Centralized Scheduling Approach to Manage Smart Charging

of Electric Vehicles in Smart Cities

Giuseppe Graber, Vincenzo Galdi, Vito Calderaro, Francesco Lamberti and Antonio Piccolo

Department of Industrial Engineering, University of Salerno, Fisciano (SA), Italy

Keywords: Charging Scheduling, Data Clustering, Electric Vehicles, Genetic Algorithms, Smart Charging.

Abstract: Electric vehicles (EVs) are emerging as the future of individual mobility systems in smart cities since they

reduce greenhouse gas emissions and fossil fuel dependence. However, the deepening penetration of battery

EVs forecasted for the incoming years could cause significant stress on distribution networks (DNs), as well

as the need to address the growing energy demand. In order to limit the negative drawbacks associated with

EVs charging demand, the paper proposes a centralized approach for the EVs smart charging, and its

performance are compared with the uncontrolled charging approach. An optimization framework is

formulated in order to reduce both the overall peak power demand and the EVs charging cost according to

the electricity prices during the day. Finally, several Monte Carlo simulations are carried out to evaluate the

benefits introduced by the proposed scheduling strategy on a real case study, in terms of charging cost for

EVs’ users, satisfaction of EVs charging needs, and flattening of the load profile.

1 INTRODUCTION

Electric vehicles (EVs) are becoming a very

interesting option to reduce both fuel consumption

and greenhouse gas emissions in the transportation

sector for the near future, (IEA, 2016). Therefore,

world governments are pushing more and more

electric mobility in the smart cities. As a result, EVs

penetration is expected to reach an amount between

10% and 25% of the overall circulating vehicles in the

2010-2030, (Mukherjee, 2015). Unfortunately,

lithium batteries of EVs still ensure a limited range of

only 150-200 km, and they often need to charge,

(Nissan, 2015). For these reasons, significant EVs

charging activities will mostly take place in users’ car

garages, public or corporate car parks and dedicated

charging stations (You, 2017 – Yu, 2016). However,

a deep penetration of EVs could increase both

average and peak load in the distribution networks,

where the charging stations are usually connected,

with a potential reduction of its reliability due to the

overloads, (Hao Xu, 2016 – Veldman, 2015). Several

studies showed that smart charging strategies could

contribute to support a deeper penetration of EVs

within distribution networks, (de Hoog, 2015 – Hao

Xing, 2016). Thus, the need for optimal charging

scheduling algorithms is becoming a relevant issue to

face with future power system planning and

management actions, (Kumar, 2015 – Qi Kang,

2016). Scheduling strategies, also, aim to optimize

EVs charging cost introducing charging options

based on real-time and/or day-ahead price

information, (Cao, 2012 – Vandael, 2015).

The smart charging problem is well studied, and

many approaches are proposed in the technical

literature. Among them, it is possible to discern two

main categories based on centralized or decentralized

approaches, (Bina, 2015 – Esmaili, 2015).

Decentralized scheduling assumes there is no central

controller and all individuals decide or optimize their

own charging profiles. Since smart grids are large-

scale systems, centralized algorithms may be

unfeasible due to lack of scalability, requirement for

global information, and expensive implementation

cost, while decentralized control algorithms are

deemed as a promising alternative, (Jiang, 2014). The

algorithms proposed in (Gan, 2013) require each EV

reporting its power demand to an aggregator. The

aggregator then broadcasts a few messages from

which each EV makes locally based and binary

charging decisions (charge or not charge). In (Wen,

2012), Authors consider a multi-layer hierarchical

power network in which each sub-aggregator can

decide the states of its associated EVs locally. On the

other hand, the centralized scheduling algorithms

238

Graber, G., Galdi, V., Calderaro, V., Lamberti, F. and Piccolo, A.

Centralized Scheduling Approach to Manage Smart Charging of Electric Vehicles in Smart Cities.

