You are on page 1of 10

19236 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 23, NO.

10, OCTOBER 2022

Fuzzy Logic Strategy for Priority Control of


Electric Vehicle Charging
George Cristian Lazaroiu , Senior Member, IEEE, and Mariacristina Roscia , Member, IEEE

Abstract— The study of future scenarios where transport proposed in [5]–[7]. A fuzzy-logic power flow controller is
systems are highly based on electric propulsion constitutes a key proposed to limit the impact of the plug-in electric vehicles
point for the world sustainable development. In particular, the on the mains supply in [8]. Fuzzy logic controller designed for
integration between electric vehicles and smart grids makes pos-
sible reaching new goals on energy savings, efficiency of energy charging/discharging electric vehicles at various buses in a test
generation and distribution. This paper proposes a controller system such that to realize voltage control and peak shaving is
integrated into the electric vehicle, based on fuzzy logic, providing proposed and implemented in [9]. A fuzzy logic proportional-
the car user with the necessary information regarding the integral controller is implemented in an islanding microgrid
charging/discharging stations located nearby. In particular, for to control the storage systems and electric vehicles to tackle
each station a priority level is reported, representing the degree
of economic convenience/necessity that the user has in using this the frequency regulation challenge [10]. Fuzzy logic controller
charging station. The final objective is to easily integrate a fleet of is proposed to control the vehicle-to-grid exchanged power to
electric vehicles into smart grids with users willing to charge and avoid voltage oscillations in the grid supplying the electric car
discharge their electric vehicles during periods and modalities in [11].
that highly favors the smart grid operation, energy savings, and Fuzzy logic controllers at distribution network level and at
economic efficiency for the car owners.
charging stations level to control the exchanged power between
Index Terms— Controller, electric vehicles, fuzzy rules, smart electric vehicles and multiple charging stations and to analyze
cities, sustainable development. the impact of communications are implemented in [12]. The
electric vehicles connected in an electrical grid managing their
I. I NTRODUCTION
charging time and supporting the mains supply through their

T HE transportation sector is one of the major sources of


pollution, as well as one of the most energy-intensive
sectors linked to human activity. For reducing the environ-
storage energy function of electricity price in presence of fuzzy
controllers is proposed in [13]. A fuzzy logic controller for dis-
tributing the aggregated power of electric vehicles among these
mental impact, new solutions for transportation systems were ins a vehicle-to-grid charging strategy is applied in [14]. Fuzzy
proposed and introduced to the market, like the electric vehi- logic controller, for power distribution between the energy
cles (EVs). The growing diffusion of this type of vehicles has sources of a plug-in hybrid electric vehicle, to determine the
led to investigate their potential for energy savings, as well best operating point of each component, to meet the driver’s
as for integration with other rapidly developing technologies, demand and to maintain the state-of-charge within admissible
like renewable energy sources (RESs) and smart grids (SGs). limits is proposed in [15]. Controller based on fuzzy logic for
A neuro-fuzzy controller for controlling the cruise to main- a fleet of electric vehicles in a distribution system considering
tain constant time between two semiautonomous vehicles is mains supply voltage profile and the electric vehicles state-
proposed and implemented in [1]. The fuzzy logic is used for of-charge is implemented in [16]. The fuzzy logic controller
routing path algorithm of electric vehicles accounting for car- ensures peak shaving, avoiding under voltages and realizing
bon emissions with centralized control difficult to implement share charging among the vehicles in the grid with distributed
for different drivers [2], price sensitive EV charging scheduling generators. An analysis of impact of fuzzy logic controller,
in regulated market from the perspective of charging station used for power sharing among engine and electric machine
operators [3], appropriate pricing rates for EV charging to in a hybrid electric vehicle, on battery lifetime and fuel
accommodate drivers economic convenience [4]. Fuzzy logic economy is conducted in [17]. The authors use the state-of-
controllers at charging station and at distribution bus levels charge level, the electric motor torque and temperature for the
implemented to realize the management of energy flows are rule-based strategy, reducing the fuel consumption and battery
capacity losses. A comparison of fuel economy of three hybrid
Manuscript received 28 August 2021; revised 31 December 2021; accepted
14 March 2022. Date of publication 31 March 2022; date of current version power systems and two adaptive energy management systems
11 October 2022. The Associate Editor for this article was J. Catalão. of plug-in hybrid commercial logistic vehicles is carried on
(Corresponding author: George Cristian Lazaroiu.) in [18]. The fuzzy logic controller energy management system
George Cristian Lazaroiu is with the Department of Power Systems, Uni-
versity POLITEHNICA of Bucharest, 060042 Bucharest, Romania (e-mail: has as inputs the state-of-charge and required power, while the
cristian.lazaroiu@upb.ro). output is the engine power.
Mariacristina Roscia is with the Department of Engineering and A fuzzy logic controller implemented for a plug-in hybrid
Applied Sciences, University of Bergamo, 24044 Dalmine, Italy (e-mail:
cristina.roscia@unibg.it). electric vehicle with inputs the battery state-of-charge, drive
Digital Object Identifier 10.1109/TITS.2022.3161398 cycle power, and the unintentional switching on/off of the
1558-0016 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
LAZAROIU AND ROSCIA: FUZZY LOGIC STRATEGY FOR PRIORITY CONTROL OF ELECTRIC VEHICLE CHARGING 19237

