old answers are replaced with new answers. This al­
• Production limits in scenarios:
gorithm is introduced in the following Sections.
PtGimn ≥ PGimin yti,mn ∀i, t, mn (22)

PtGimn ≤ PGimax yti,mn ∀i, t, mn (23) 4.1. Initial population

where yti,mn is a binary variable. First, the initial population of the algorithm with several members
(p) is randomly selected so that each member is part of the problem
• DR reserve:
search space. Moreover, the location of the member i is denoted through
∑
NSL Xi=(x1,i, x2,i,…, xs,i) [35].
drLt mn = ΔsLt νsLtk ∀t, k, L ∈ SDRP
ru rd
orSDRP (24)
s=1

4.2. Frogs distribution

∑
NSL
rEDRtLmn = ΔsLt π sLt νsLtk ∀t, k, L ∈ SDRP
ru
orSrd (25)
At this stage, the fitness value is used to rank the members so that the
DRP
s=1
members are sorted in descending order. Next, the considered popula­
The reserve used from the DRP L at time slot t and scenario mn is
tion is distributed within m memeplexes such that each memeplex
denoted by drLt mn while νsLtk indicates a binary variable assigned to point S
contains n frogs, i.e. p = m × n. The first member is called memeplex 1
of the DRP L in time slot t and scenario mn. and it carries on correspondingly to the end. Consequently, the m + 1
• •Load shedding and surplus renewable power frog is allocated to the first memeplex and it continues till the last frog
0 ≤ shed ≤ LCLt mn (26) [36].

⎛ ⎞
∑ ∑ 4.3. Local search
LCLt mn =⎝ DRtL ± drLt mn ⎠∀L ∈ SPBDR , t, k
∕ (27)
e
L∈SDRP ru orSrd
L∈SDRP
Based on the operation of the proposed algorithm, each frog with a
DRP

0 ≤ Ptwsmn ≤ Ptwmn (28) desired position is replaced by another member with an undesired po­
sition in the memeplex. The position of frogs is modified according to the
0 ≤ PtPVsmn ≤ PtPVmn (29) mean value of the fitness. The following steps are required for every
memeplex in the given iterations:
According to Eq. (27), for DR providing up reserve), negative sign
should be applied. 1- The best and the weakest locations of frogs needs to be identified and
denoted by Xb and X respectively
3.5. Constraint of two stage modeling 2- A modification is required at this stage for the location of the worst
frog, i.e. Xw, considering the best frog, i.e. Xb in the memeplex as
• Disintegration of units stated in Eqs.(37) and (38).

PtGimn = PtGi + rGi.u

t
mn
t
− rGi.d mn
∀i, t, k (30) D = rand × (Xb − Xw ) (37)

t
where rGi.u t
and rGi.d are the up and down reserve. Xwnew = Xwold + D (38)
mn mn

• Spinning reserve
where D is frogs’ mutation vector rand displays a random number [0
t
0 ≤ rGi.umn
≤ RtGi.u ∀i, t, k (31) 1].

t
0 ≤ rGi.dmn
≤ RtGi.d ∀i, t, k (32) If a dominant answer can be achieved, a novel member (frog) is
replaced with the scheduled member and the process continues to
•DR reserve
Step 5. On the other hand, step 3 is performed. Xwnew and Xwold display
0 ≤ drLt mn ≤ DRtL ∀i, t, k, L ∈ SDRP
ru rd
orSDRP (33) the locations of the most inferior frog at the earlier instant and the
new one.
As inequalities (33) and (34) indicate, the reserve in every scenario
should be less than or equal to the reserve scheduled in the first stage.
1- Xg is replaced by Xb in Eq. (37) and then, the new frog is gotten, using
• The start-up cost-adjustment of units:
Eq. (38). If better result is achieved, the frog is replaced with the old
Csit mn = CCsit mn − Csit ∀i, t, k (34) member and step 5 is performed. Otherwise, the process gates with
the fourth step.
)
CCsit mn ≥ λSU t t− 1
it (yimn − yimn ∀i, t, k (35) 2- Based on the random manner another frog is generated and replaced
with the worst member within the memeplex.

