View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Institute of Advanced Engineering and Science

International Journal of Electrical and Computer Engineering (IJECE)

Vol. 10, No. 5, October 2020, pp. 5093~5107
ISSN: 2088-8708, DOI: 10.11591/ijece.v10i5.pp5093-5107  5093

Optimal scheduling of smart microgrids considering electric

vehicle battery swapping stations

J. Garcia-Guarin1, W. Infante2, J. Ma3, D. Alvarez4, S. Rivera5

1,4,5Departmentof Electrical Engineering, Faculty of Engineering, Universidad Nacional de Colombia, Colombia
2,3Department of Electrical Engineering, Faculty of Engineering, University of Sydney, Australia

Article Info ABSTRACT

Article history: Smart microgrids belong to a set of networks that operate independently.
These networks have technologies such as electric vehicle battery swapping
Received Mar 22, 2020 stations that aim to economic welfare with own resources. Improper handling
Revised Apr 24, 2020 of electric vehicles services represents overload, congestion or surplus
Accepted May 3, 2020 energy. This study addresses both management and support of electric
vehiculos battery swapping stations. The formulation of a decision matrix
examines economically viable alternatives that contributes to scheduling
Keywords: battery swapping stations. The decision matrix is implemented to manage
the swapping, charging and discharging of electric vehicles. In addition,
Electric vehicle the smart microgrid model evaluates operation issues. The smart microgrid
Microgrids model used extends with considerations of demand response, generators with
Optimization renewable energies, the wholesale market, local market and electricity spot
Scheduling price for electric vehicles. Additionally, uncertainty issues related to
Swapping stations the planning for the infrastructure of the electric vehicle battery swapping
station, variability of electricity prices, weather conditions and load
forecasting are used. Mentioned stakeholders must maximize their economic
benefits optimizing their uncertain day-ahead resources. The proposed hybrid
optimization algorithm supports aggregator decision making. This algorithm
achieves a 72% reduction in total cost. This percentage of feasible reduction
is obtained by calculating the random solution with respect to the suboptimal
solution.
Copyright © 2020 Institute of Advanced Engineering and Science.
All rights reserved.

Corresponding Author:
Sergio Rivera,
Department of Electrical Engineering,
Universidad Nacional de Colombia,
Carrera 30 Número 45-03, Bogotá, Colombia.
Email: srriverar@unal.edu.co

NOMENCLATURE
Indexes:
f Time zero for EV batteries L Loads

i Distributed generation (DG) units M Markets

j PV units S Scenarios

k External suppliers t Periods

E Energy storage systems

Variables:
Active power generation (kW) Discharge power of ESS (kW)
PDG PESS 
External supplied power (kW) Discharge power of EV (kW)
Pext PEV 

Journal homepage: http://ijece.iaescore.com/index.php/IJECE

5094  ISSN: 2088-8708

Charge power of ESS (kW) Number of fully-charged batteries

PESS  N fc
Charge power of EV (kW) Number of batteries to charge
PEV  Nc
Power reduction of load (kW) Number of batteries to discharge
Pcurt Nd
Unsatisfied power for load (kW) Number of batteries to swap
Pimb  N sw
Exceeded power of DG unit (kW) Number of initial batteries in the station
Pimb Ni
Power bought to the market (kW) Initial EV station visits (EV demand)
Pbuy ND
Power sold to the market (kW) Number of EVs
Psell N ev
Photovoltaic generation (kW) J Demand welfare
PPV
Parameters
 Full charging battery interval
Cbuy Cost for buying energy (m.u./kWh)

 Full discharging battery interval

Csell Cost for selling energy (m.u./kWh)

N DG Number of DG
 s Probability of scenario 𝑠

Number of external suppliers Photovoltaic generation (kW)

Nk PPV
Number of ESSs Forecasted load (kW)
Ne Pload
Number of loads MP Market price in wholesale and local markets
NL (m.u./kWh)
Number of markets Price for swapping a batterie (m.u.)
Nm Psw
Number of scenarios Price for missing a battery (m.u.)
Ns Pm
T Number of periods EVs market price (m.u.)
MPe
CDG Generation cost of DG (m.u./kWh)
d Efficiency for discharging

Cext Cost of external supplier (m.u./kWh)

c
CPV Cost of PV generation (m.u./kWh)

CESS Cost of ESS (m.u./kWh)
 Charger rate of the on-board EV (kW)

Cload Cost of load (m.u./kWh)

 Charger rate from the charger level specification
(kW)
Cimb Grid imbalance cost (m.u./kWh) ev Type of EV

1. INTRODUCTION
Electric vehicles (EVs) have shown additional benefits compared with their fossil fuel vehicle
counterparts [1]. They produce fewer emissions even when considering their whole process of energy
production, independently of their energy source [1]. In addition to lower emissions, EVs may also use
renewables energies as a source of power [1]. However, one of the greatest deficiencies of EVs is that
conventional vehicles can easily be refueled, while EVs require long charging times that need planning.
This is combined with the difficulties in setting up an infrastructure with specialized equipment to facilitate
the use of EVs [1]. Given these scenarios, the sustainable EV operation must rely on the efficient EV
scheduling, among others.
There are still challenges in the deployment of EV charging infrastructure. First, EVs increase
the power demand on the grid [2]. Second, when distributed generation is available, uncertainty over
the photovoltaic and wind generation, electricity prices, load forecasting, and swapping demand can
influence the operation of the electrical network [3]. Furthermore, demand management considers strategies
for flatting the demand of EV´s and load with demand response (DR), such as moving the peak load to valley
hours [3]. Third, time role is crucial to schedule events, such as swapping, charging, and discharging of EVs
batteries [4]. Fourth, the longevity of battery life is a concern for end users given the level of charging and
discharging that users must undergo [4]. Fifth, consumers may have range anxiety on whether EVs batteries
can last all their trips. These barriers can also represent restrictions to the proper use of the microgrid with
EVs. As an illustration, if the customers distrust both EV technology and the associated infrastructure.