internal combustion engine, is proposed to integrate the


vehicle-to-home and home-to-vehicle capabilities in [19].
Fuzzy logic energy control for realizing power sharing
between batteries and ultracapacitor in a hybrid electric vehicle
is proposed in [20]. A fuzzy logic controller with demanded
power, derivation of demanded power and battery state-of-
charge and implemented for a fuel cell-based hybrid electric Fig. 1. Key role of fuzzy logic.
vehicle for the energy management system of the installation is
proposed in [21]. An intelligent energy management system The classification of a variable with fuzzy logic allows
with fuzzy logic controller of a hybrid electric autonomous a different approach, more qualitative than quantitative. The
vehicle to modify its torque function of battery state of charge fuzzy logic allows passing from one class to the adjacent ones,
and throttle signal of the motor is proposed in [22]. A fuzzy not fixed but fuzzy, more closely to the human behaviour. This
logic control for gear shifting and power management between consideration determined the use of fuzzy logic. The black box
combustion engine and electric motor was applied on a plug-in of Fig. 1, which represents what is between the input space and
hybrid electric vehicle through simulations in [23]. The power the output space, can contain linear systems, neural networks,
management of a battery-supercapacitor system coupled with a differential equations, and many others. The advantage of
permanent magnet synchronous motor to ensure through fuzzy using the fuzzy logic is represented by the possibility for
logic controller an optimal load power sharing is proposed and automating the process of decision making, finalized with opti-
analyzed in [24]. mizing the operating conditions. This is in general managed
The current paper is proposing a controller on the by human operators based on their knowledge and experience,
“end-user side”, based on fuzzy logic integrated within the through a modeling system used for representing the vague
electric vehicle. Based on the information from charging nature and not precisely quantifiable of expressions used for
stations (costs, charging time, etc.), the end user can evaluate, human reasoning. Ideally would be to have all information
function of the necessities, to each charging station to head. that characterize the optimum choice: speed limits, territory
The current paper deals with the charging strategies function topography, street layer characteristics, traffic situation, state
of management, of optimizing in choosing the charging station of charge of the battery, etc. The proposed strategy accounts
most suitable for the vehicle based on the distance and choices for the power demand and other information obtained in real
of the user, efficiently contributing to the management of the time like SoC, analysis of the driver commands, efficiency,
charging and discharging processes and power exchanges with variation of the charging price etc. The control system evalu-
the mains supply. In this paper, in addition to the variables ates all incoming parameters and decides how to choose the
bus-voltage and state of charge (SoC), the choice exercised charging point among more charging stations, such that to
by the vehicle owner function of some determining factors is realize an optimum choice function of all input variable like
considered, as: distance, price, SoC, etc. In this way the energy flows with
• the maximum price that the user is willing to pay for
the upstream grid could be optimized.
charging the EV, Pcmax ;
III. P ROPOSED C ONTROLLER D ESIGN
• the minimum price that the user is willing to receive for
discharging the EV to support the grid, Pdmax ; Many analyses of EVs’ operation in the distribution system
• the minimum SoC that the user is willing to reach when made the hypothesis that, at given time instant, the fleet of
the EV interfaces with the mains supply, SoCmin . EVs is connected to the mains supply. However, in a real
The proposed controller design, with the help of fuzzy logic, case, a peak of production or demand can occur at a random
was implemented in a straight manner in Matlab/Simulink. time instant. The question to answer is how to make the fleet
The membership functions used in this paper are triangular of EVs, which could help in peak shaving, to be connected
and trapezoidal ones. The controller efficiency is demonstrated to the mains supply at that time instant? The fact that a fleet
by the conducted case studies. The controller could be in of EVs is connected to the mains supply depends on a large
the future integrated with genetic algorithms and applied to number of variables, and the displacement of the EVs can also
electric vehicles and choosing the charging station within the be considered random.
vicinity, function of prechosen parameters and imposed within In addition to the study and the forecast of EVs move-
the on-board navigator. ment, there is also the possibility to incentive/discourage
users heading towards the charging stations by providing real-
time information on the distance, and on the current and
II. F UZZY L OGIC FOR C HARGING E LECTRIC V EHICLES forecasted charging/discharging prices. By using also the data
Some traditional control strategies for electric vehicles are concerning the SoC, the onboard computer capable to provide
based on monitoring variables and using the Boolean logic to the aforementioned information can constitute a tool able,
determine operating parameters, defined with pre-established if not to govern, at least to heavily address user behavior.
fixed threshold values. For example, switching to traditional The assumption that there are no cyber threats in the network
supply when the SoC decreases below a certain value and to influence the decisions is made [25].
commuting to battery when SoC increases over a certain value The proposed FLC integrated in the satellite navigator of
in hybrid vehicles. the electric vehicle is delivering to the user indications on

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
19238 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 23, NO. 10, OCTOBER 2022