5
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig 2. ISLFA method.

3- For a given number of numbers, the first four stages can be per­ {
formed in the loop. The search process is continued till the termi­ new
Xwj =
pre − Xwj if rand < CRg or j = rm
(40)
nation standard is fulfilled.
old
Xwj otherwise

4.4. Improved SFLA where the generic constant is stated by Crg, rm is a random value
old
showing one of the Xwj parameters among pre − Xwj . If a better answer
Even though the SFLA has good performance with high convergence is obtained, the new answer replaces the previous one. On the other
speed, it is faced some limitations. For increasing search capability of hand, if the better answer is not obtained, Eqs. (39) and (40) are per­
this algorithm, the MSFLA is suggested in this study. The position of the formed by replacing the most superior outcome of every memeplex with
Xw within every memeplex is improved in every iteration. Xg, and CRb, which is the local crossing constant, with CRg. If a better
For undesired result, the vector of mutation is created based on (39): answer is not obtained, one of the frogs is randomly selected and re­
pre Xw = Xg + F × (Xr1 − Xr2 ) (39) places the previous frog. The operation process of the proposed algo­
rithm is similar to the original algorithm, and it is described in this
Xr1 and Xr2 are selected randomly from the current members. Xg is the section. At the beginning of the process, a frog position is created
best answer in the current iteration and F is a parameter for mutation randomly. The vector of Xj is the location of the frog j. By starting the
factor representing the discrepancy between Xr1 and Xr2. The value j of algorithm procedure, the position of the frogs should be determined so
the vector Xwnew within the subsequent iteration can be given as: that all the constraints of the problem are satisfied. The fitness of each
member is identified after determining the objective function for each

6
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Table 1
The information of DG sources.
Type Min power Max power Bid Reserve cost (up and down) Start-up cost CO2 (kg/MW NOx (kg/MW SO2 (kg/MW
(kW) (Kw) ($/kW) ($/kW) ($) h) h) h)

DSG 2 50 0.187 0.11 1.82 840 0.2 0.0071

MT 6 30 0.457 0.316 0.96 720 0.1 0.0036
FC 3 30 0.294 0.19 1.65 460 0.0075 0.003
WT 0 15 1.073 0 0 0 0 0
PV 0 30 2.548 0 0 0 0 0
EVa ( 0 1.5 0.83 0.07 0 0 0 0
×40)

5. Simulation results
Table 2
DRP and controllable loads.
The test system of this study is based on Ref [7]. The limitation of DG
Value reserve of responsive loads Reserve cost (πS Lt) Energy cost (π eS Lt) power generation, bid coefficients (€ct/kWh) and emission coefficients
33% of total load 2.5 20 of DGs (kg/MWh) are given in Table 1 [17–19]. Table 2 also gives in­
66% of total Load 3.5 28 formation of the DR program. It is considered that 10% of loads are
100% of total load 4.5 36 responsive and can participate in the DR program. Table 3 also shows
the data of WT and PV systems as well as load demand. The Matlab
member. Based on the fitness, the frogs are ranked is ranked. Then, for software is applied to confirm the effectiveness of the suggested
every memeplex, the local search is accomplished. During the genera­ framework. MG structure as shown Fig. 3(a). The renewable energies
tion of new situations, many of the constraints of the problem may be data as shown Fig. 3(b) and 3(c). The microgrid includes numerous DERs
out of bounds for unacceptable members during the local search process. such as MT, FC, PV, WT and battery storage systems (BSS). The optimal
The algorithm stops solving the problem when the iteration constraints management is done for day-ahead in a 24 h period hourly. CO2, SO2
are satisfied. The final solution is the position of the best selected frog for and NOx are considered to be minimized here. A microgrid central
solving the problem of optimal operation of the microgrid. Fig. 2 depicts controller (MGCC) is in charge of controlling the system and the
the procedure of the ISLFA. microgrid can export/import power with the utility grid. The total en­
ergy consumption of the microgrid in one day is 4034 kWh. Moreover,
20% of the highest marginal cost of energy generation is considered for
spinning reserve cost.