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5095

This risky behaviour may result in the overuse of charging stations [4]. Finally, the overall inherent
uncertainties that the aggregator must manage when implementing the smart microgrid (SMG) is significant.
In fact, SMG is defined as a group of loads with DR and energy storage systems (ESSs), as much as they act
as a unified entity. SMG operates in both connected and isolated modes [5]. However, demand of users,
market prices, and uncertain renewable energy production may lead to barriers in creating efficient charging
station infrastructures [6]. To countermeasure these challenges, an aggregator can be formulated [6, 7].
Energy aggregators can be made by integrating EVs and ESSs. Furthermore, battery swapping
stations (BSSs) have more competitive visit times of EVs in comparison with traditional stations, which only
have the option of charging batteries [6]. Regarding the schedule, public transport would prefer swapping
the battery instead of charging batteries due to the critical operating time [6, 7]. In SMGs, a forecasting issue
is the prediction of EVs visits [6]. An aggregator may deal with this problem, as it manages components such
as photovoltaic panels, loads with DR, ESSs, and EVs [8]. The last two can both discharge (or charge) power
and buy (or sell) energy in SMGs [9]. The aggregator coordinates the service reliability among
the stakeholders, EVs scheduling, hours of load shifting, or valley filling of EVs. Aggregators should also
consider uncertainties such as EVs trips planning, forecasting load, electricity prices, and generation with
renewable energies [10]. Finally, the increase in SMG profits has been addressed with metaheuristic
optimization tools, producing outstanding solutions in acceptable times [9].
Under this context, the proposed SMG model introduces an aggregator, who optimizes EV BSS.
Equally notable, this approach is formulated using a decision matrix, which aims to simulate operating
conditions such as charging (or discharging) times, stochastic visits, and batteries swapping of EVs. Without
overlooking other elements, this study formulates the programming of traditional and renewable energies,
ESSs, electricity markets, and loads with DR. The main contributions of this paper are described below:
The SMG model introduces stochastic scheduling due to the EV station visits based on the K-means
clustering approach, and price variability in electricity markets. Both aspects are formulated according to
the operation approach of EV BSS. To be more comprehensive, the SMG model includes uncertainty
conditions, such as weather forecast, load prediction, and wholesale and local markets.
A decision matrix to EV BSS is scheduled introducing significant elements in EVs, such as state of
charge (SOC) of EVs, type of vehicle, probability of EVs visiting the BSS, state of the battery (swap or do
nothing) and idle spaces (per hour). The idle spaces are generated when the charge of a battery must be
mandatory, to exemplify a reserve of batteries that guarantees a good service in the case of unexpected visits
of EVs in the BSS. The proposed SMG model has continuous and discrete variables, as well as a decision
matrix that introduces a new problem related to integer numbers, sequence of numbers and idle spaces.
Therefore, the optimization algorithm “variable neighbourhood search-differential evolutionary particle
swarm” (VNS-DEEPSO) is implemented in order to reduce the cost in SMGs with uncertainty environments.
This document is organized as follows: section 2 presents the state of the art of SMG models that
include EV BSS. Section 3 formulates the SMG model and emphasizes the constraint of the SMG with EV
BSSs. In section 4, the uncertainty sources are highlighted and the EV BSS matrix with uncertainty is
presented, showing the operational strategy. In section 5, the case study and results are discussed. Finally,
section 6 outlines the conclusions.

2. STATE OF THE ART

SMGs have challenges related to the planning of the day-ahead operation of EVs. At the same time,
other elements (load with demand response - DR, generation - Gen and ESSs) must be coordinated [7].
Six elements are tackled as main contribution of this study: (1) load with DR, (2) Gen, (3) ESSs,
(4) EV-prosumer (EVP), (5) BSS, and (6) uncertainty of EV BSS, Gen, loads with DR and energy markets.
Hereinafter, the six terms are called uncertainty sources. According to the information collected, Table 1
organizes SMG models from 1 to 10, based on their main characteristics.
SMG model 1 schedules the day ahead, where the optimization problem is solved by assembling
the genetic algorithm with the feasible solution region [3]. Therefore, an economical operation of BSS is
proposed when it chooses the level of tolerable risk. This BSS model operates as an isolated network, that is,
it does not integrate other elements of the grid and eliminates the discharged function of EVs batteries [3].
SMG model 2 coordinates the charging, discharging, and swapping of batteries. When EVs can
buy/sell energy to the grid, this role is called EVP. The swapping prices and charging intervals are compared
with traditional strategies. Then, the results demonstrate that the profit of BSS coordination is acceptable in
comparison with other strategies. Nevertheless, the associated cost does not take into account other elements
of the grid [6].
SMG model 3 includes other elements such as EV operation, photovoltaic (PV) generation,
electricity market pricing, and DR. In this case, the EV model has no comprehensive description of scenarios

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)

5096  ISSN: 2088-8708

and constraints [9]. SMG model 4 proposes the dispatch of energy for EVs, wind and photovoltaic power
generation. In fact, this EV model associates EVs with each node and only the load power varies. Therefore,
the model lacks an approach of discharging, swapping batteries, and constraints [11].
BSS model 5 gathers the EVs in programs of DR. The case study demonstrates that ignoring
the uncertainty of EVs traffic and PV generation leads to an inappropriate planning. Furthermore, BSS model
5 assumes simplifications without considering other elements of the grid and the uncertainty of the market
prices [12]. The model 6 for battery charging stations considers the uncertainty over charging EVs.
EV uncertainties are modeled through a Gaussian distribution model. In addition, the EVs are connected to
the grid, PV generation, and battery-ESSs. In spite of this, the last variables have uncontrolled stochastic
uncertainties, since there are no details about these uncertainty models [13].
BSS model 7 forecasts wind-power generation and uses real time prices to plan in short time.
However, the EV model takes into account the uncertainty over charging EVs [14]. BSS model 8 considers
the stochasticity in energy markets and demand management of batteries. Nevertheless, model 8 simplifies
the interaction between stakeholders and EV BSS [15].
BSS model 9 purchases power in an upstream network and maximizes the day-ahead incomes.
This bi-level model is divided into lower and upper levels. This model has acceses to the mechanism of
real-time prices. The BSS model includes a micro-turbine, PV and wind generation. The uncertainty of
the last two is considered together with the uncertainty of the load demand and the arrival time of EV.
However, this model includes neither the ESS nor the load shedding restrictions [16]. A framework is
proposed to restrict the load shedding in [17]. Other research also considers the uncertainty of PV and wind
generation and introduces an index of probability of islanding operation [18].
BSS model 10 is formulated in this paper and integrates constraints to EVs through BSS, where BSS
uncertainty is simulated by means of a K-means clustering approach. Moreover, it integrates SMG with
uncertainty elements such as PV generation, electricity markets, and loads with DR. SMG planning is
performed by the aggregator, which tries to increase the profits of the SMG. Then, the aggregator minimizes
the operational cost while increasing the incomes. Indeed, the transactions in electricity markets generate
profit. Besides, the aggregator has assets, such as EVs BSSs and ESSs, which become prosumers as they can
either buy or sell energy.