DOI: 10.5220/0006761002380245

In Proceedings of the 7th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2018), pages 238-245

ISBN: 978-989-758-292-9

provide a higher EV integration level in the existing

grid, (Wanrong, 2014 – Zhou, 2014): in fact, a single

operator controls precisely time duration and rates of

all charging actions. Each EV submits detailed

information to a central processing unit, which

produces the charging schedule for each EV,

considering various objectives, such as power loss

and/or load variance minimization, or maximization

of the EVs penetration level (Veldman, 2015 – Qi

Kang, 2016). Within the literature, (Rezaee, 2013)

presents a global scheduling optimization problem in

which the charging events are chosen to minimize the

total EVs charging cost during the day. (Cao, 2012)

proposes optimized EV charging strategies in

response to time-of-use (ToU) prices in a regulated

market. Finally charging scheduling algorithms

present an opportunity also to provide electric energy

storage (EES)-based ancillary services, e.g.,

smoothing intermittency due to renewable energy

sources (RESs) and supporting grid-widefrequency

stability, (Zhang, 2014 – Falahati, 2016).

Here, we want to hightlight that a centralized

approach to manage the scheduling of EVs,

represents a good solution for two important aspects:

i) by means of adequate optimization strategies it is

possible to take into account the requirements of

users, distribution systems, and aggregators; ii) in

perspective, more different aggregators can represent

controlled buses for a smart grid, and a centralized

approach allows using the potentiality of the bus as

generator, load or storageto support the distribution

systems.

In order to highligth these potentialities, we

propose a contribution that mainly consists of two

parts: i) a formulation of an optimization problem to

centralize the management of EVs charge, by

flattening the demand load profile and minimizing the

EVs charging costs, according to the electricity prices

during the day; ii) the assessesment of the impact of

EVs on a real microgrid to evaluate the benefits

introduced by the proposed smart charging method.

The management proposal is tested by using

measured data and identifying typical load cluster

for the EVs charging demand.

The remainder of paper is organized as follows:

Section 2 describes the data acquisition and clustering

methods employed in the CO.S.MO. research project.

Section 3 points out the mathematical formulation of

the proposed scheduling problem, while results of

several simulations based on the case study

represented by the University of Salerno (UniSA)

campus are discussed in Section 4. Finally, Section 5

ends the paper with concluding remarks.

2 MODELING AND PROBLEM

FORMULATION

Here we describe the charging stations (CSs) model,

the proposed approach to the centralized EVs smart

charging and some details on its implementation.

2.1 Charging Mode

According to the Nissan Leaf specifications (Nissan,

2015), we consider two values of CSs’ rated power

for the incoming EVs in the parking areas: AC1 (230

V, 3.3 kW) for domestic use charging and AC2 (230

V, 6.6 kW) for public use charging, (de Hoog, 2015).

However, the proposed methodology can be applied

even in the case of other types of charging stations.

For each CS

, given the residual SoC value of the EV

connected to it, SoC

(t), its variation at time t+dT is

described as follows:

)()()(

3600

)(

)()(

tItVPtP

dTtP

tSoCdTtSoC

Batt









(1)

Where dT is the time-step,

)(tP

represents the

charging power at charging station CS

, and η is the

charging efficiency.

)(tP

Batt

, V

(t) and I

(t) are the

battery charging power, voltage and current,

respectively, computed according to the constant

current (ccm) and constant voltage (cvm) charging

mode of the EV battery, (Cao, 2012): they are

approximated by (2) and depicted in Figure 1.

 























































SoC

uVu

IuII

1exp1

1exp

(2)

Here

and

represent the current values and

the duration, respectively, of the battery in ccm,

while,

and

represent the minimum voltage,

the maximum voltage and time constant of the

battery, respectively, in cvm. Finally,

SoC

is a binary

variable and its value is defined by the SoC value. In

particular, we assume the ccm for SoC values below

80% and thus we set

SoC

= 1, while

SoC

= 0 in the

cvm where SoC > 80%.

Centralized Scheduling Approach to Manage Smart Charging of Electric Vehicles in Smart Cities

239

Figure 1: Charging modes of the battery pack.

We define the effective charging time

for the

EV user connected to the CS

as the minimum value

between the expected parking time

PARKING

and the

required time to fully charge the EV,

CHARGE

. If

PARKING

CHARGE

, then

is set to the time

required to the CS

to bring the SoC to such a value as

to allow the EV user coming back to its departure site.

In this case, it is necessary to wait for an extra time

before the charging session is complete.

To address the worst case in terms of additional

load required at the point of common coupling (PCC),

we assume the number of available CSs equal to the

capacity of the considered parking areas.