Table 3
The data for wind and PV units.
Time 1 2 3 4 5 6 7 8 9 10 11 12

Load demand 52 50 50 51 56 63 70 75 76 80 78 74
Time 13 14 15 16 17 18 19 20 21 22 23 24
Load demand 72 72 76 80 85 88 90 87 78 71 65 56

Fig 3. Data of renewable energies, 3(a) Wind power; 3(b) PV power.

7
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig 3. (continued).

The WT and PV systems are operated in their maximum possible needs to be provided. In compared with WT, the PV includes higher cost;
power. The data of load demand and the market prices during the study therefore, the PV unit is operated at predicted value to minimize the
for the microgrid are given in [17,18]. For investigating the effects of reserve power essential by the main grid. Same as WT, the PV is set at
EVs and DR programs on microgrid scheduling, three scenarios are maximum value in peak load conditions to achieve the low emotions.
considered for simulations: According to the results, the generation, startup, and reserve generator
Scenario 1: Microgrid scheduling without EVs and DR costs are $847.3967, $15.69, and $24.7746, correspondingly. Moreover,
Scenario 2: Microgrid scheduling considering only DR the emission is 1169.278 kg. The output power and reserve are pre­
Scenario 3: Microgrid scheduling considering both DR and EVs sented in Figs. 4 and 5.
The ISFLA is used for optimization and the outcomes are displayed in
5.1. Scenario 1: microgrid scheduling without EVs and DR Table 4. The ISFLA performance is compared with GA, PSO and ICA, in
terms of the best solution (BS), the average solution (AS), the worst
The microgrid load demand is supplied by MT, FC, diesel generator, solution (WS), and the standard deviation (SD). According to the results,
PV and WT. Because of the stochastic nature of WT and PV systems, MT, the ISFLA has an edge over other algorithms. The simulations shown
FC, diesel generator are responsible for proving the required reserve. that the total operating cost of the microgrid in this scenario is 845.3254
The output power of different sources are given in Figs. 4 and 5. Ac­ $. In addition, the WS of the ISFLA is noticeably lower than those gotten
cording to the results, diesel generator is operated here because of low by conventional algorithms. The average solution time of the ISFLA al­
cost. The WT is operating in maximum possible power. For minimizing gorithm with that of other algorithms to assess the computational effi­
the cost, it is required to use diesel generator in most hours of the day. cacy that shows of good performance of the suggested optimization
Nevertheless, the use of this conventional unit leads to high emissions. algorithm.
To deal with this problem, it is necessary to use emission free units such
as MT, FC, PV and WT. However, the use of MT and FC increases the cost 5.2. Scenario 2: microgrid operation considering only DR
of the microgrid. As the defined objective function is based on Eco­
nomic– Emission despatch, a trade-off between these two objectives By raising the power generation through WT, the requirement of the

8
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 4. The power output of units_Scenario 1.

Fig. 5. Reserve procurement by units_Scenario 1.

The use of responsive loads in this case study is very effective and can
Table 4 be taken into account as a grid reserve resource in the market by
Comparing of cost function. contributing to the DR program. Not that 10% of loads are responsive.
Method SD($) AS($) WS($) BS($) Mean This strategy provides ancillary services for the system and the use of PV
simulation and WT is reasonable to minimize emissions. The results of optimal
time scheduling in this scenario are given in Figs. 6 and 7. The use of DR and
GA 16.6365 883.6174 901.2851 861.3371 12.54 other controllable units can cope with the uncertainties of WT and PV. In
PSO 12.1344 876.517 897.3917 859.6317 10.91 other words, the DR and other generators provide a balance between
DE 9.4261 873.6617 890.3361 854.3691 8.65
generation and consumption when the WT and PV are in low-level
ICA 6.3789 870.3197 887.9641 851.9154 7.57
Suggested 3.6679 860.34 865.3378 845.3254 5.78 power generation. As seen, the DR program is accomplished at a
method maximum rate at 10:00, 11:00, and 14:00 h during the day. On the other
hand, it is reduced to 66% at 13:00 h and to 33% in the other hours. The
DR program provides the possibility of cheaper reserves than generators
grid for spinning and nonspinning reserves will be amplified. Therefore, and resulting in using PV and WT to minimize the emission. According
the operational cost is increased. For minimizing the cost, the best so­ to simulation results, the generation, reserve, and startup costs are
lution is reducing the cost of reserve that is considered for addressing $820.3617, $18.0522, and $14.52. In comparison with the previous
uncertainties.