Table 1. Review of microgrids with EVs BSS

No DR Gen ESS EVP BSS Uncertainty sources
1 No Yes No No Yes EVs and PV generation [3]
2 No No No Yes Yes EVs and electricity markets [6]
3 Yes Yes Yes Yes No EVs, PV generation, electricity markets, and loads with DR [9]
4 Yes No No No No EVs, wind-power, and PV generation [11]
5 Yes Yes No No Yes EVs and PV generation [12]
6 No Yes Yes No No EVs [13]
7 No Yes No No Yes Electricity markets and wind-power generation [14]
8 No Yes No Yes Yes Battery demand and electricity markets [15]
9 No Yes No Yes Yes PV and wind-power generation, load demand and EVs [16]
10 Yes Yes Yes Yes Yes EVs, PV generation, electricity markets, and loads with DR (This SMG
model is formulated in this paper)

The BSS model formulation can be applied for DR programs [19]. The load is usually involved in
DR programs [9]. In the BSS model, the aggregator negotiates energy in the electricity spot price. In the BSS,
one available asset is the battery bank in the BSS that can be encoded with a EVs BSS matrix.
The aggregator deals with the uncertainty in the arrival time of EVs and the volatility of market prices.
For both, the uncertain scenarios are estimated with the K-means grouping method. In the case study, vehicle
driving patterns are forecast based on NWS traffic [6]. Load management for EVs should tackle congestion
times and relieve the scarcity of energy [20]. The SOC eases the scheduling the batteries charge or discharge
as indicator [20].
In this way, SMG model 10 is the most comprehensive, taking into account the previous revision in
Table 1. Satisfactory solutions of vital importance are inspected, through the formulation of heuristic
methods [9, 21]. These methods have been widely studied, however, there are still development gaps related
to yields, and exploration and exploitation methods [21]. A vast majority of metaheuristics proposes
a sequential combination of heuristics to be more successful [21]. For example, a hierarchy analysis is set out
among several algorithms such as enhanced velocity differential evolutionary particle swarm optimization
algorithm. In Figure 1, the main algorithms reported in the literature are organized according to the hierarchy
of a better suboptimal solution [9, 22]. In summary, VNS-DEEPSO algorithm turns out to have a greater

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5097

hierarchy regarding the SMG models proposed in [9, 22] and it is described in two (VNS and DEEPSO) main
components below. VNS-DEEPSO is widely tested in a robust optimization approach [9]. This research finds
out a robust solution with low sensibility to some uncertainty parameters, such as PV power forecast, load
prediction, EVs planning, and wholesale and local markets [9].

First
Variable neighborhood search differential evolutionary particle swarm (VNS-DEEPSO)
hierarchy

Second
Evolutionary particle swarm optimization (EVDEPSO) hierarchy

Ant colony optimization (ACO), chaotic evolutionary swarm optimization (CESO), cuckoo
search (CS), evolutionary particle swarm optimization (EPSO), flower pollination (FP), Third
differential evolution (DE), genetic algorithm (GA), hybrid simulated annealing (HSA), hierarchy
hybrid differential search algorithm (HDSA), particle swarm optimization (PSO), quantum
particle swarm optimization (QPSO), simulated annealing (SA) and Tabu search (TS).

Figure 1. Hierarchy of the best suboptimal solutions based on [9, 22]

2.1. Variable neighborhood search-VNS

VNS is used in nonlinear optimization, as it can find a local optimum with quality solutions [23].
The algorithm makes automatics neighborhood changes and aims to extend the local search. The VNS
algorithm is defined in two main stages: initialization and repetition. The initialization stage consists in
defining the neighborhood structures. In the repetition is carried out in three steps. First step, the initial
solution is found in the neighborhood. Second step, the search method looks for an initial solution called
local optimum. In the third step, the local optimum is stored. Afterward, when the new solution improves
the local optimum solution, then the local optimum solution will be updated. Two cases can happen:
first, the new solution found in the process is worse than the local optimum or second, the new solution found
is the same as the local optimum. In both cases, the search continues for the next neighborhood structure.
These steps are repeated until the last neighborhood structure [23, 24].

2.2. Differential evolutionary particle swarm optimization-DEEPSO

DEEPSO algorithm was developed based on three metaheuristics: differential evolution (DE),
evolutionary algorithm (EA) and particle swarm optimization (PSO). These methods try to find the global
correct direction and jointly search for a robust solution [25]. EA algorithm uses biological operators such as
reproduction, mutation, recombination, and selection. Therefore, the population evolves through biological
techniques. Furthermore, the fitness determines the solution quality [26].
PSO algorithm is a set of candidate solutions as swarm particles. They flow in tracks of the search
space. The search motivation is the best performance of them and their neighbors [27]. DE algorithm has
three or four operational parameters and in roughly 20 lines, it includes a simple and adaptable mutation
operator [26]. Finally, DEEPSO algorithm was introduced as a metaheuristic technic and is a hybrid among
EA, PSO, and DE. It has best outcomes than other techniques, however, it does not provide a deductive
demonstration [25].

3. SMART MICROGRID MODEL

This section presents the main assumptions in the BSS model 10 of SMG that represents the profits
(including operation costs and incomes). Figure 2 represents the main elements of the SMG (ESS, EV BSS,
load with DR, day-ahead markets, PV generation and DG). The arrows mean buy or sell energy: selling is
symbolized by the line entering to the aggregator and buying by the line coming out from the aggregator.
The red line highlights the elements with uncertainty. The SMG model is obtained by codifying a black box
model of SMG from [28] and a decision matrix is developed by collecting available data from [29, 30],
together with the EV uncertainty model based on the methodology from [6]. The main SMG assumptions are
listed below.

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)

5098  ISSN: 2088-8708

- SMGs is a component of the smart grid, but also SMGs operate independently and self-sufficiently.
By the same token, SMGs consist of a set of loads and generators operating as a unique system [31].
In this context, SMGs have aggregators that try to increase the SMG profits.
- SMGs are flexible in connecting distributed generation (DG) and PV generation [32].
- Two electricity markets are considered wholesale and local, and the EV prices are estimated with
the Australian Market Operator from January 2016 to June 2019. Prices from electricity markets lead to
revenues for selling or expenses for buying energy [28].
- A realistic approach of SMG includes energy resources with uncertainty. The uncertainty sources come
from (a) PV renewable generation, (b) load profiles, (c) visit of EVs to BSSs, and (d) electricity market
prices for wholesale, local, and electricity spot for EVs [6, 28].
- The aggregator uses ESS and BSS as their own assets and supports the programming of energy
resources of SMG. Indeed, ESS has constraints related to the state of the battery, while EVs have
constraints related to stochastic visits to BSS, brand, and plug type of EV.
- The improvement of SMG profits is formulated with a heuristic optimization algorithm. Specifically,
the VNS-DEEPSO algorithm is a recently studied technique that has demonstrated good performance
for these types of problems [21].