The source of these input variables changes function of the


variable. The SoC and D are internal data of the electric
vehicle, while Pc and Pd are data to be provided instant
by instant by the station owner.
1) Distance Variable: The variable D represents the dis-
tance, in [km], between the current location of the electric
vehicle and the position of the charging station. The distance
is calculated based on the route length showed by the satellite
navigator equipping the electric vehicles. Two operating cases
can be considered function of the user imposing or not a
navigation route. Former, the on-board computer knows the
Fig. 2. Layout of proposed fuzzy logic controller.
route chosen by the user and the distance between electric
vehicle and charging station. In addition, the distance between
the charging stations available nearby. In particular, for each
a random point of the ongoing itinerary and the charging
charging station a priority level “priority”, P, is signalled,
station can be considered: a charging station located distantly
representing the degree of economic convenience and oper-
at a certain moment, will become a nearby station during the
ation necessity that the user has in heading towards the
navigation route and can represent a possible charging station
charging station. The final purpose is to make a fleet of
for the user. Lastly, as on-board computer does not know the
electric vehicles integrating in a virtuous manner in a smart
route chosen by the user, the considered distance is the one
grid. The user would be headed to charge and discharge
that instant to instant is between the location of electric vehicle
their vehicles at time instants and modalities to support the
and position of charging station.
electric network, that determine energy savings and economic
The distance D is determined by the navigator for a set
advantages for the end users. Thus, the probability for two
of charging stations, resulting in a plurality of distances. The
vehicles to be at the same distance, with the same state of
question to be answered is: for which charging station, the
charge, same requests and constraints is limited. However, the
distance D as input for the controller should be evaluated
controller would be integrated with the reservation system for
considering all other input variables? Hence, the distance D
the charging station and with the control system of the car
to be considered in determining the priority P must fulfill a
and the internet application. This typology of procedure is fast
constraint on maximum admissible distance. The parameter
and the user, in few instants, can know if the made reservation
for the maximum permissible distance is denoted Dmax and
is successfully concluded. The FLC is used to reproduce the
considered as 5 km. This value is considered to be reason-
objective “answer” of an EV user when it is demanded: which
able for cities, providing priority information to the user.
degree of attractivity would you give to this charging station in
An electric vehicle user would hardly find reasonable heading
the current case? The degree it is supposed to be expressed in
to a charging station located more than 5 km away. For a
a normalized range between zero and ten, where zero represent
charging station located far away, even with very advantageous
lowest attractivity and ten the highest. For objective judgement
charging/discharging rates, it would be not reasonable for the
or answer, only operation convenience is considered, and not
user to deviate from the chosen itinerary. In addition to a
sympathy towards an EV operator. The proposed fuzzy model
maximum admissible distance, a minimum admissible distance
is a human objective simulator, a decision-making process with
Dmin was chosen as 0.1 km. This lower limit was chosen
elements weighted and reported among them for support to the
considering that 0.1km is a distance such that, if there is
specific EV user.
a charging station nearby, the user would already be aware
Thus, the user is continuously informed on the degree of
of it. This nearby station would be already signaled and
economic convenience/operation necessity that has in heading
at 0.1km the user would have this location in his visual
towards a nearby charging station and can decide based on own
field.
requirements, which often coincides with the requirements of
The input variable D and its max/min levels are used to
electrical grids.
define the relative membership function. As shown in Fig. 3,
The design of the fuzzy set requires establishing the input
the range [Dmin , Dmax ] has been divided into three fuzzy sets
and output variables. Four input variables were defined,
represented by the membership functions Short, Medium, and
respectively:
Long. Two trapezoidal and one triangular curves were chosen,
• distance (D);
symmetrical with respect to the range of D. The subdivision
• state-of-charge (SoC);
has the following characteristics: starting from the minimum
• variation of the charging price Pc ;
limit of 0.1km, the membership function Short extends up to
• variation of the charging price Pd .
2.06km, but overlaps with the membership function Medium
A single output variable was defined, respectively the priority at 1.08km. The membership function Medium extends up to
P. Fig. 2 illustrates the layout of the fuzzy logic controller. 4.02km, but overlaps with the membership function Long at
3.04km. The membership function Long extends up to 5km.
A. Input Variables The fact that the membership functions overlap, that is that
The input variables are the required data for the controller there are no well-defined outlines, is very representative of
for the fuzzy inference process to generate the output variable. the concept of fuzzy sets that underlies the fuzzy logic.

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
LAZAROIU AND ROSCIA: FUZZY LOGIC STRATEGY FOR PRIORITY CONTROL OF ELECTRIC VEHICLE CHARGING 19239

Fig. 3. Membership functions for D.


Fig. 5. Membership functions for Pc .
TABLE I
PART OF THE S ERIES W ITH R EPETITIONS OF L INGUISTIC VALUES

Fig. 4. Membership functions for SoC.

2) State of Charge: The variable SoC is expressed in per-


centage of the available residual energy. The variation interval
for this variable is established based on practice operating
assumptions. The maximum value SoCmax is 100%, while the required. The variation interval [Pcmin , Pcmax ], was divided
minimum limit SoCmin was established at 5%. A minimum in three fuzzy sets represented by the membership functions
limit of 0% would not make sense to be considered, as in that LowPositive, MediumPositive, and HighPositive, as illustrated
instant, the EV would not have the required energy for heading in Fig. 5. Starting from the minimum limit of 1%, the
towards a charging station. The variation interval [SoCmin , LowPositive extends up to 12.6% and begins to overlap the
SoCmax ] was divided into three fuzzy sets represented by the MediumPositive at 6.8%. The MediumPositive extends up to
membership functions: Low, Medium, and High, as illustrated 24.2%, but overlaps the HighPositive at 18.4% and then
in Fig. 4. The two trapezoidal and one triangular curves are extends to the maximum limit of 30%.
symmetrically located with respect to the SoC range. Starting 4) Variation of the Discharging Price: Analogous definition
from the minimum limit of 5%, the Low membership function can be made also for the variable Pd , expressed as
extends up to 43% and it begins to overlap the Medium
Pd = (Pd − Pdmin )/Pd · 100,
membership function at 24%.
The Medium one extends up to 81%, but begins to overlap where Pdmin is the minimum discharging price that the user
the High membership function at 62% and then extends to the is willing to accept, and Pd is the discharging price offered
maximum limit of 100%. by the station owners at a given time instant. The user
3) Variation of the Charging Price: The variable Pc is can discharge the EV at an energy price that is economic
measured in percentage and is expressed as: advantageous as higher the value of Pd . The interval of
variation [Pdmin , Pdmax ] and membership functions of these
Pc = (Pcmax − Pc )/Pc · 100,
variable are similar with those of Pc , as illustrated in Fig. 6.
where Pcmax is the maximum charging price that the user
B. Output Variables
is willing to accept and Pc is the charging price offered
by the station owner at a given time instant. As previously The output variable is the priority P. The variable P [%]
defined, Pc represents the standardized difference, expressed represents the degree of economic convenience/necessity of
as percentage between the maximum charging price the user is the user heading to a certain charging station. The variation
willing to accept and the charging price offered by the station interval of the priority variable P was set between 100% and
owners. Therefore, the user is charging the EV at an energy 0%. The variation interval [Pmin , Pmax ], was divided into three
price that leads to economic savings higher the value of Pc . fuzzy sets represented by the membership functions: Low,
The variation of Pc is between Pcmax , equal to 30%, Medium, and High, as shown in Fig. 7. Two trapezoidal and a
and Pcmin , equal to 1%. The maximum limit was chosen triangular curve were chosen, symmetrical with respect to the
considering that Pc depends on the fluctuations of Pc and, variation interval of P. Starting from the minimum value of
in particular, of the Pcmax chosen by the user. This latter 0%, the Low membership function extends to 40%, superpos-
value depends on the free choice of a user and could vary ing to Medium one at 20%. The Medium membership function
substantially. The 30% value was chosen considering for extends to 80% and superposes to High membership function
price volatility and the limit values. However, in most cases, at 60%, and then extends to the maximum limit of 100%.
it will be difficult to reach values close to this maximum.
The minimum limit of 1% was set for excluding the 0% C. Fuzzy Rules
values, as the economic advantage for the customer would The fuzzy rules are the core elements of the fuzzy controller.
miss and the operation of fuzzy logic controller would not be These rules define the way in which the controller manages