9
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 6. Energy management of scenario 2.

Fig. 7. Reserve management of scenario 2.

scenario, these values are considerably reduced. Moreover, the emission other algorithms.
is 1090.658 kg, which is reduced by 25.5406 kg in the previous scenario.
The output power and reserve are shown in Figs. 6 and 7.
5.3. Scenario 3: microgrid operation considering both DR and EVs
Fig. 8 indicates that the controllable loads would procure a sub­
stantial amount of the reserve required which in turn mitigate the sys­
In this part, both DR and EVs are taken into account. The use of these
tem’s total operating cost. Fig. 9 gives a detailed comparison in terms of
technologies can enhance the efficacy of the microgrid. The role of the
spinning reserve deployment of units where it can be concluded that the
DR is to provide the reserve while the role of EVs is like as controllable
controllable loads could successfully alleviate the dependence on the
loads. The results of this part based on the defined Economic–Emission
reserve procured by the units. Moreover, a comparison is made in Fig. 10
despatch is shown in Figs. 11 and 12. As given in Fig. 12 and he hourly
between the first two scenarios in terms of the reserve cost.
power and reserve despatch have been depicted in Figs. 11 and 12,
The ISFLA is used for optimization and the results are shown in
respectively.
Table 5. The ISFLA performance is compared with GA, PSO, and ICA.
EVs can be used in peak load conditions through discharging instead
According to the results, the ISFLA has an edge over other algorithms. In
of MT to minimize operational costs. According to the load pattern, peak
addition, the WS of the ISFLA is noticeably lower than those gotten by
hours are from 9:00 to 12:00 and from 18:00 to 20:00. On the other
conventional algorithms. The mean simulation time show that the ISFLA
hand, the EVs start to charge in light load conditions during the evening
method is good performance than other algorithms. The mean simula­
hours to be used for peak load conditions. By using this strategy, the
tion time represents that the ISFLA algorithm is good performance than
requirement for switching off the base load generators is omitted and the

10
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 8. Shares of loads and units in the reserve procurement.

Fig. 9. Performance of Scenarios 1 and 2 in terms of reserve amounts.

Fig. 10. Performance of Scenarios 1 and 2 in terms of reserve costs.

11
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Table 5 more WT and PV powers. The results for generation, reserve, and startup
Comparing of cost function. costs are $745.6913, $10.5278, and $6.35, respectively. As seen, these
Method SD($) AS($) WS($) BS($) Mean results are improved in the previous two scenarios. The amendment of
simulation the load curve using EVs and cancelling the avoidable switching on/off
time of certain units can reduce the costs. Moreover, the emission is
GA 15.267 861.6174 879.2851 840.9731 13.64 1081.256 kg. The comparison of different solution methods is provided
PSO 10.7649 854.517 875.3917 839.2677 11.31 in Table 6. According to the results, the ISFLA has an edge over other
DE 8.0566 851.6617 868.3361 834.0051 9.34 algorithms as it signifies less computation time than other approaches.
ICA 5.0094 848.3197 865.9641 831.5514 6.88
Suggested 2.2984 837.9742 843.3378 820.3617 4.37
Time used by each technique to solve the problem is represented in
method Table 6 that shows good performance of the proposed method than other
methods.

cost of the system will be minimized. Fig. 13 shows the performance of 6. Conclusions
the generating units considering EVs. The effect of EVs on load curve
correction is shown in Fig. 14. The values of the two objectives obtained The optimal scheduling of a microgrid considering EVs and respon­
in the three scenarios are compared in Figs. 15 and 16. sive loads is investigated in this study. EVs are responsible for coping
The DR is also used here like in the previous scenario to provide a with the peak load conditions. In addition, the DR program is employed
section of the essential reserve power and provide the possibility to use to handle a part of the needed grid reserve to cope with WT and PV

Fig. 11. Scheduling of energy.