Decision matrix for EV BSS, data from [29, 30]

Electricity spot price for EVs

ESS BSS
Stochastic visit of EVs

DG Load with
Aggregator optimizes energy
DR
PV transactions of SMG
generation

Day-ahead markets

Uncertainty Bidirectional transaction One-directional transaction

Figure 2. Depiction of energy transactions in SMG

According to the main aspects in the SMG operation, the objective function aims to minimize Z,
that is, the negative of the profits (1), which means that profits are maximized. It is defined as the addition
and subtraction of six terms. The first term introduces the value of buying energy to load with DR.
The second term, Gen can be negative when the energy of DGs and PV generation are sold. The third, fourth
and fifth terms are the prosumers, represented by the ESS, BSS of EVs, and market transactions (MTs).
These costs or revenues are quantified based on the type of transfer. In the first place, the incomes are
represented with a negative value and, in the second place, costs are estimated with a positive value.
The sixth term of (1) is penalties (Pen), which are defined for an imbalance, when the demand is not satisfied
or the generation is exceeded [28]. Fixed costs are related to the purchase of batteries, degradation of
components, and maintenance. As in math, the derivative of a constant is zero. Fixed costs are also constant
(they do not vary regardless of the decision variables), therefore, they are not included in the objective
function, because their change rate is zero (1) [6]. For explanations purposes, the six terms on the right side
of (1) are formulated from (2) to (7).
Equation (2) represents the prices (𝑃) per capacity (𝐶) of loads with DR and the scenarios (s) have
a probability (𝜋(𝑠)). In (3), the total generation is denoted as the sum of DGs, external supplier, and PV
generation. Equation (4) details the operation of the ESS. In (5), BSS revenue is calculated from allocating
a swapped battery 𝑁𝑆𝑊 per price 𝑃𝑆 . If EVs visit the station, but there is no battery to swap, then BSS owners
will receive penalties. The penalty is calculated based on the subtraction of an assigned battery 𝑁𝑠𝑤 from EVs
battery demand 𝑁𝐷 [15]. In (5), the efficiencies (𝜂 𝑑 and 𝜂 𝑐 ) refer to charging and discharging of batteries,

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5099

while 𝑁𝑐 and 𝑁𝑑 represent the number of batteries that are charged and discharged in the electricity
spot price for EVs (𝑀𝑃𝑒 ). In (6), MTs represent the sale (𝐶𝑏𝑢𝑦 ) and purchase (𝐶𝑠𝑒𝑙𝑙 ) of energy. Finally,
the penalizations (Pen) are made such that the demand is not satisfied, or the generation is exceeded as shown
in (7). This model is structured with mixed-integer linear programming. However, in (9) the constraint for
demand with DR is nonlinear, so it can be classified as mixed-integer nonlinear programming.

minimize Z  DR  Gen  ESS  BSS  MT  Pen (1)

N s T  24 N L
DR  Pload l ,t ,s   Cload l ,t     s  (2)
s 1 t 1 l 1

T  24 N DG T  24 N k

 P
t 1 i 1
DG  i ,t 
 CDG  i ,t    P
t 1 k 1
ext  k ,t 
 Cext  k ,t 
Gen  N s T  24 N PV
(3)
   PPV  j ,t , s   CPV  j ,t     s 
s 1 t 1 l 1

N s T  24 Ne
ESS  PESS  j ,t ,s   CESS  j ,t     s  (4)
s 1 t 1 e 1

 
 
  N sw t , s   Psw t  
N s T  24  

BSS      Pm t  N D t , s   N swt ,s 
s 1 t 1  

   s  (5)

  d N c t , s   
  MPe t , s   N d t ,s   c  
   

 Nm 
 
N s T  24
MT      Cbuy  m ,t   Csell  m ,t    MP m ,t , s     s  (6)
s 1 t 1  m 1 

 NL 
N s T  24   imb   j ,t , s 
P  Cimb t  
Pen     N    s 
l 1
(7)
 
  Pimb  j ,t ,s   Cimb t  
DG
s 1 t 1

 i 1 

The energy balance has constraints shown in (8). The conservative balance should be zero.
The energy balance is composed of the total generation (DG, an external supplier, and PV), ESS, load with
DR, transfer in the market, and imbalance in the generation of energy [28]. EVs BSS has another analysis
equal to the energy balance shown from (10) to (17). This model also considers that EVs and loads are
involved in DR programs. Equation (9) formulates a constraint for the load DR that they should be different
from zero and the demand welfare 𝐽 represents at least 10 % of the total energy being supplied for all
periods [32]. Equation (1) is subject to (8) and (9). EVs encompass the DR program and schedule the BSS
arrival time with the following assumptions. The amount of visits of EVs is expected to be fixed [6].
However, visiting hours are estimated with the K-means clustering method [6]. Therefore, the visiting hours
of EVs is uncertain [16]. The priority in the BSS is that the swapping batteries must be charged. Once
the batteries are charged, the aggregator decides whether to discharge or swap them. This option is only
possible when one or more EVS eventually demand to swap a battery.

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)

5100  ISSN: 2088-8708

t T , s S
 
N DG Nk N PV  DG Ne (8)
P
i 1
DG  i ,t 
 Pext  k ,t  
k 1
Pj 1
PV  j ,t , s 
  PESS   e,t , s   PESS   e,t , s 
e 1

   
NL Nm
  Pcurt l ,t , s   Pload l ,t , s    Pbuy  m,t , s   Psell  m ,t ,s 
l 1 m 1
N DG NL
 Pimb i ,t , s   Pimb l ,t , s   0
i 1 l 1