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
19240 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 23, NO. 10, OCTOBER 2022

• as Pc increases, P increases: as the advantage on the


charging price augments, regardless of the distance and
SoC, the utility of the user heading towards charging
station increases;
• as Pd increases, P increases: as the advantage on the
discharging price augments, regardless of the distance
and SoC, the utility of the user heading towards charging
Fig. 6. Membership functions for Pd . station increases.
Once the variable and fuzzy rules were defined, the design
of the fuzzy logic controller can be considered completed.
The method chosen for aggregating the outputs of the fuzzy
rules is the “maximum” function and the method used for
defuzzification is the barycenter method.

IV. C ONTROLLER T ESTS AND S IMULATIONS


The simulation model was built around the main “Fuzzy
Fig. 7. Membership functions for P.
Logic Controller (FLC)” block which integrates the newly
designed FLC, as shown in Fig. 8. The block was created as a
the input variables and determines the output variable value. simulation environment suitable to provide instant by instant
In the first stage, the number of fuzzy rules for covering all to the FLC the data corresponding to the input variables and
operation cases and for managing the fuzzy controller reaction to store the data corresponding to the output variable.
in a predictable way to all possible cases was defined. The total
number of fuzzy rules is N F R = Nc1 × Nc2 × . . . × Nc j , where
A. Model Description
N F R is the number of fuzzy rules and Nc1 , Nc2 , . . . , Nc j are
the numbers of membership functions for each input variable. The input variables, whose data are stored in an external file,
For the current case, the set of 81 fuzzy rules is defined as are submitted to the controller. This file contains the simulation
N F R = NcNv , and Nv is the number of input variables. time T expressed in seconds, as well as the D [km], Pc [%],
For defining the fuzzy rules, an explicit writing of all ele- Pd [%], SoC [%]. As illustrated in Fig. 8, first of all there
ments of the series with repetitions is required. Table I reports is the block assigned to provide instant by instant the data
a part of the list of elements of the series with repetitions, corresponding to the four input variables.
which represent all possible configurations of linguistic values The four input variables are multiplexed into a single vector
that input variables may assume. entering the fuzzy logic controller. The input variable signals
The rule editor is formulating the fuzzy rules. The fuzzy are also sent to a conditional block that implements an if -else
rules to be defined where the ones that have in precedency logic. Considering generic inputs u1 , u2 , u3 , and u4 , the block
the union of fuzzy sentences and input variable and thus, has the conditions:
as consequence, a sentence fuzzy for a unique output variable. (u 1 ≥ 0.1 & u1 ≤ 5) & (u 2 ≥ 1 & u 2 ≤ 30)
The type of fuzzy rule chosen to be used has the following
expression: &(u 3 ≥ 1 & u3 ≤ 30) & (u 4 ≥ 5 & u 4 ≤ 100)
“If (D is Short) and (Pc is LowPositive) and (Pd is Thus, these expressions constraint the execution of the sub-
LowPositive) and (SoC is Low), then (P is Average)”. system “If action subsystem 1” to all variables having values
In addition, to each fuzzy rule a weight equal to 1 was within the ranges set during the design phase of the FLC. If an
assigned. Thus, all the fuzzy rules have the same importance in out-of-range value arrives at the controller, the output value is
determining the output value. For reporting in fuzzy terms the a random one or a meaningless value for the operation of
operator “and” and the fuzzy inference “then”, the “minimum” the system, as the functioning of the controller would become
function was chosen. unpredictable. If an out-of-range value arrives at the controller,
The fuzzy rules were defined based on the operation of the returned output in that time instant is filtered as it is has
smart grids with electric vehicles, on the constraints to be no meaning.
fulfilled, and on the necessities to satisfy for the safe operation The “Merge” block has the function of unifying the sig-
of entire systems. The applied base principles are: nals coming from the two subsystems, executed alternately.
• as D decreases, P increases: this is intuitively as shorter As shown in Fig. 8, the “subsystem 1” is executed when the
the distance between the EV and charging station, the If condition of “block 1” is fulfilled, while the “subsystem 2”
higher will be for the user the convenience to reach this is executed on the contrary. Thus, at a time instant, all input
station as less energy will be consumed, regardless of variables are within the imposed ranges, “subsystem 1” is
prices and SoC; executed, while if a single variable are exceeding the range
• as SoC decreases, P increases: regardless of prices and “subsystem 2” is executed. The two subsystems have as inputs
distance, the user must be protected against the risk of the same signals: the output of the controller and the signal
remaining without available energy; Pnull [%] (a constant signal of value 0).

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
LAZAROIU AND ROSCIA: FUZZY LOGIC STRATEGY FOR PRIORITY CONTROL OF ELECTRIC VEHICLE CHARGING 19241

Fig. 8. Block diagram representation of the simulation model of EV charging with FLC.

