Received November 20, 2019, accepted December 4, 2019, date of publication December 9, 2019, date of current version January

22, 2020.
Digital Object Identifier 10.1109/ACCESS.2019.2958403

Dynamic EV Charging Pricing Methodology for

Facilitating Renewable Energy With
Consideration of Highway
Traffic Flow
SUYANG ZHOU 1 , YUE QIU 1 , FENGHUA ZOU 1, DI HE 2 , PENG YU 3 , JINQIAO DU 3 ,
XINER LUO 3 , CHENGLIANG WANG4 , ZHI WU 1,5 AND WEI GU 1,2 , (Senior Member, IEEE)
1 School of Electrical Engineering, Southeast University, Nanjing 210096, China
2 School of Cyber Science and Engineering, Southeast University, Nanjing 211189, China
3 Shenzhen Power Supply Bureau Company, Ltd., Shenzhen 518001, China
4 Jiangsu Fangtian Electric Technology Company, Ltd., Nanjing 211106, China
5 Jiangsu Key Laboratory of Smart Grid Technology and Equipment, Nanjing 210096, China

Corresponding author: Zhi Wu ([email protected])

This work was supported in part by the National Natural Science Foundation of China under Grant 51807024, and in part by the Jiangsu
Key Laboratory of Smart Grid Technology and Equipment.

ABSTRACT Electric vehicles (EVs) are expected to play a critical role in future transportation systems.
A number of countries have published roadmaps aiming to facilitate the adoption of EVs on the road. It is
estimated that existing charging facilities will not be able to satisfy the tremendous charging demands of
a dramatically increasing number of EVs. Following the rapid development of artificial intelligence and
mobile communication technology, certain charging pricing mechanism is expected to influence the charging
behavior of EV drivers. In order to maximize the working efficiency of highway charging facilities and the
consumption of the renewable energy near charging facilities, this paper proposes a pricing methodology
taking into account the charging facility service ratio, traffic flow and renewable energy generation. To sup-
port the adoption of the proposed pricing methodology, forecasts of hour-by-hour traffic flow and renewable
generation as well as calculation of the shortest paths to different charging stations (CSs) are investigated.
A road network testbed based on the Dublin traffic network is established to evaluate the proposed pricing
methodology. It is discovered that for certain wind-rich CSs, the proposed pricing methodology can increase
the consumption rate of wind energy by up to 82.97%, with an average improvement of 30.73%; for certain
solar-rich CSs, it can improve the level of solar energy consumption by up to 59.50% and an average increase
of 29.28% is achieved. The proposed pricing methodology can also reduce traffic jams to some extent at both
peak and off-peak times.

INDEX TERMS Electric vehicle, charging stations, charging pricing methodology, renewable energy
consumption.

NOMENCLATURE nSC
i,t Number of slow chargers in use within the ith CS
Symbol Quantity at time t
OSC
i,t Operational state of slow charging facilities nFC
i,t Number of fast chargers in use within the ith CS
within the ith CS at time t at time t
OFC
i,t Operational state of fast charging facilities within NiSC Total number of slow chargers within the ith CS
the ith CS at time t NiFC Total number of fast chargers within the ith CS
Rm,t Traffic flow rate on the mth road at time t
nRm,t Amount of vehicles on the mth road at time t
The associate editor coordinating the review of this manuscript and Nmmax Maximum capacity of the mth road
approving it for publication was Chandan Kumar . A EV charging probability considering the SoC

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
VOLUME 8, 2020 13161
 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