 t 1 l L1 Pload l ,t ,s  Cload l ,t 

T N
10

max    Pload  l ,t ,s  Cload  l ,t  
T NL t T , s S (9)
J  1 e t 1 l 1

Fully charged batteries are available per hour. At least one fully charged battery is guaranteed every
period in the BSS. In the event that electric vehicles visit the substation, two cases may occur. In the first
case, the battery is not swapped because the aggregator does not consider it compelling or there are no
batteries available. These assumptions contemplate that the aggregator is penalized, so these situations are
undesirable [15]. In the second case, the battery is swapped. The day begins with a fully charged battery
bank. EVs arriving at the BSS have completely depleted batteries [6]. The charging / discharging capacity of
EVs batteries is constant. This is estimated with the average charging time according to the type of charger
and the EV brand [6]. The battery swapping time is considered negligible [16].
The substation has a constant number of batteries [33]. However, the SOC depends on factors, such
as the visits of EVs and the decisions of the aggregator [16]. Battery degradation is estimated as a fixed cost.
In other words, the variability of costs for battery degradation is assumed negligible for day-ahead
planning [33]. The aggregator can swap several batteries at the same time [33]. In accordance with
the aforementioned considerations, the restrictions for the programming of EVs are formulated.
In (10), fully-charged batteries (𝑁𝑓𝑐 ) include the previous fully-charged batteries in each period, new
batteries that are fully-charged (𝑁𝑐 ) and depleted batteries that are swapped for fully-charged batteries (𝑁𝑠𝑤 ).
An additional function of EV batteries consists in supplying the SMG demand. In (11), the waiting time for
charging or discharging is represented by the interval charging time (𝛿) or interval discharging time (𝜁).
𝛿 is calculated as the ratio between the vehicle battery capacity (𝜐 𝛽 ) and the minimum between the power of
the on-board EV (𝜐 𝛼 ) and the allowed power from the charger level specification (𝜐 𝛾 ).
The acceptance rate is limited by the EV technology for charging or discharging in BSS.
EV charging levels are frequently grouped into three categories as shown in Table 2 [34]. Level 1 chargers
are required either at home or in overnight process. Level 2 chargers have two uses in public or private
facilities. Level 3 chargers are developed with technology that allows a quick charge in a short time;
however, these advances are at a premature phase yet [34].
The next EV restriction (12) establishes the initial batteries in the station (𝑁𝑖 ). Fully-charged
batteries (𝑁𝑓𝑐 ) should maintain this limit. Furthermore, discharging (𝑁𝑑 ) and swapping (𝑁𝑠𝑤 ) batteries are
allowed without surpassing the number of initial batteries (13). In addition, EV customers demand (𝑁𝐷 )
should be higher than the batteries being swapped (14).
The batteries to be charged (𝑁𝑐 ) depend (𝑁𝑖 − 𝑁𝑓𝑐(𝑡,𝑠) ) on the battery availability and the allocated
batteries to swap 𝑁𝑠𝑤 in the actual period (15). In (16), the SOC is represented by the succession of
𝜁−1
the charge (∑𝛿−1𝑓=0 𝑁𝑐(𝑡−𝑓,𝑠) ) and discharge (∑𝑓=0 𝑁𝑑(𝑡−𝑓,𝑠) ) periods. The sum of these two terms cannot exceed
the initial number of batteries. In (17), decision variables are greater than or equal to zero.
Equation (1) is then subject to

N fct 1,s   Nct  ,s   Nd t  ,s   Nswt ,s   N fct ,s  t T , s S (10)

 (11)
 or  
min   ,  

0  N fct ,s   Ni t T , s S (12)

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5101

Nd t ,s   Nswt ,s   Ni t T , s S (13)

N swt ,s   N Dt ,s  t T , s S (14)

Nct ,s   Ni  N fct ,s   N swt ,s  t T , s S (15)

 1  1 t T , s S (16)
Nct  f ,s   Nd t  f ,s   Ni
f 0 f 0

Ni , Nswt ,s  , Nct ,s  , Nd t ,s   0 t T , s S (17)

N i , N swt ,s  , Nct ,s  , Nd t ,s  ℤ

Table 2. Specifications of the charger power level [34]

Power Level Typical use Charger rate (𝜐 𝛾 ) Variant
Level 1 Charging at home or office 1.4 kW (12 A) A
1.9 kW (20 A) B
Level 2 Charging at private or public outlets 7.7 kW (32 A) A
19.2 kW (80 A) B
Level 3 Commercial Up to 50 kW A
Up to 100 kW B

4. UNCERTAINTY SOURCES
The uncertainty is represented in the three groups presented below.

4.1. Uncertainty of PV generation, electricity markets, and loads with DR

The case study is created by using 5000 scenarios based on [28]. This research looks for robust
solutions to the uncertainty parameters, such as PV power forecast, load prediction, and wholesale and local
markets [28]. The PV power geration, the loads with DR and the variations of energy markets (wholesale and
local) have forecast errors of 15%, 10% and 20% respectively. In (18), the Monte Carlo method generates
error , s
the scenarios. A normal distribution function is used to forecast the error ( x ) through historical data.
forecast
The forecast errors ( x ) with uncertainty are represented by a normal distribution function. In the second
step, 5000 scenarios are reduced to 100 using a reduction technique based on low probabilities [35].

X S  t   x forecast  t   xerror ,s  t  (18)

4.2. Uncertainty of driven pattern and electricity spot price for EVs
Data on EV trips and electricity spot prices for EVs use the transportation and electricity sources
from New South Wales (NSW) [30]. The data on electricity spot prices is obtained from January 2016 to
June 2019. They are estimated with the Australian Energy Market Operator. The time resolution is per hour
and the K-means clustering is employed to forecast electricity spot prices from hour 1 to hour 24 [6].
Electricity spot prices for EVs are grouped into 10 clusters together with their probability percentages,
as shown in Figure 3. The percentages represent the probability of occurrence. A similar feature is called
a scenario and the prices are normalized so that the sum is unity for each day [6].
The cluster dataset identifies the EV expected visits in a specified region [6]. It is illustrated in
Figure 4, EVs are group by arrival time. Then, the traffic data includes the amount of cars for a 24-hour time
interval. The numer of vehicles and market prices have been normalized with a value of 0.1. Cars passing in
southbound and northbound areas are separated. The scenarios are produced through probability of
occurrence, where each cluster has some associated feature sets. In addition, the probability of occurrence
depends on the number of characteristic patterns [6].

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)

5102  ISSN: 2088-8708

Figure 3. Clustering of scenarios to EV market

Figure 4. Clustering of scenarios for the driven pattern of EV

4.3. EVs BSS matrix with uncertainty

Scenarios of EVs arrival time from January 2016 to June 2019 in NSW are modeled and reduced to
100 stochastic scenarios. For example, the EV BSS matrix illustrates a stochastic scenario and a feasible
solution. Table 3 shows the encoding for a random scenario. The charging (𝛿) and discharging (𝜁) time are
computed with Table 2. The time for charging and discharging are 8, 12, and 9 hours for Toyota, Nissan, and
Mitsubishi at levels 1 A, 1 B, and 2 A, respectively. To emphasize, the decision matrix is formulated to
quantify the number of fully charged batteries (Nfc), the number of batteries to charge (Nc), the number of
batteries to discharge (Nd) and the number of batteries to swap (Ns). Fully charged batteries are estimated for
the current (t) and previous (t-1) instant.