Fig. 12. Reserve management of scenario 3.

12
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 13. Operating cost of sources with EVs.

Fig. 14. Load curve considering EV.

uncertainties. The operation costs and emissions are two important 3 Assessing the performance of integrated demand response programs
challenges that should be minimized in the optimization problem. The for multi-carrier energy systems.
use of this model can consirablely minimize the cost and emissions. The 4 Proposing a hybrid robust chance-constrained optimization frame­
cost is obtained using a two-stage optimization model. Firstly, reserve work to approach the problem.
and generation power costs are determined. Next, the cost of re-
scheduling because of the deviations in WT and PV powers is calcu­ CRediT authorship contribution statement
lated. Simulation results are provided in different case scenarios. The
ISFLA algorithm is used for minimizing the total operating cost and Tao Hai: Methodology, Writing – original draft, Software, Valida­
emissions. The ISFLA algorithm performance is compared with other tion, Formal analysis, Writing – review & editing. Jincheng Zhou:
well-known algorithms. The results confirm the efficacy of the recom­ Methodology, Writing – original draft, Software, Validation, Formal
mended structure in minimizing the total operating costs and emissions. analysis, Writing – review & editing. Alireza Rezvani: Methodology,
The use of this model can reduce the operation costs and emissions while Writing – original draft, Software, Validation, Formal analysis, Writing –
the interest in using DR programs, EVs and renewable-based sources are review & editing. Binh Nguyen Le: Methodology, Writing – original
increased. The future work may include the following items to further draft, Software, Validation, Formal analysis, Writing – review & editing.
extend the current research study: Hitoshi Oikawa: Methodology, Writing – original draft, Software.

1 Assessing the operation of the utility grid with multiple microgrids.

2 Extending the current model to multi-carrier energy systems opera­ Declaration of Competing Interest
tion including heating and cooling load demands beside the elec­
trical load demands. The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence

13
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

Fig. 15. Evaluation of three scenarios.

Fig. 16. Comparison of the emissions.

the work reported in this paper.

Table 6
Comparing of cost function.
Data availability
Method Std($) Mean($) WS($) BS($) Mean
simulation
No data was used for the research described in the article.
time

GA 17.6677 784.2557 799.9041 761.703 14.66

PSO 13.1656 777.1553 796.0107 759.9976 13.74
DE 10.4573 774.3 788.9551 754.735 10.72 Acknowledgments
ICA 7.4101 770.958 786.5831 752.2813 8.91
Suggested 4.6991 760.9783 763.9568 745.6913 6.81 This work was supported by Key Laboratory of Advanced
method
Manufacturing Technology of the Ministry of Education of Guizhou
University (GZUAMT2022KF[07]), Foundation of State Key Laboratory
of Public Big Data(No.2023004), National Natural Science Foundation
of China (No.61862051), the Science and Technology Foundation of
Guizhou Province (No. ZK[2022]549), the Natural Science Foundation
of Education of Guizhou province (No. [2019]203, No. KY[2019]067),