2.2 Scheduling Optimization Problem

The main idea of the proposed scheduling approach

is to break the required charging time into many small

charging intervals within the parking time, (Rezaei,

2014). In each scheduling-slot, Δt, if enabled, a CS

can charge the EV connected to it only for the

duration of the scheduling-slot providing a charging

packet. During each Δt, the algorithm assigns the

starting time of each charging packet for all EVs in

the parking areas minimizing the peak demand and

the charging cost at the same time. Each EV is

charged according to a FCFS - first come first serve -

service policy. The proposed scheduling technique is

mathematically formulated as follows:

 



















 



PARKING

PACK

START

CHARGE

BASESTART

dtttptpttcJ

1 1

)()()(

min

(3)

The problem (3) is subject to the constraints (4):

Where NCS is the number of CSs, represents the

number of charging packet for the EV at CSi (it is the

first integer greater than /Δt). Moreover, c(t)

represents the electricity unit price depending on the

hours of day, is the starting time of the j-th charging

packet related to the EV connected to the CSi. Finally,

(t) is the forecasted base load including production

from RES whereas (t) is the charging power supplied

by CSi to the EV.

Finally,

BASE

(t) is the forecasted base load

including production from RES whereas

CHARGE

(t) is

the charging power supplied by CS

to the EV.



























Batt

CHARGE

BattSTART

CHARGE

PARKING

START

PACK

START

NiCSoCdttp

NiTtt

Nittt

PARKING

PACK

,...,1

)(

)(0

,...,1)1()(

,...,1

,...,1,

,1,



(4)

In (4),

START

SoC

and

)(tP

Batt

are the residual SoC

values of the EV at the CS

and its required charging

power, respectively.

2.3 Method of Solution

A heuristic algorithm based on genetic algorithm

(GA) is proposed to solve the scheduling problem.

The GA based scheduler for each EV in the parking

areas computes the scheduling-slot sequence in which

each EV is enabled to receive a charging packet. We

use a binary chromosome changing its length

according to the

PARKING

value, for each EV in the

parking areas. Thus, they are aggregated into a single

2-dimensional chromosome representing the

scheduling solution.

The structure of the scheduling chromosome

composed by several EV-chromosome with different

length is shown in Figure 2. The scheduling is a

Figure 2: Genetic representation of the scheduling.

SMARTGREENS 2018 - 7th International Conference on Smart Cities and Green ICT Systems

240

is a 2-dimensional matrix having number of row

equal to the EVs in the parking areas, N, and number

of column depending on the maximum parking time

(in scheduling slot), M, among all the EVs. The

number and the length of the scheduling chromosome

also affects the simulation speed performance and the

reliability of the final solution. Therefore, we impose

a limit of 10.000 iterations and a population of 20

scheduling chromosomes.

In Figure 3 is shown the proposed scheduling

architecture: a central controller (CC) is assumed

receiving the forecasted base load and the production

from RES for the day. Each charging station

communicates with a local controller (LC) - one for

each parking area - to send the charging

update/request of the EV connected to it. LCs acquire

and aggregate data sent by the CSs located in their

parking area and transmit them to the CC. According

to data provided by LCs, CC performs the scheduling

optimization deciding when to allow each CS to

charge the EV connected to it.

Figure 3: Architecture of the coordinated scheduler.

In the following, the CC and LC routines,

performed on a daily base, are shown.

Initialize: The CC forecasts the base

load for the current day.

Loop (for each scheduling slot):

Begin LC routine

1. Receive the scheduled charging

sessions for the current time slot

from the CC.

2. Acquire new charging request.

3. Update charging vehicles status.

4. Send data to the CC.

End

Begin CC routine

1. Receive the new charging request

and the charging vehicles status

for the current time slot.

2. Run the EVs scheduling strategy

to all the vehicles in the parking

areas.

3. Update base load profile

addicting in charge EVs.

4. Update the base load forecast,

including the RES

5. Send data to the LCs.

End

3 SIMULATION AND RESULTS

To show the effectiveness of the presented

methodology, a Monte Carlo simulation framework is

implemented to calculate the EVs charging needs

starting from given statistical distribution of

incoming EVs, their parking times and residual SoC

values, (Calderaro, 2014). We use real data acquired

during the fulfilment of CO.S.MO. research project,

and processed through clustering operations.

3.1 Case Study

The UniSA microgrid is a 12 bus 20 kV distribution

system with two feeders configured in closed loop.

Connected to the grid, there are several distributed

generators (DG): two cogeneration (CHP) units, with

a rated power of 580 kW each one, and eight

photovoltaic (PV) power plants for a total PV rated

power of 1076 kW installed on the roof of the campus

buildings. CHP units produce both electricity used to

supply the loads and thermal energy used to heat

water of the campus sport facilities.