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
 Int J Elec & Comp Eng ISSN: 2088-8708  5103

Table 3. Sample of EVs BSS matrix

State of change of 15 EVs Batteries EVs Batteries counter
Nfc Nfc Ni-Nfc
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Nd Nc Ns
(t-1) (t1) (t-1)-Ns
1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 14 0 0 0 1 15
2 1 2 1 1 3 3 3 3 3 3 3 3 3 3 3 11 2 0 2 3 13
3 8 1 1 1 2 4 4 4 4 4 4 4 4 4 4 10 2 1 2 3 12
4 7 8 9 1 1 3 3 3 3 3 5 5 5 5 5 5 7 2 1 2 12
5 6 7 8 1 1 2 2 2 2 2 4 4 4 6 6 2 11 0 2 2 13
6 5 6 7 8 1 1 1 1 1 1 3 3 3 5 5 0 13 1 1 6 13
7 4 5 6 7 8 9 8 8 9 1 2 2 2 4 4 0 9 5 1 1 9
8 3 4 5 6 7 8 7 7 8 8 1 1 1 3 3 0 14 1 0 3 14
9 2 3 4 5 6 7 6 6 7 7 8 9 1 2 2 0 12 2 1 1 12
Planning day - ahead

10 1 2 3 4 5 6 5 5 6 6 7 8 1 1 1 0 14 0 1 4 14
11 2 1 2 3 4 5 4 4 5 5 6 7 8 12 1 1 11 2 1 2 12
12 1 1 1 2 3 4 3 3 4 4 5 6 7 11 9 0 13 1 1 3 13
13 8 8 1 1 2 3 2 2 3 3 4 5 6 10 8 0 12 2 1 2 12
14 7 7 9 1 1 2 1 1 2 2 3 4 5 9 7 0 13 1 1 4 13
15 6 6 8 8 2 1 1 1 1 1 2 3 4 8 6 1 11 1 2 5 12
16 5 5 7 7 1 2 2 2 2 2 1 2 3 7 5 5 10 0 0 2 15
17 4 4 6 6 1 1 1 1 3 3 2 1 2 6 4 3 11 0 1 5 14
18 3 3 5 5 9 1 1 1 2 2 3 2 1 5 3 2 9 1 3 4 11
19 2 2 4 4 8 12 8 8 1 1 2 1 2 4 2 1 11 3 0 3 12
20 1 1 3 3 7 11 7 7 9 8 1 1 1 3 1 0 12 2 1 6 12
21 2 2 2 2 6 10 6 6 8 7 8 9 8 2 1 2 9 3 1 1 11
22 3 3 1 1 5 9 5 5 7 6 7 8 7 1 12 2 12 1 0 3 14
23 2 2 2 2 4 8 4 4 6 5 6 7 6 1 11 2 12 0 1 1 14
24 1 1 1 1 3 7 3 3 5 4 5 6 5 9 10 0 14 1 0 4 14

Free spots: Swapping batteries: Penalties: 4 Total swapping: 30

Nomenclature
Nd: Number of batteries to dischange Nc Number of batteries to change
Nfc Number of fully changed batteries Ns Number of batteries to swap

The SOC is calculated each hour for 24 hours with 15 batteries to swap. Vehicles such as Toyota,
Nissan, and Mitsubishi use the services of BSS. The countdown decreases from 8, 12, and 9 hours (battery
completely discharged) to 1 hour (battery fully charged). Regarding the charging of EV batteries,
it is assumed that when the battery is swapped in BSS, it will be connected to the BSS charging point.
Swapping batteries are represented by the red spots, while the free spots (blue spots) can make any decision:
charging, discharging, or nothing. Regarding swapping, the aggregator does not know when the EV visits
the BSS, because it depends on a random scenario. However, it can decide if it accepts or not the swapping
battery. This BSS model assumes that vehicles arrive with depleted batteries. Nevertheless, the EV BSS
matrix can also quantify different SOCs. For example, for the Toyota brand that requires 8 hours of charge,
the SOC can be replaced by a different value between one and seven for the remaining hours. The arrival of
vehicles with different SOCs into the BSS- can be further explored in future research.

5. RESULTS AND DISCUSSION

This section develops the case study and the results are discussed after its formulation.

5.1. Development of the case study

The structure of the problem is based on the modified case study proposed in the IEEE-
CEC/GECCO 2019 competition [28]. This case study is carried out in Matlab 2016. Figure 5 shows the SMG
components. Therefore, a 26-bus SMG has 5 DGs, 1 external supplier, 1 PV generation, 15 swapping
batteries for EVs, 34 EVs, 90 loads with DR, and 2 ESSs. Moreover, the wholesale and local markes, and
electricity spot price for EVs are considered. The capacity of energy resources is represented in Table 4.
Prices of the wholesale and local markets are taken from [28], while the prices for EVs are taken from
the New South Wales (NSW) electricity market given in monetary units (m.u.) per kWh [30].
The capacity of energy resources is representedn Table 4. Prices of the wholesale and local markets
are taken from [28], while the prices of the EVs market are taken from the New South Wales (NSW)
electricity market given in monetary units (m.u.) per kWh [30]. EVs are selected based on the market reach
of top EV manufacturers. Under this context, three types of vehicles are chosen: Toyota, Nissan, and
Mitsubishi. In Table 5, surveys of EVs sales allow to compute the probability of EV transit [29].
Additionally, the battery characteristics are provided in [36].

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)

5104  ISSN: 2088-8708

External
PV 5 DGs G
supplier

4
23 25 18 21 14 15 16 17 11 26
BSS

22 24 19 20 3 6-9
12 13 10 5

90 Controllable loads ; 34 Electric vehicles; 2 Energy storage systems;

26-Bus Microgrid; 15 Swapping batteries to EVs

Figure 5. Representation of a 26-bus SMG

Table 4. Energy resources [28]

Energy resources Prices (m.u.*/kWh) Capacity (kW) Units
Dispatchable DGs 0.07-0.11 10-100 5
External suppliers 0.074-0.16 0-150 1
Charge - 0-16.6
ESS:
Discharge 0.03 0-16.6 2
Loads with DR 0.0375 4.06-8.95 90

Forecast (kW)
Photovoltaic - 0-106.81 1
Load - 35.82-83.39 90

Limits (kW)
Market 1 (WS) 0.021-0.039 0-100 1
Market 2 (LM) 0.021-0.039 0-10 1
Electricity spot price (EVs) 0.014-0.119 0.0476-4.076 1
* Monetary units (m.u)

Table 5. Characteristic and transiting probability

Price sales Percentage Cars Battery capacity (𝜐 𝛽 ) Charger rate (𝜐 𝛼 )
[29] (%) (kWh) [36] (kW)
Toyota 27,595 59 20 10 1.25
Nissan 14,715 32 11 23 [34] 3
Mitsubishi 4,166 9 3 16 1.78
Total 46,476 100 34

5.2. Discussion of case study results

In this study case, the SMGs are optimized through the VNS-DEEPSO algorithm with 100
probabilistic scenarios. The cost is reduced by 72 % in the overall cost, which is obtained by calculating
the reduction percentage of the random solution with respect to the suboptimal solution. Additionally,
according to [9], the VNS algorithm is selected (simple version) and the DEEPSO algorithm is formed by
the following parameters: the mutation rate, the communication probability, the population and the local
DEEPSO search probability as 0.8, 0.8, 2 and 0.1, respectively. The number of evaluations is restricted to
a maximum of 50,000 [28].
Results are summarized in Figure 6, where the generation is presented for a period of 24 hours
(in day-ahead markets). For this period, the maximum generation is supplied by the external supplier. DG 5
has relevant participation followed by PV generation. The other generators (DG 1, DG 2, DG 3, and DG 4)
have a lower participation rate. To summarize, the generation presents a stochastic generation. Load with DR
is calculated by using a coefficient 10 from (9), as suggested by [32]. However, it is adjusted to 20 to avoid
consumption equal to zero. The peak consumption in Figure 7 is 0.0722 kW at midday. Other peaks are
found at 10 a.m., 7 p.m., and 9 p.m. with consumption of 0.0147 kW, 0.0198 kW, and 0.0117 kW,
respectively. In [9], they report that the load is reduced because it is more convenient for the aggregator to
carry out load shedding.