As shown in Fig. 8, the system operates as: to reduce the range for the input variables. This allows
• If at a given instant, the condition of block “If ” is to consider the case when the user is interested in having
fulfilled and the output of the controller is valid, the advantageous price higher than 15%. Thus, Pc and Pd
“subsystem 1” is executed. The input for the block vary from 15% to 30%, instead from 1% to 30%. To simulate
“Merge 1” is the output of the controller with the valid the operation of the controller under these conditions it is
P outgoing from “subsystem 1” and this one is sent to sufficient to modify the “textit if expression” as:
its output. In the same time, the signal outgoing from
(u 1 ≥ 0.1 & u1 ≤ 5) & (u 2 ≥ 15 & u 2 ≤ 30)
“subsystem 1” is a signal Pnull entering the “Merge 2”
block. Thus, in this condition, the output of block “Merge &(u 3 ≥ 15 & u3 ≤ 30) & (u 4 ≥ 5 & u 4 ≤ 100)
1” is Pok , while the output of ‘Merge 2” is Pnot −ok . The same operation can be done for other variables to
• If at a given instant, the condition of block “If ” is not
accommodate all possible operating cases.
fulfilled and the output of the controller is not valid, The input variables are combined in a structure with time
the “subsystem 2” is executed. The input for the block step of 1 second. Thus, the simulation is done in discrete
“Merge 2” is the output of the controller with the non- environment with fixed time step.
valid P outgoing from subsystem 2 and this one is sent A fixed-step solver, i.e. with a constant step, and a discrete
to its output. In the same time, the signal outgoing from type one, which solves the model at regular intervals of time
subsystem 2 is a signal Pnull entering the “Merge 1” from the beginning to the end of the simulation were chosen.
block. Thus, the output of block “Merge 2” is Pnot −ok , The characteristics of the solvers are well coupled with the
while the output of ‘Merge 1” is Pnull . structure of the data set provided in input (Table II) which
This type of system was required as, if at a given instant in fact is organized according to a discrete time structure
a subsystem is not executed, for the intrinsic operation of the with time steps of 1 s. Consequently, always within the
block, its output would be equal to the previous step. Thus, configuration parameters, a step size of 1 s was chosen.
the output of the subsystem would be a false value of P, that Finally, the simulation start time has been set to 0 s, while
does not highlight the real input conditions. When a subsystem the simulation end time varies according to the type of test to
is not executed, the output of block ‘Merge” is a null value be performed.
highlighting the not-validity of the output at that instant. Thus, The first test is aimed at analyzing the functioning of the
the system previously described determines the output of block FLC in the most extensive and complete way possible to verify
“Merge 1” the values Pok when the variation constraints are that it reacts in a predictable way in all situations and does
fulfilled, and Pnull when these constraints are not respected. not have unexpected behaviors. To this end, it was decided
Vice versa, the output of block “Merge 2” are the values to provide a set of data that covers the set of all possible
Pnot −ok when the variation constraints are not fulfilled, and combinations of values that the FLC can receive as input.
Pnull when these constraints are respected.
The output variables Pok of the block “Merge 1”, respec-
tively Pnot −ok of the block “Merge 2”, are stored for future B. Tests
manipulation. The “If ” block also offers the possibility to Tests were carried on to analyze the correct operation of the
expand the simulation capacity of this proposed model. controller by combining all input variables based on the data
By modifying the “if expression”, the user has the possibility reported in Table II.

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
19242 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 23, NO. 10, OCTOBER 2022

TABLE II
VALUES FOR THE I NPUT VARIABLES

Starting from these data set, four simulations were carried


on by:
• decreasing the values of D;
• decreasing the values of SoC;
• increasing the values of Pc ;
• increasing the values of Pd .
As illustrated in Fig. 9, the values of P increase as D
decreases, considering all present oscillations determined by
the continuous spacing of the other variables inside their own
ranges. As SoC is decreasing, the priority P is increasing.
The variation of P is less evident as the fuzzy rules were built
considering D as priority variable for determining P. As SoC
decreases, a higher density of P values in the upper part of its
variation range occurs. This trend is determined by evaluating
a system dependent on four variables. As the price advantages
Pc and Pd increase, the density of values in the upper
part of P range increases. In conclusion, the test demonstrated
that the fuzzy controller operates in the correct way over all
possible combinations of input values.
1) Test With Values Out of Range: This test analyzes how
the proposed model operates when there are input values that
do not fulfill the range established during the design phase.
The values for the input data are reported in Table III. Fig. 9. Variation of valid priority P function of the input variable.
The data reported in Table III was randomly generated,
with a value for each variable out of range. It can be seen network or function of extraordinary events, for encouraging
that at every time instant when a value out of range occurs, or discouraging users to direct towards a charging station.
the obtained value for P is filtered. This value is stored as During the current tests, the variations of Pc and Pd are
Pnot −ok . Instead of the value not valid for P, in the column not based on real values, but were alternatively increased or
Pok is memorized a cero value. This proves that at the ana- maintained constant. Thus, the incidence of each input variable
lyzed time instant, the fuzzy controller did not return a valid on the operation of fuzzy logic controller and determination
value. of P is investigated. For example, when an electric vehicle
2) Tests With Quasi-Realistic Variations of Input Variables: is approaching a charging station, the variables which are
The tests conducted for investigating the correct operation varied are:
of the fuzzy logic controller and the results were previously
• D linearly decreasing;
presented. Further, the evaluation of the controller operation
• SoC linearly decreasing;
when two or three variables are varying is conducted. This is
determined by the variation of these variable sin real practice. • Pc and Pd alternatively increasing or maintained
The hypothesis of a realistic trend for the distance D is highly constant.
dependent on the type of routing for the electric vehicle and Hence, in the simulation results reported in this section,
the driving style of the user. The SoC is always decreasing with the variables are varied two or three simultaneously. First,
a ramp dependent on the route and driving style. The trends the hypotheses that the vehicle was moving on a mix route
for Pc and Pd are based on the costs of electrical energy, urban/extra-urban and the average velocity was 29 km/h was
varying daily and yearly. The time variation of price variables made [26]. Starting from this data, a linear increase or decrease
will be in correlation with load demand in the electrical trend for D was established. For the SoC variation, a real

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
LAZAROIU AND ROSCIA: FUZZY LOGIC STRATEGY FOR PRIORITY CONTROL OF ELECTRIC VEHICLE CHARGING 19243

TABLE III
I NPUT AND O UTPUT D ATA FOR T ESTING THE O PERATION W ITH VALUES O UT OF R ANGE

Fig. 10. Results obtained in the case study 1. Fig. 11. Results obtained in the case study 2.

practice electric vehicle was chosen as basis for the autonomy.