Figure 4: Active power drown from the UniSA microgrid.

In Figure 4 is shown the typical daily profiles of

the net active power drown from the main external

PCC by the UniSA network. Blue and green lines

depict the active power absorption with and without

internal PVs and CHPs, respectively. Finally, red and

Centralized Scheduling Approach to Manage Smart Charging of Electric Vehicles in Smart Cities

241

pink lines show the average (calculated every 15

minutes) active power generated by the PV and CHP

units.

Table 1: Daily price of the electricity cost.

Price category

Hours

Days

F1 – Peak

10:00 - 15:00 /

18:00 - 21:00

Monday to

Friday

F2 – Mid-level

07:00 - 10:00 /

15:00 - 18:00 /

21:00 - 23:00

Monday to

Friday

07:00 -23:00

Saturday

F3 – Off-peak

23:00 - 07:00

All the week

and holidays

Table 1 shows the three hourly price categories for

the electricity cost in the UniSA campus. Adopting

the actual prices of peak, mid-level and off-peak load

period in the city are 0.18 €/kWh, 0.14 €/kWh and

0.10 €/kWh, respectively, (Enel Distribuzione, 2017).

3.2 Data Acquisition

CO.S.MO. (Cooperative Systems for Sustainable

Mobility and Energy Efficiency) was a 32-months

pilot project co-founded by the European

Commission involving the installation of advanced

intelligent transportation system (ITS), in three pilot

sites: Göteborg (SE), Vienna (AT), and the UniSA

campus in Salerno (IT), (Alcaraz, 2013). The scope

of CO.S.MO. was to prove and quantify the benefits

of cooperative mobility services for increasing the

energy efficiency of infrastructures and vehicles.

Figure 5: Parking areas at UniSA - Google Maps view.

Several cameras and antennas systems were

installed at both entrances and exits of UniSA parking

areas (outlined in Figure 5), in order to record data

related to their occupancy level. They were able to

read vehicles plates, date, time and parking time: all

data was stored in a database. It consists of more than

200.000 parking events over a one-year period of

observation. The collected data are representative of

the parking areas used by a significant number of

students with their own car.

3.3 Data Clustering Results

In order to assess the different state of charge (SoC)

values for trips made with EVs, we evaluate the

origin-destination routes of students enrolled at the

UniSA. The data analysis allows classifying all

different paths to reach UniSA campus from different

departure points in terms of urban, extra urban and

highway routes. According to the studies published

by the Idaho National Laboratory, (INL, 2015) real

users of EVs show a strong preference to charging in

the evening and driving during the day. Thus we

assume the students’ EVs leaving fully charged from

each departure point (e.g., students’ homes), and we

derive the arrival residual SoC by using Eq. 5.

Batt

HHEEUU

dcdcdc

SoCSoC

)()()( 



(5)

Where SoC

e SoC

are the arrival and departure

SoC values, respectively, and C

Batt

[kWh] is the EV’s

battery capacity. Here, c

, c

[kWh/km] represent

the energy consumptions coefficients, whereas d

, d

[km] are the distances covered in the urban, extra

urban, and highway route, respectively.

We compute the electric energy consumption of

the EVs associated to each travel, assuming as

reference EV the Nissan Leaf model, having battery

capacity of 24 kWh. The manufacturer provides the

average consumptions for different route types and

they are shown in Table 2, (Nissan, 2015).

Table 2: Nissan Leaf Energy Consumption.

Route Type

Value [kWh/km]

Urban

0.160

Extra urban

0.126

Highway

0.185

Mixed

0.169

By analysing the departure points of the UniSA

students considering the shortest path to reach the

campus, 50 clusters are derived, each one with a

different residual SoC value at the arrival in the

parking areas (Figure 6). The average residual SoC

value of EVs arriving at the UniSA is approximately

equal to 70%.

Figure 7 illustrates the statistical distribution of

the average EVs parking time. In particular, the most

of the students arrive at the UniSA campus around the

8:30 a.m. and they are characterized by an average

parking time of 4 hours. This fact leads to an expected

SMARTGREENS 2018 - 7th International Conference on Smart Cities and Green ICT Systems

242

peak demand for charging between 9:00 a.m. and

12:00 a.m. that can be flattened by the scheduler

according to the parking time declared by each EV.