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5105

EV batteries have events such as swapping, charging, and discharging as shown in Figure 8. A BSS
of EVs has 15 EV batteries and it has scheduled visits of 34 EVs at any given time. The EV batteries can be
partially charged. Figure 8 shows a smooth increase in the number of battery discharge at 2 a.m. and
the number of discharging batteries decreases close to zero at 9 a.m. Afterward, from 2 p.m. the number of
discharging batteries increases smoothly until 11 p.m. and ends steadily.
The number of swapping batteries have stochastic behavior. It depends on the EVs visit and the EVs
batteries that are accepted. The number of charging EVs batteries increases from 2 a.m. to 5 a.m. as shown in
Figure 8. Then, it remains steady until 7 a.m. and decreases close to zero at 1 p.m. ESS 1 as shown in
Figure 9 has a discharging peak with 0.1986 kW at 4 a.m. and a charging peak with 0.3144 kW at 9 p.m.
ESS 2 has lower participation, presumably due to the use of EVs. In addition, [22] reports ESS 2 with
a suboptimal solution of 0.87 kW. Transfer power in electricity spot price for EVs is computed as result of
EVs matrix per average charger rate (19). This equation takes into account the average type of electric
vehicle and type of charger rate.

Figure 6. Power generated for 24 hours Figure 7. Consumption time of load with DR

Figure 8. EV battery event programming for Figure 9. Planning of the energy storage system
an average of 100 scenarios for one day

𝑁𝑠 min(𝜐𝛼 ,𝜐𝛾 )∗𝑒𝑣

∑𝑒𝑣=1
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑐ℎ𝑎𝑟𝑔𝑒𝑟 𝑟𝑎𝑡𝑒 = (19)
𝑁𝑒𝑣

Electricity markets in Figure 10 show a tendency to sell and not to buy energy for higher profits.
The wholesale market has the higher participation with transactions close to 85 kW and some fluctuations.
The lowest energy sale takes place at 12 p.m. Sales in the local market have fluctuations around 36 kW.
The electricity spot price for EVs carries out a small percentage of transactions per day. In the ESS, this is
Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)
5106  ISSN: 2088-8708

highlighted when there are batteries charging and discharging. The transfer power is not carried out in
the EVs market, but it is performed between the same batteries of the BSS. Therefore, the energy transaction
of electricity spot price for EVs shown in Figure 10 is lower.

Figure 10. Electricity markets

6. CONCLUSIONS
The optimization of EVs in SMG with uncertain scenarios has not been fully explored, thus this
study proposes a new methodology that involves uncertainty decisions in battery swapping stations.
In the proposed method, a decision matrix is formulated to schedule EVs BSS. Furthermore, the stocastic
visit is based on the K-means clustering approach. Once the visit schedule to EV BSS has been generated,
idle spaces appear in the probabilistic scenarios within a decision matrix. This means that the idle spaces are
generated when the charge of a battery must be mandatory.
As a result, optimization is only allowed for free spaces, that is, charging, discharging, or no action.
In addition, other elements of the SMG are included in the model, such as DG, PV generation, ESS, loads
with DR, and day-ahead markets. Finally, the best schedule of charging, discharging, and swapping for EVs
batteries in BSS, and the transfer of power in the market are identified satisfactorily. The generation, loads
with DR, and ESS are stochastic, however, the patterns are recognized of the main generators, peak hours in
loads with DR (at midday), and ESS participation.

APPENDIX
Supplementary Materials
- The SMG is available online at: http://www.gecad.isep.ipp.pt/WCCI2018-SG-
COMPETITION/ and http://www.gecad.isep.ipp.pt/ERM2019-Competition/
- EV brands are available online at: https://evadoption.com/ev-sales/evs-percent-of-vehicle-sales-by-
brand/. Electricity spot price for EVs is available online at:
https://www.aemo.com.au/Electricity/National-Electricity-Market-NEM/Data-dashboard

REFERENCES
[1] H.-Y. Mak, Y. Rong, and Z.-J. M. Shen, “Infrastructure Planning for Electric Vehicles with Battery Swapping,”
Manage. Sci., vol. 59, no. 7, pp. 1557-1575, Jul. 2013.
[2] L. Zhang, S. Zhou, J. An, Q. K.-I., “Demand-Side Management Optimization in Electric Vehicles Battery
Swapping Service,” IEEE Access, vol. 7, pp. 95224-95232, 2019.
[3] H. Liu, Y. Zhang, S. Ge, C. Gu, and F. Li, “Day-Ahead Scheduling for an Electric Vehicle PV-Based Battery
Swapping Station Considering the Dual Uncertainties,” IEEE Access, vol. 7, pp. 115625-115636, 2019.
[4] J. Yang, F. Guo, and M. Zhang, “Optimal planning of swapping/charging station network with customer
satisfaction,” Transp. Res. Part E Logist. Transp. Rev., vol. 103, pp. 174-197, 2017.
[5] A. Valibeygi, et al., “Robust Power Scheduling for Microgrids with Uncertainty in Renewable Energy Generation,”
2019 IEEE Power Energy Soc. Innov. Smart Grid Technol. Conf., pp. 1-5, Feb. 2019.
[6] W. Infante, J. Ma, X. Han, and A. Liebman, “Optimal Recourse Strategy for Battery Swapping Stations
Considering Electric Vehicle Uncertainty,” IEEE Trans. Intell. Transp. Syst., pp. 1-11, 2019.