14
T. Hai et al. Electric Power Systems Research 221 (2023) 109370

and the Funds of Qiannan Normal University for Nationalities (No. [17] T. Niknam, R. Azizipanah-Abarghooee, M.R. Narimani, An efficient scenario-based
stochastic programming framework for multi-objective optimal micro-grid
qnsy2019rc09).
operation, Appl. Energy 99 (2012) 455–470. Nov 1.
[18] S. Mohammadi, B. Mozafari, S. Solimani, T. Niknam, An adaptive modified firefly
References optimisation algorithm based on Hong’s point estimate method to optimal
operation management in a microgrid with consideration of uncertainties, Energy
[1] E. Rahmani, S. Mohammadi, M. Zadehbagheri, M. Kiani, Probabilistic reliability 51 (2013) 339–348. Mar 1.
management of energy storage systems in connected/islanding microgrids with [19] A. Kavousi-Fard, A. Abunasri, A. Zare, R. Hoseinzadeh, Impact of plug-in hybrid
renewable energy, Electr. Power Syst. Res. 214 (2023), 108891. Jan 1. electric vehicles charging demand on the optimal energy management of
[2] T. Hai, J. Zhou, K. Muranaka, Energy management and operational planning of renewable micro-grids, Energy 78 (2014) 904–915. Dec 15.
renewable energy resources-based microgrid with energy saving, Electr. Power [20] C. Battistelli, L. Baringo, A.J. Conejo, Optimal energy management of small electric
Syst. Res. 214 (2023), 108792. Jan 1. energy systems including V2G facilities and renewable energy sources, Electr.
[3] A. Ghasempour, J. Lou, Advanced metering infrastructure in smart grid: Power Syst. Res. 92 (2012) 50–59. Nov 1.
requirements challenges architectures technologies and optimizations. Smart [21] J. Soares, H. Morais, T. Sousa, Z. Vale, P. Faria, Day-ahead resource scheduling
Grids: Emerging Technologies, Challenges and Future Directions, Nova Science including demand response for electric vehicles, IEEE Trans. Smart Grid 4 (1)
Publishers, Hauppauge, NY, USA, 2017, pp. 1–8. Oct. (2013) 596–605. Jan 14.
[4] A. Ghasempour, Internet of things in smart grid: architecture, applications, [22] T. Sousa, H. Morais, J. Soares, Z. Vale, Day-ahead resource scheduling in smart
services, key technologies, and challenges, Inventions 4 (1) (2019) 22. Mar 26. grids considering vehicle-to-grid and network constraints, Appl. Energy 96 (2012)
[5] A. Ghasempour, Optimum number of aggregators based on power consumption, 183–193. Aug 1.
cost, and network lifetime in advanced metering infrastructure architecture for [23] Han S., Han S., Sezaki K. Optimal control of the plug-in electric vehicles for V2G
Smart Grid Internet of Things, in: Proceedings of the 2016 13th IEEE Annual frequency regulation using quadratic programming InISGT 2011 2011 Jan 17 (pp.
Consumer Communications & Networking Conference (CCNC), IEEE, 2016, 1–6). IEEE.
pp. 295–296. Jan 9. [24] E. Sortomme, M.A. El-Sharkawi, Optimal scheduling of vehicle-to-grid energy and
[6] A. Ghasempour, Optimized advanced metering infrastructure architecture of smart ancillary services, IEEE Trans. Smart Grid 3 (1) (2011) 351–359. Sep 19.
grid based on total cost, energy, and delay, in: Proceedings of the IEEE Power & [25] H. Morais, T. Sousa, Z. Vale, P. Faria, Evaluation of the electric vehicle impact in
Energy Society Innovative Smart Grid Technologies Conference (ISGT), IEEE, 2016, the power demand curve in a smart grid environment, Energy Convers. Manag. 82
pp. 1–6. Sep 6. (2014) 268–282. Jun 1.
[7] C. Liu, S.S. Abdulkareem, A. Rezvani, S. Samad, N. Aljojo, L.K. Foong, K. Nishihara, [26] W. Yao, J. Zhao, F. Wen, Y. Xue, G. Ledwich, A hierarchical decomposition
Stochastic scheduling of a renewable-based microgrid in the presence of electric approach for coordinated dispatch of plug-in electric vehicles, IEEE Trans. Power
vehicles using modified harmony search algorithm with control policies, Sustain. Syst. 28 (3) (2013) 2768–2778. Apr 23.