The autonomy declared by the manufacturer is 270 km. The
discharge of the vehicle was considered linear with distance,
and the SoC (percentage of energy used by the vehicle
approaching the charging station) was obtained. Hence, the
linear decreasing trend for SoC was achieved.
Four simulations, with the vehicle heading towards the
charging station, were carried on. In this case, the distance
is linearly decreasing, the SoC is linearly decreasing, while
Pc and Pd linearly increase or remain constant. The four
case studies are:
Fig. 12. Results obtained in the case study 3.
1) D and SoC linearly decreasing, Pc linearly increasing,
Pd constant; increases from 15% to 50% as D decreases, but remains at
2) D and SoC linearly decreasing, Pc and Pd constants; a value much lower with respect to case study 1). This due
3) D and SoC linearly decreasing, Pc and Pd constants, to the Pc remaining constant at Pcmin . The results show
but with a value for Pc much higher than in case 2); the influence of Pc on the priority compared to variation
4) D and SoC linearly decreasing, Pc linearly increasing, of D. The distance is decreasing but the prices is not becoming
Pd constant, but with an initial value for SoC much economically advantageous during the driving period, as in the
higher than in case 1). case study 1) and the priority reaches and remains at a much
In the case study 1), the distance D linearly decreases from lower value.
Dmax to Dmin , SoC decreases linearly from 20% to 18%, In the case study 3), the distance D linearly decreases from
Pc increases linearly from Pcmin to Pcmax , while Pd Dmax to Dmin , SoC decreases linearly from 20% to 18%,
is constant at 3%. Fig. 10 shows the variations of the input Pc is constant at Pcmax , while Pd is constant at 3%.
variable and of the output variable P function of simulation Fig. 12 illustrates the variations of the input variable and of
time. The output P increases from 15% to 85% as D decreases the output variable P function of simulation time. The output
and Pc increases. The important augmentation of the priority P increases from 50% to 85% as D decreases, but remains at a
on the route is determined by the reduction of distance and value much higher with respect to the case study 2). This is due
charging price more advantageous. to the Pc remaining constant at Pcmax , which is constant
In the case study 2), the distance D linearly decreases from at a value more economically advantageous compared to the
Dmax to Dmin , SoC decreases linearly from 20% to 18%, case study 2).
Pc is constant at Pcmin , while Pd is constant at 3%. In the case study 4), the distance D linearly decreases from
Fig. 11 shows the variations of the input variable and of the Dmax to Dmin , SoC decreases linearly from 80% to 78%,
output variable P function of simulation time. The output P Pc increases linearly from Pcmin to Pcmax , while Pd is

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
19244 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 23, NO. 10, OCTOBER 2022

The simulation was carried on considering that the distance


D linearly decreases from 6km to Dmin , SoC decreases linearly
from 70% to 68%, Pc varies between a constant value of
5% and peaks of Pcmax , while Pd is always constant at
5% except when it increases to Pdmax .
The initial value of the distance in this simulation is exceed-
ing the maximum distance Dmax . This assumption was made
to test the filtering mechanism of false values of priority P
due to input variables out-of-range. During the first kilometer
travelled, the fuzzy logic controller should return as output a
Fig. 13. Results obtained in the case study 4. null value. In addition, the distance was split in 3 km of urban
driving and 3km of extra-urban driving. In this simulation the
driving of the electric vehicle is considered to be different
compared to previous cases. Previously, the linear decrease
of the distance was based on an average speed related to a
mix driving urban/extra-urban. In the current simulation, two
distinct velocities are considered: an average velocity vmu for
urban driving and an average velocity vmeu for extra-urban
driving. The average velocity vmeu for extra-urban driving
for Italy is 47km/h [27], while the average velocity Vmu for
urban driving is 11km/h. When the distance D decreases more
rapidly, the fuzzy logic controller delivers as output a priority
Fig. 14. Simulations with peaks of price variables. P that rapidly increases.
Fig. 14 illustrates the obtained simulation results under the
constant at 3%. Fig. 13 illustrates the variations of the input made assumptions. In the first kilometer travelled, when the
variable and of the output variable P function of simulation variable D is out of range, the returned priority P is null.
time. The output P increases from 15% to 50% as D decreases From the second kilometer travelled, the controller operates
and Pc increases, but rests at a value much lower with and generates the steps of priority P corresponding to the
respect to case study 2). This is due to the SoC decreasing peaks of Pc and Pd . The fuzzy logic controller has an
from 80%. Hence, the necessity to recharge the vehicle is important role in directing the users towards the charging
much lower, even if the charging price is becoming more stations during the periods with extreme events in the electrical
advantageous. The level reached by the priority P remains network. When the electric vehicle is passing to the extra-
at lower maximum value compared to previous cases. urban travelling (the distance D is decreasing rapidly), the
The simulation results demonstrate clear adherence between priority P increases with a steep ramp-up.
the design phase and the operation of the fuzzy controller.
In the previous case studies, the modification was mainly on V. C ONCLUSION
the Pc , but the results can be extended analogously for the Nowadays, electric vehicles are more and more present in
variation of Pd . This is due to the characteristics of these the transportation sector due to environmental constraints and
two variables, with similar weights in the determination of P. governmental incentives. The design and operation of systems
Small differences, determined by the prioritization of charging for electric vehicles to support the synergy between these and
price with respect to the discharging price in the elaboration smart grids can lead to both economic and energy savings.
of the rules for obtaining P could occur. At parity of economic This paper deals with a controller based on fuzzy logic
advantages, the discharging operation is a facultative action, capable to direct the users’ choices towards the operations
while the charging (when SoC is low) is a forced necessary of charging/discharging in the instants mostly favorable for
action. distribution network support, energy efficiency and economic
3) Tests With Large Values for Prices: Several tests were advantages for the user itself. The fuzzy logic controller
carried on analyzing the operation of the fuzzy controller with informs the users, instant by instant, to which charging station
inputs high values of price variables, as shown in Fig. 14. to direct and with which degree of convenience/necessity.
This type of trend for Pc and Pd can be associated in A local controller can be inserted in each EV for optimizing
real practice with outages in the mains supply, like failures the various decisions, considering the power system con-
in generation, disconnection of distribution lines, fast demand straints, the battery charging necessities, allowing the user to
variations determined by the operation of air conditioning obtain a price for the recharging based on electricity market
units, or even production peaks of renewable sources due information and reducing peal loads. The proposed system is
to atmospheric conditions. The fuzzy logic controller, when dealing with the smart charging, through an optimal choice
a peak of a price variable occurs, supplies a higher output made with the support of the fuzzy logic controller. The smart
variable P. Thus, the user could direct to the charging station charging must be implemented considering the price variations
and can support the affected electrical network, possibly determined by the upstream network and station management,
obtaining economic advantages. electricity market etc., with real-time price updates and with