Int J Elec & Comp Eng, Vol. 10, No. 5, October 2020 : 5093 - 5107
Int J Elec & Comp Eng ISSN: 2088-8708  5107

[7] B. Sun, X. Sun, D. Tsang, and W. Whitt, “Optimal Battery Purchasing and Charging Strategy at Electric Vehicle
Battery Swap Stations,” Eur. J. Oper. Res., vol. 279, no. 2, pp. 524-539, 2019.
[8] J. Garcia-Guarin, S. Rivera, and H. Rodriguez, “Smart grid review: Reality in Colombia and expectations,” IOP
Conf. Series: J. Phys. Conf. Ser., vol. 1257, 012011, pp. 1-7, Jun. 2019.
[9] J. Garcia-Guarin et al., “Smart Microgrids Operation Considering a Variable Neighborhood Search:
The Differential Evolutionary Particle Swarm Optimization Algorithm,” Energies, vol. 12, no. 16, pp. 1-13,
Aug. 2019.
[10] J. Garcia Guarin, F. Lezama, J. Soares, and S. Rivera, “Operation scheduling of smart grids considering stochastic
uncertainty modelling,” Far East J. Math. Sci., vol. 115, no. 1, pp. 77-98, 2019.
[11] J. Arévalo, F. Santos, and S. Rivera, “Uncertainty cost functions for solar photovoltaic generation, wind energy
generation, and plug-in electric vehicles: mathematical expected value and verification by Monte Carlo simulation,”
Int. J. Power and Energy Convers., vol. 10, no. 2, pp. 171-207, 2019.
[12] Y. Cheng and C. Zhang, “Configuration and operation combined optimization for EV battery swapping station
considering PV consumption bundling,” Prot. Control Mod. Power Syst., vol. 2, pp. 1-18, Dec. 2017.
[13] T. Li et al., “An optimal design and analysis of a hybrid power charging station for electric vehicles considering
uncertainties,” in Proceedings: IECON 2018 - 44th Annual Conference of the IEEE Industrial Electronics Society,
pp. 5147-5152, 2018.
[14] J. Li, X. Tan, and B. Sun, “Optimal Power Dispatch of a Centralized Electric Vehicle Battery Charging Station with
Renewables,” IET Communications, vol. 12, no. 5, pp. 579-585, 2018.
[15] M. R. Sarker, H. Pandžić, and M. A. Ortega-Vazquez, “Optimal operation and services scheduling for an electric
vehicle battery swapping station,” IEEE Trans. Power Syst., vol. 30, no. 2, pp. 901-910, Mar. 2015.
[16] M. Hemmati, M. Abapour, B. Mohammadi-ivatloo, “Optimal scheduling of smart Microgrid in presence of battery
swapping station of electrical vehicles,” Electric Vehicles in Energy Systems, Springer, pp. 249-267, 2020.
[17] M. Hemmati, S. Ghasemzadeh, and B. Mohammadi-Ivatloo, “Optimal scheduling of smart reconfigurable
neighbour micro-grids,” IET Gener. Transm. Distrib., vol. 13, no. 3, pp. 380–389, Feb. 2019.
[18] M. Hemmati, B. M.-Ivatloo, M. Abapour, and A. Anvari-Moghaddam, “Optimal chance-constrained scheduling of
reconfigurable microgrids considering islanding operation constraints,” IEEE Systems Journal, pp. 1-10, 2020.
[19] F. Moaidi and M. A. Golkar, “Demand response application of battery swap station using a stochastic model,”
in 2019 IEEE Milan PowerTech, PowerTech 2019, pp. 1-6, 2019.
[20] W. Sutopo et al., “A technical review of BMS performance standard for electric vehicle applications in Indonesia,”
TELKOMNIKA (Telecommunication, Computing, Electronics and Control), vol. 16, no. 2, pp. 544-549, 2018.
[21] P. J. García-Guarín, et al., “Implementación del algoritmo VNS-DEEPSO para el despacho de energía en redes
distribuidas inteligentes - Implementation of the VNS-DEEPSO algorithm for energy dispatch in intelligent
distributed networks (in English),” INGE CUC, vol. 15, no. 1, pp. 142-154, 2019.
[22] D. Dabhi and K. Pandya, “Enhanced velocity differential evolutionary particle swarm optimization for optimal
scheduling of a distributed energy resources with uncertain scenarios,” IEEE access, vol. 8, pp. 27001-27017, 2020.
[23] E. de V. Fortes, et al., “A VNS algorithm for the design of supplementary damping controllers for small-signal
stability analysis,” Int. J. Electr. Power Energy Syst., vol. 94, pp. 41-56, 2018.
[24] P. Hansen and N. Mladenović, “Variable neighborhood search: Principles and applications,” Eur. J. Oper. Res.,
vol. 130, no. 3, pp. 449–467, May 2001.
[25] V. Miranda and R. Alves, “Differential evolutionary particle swarm optimization (deepso): A successful hybrid,”
IEEE BRICS Congr. Comput. Intell. 11th Brazilian Congr. Comput. Intell., pp. 368-374, 2013.
[26] D. G. Mayer, B. P. Kinghorn, and A. A. Archer, “Differential evolution – an easy and efficient evolutionary
algorithm for model optimisation,” Agric. Syst., vol. 83, no. 3, pp. 315–328, Mar. 2005.
[27] F. Marini and B. Walczak, “Particle swarm optimization (PSO): A tutorial,” Chemom. Intell. Lab. Syst., vol. 149,
pp. 153–165, Dec. 2015.
[28] F. Lezama, J. Soares, Z. Vale, and J. Rueda, “Evolutionary computation in uncertain environments: a smart grid
application,” IEEE World Congress on Computational Intelligence 2018 – WCCI 2018, 2018.
[29] Automaker, “EVAdoption,” Analyzing key factors that will drive mass adoption of electric vehicles, 2019.
Accessed: 29 Dec 2019. [Online]. Available: https://evadoption.com/ev-sales/evs-percent-of-vehicle-sales-by-
brand/.
[30] AEMO, “Electricity price and demand,” Australian Energy Market Operator-, 2019. Accessed: 29 Dec 2019.
[Online]. Available: https://www.aemo.com.au/Electricity/National-Electricity-Market-NEM/Data-dashboard.
[31] J. P. Fossati, “Revisión bibliográfica sobre micro redes inteligentes,” Lit. Rev. microgrids, vol. 9, pp. 13-20, 2011.
[32] J. Garcia-Guarin, S. Rivera, and L. Trigos, “Multiobjective optimization of smart grids considering market power,”
J. Phys. Conf. Ser., vol. 1409, 012006, Nov. 2019.
[33] M. Mahoor, Z. Hosseini, A. K., “Least-cost operation of a battery swapping station with random customer
requests,” Energi, Elsevier, vol. 172, pp. 913-921, Apr. 2019.
[34] M. Yilmaz and P.T. Krein, “Review of battery charger topologies, charging power levels, and infrastructure for
plug-in electric and hybrid vehicles,” IEEE Transactions on Power Electr., vol. 28, no. 5. pp. 2151–2169, 2013.
[35] J. Soares, et al., “Two-Stage Stochastic Model Using Benders’ Decomposition for Large-Scale Energy Resource
Management in Smart Grids,” IEEE Trans. Ind. Appl., vol. 53, no. 6, pp. 5905–5914, Nov. 2017.
[36] J. Ferreira, “Mobi-System: towards an information system to support sustainable mobility with electric vehicle
integration,” Doctoral Thesis, University of Minho, Portugal, 2013.

Optimal scheduling of smart microgrids considering electric vehicle… (J. Garcia-Guarin)