Cities Soc. 59 (2020), 102183. Aug 1. [27] L. Jian, Y. Zheng, X. Xiao, C.C. Chan, Optimal scheduling for vehicle-to-grid
[8] M.S. Javadi, M. Gough, A.E. Nezhad, S.F. Santos, M. Shafie-khah, J.P. Catalão, Pool operation with stochastic connection of plug-in electric vehicles to smart grid,
trading model within a local energy community considering flexible loads, Appl. Energy 146 (2015) 150–161. May 15.
photovoltaic generation and energy storage systems, Sustain. Cities Soc. 79 (2022), [28] B. Kabatepe, M. Türkay, A bi-criteria optimization model to analyze the impacts of
103747. Apr 1. electric vehicles on costs and emissions, Comput. Chem. Eng. 102 (2017) 156–168.
[9] X. Zheng, H. Wu, Q. Ye, A cloud fog intelligent approach based on modified Jul 12.
algorithm in application of reinforced Smart Microgrid management, Sustain. [29] P. Beraldi, F. De Simone, A. Violi, Generating scenario trees: a parallel integrated
Cities Soc. 76 (2022), 103455. Jan 1. simulation–optimization approach, J. Comput. Appl. Math. 233 (9) (2010)
[10] Y. Li, S. Samad, F.W. Ahmed, S.S. Abdulkareem, S. Hao, A. Rezvani, Analysis and 2322–2331. Mar 1.
enhancement of PV efficiency with hybrid MSFLA–FLC MPPT method under [30] H. Falsafi, A. Zakariazadeh, S. Jadid, The role of demand response in single and
different environmental conditions, J. Clean Prod. 271 (2020), 122195. Oct 20. multi-objective wind-thermal generation scheduling: a stochastic programming,
[11] E. Alvarez, A.M. Campos, P. Arboleya, A.J. Gutiérrez, Microgrid management with Energy 64 (2014) 853–867. Jan 1.
a quick response optimization algorithm for active power dispatch, Int. J. Electr. [31] H. Morais, T. Sousa, Z. Vale, P. Faria, Evaluation of the electric vehicle impact in
Power Energy Syst. 43 (1) (2012) 465–473. Dec 1. the power demand curve in a smart grid environment, Energy Convers. Manag. 82
[12] A. Chaouachi, R.M. Kamel, R. Andoulsi, K. Nagasaka, Multiobjective intelligent (2014) 268–285. Jun 1.
energy management for a microgrid, IEEE Trans. Ind. Electron. 60 (4) (2012) [32] P. Beraldi, F. De Simone, A. Violi, Generating scenario trees: a parallel integrated
1688–1699. Feb 24. simulation–optimization approach, J. Comput. Appl. Math. 233 (9) (2010)
[13] S. Abedi, A. Alimardani, G.B. Gharehpetian, G.H. Riahy, S.H. Hosseinian, 2322–2332. Mar 1.
A comprehensive method for optimal power management and design of hybrid [33] R.A. Hooshmand, M. Parastegari, M.J. Morshed, Emission, reserve and economic
RES-based autonomous energy systems, Renew. Sustain. Energy Rev. 16 (3) (2012) load dispatch problem with non-smooth and non-convex cost functions using the
1577–1587. Apr 1. hybrid bacterial foraging-Nelder–Mead algorithm, Appl. Energy 89 (1) (2012)
[14] Y.Y. Hong, J.K. Lin, Interactive multi-objective active power scheduling 443–453. Jan 1.
considering uncertain renewable energies using adaptive chaos clonal evolutionary [34] M. Eusuff, K. Lansey, F. Pasha, Shuffled frog-leaping algorithm: a memetic meta-
programming, Energy 53 (2013) 212–220. May 1. heuristic for discrete optimization, Eng. Optim. 38 (2) (2006) 129–154. Mar 1.
[15] M. Motevasel, A.R. Seifi, Expert energy management of a micro-grid considering [35] A. Ghasempour, Using a genetic-based algorithm to solve the scheduling
wind energy uncertainty, Energy Convers. Manag. 83 (2014) 58–72. Jul 1. optimization problem for long-range molecular communications in nanonetworks,
[16] S. Mohammadi, S. Soleymani, B. Mozafari, Scenario-based stochastic operation in: Proceedings of the IEEE 26th Annual International Symposium on Personal,
management of microgrid including wind, photovoltaic, micro-turbine, fuel cell Indoor, and Mobile Radio Communications (PIMRC), IEEE, 2015, pp. 1825–1829.
and energy storage devices, Int. J. Electr. Power Energy Syst. 54 (2014) 525–535. Aug 30.
Jan 1.

15