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.
LAZAROIU AND ROSCIA: FUZZY LOGIC STRATEGY FOR PRIORITY CONTROL OF ELECTRIC VEHICLE CHARGING 19245

the previous monetary flows. This would allow not only [16] S. Faddel, A. A. S. Mohamed, and O. A. Mohammed, “Fuzzy logic-
to define a dynamic price, but to determine the users to based autonomous controller for electric vehicles charging under differ-
ent conditions in residential distribution systems,” Electric Power Syst.
spread towards different charging stations with competitive Res., vol. 148, pp. 48–58, Jul. 2017, doi: 10.1016/j.epsr.2017.03.009.
prices for helping the power system. The existing models, [17] S. Sarvaiya, S. Ganesh, and B. Xu, “Comparative analysis of hybrid
based on various algorithms, for charging the EV and their vehicle energy management strategies with optimization of fuel economy
and battery life,” Energy, vol. 228, Aug. 2021, Art. no. 120604, doi:
corresponding costs should be integrated in the proposed 10.1016/j.energy.2021.120604.
controller that performs a non-economic analysis. The FLC [18] C. Wei et al., “Comparison of architecture and adaptive
energy management strategy for plug-in hybrid electric logistics
can be integrated with smart agents able to learn in time the vehicle,” Energy, vol. 230, Sep. 2021, Art. no. 120858, doi:
habits of the driver and adapting to the exigencies. The fuzzy 10.1016/j.energy.2021.120858.
logic interface was chosen also because it is user friendly. [19] F. Berthold, A. Ravey, B. Blunier, D. Bouquain, S. Williamson, and
A. Miraoui, “Design and development of a smart control strategy for
This represents a point for future researches where will be plug-in hybrid vehicles including vehicle-to-home functionality,” IEEE
integrated with blockchain systems. In this way, the process Trans. Transport. Electrific., vol. 1, no. 2, pp. 168–177, Aug. 2015, doi:
flows will be managed not only based on the derivation from 10.1109/TTE.2015.2426508.
[20] H. Yin, W. Zhou, M. Li, C. Ma, and C. Zhao, “An adaptive fuzzy
charging station, but also from the cash flows of the renewable logic-based energy management strategy on battery/ultracapacitor hybrid
energy sources (REScoin). electric vehicles,” IEEE Trans. Transport. Electrific., vol. 2, no. 3,
pp. 300–311, Sep. 2016, doi: 10.1109/TTE.2016.2552721.
[21] M. Kandidayeni, A. O. M. Fernandez, A. Khalatbarisoltani, L. Boulon,
R EFERENCES S. Kelouwani, and H. Chaoui, “An online energy management strategy
for a fuel cell/battery vehicle considering the driving pattern and
[1] L. Cai, A. B. Rad, and W. L. Chan, “An intelligent longitudi- performance drift impacts,” IEEE Trans. Veh. Technol., vol. 68, no. 12,
nal controller for application in semiautonomous vehicles,” IEEE pp. 11427–11438, Dec. 2019, doi: 10.1109/TVT.2019.2936713.
Trans. Ind. Electron., vol. 57, no. 4, pp. 1487–1497, Apr. 2010, doi: [22] D. Phan, A. Bab-Hadiashar, M. Fayyazi, R. Hoseinnezhad, R. N.
10.1109/TIE.2009.2029571. Jazar, and H. Khayyam, “Interval type 2 fuzzy logic control for
[2] H. Yang, S. Yang, Y. Xu, E. Cao, M. Lai, and Z. Dong, “Electric vehicle energy management of hybrid electric autonomous vehicles,” IEEE
route optimization considering time-of-use electricity price by learnable Trans. Intell. Vehicles, vol. 6, no. 2, pp. 210–220, Jun. 2021, doi:
Partheno-genetic algorithm,” IEEE Trans. Smart Grid, vol. 6, no. 2, 10.1109/TIV.2020.3011954.
pp. 657–666, Mar. 2015, doi: 10.1109/TSG.2014.2382684. [23] S. F. Da Silva et al., Energy Convers Manage, vol. 234, p. 113909,
[3] Y. Cao et al., “An optimized EV charging model considering TOU price 2021, doi: 10.1016/j.enconman.2021.113909.
and SOC curve,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 388–393, [24] M. Sellali et al., “Novel control implementation for electric vehicles
Mar. 2012, doi: 10.1109/TSG.2011.2159630. based on fuzzy-back stepping approach,” Energy, vol. 178, pp. 644–655,
[4] M. L. Crow, “Cost-constrained dynamic optimal electric vehicle charg- Jul. 2019, doi: 10.1016/j.energy.2019.04.146.
ing,” IEEE Trans. Sustain. Energy, vol. 8, no. 2, pp. 716–724, Apr. 2017, [25] S. Qureshi et al., “A hybrid DL-based detection mechanism for cyber
doi: 10.1109/TSTE.2016.2615865. threats in secure networks,” IEEE Access, vol. 9, pp. 73938–73947,
[5] M. Singh, P. Kumar, and I. Kar, “Implementation of vehicle to grid 2021, doi: 10.1109/ACCESS.2021.3081069.
infrastructure using fuzzy logic controller,” IEEE Trans. Smart Grid, [26] Quattroruote. (Aug. 25, 2021). Aumentano i km Percorsi e il Tempo