Figure 6: Residual SoC of incoming EVs.

Figure 7: Average parking time of incoming EVs.

Data related to the parking areas occupancy and

hourly rate of incoming EVs, collected during the

observation period, on a day-by-day base, are

analysed in order to find common features concerning

the days of the week, months, and seasons. Thus, they

are taken into account by splitting the observation

period into different clusters.

Figure 8: Arrival rate of incoming EVs (courses period).

The clustering function is implemented by using

the k-means algorithm because it ensures a very quick

convergence and it minimizes the total intra-cluster

variance. All the observed days are divided in two

main clusters: the first one concerning the

institutional courses period and the second one related

to the examination period. The courses period cluster

is characterized by days in which the EVs’ arrival rate

is significantly greater than those are in the

examination period cluster. It is possible to subdivide

the first main cluster in three other different sub-

clusters: from Monday to Wednesday, Thursday and

Friday respectively. In Figure 8 is shown the EVs’

arrival rate cluster to one parking area for each day of

the week, during the institutional courses periods. The

examination period is considered as a single cluster,

because of the absence of relevant differences among

the weekdays. Thus, four different daily clusters are

assumed adequate to describe arrival rate concerning

the three parking areas.

3.4 Scheduled Ev Charging Results

Figure 9 and Figure 10 show the comparison between

the uncontrolled charge and the proposed scheduling

strategy assuming AC1 and AC2 charging stations,

respectively. It is worth to note a flattened load profile

and a significant reduction of the peak power

absorption in the hours between 8:00 a.m. and

Figure 9: Comparison between uncontrolled and scheduled

charging (AC1 CSs).

Figure 10: Comparison between uncontrolled and

scheduled charging (AC2 CSs).

Centralized Scheduling Approach to Manage Smart Charging of Electric Vehicles in Smart Cities

243

12:00 p.m. compared to the uncontrolled charging

Specifically, the proposed scheduling algorithm

shows, at 9:45 a.m., a peak reduction of 232 kW and

326 kW for AC1 and AC2 charging stations,

respectively. On the other hand, the scheduler

significantly increases the load profile during the

hours of low charging demand assuming the

uncontrolled charging, e.g. between 2:00 p.m. and

4:00 p.m. The charging peak demand by using AC1

CSs is slightly higher than the peak of the base load

(at 17:00 p.m.) while it is always lower by using AC2

CSs. Furthermore, Table 3 shows the comparison on

battery SoC level reached by EVs between AC1 and

AC2 CSs. In particular, the number of EVs having

battery SoC level able to get home the user, and the

number of fully charged EVs compared to the overall

EVs are assumed as performance indices.

Table 3: SoC level (AC1 and AC2 CSs).

Residual SoC [%]

Number of EVs

AC1 mode

AC2 mode

SoC ≥ 80

100 (87)

100 (94)

65 ≤ SoC < 80

100 (74)

100 (83)

50 ≤ SoC < 65

92 (50)

99 (92)

30 ≤ SoC < 50

62 (41)

80 (72)

By using the AC1 CSs the scheduler ensures to all

users with residual SoC value higher than 65% to

come back to their departure point, but not to fully

charge their vehicle. For users with residual SoC

value less than 65% (about 20% of all EVs in a single

day), the performance indices are higher than 70%.

However, AC2 CSs ensure that all incoming EVs in a

day are able to come back to their starting point, but

only 90% can fully charge their vehicle. Table 4

shows the comparison concerning the charging extra

time by using the proposed scheduling strategy

between AC1 and AC2 modes. The average extra

time required to complete the EVs charge is

considered when T

PARKING

< T

CHARGE

: it is due to EV

user expected parking time and it is minimized by

scheduling operations. By using AC1 CSs, EVs with

the lowest residual SoC value have to wait an average

extra-time of about 2 hours, whereas AC2 CSs lead to

an average extra-time less than 1 hour.

Table 4: Charging extra time (AC1 and AC2 CSs).

Residual SoC [%]

Average extra time [min]

AC1 mode

AC2 mode

SoC ≥ 80

65 ≤ SoC < 80

50 ≤ SoC < 65

30 ≤ SoC < 50

121

Finally, Table 5 shows the comparison concerning

the average charging cost by using the proposed

scheduling strategy. In particular, EV user benefits of

a slightly reduction in the average charging cost

because the chronological shift of the charging

packets allow to obtain a lower electricity cost. In

particular, the average charging cost is reduced by

16.2% and 18.5% compared to uncontrolled charge

and considering AC1 and AC2 CSs, respectively.