vol. 3, no. 1, pp. 565–577, Mar. 2012, doi: 10.1109/TSG.2011.2172697. Trascorso Alla Guida. [Online]. Available: https://www.quattro
[6] M. Singh, K. Thirugnanam, P. Kumar, and I. Kar, “Real-time coordina- ruote.it/news/curiosita/2017/12/06/italiani-in-auto-aumentano-i-km-
tion of electric vehicles to support the grid at the distribution substation percorsi-e-il-tempo-trascorso-alla-guida.html
level,” IEEE Syst. J., vol. 9, no. 3, pp. 1000–1010, Sep. 2015, doi: [27] TGcom24. (Aug. 25, 2021). Meno Spostamenti città, Aumenta
10.1109/JSYST.2013.2280821. la Velocità Media di Auto e Moto. Doppiati i Mezzi Pubblici.
[7] A. Dubey and S. Santoso, “Electric vehicle charging on residential [Online]. Available: http://stradafacendo.tgcom24.it/2017/04/27/meno-
distribution systems: Impacts and mitigations,” IEEE Access, vol. 3, spostamenti-in-citta-aumenta-la-velocita-media-di-auto-e-moto-
pp. 1871–1893, 2015, doi: 10.1109/ACCESS.2015.2476996. doppiati-i-mezzi-pubblici/
[8] T. Ma and A. O. Mohammed, “Optimal charging of plug-in electric
vehicles for a car-park infrastructure,” IEEE Trans. Ind. Appl., vol. 50,
no. 4, pp. 2323–2330, Jan. 2014, doi: 10.1109/TIA.2013.2296620. George Cristian Lazaroiu (Senior Member, IEEE)
[9] R. Azizipanah-Abarghooee, V. Terzija, F. Golestaneh, and A. Roosta,
received the Ph.D. degree in electrical engineering
“Multiobjective dynamic optimal power flow considering fuzzy-based
from the Politecnico di Milano, Italy, in 2006. He is
smart utilization of mobile electric vehicles,” IEEE Trans. Ind. Informat.,
a Professor with the Department of Power Systems,
vol. 12, no. 2, pp. 503–514, Apr. 2016, doi: 10.1109/TII.2016.2518484.
[10] M. U. Jan, A. Xin, H. U. Rehman, M. A. Abdelbaky, S. Iqbal, and University POLITEHNICA of Bucharest, Romania.
M. Aurangzeb, “Frequency regulation of an isolated microgrid with elec- He was the General Co-Chair of the IEEE 16th
tric vehicles and energy storage system integration using adaptive and International Conference on Harmonics and Qual-
model predictive controllers,” IEEE Access, vol. 9, pp. 14958–14970, ity of Power (ICHQP 2014) and the 2019 IEEE
2021, doi: 10.1109/ACCESS.2021.3052797. PES Innovative Smart Grid Technologies Europe
[11] B. Sah, P. Kumar, and S. K. Bose, “A fuzzy logic and artificial (ISGT-Europe). His research interests include energy
neural network-based intelligent controller for a vehicle-to-grid sys- systems development, smart cities, monitoring and
tem,” IEEE Syst. J., vol. 15, no. 3, pp. 3301–3311, Sep. 2021, doi: control of power quality, electricity markets, and optimization of power
10.1109/JSYST.2020.3006338. systems.
[12] K. P. Inala, B. Sah, P. Kumar, and S. K. Bose, “Impact of V2G
communication on grid node voltage at charging station in a smart grid
scenario,” IEEE Syst. J., vol. 15, no. 3, pp. 3749–3758, Sep. 2021, doi: Mariacristina Roscia (Member, IEEE) received the
10.1109/JSYST.2020.3007320. Ph.D. degree from the University of Naples Federico
[13] H. S. Salama, S. M. Said, M. Aly, I. Vokony, and B. Hartmann, II in 2004. Since 2019, he has been an Associate
“Studying impacts of electric vehicle functionalities in wind energy- Professor for the courses of “electrical systems” and
powered utility grids with energy storage device,” IEEE Access, vol. 9, “smart home into smart city” with the University
pp. 45754–45769, 2021, doi: 10.1109/ACCESS.2021.3066877. of Bergamo. The main research fields are electrical
[14] M. F. Bandpey and K. G. Firouzjah, “Two-stage charging strategy of systems for transport and smart grids in the presence
plug-in electric vehicles based on fuzzy control,” Comput. Oper. Res., of distributed generation. The research activity deals
vol. 96, pp. 236–243, Aug. 2018, doi: 10.1016/j.cor.2017.07.014. with the implementation of pilot projects related to
[15] M. Mahmoodi-K, M. Montazeri, and V. Madanipour “Simultaneous the creation of a smart city, and in the drafting of the
multi-objective optimization of a PHEV power management system and Sustainable Energy Action Plan (SEAP) of European
component sizing in real world traffic condition,” Energy, vol. 233, Commission, for the Municipality of Naples, both in participating in the
Oct. 2021, Art. no. 121111, doi: 10.1016/j.energy.2021.121111. Carbon Disclosure Project (CDP) of European Commission.

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 24,2023 at 08:27:58 UTC from IEEE Xplore. Restrictions apply.

You might also like