Table 5: Charging cost (AC1 and AC2 CSs).

4 CONCLUSIONS

The paper presents a centralized scheduling algorithm

for optimizing EVs charge in smart grids in order to

minimize the EVs charging costs and reduce the peak

power demand. The UniSA parking areas are

characterized in terms of incoming EVs rate and

hourly occupancy level, according to a large

measured database, in order to evaluate the EVs

charging demand. Several Monte Carlo simulations

are performed to test the proposed scheduling

algorithm. Obtained results confirm the effectiveness

of the proposed scheduling algorithm: by using AC1

CSs, it ensures up to 75% of users the minimum

charge required to come back to their departure point,

whereas by using AC2 CSs it ensures the fully charge

to over 90% of users. Finally, the EV user’s further

benefits by a lower charging cost compared to the

uncontrolled charge. In conclusion, the obtained

results proof that by means of optimization strategies

it is possible to take into account the requirements of

users, distribution systems, and aggregators;

furthermore, with regard to distribution system

impact, the results show that an aggregator is a

potential smart bus that can bring many benefits to

distribution systems.

ACKNOWLEDGEMENTS

The authors gratefully thank UE and all technological

partners who have contributed to the success of the

CO.S.MO. research project.

PEV charge

Average charging cost [€]

AC1 mode

AC2 mode

Uncontroled

1.78

1.62

Scheduled

1.49

1.32

SMARTGREENS 2018 - 7th International Conference on Smart Cities and Green ICT Systems

244

REFERENCES

International Energy Agency (IEA), 2016. World Energy

Outlook. Available on line at url

http://www.worldenergyoutlook.org /weo2016/

Mukherjee, J.C., Gupta, A., 2015. A Review of Charge

Scheduling of Electric Vehicles in Smart Grid. In IEEE

Systems Journal, vol. 9, pp. 1541-1553.

Nissan, 2015. Nissan Leaf prices and specifications.

Available on-line at: <http://media.nissan.eu/content

/dam/pes/GB/leaf/prices_and_equipment/pdf/New-

LEAF-Tech-Spec-and-Pricesv3.pdf>, pp. 1-7.

You P., Low S. Tushar W., Geng G., Yuen C., Yang Z., Sun

Y., 2017. Scheduling of EV Battery Swapping, I:

Centralized Solution. In IEEE Trans. on Control of

Network Systems, vol. PP.

Rong Yu R., Zhong W., Xie S., Yuen C. Gjessing S., Zhang

Y., 2016. Balancing Power Demand through EV

Mobility in Vehicle-to-Grid Mobile Energy Networks.

In IEEE Trans. on Industrial Informatics, vol. 12, pp.

79-90.

Hao Xu, Fan Ouyang, Haifeng Liu, Shihong Miao, Haiguo

Tang, Wenwu Liang, Hui Li, Dongyuan Shi, 2016. A

probability load modeling method for the charging

demand of large-scale PEVs accounting users’ charging

willingness. In International Journal of Electrical

Power & Energy Systems, vol. 82, pp. 331-338.

Veldman, E., Verzijlbergh, R.A., 2015. Distribution Grid

Impacts of Smart Electric Vehicle Charging From

Different Perspectives," IEEE Trans. on Smart Grid,

vol. 6, pp. 333-342.

de Hoog, J., Alpcan, T., Brazil, M., Thomas, D.A., Mareels,

I., 2015. Optimal Charging of Electric Vehicles taking

Distribution Network Constraints into Account. In

IEEE Trans. on Power Systems, vol. 30, pp. 365-375.

Hao Xing, Minyue Fu, Zhiyun Lin, Yuting Mou, 2016.

Decentralized Optimal Scheduling for Charging and

Discharging of Plug-In Electric Vehicles in Smart

Grids. In IEEE Trans. on Power Systems, vol. 31, pp.

4118- 4127.

Kumar K.N., Sivaneasan B., Lam So, P., 2015. Impact of

Priority Criteria on Electric Vehicle Charge

Scheduling. In IEEE Trans. on Transportation

Electrification, vol. 1, pp. 200-210.

Kang, Q., Wang, J., Zhou, M., Ammari, A.C., 2016.

Centralized Charging Strategy and Scheduling

Algorithm for Electric Vehicles Under a Battery

Swapping Scenario. In IEEE Trans. on Intelligent

Transportation Systems, vol. 17, pp. 659-669.

Cao, Y. Tang, S., Li, C., Zhang, P., Tan, Y., Zhang, Z., Li,

J., 2012. An Optimized EV Charging Model

Considering TOU Price and SOC Curve. In IEEE

Trans. on Smart Grid, vol. 3, pp. 388-393.

Vandael, S., Claessens, B., Ernst, D., Holvoet, T.,

Deconinck, G., 2015. Reinforcement Learning of

Heuristic EV Fleet Charging in a Day-Ahead Electricity

Market. In IEEE Trans. on Smart Grid, vol. 6, pp. 1795-

1805.

Bina, T., Ahmadi, D., 2015. Stochastic modelling for

scheduling the charging demand of EV in distribution

systems using copulas. In International Journal of

Electrical Power & Energy Systems, vol. 71, pp. 15-25.

Esmaili, M., Goldou, A., 2015. Multi-objective optimal

charging of plug-in electric vehicles in unbalanced

distribution networks. In Int. Journal of Electrical

Power & Energy Systems, vol. 73, pp. 644-652.

Jiang, T., Putrus, G., Gao, Z., Conti, M., McDonald, S.,

Lacey, G., 2014. Development of a decentralized smart

charge controller for electric vehicles. In International

Journal of Electrical Power & Energy Systems, vol. 61,

pp. 355-370.

Gan, L., Topcu, U., Low, S.H., 2013. Optimal decentralized

protocol for electric vehicle charging. In IEEE Trans.

on Power Systems, vol. 28, pp. 940-951.

Chao-Kai Wen, Jung-Chieh Chen, Jen-Hao Teng, Pangan

Ting, 2012. Decentralized energy management system

for charging and discharging of plug-in electric

vehicles. In Wireless Communications & Signal

Processing Int. Conference, pp. 1-6.

Wanrong T., Suzhi Bi, Y.J.A. Zhang, 2014. Online

Coordinated Charging Decision Algorithm for Electric

Vehicles without Future Information. In IEEE Trans.

on Smart Grid, vol. 5, pp. 2810-2824.

Kan Zhou, Lin Cai, 2014. Randomized PHEV Charging

Under Distribution Grid Constraints. In IEEE Trans. on

Smart Grid, vol. 5, pp. 879-887, March 2014.

Rezaee, S., Farjah, E., Khorramdel, B., 2013. Probabilistic

analysis of plug-in electric vehicles impact on electrical

grid through homes and parking lots. In IEEE Trans. on

Sustainable Energy, vol.4, pp. 1024-1033.

Tian Zhang, Wei Chen, Zhu Han, Zhigang Cao, 2014.

Charging Scheduling of Electric Vehicles with Local

Renewable Energy under Uncertain Electric Vehicle

Arrival and Grid Power Price. In IEEE Trans. on

Vehicular Technology, vol. 63, pp. 2600-2612.

Falahati, S., Taher, S.A., Shahidehpour M., 2016. A new

smart charging method for EVs for frequency control of

smart grid. In International Journal of Electrical Power

& Energy Systems, vol. 83, pp. 458-469.

Alcaraz, G., Galdi, V., Trapuzzano, A., 2013. Cooperative

Systems for Energy Efficiency and driver behaviour

change: University of Salerno Pilot Site Results. In ITS

European Congress, pp. 1-6.

Idaho National Laboratory (INL), 2015. Plugged In: How

Americans Charge Their Electric Vehicles. Available

on-line at: <https://avt.inl.gov/sites/default/files/pdf/

arra/SummaryReport.pdf>.

Rezaei, P., Frolik, J., Hines, P.D.H., 2014. Packetized Plug-

In Electric Vehicle Charge Management. In IEEE

Trans. on Smart Grid, vol. 5, pp. 642-650.

Calderaro, V., Galdi, V., Graber, G., Massa, G., Piccolo, A.,

2014. Plug-in EV charging impact on grid based on

vehicles usage data. In IEEE International Electric

Vehicle Conference, pp. 1-7.

Enel Distribuzione, 2017. Price categories for the electricity

cost. Available on-line at: <https://www.enelservizio

elettrico.it/itIT/tariffe/altriusi/bta-6-trioraria>.

Centralized Scheduling Approach to Manage Smart Charging of Electric Vehicles in Smart Cities

245