τ Charging probability parameter when SoC packs in the past few years, electric vehicles (EVs) are
is larger than 20% and smaller than 100% expected to be cost-competitive compared with conventional
wci,j,t Willingness of the jth EV driver to choose fossil fuel based vehicles. It was estimated in [4] that EVs
the ith CS considering the price difference will be commercially competitive in nearly half of the US
at time t car market by 2020 if the battery pack price follows the
wdi,j,t Willingness of the jth EV driver to choose decreasing price trend. Thus, it is predictable that EVs will
the ith CS considering the distance differ- become extremely important in future transportation sys-
ence at time t tems. To accommodate the charging demand of EVs, it is
lj Loyalty level of the jth EV driver essential to install a significant number of charging facilities/
Si,j Indicator that indicates if the jth EV driver stations and implement an effective management and pricing
has previously used the ith CS mechanism.
pi,t Charging price per kWh of the ith CS at
time t B. RELATED WORKS
p¯t Regional average charging price at time t The penetration of EVs bring opportunities and challenges to
α0 , α1 Constant related to wci,j,t the power system and transportation network. Mobile energy
oc Offset related to wci,j,t storage units, peak-valley load fluctuation, traffic congestion
di,j,t Distance of the shortest path to the ith CS management, renewable energy consumption and other topics
for the jth EV driver have gradually become the research focus of EVs [5]–[11].
α2 , α3 , α4 , α5 Constant related to wdi,j,t The flexible nature of EVs (spatiotemporal demand flexibil-
od Offset related to wdi,j,t ity, autonomous charging station selection and path planning
ω1 Weight of the willingness attributed to the et al.) can be systematically leveraged to improve the oper-
price ation of both power system and transportation system. [7]
ω2 Weight of the willingness attributed to the discussed the feasibility of deploying the EVs into the smart
distance grid system. Both the ancillary services that EVs can provide
λi,t Charging price of the ith CS with renewable to the grid and the side effects of the EV related technologies
generation nearby or onsite are presented in an intensive review. [8] took into account the
pt Charging price of the CSs without renew- interaction between the locational marginal prices determined
able generation nearby or onsite by the independent system operator (ISO) and the charging
λFIT Renewable energy Feed-in-Tariff demand influenced by the decision of EVs and Charging Sta-
pG Renewable generation at the ith CS at time tions (CSs). The collaborative scheduling strategy can inte-
i,t
t grate electrified transportation into power system to achieve
a globally optimum solution. [10] presented the benefits of
pDi,t Charging demand at the ith CS at time t
Plug-in Hybrid Electric Vehicles (PHEVs) from the perspec-
ai,t , bi,t , ci,t Pricing parameters of the ith CS at time t
tive of pollution output, energy efficiency and sustainability
L Square error loss function
of the transportation energy sector. In addition, the capacity
n True value of the target in the ANN model
of the grid to meet large adoption of PHEVs was also ana-
n̂ Predicted value of the target in the ANN
lyzed. [12] modelled the interdependency between the power
model
systems and the electrified transportation networks and the
W Weights of the input layer and the hidden
Plug-in Electric Vehicles (PEVs) served as coupling agents to
layer in the ANN model
realise the interdependencies of the power and transportation
β A non-negative hyper parameter that con-
networks. [13] proposed a two-stage approach to allocate EV
trols the magnitude of the penalty
parking spaces and distributed renewable resources (DRRs),
R2 Evaluation score for each training
which could reduce the distribution network loss. [14] devel-
N Number of the whole target
oped a novel dynamic demand control method to coordinate
yi True value of the ith target the charging and discharging behaviors of EVs in the grid
ŷi Predicted value of the ith target with high penetration level of renewable energy.
ȳi Mean value of the whole target A number of researches have been carried out in the field
of EV charging pricing methodology [15]–[17] and price-
I. INTRODUCTION driven EV charging management [18]–[23]. [16] proposed
A. MOTIVATION a retail pricing mechanism implemented by charging net-
Vehicle electrification has been considered as an effec- work operators (CNOs), considering the interaction among
tive way to achieve low-carbon transportation systems. CNOs, EV drivers and power system operator. In this work,
A number of countries and areas have published roadmaps optimal retail prices for CSs are designed to minimize the
for low-carbon transportation (e.g. Scotland announced its CNOs’ electricity costs whilst the convenience of EV drivers
commitment to phasing out petrol and diesel vehicles by can be improved. A route optimization approach for EVs
2032) [1]–[3]. Along with continuous price drop of battery was presented in [17] considering the time-of-use (TOU)

13162 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

electricity tariff. The proposed approach modelled the routing C. CONTRIBUTION

problem based on travelling salesman problem and miti- In this paper, a dynamic pricing methodology for the EV CSs
gated the system peak load in the network. Schneider et al. to facilitate the renewable energy consumption is proposed,
investigated a routing method of EVs emphasizing on time which takes into account the traffic congestion information,
windows and recharging stations, which ensured that the CS service ratio, renewable generation and charging demand.
EVs could arrive at recharging stations within the limited The contributions of this paper are summarized below.
battery capacity [18]. In [19], a congestion management (1) A pricing methodology for EV CSs that aims to facili-
scheme for EV charging in urban areas was proposed. The tate the consumption of renewable energy is proposed.
proposed approach specified the charging price of charging (2) The willingness of EV drivers to select a CS is mod-
facilities depending on the congestion level and available elled based on the charging price, distance to CSs and traffic
charging slots. A charging management scheme was mod- congestion information.
elled in [20] as the CS selection problem where the EV (3) Forecasting methods for traffic flow and renewable
driver would select the CS with shortest distance or the generation are evaluated.
shortest waiting time. Another study in [21] focused on EV (4) The performance of the proposed pricing methodology
charging management taking into account the trip duration is evaluated using a traffic network model that is established
and uncertainties of EV drivers. Most of the aforementioned based on the Dublin traffic system.
researches modelled the EV charging management scheme
as a logistics application where the willingness of EV drivers
D. ORGANIZATION
to get access to charging facilities, the renewable genera-
The remainder of this paper is organized as follows: Section II
tion supply in the area and traffic congestion were not fully
describes the pricing methodology of CSs and how it is
considered.
operated. Section III presents forecasting methodologies for
From the perspective of human behavior, certain
researches have been conducted to measure its impact on highway traffic flow and renewable energy which support
the EV charging behavior. The major factors that could the operation of the CS pricing scheme. Section IV evaluates
influence the EV driving patterns and charging behavior the performance of the proposed pricing scheme within the
traffic network based on the real traffic network of Dublin.
were investigated in [24]. It was found that the power of
The conclusion is presented in Section V.
charging systems and the charging cost were two major
factors affecting EV charging behavior. The Tokyo Electric
Power Company presented a study on the impact of charging II. DYNAMIC PRICING METHODOLOGY OF EV
facility power on EV charging behavior [25]. The results CHARGING STATIONS
showed that the charging durations were similar regardless Following the predictable increasing penetration of EVs, CSs
of the power rating of the charger when EV drivers used will become the major energy consumers in the electricity
non-residential chargers. Different EV owner groups also domain. Maximizing the usage of CSs and renewable gen-
behaved differently when selecting CSs. More specifically, erations within a flexible pricing environment will be an
both private and commercial users preferred charging at essential way to promote EVs and reduce surplus renewable
stations that required a shorter detour whilst the charging energy. In this paper, a pricing methodology and the asso-
cost had a larger influence on private users than commercial ciated steps to implement such a pricing scheme in trans-
users [26]. Part of the findings presented in [27] could support portation networks is presented. The working principle of the
the results of [26] that EV owners preferred the nearby CSs, pricing methodology is shown in Figure 1.
and [27] also indicated that EV drivers inclined to choose CSs Advanced IT technologies will enhance the availability of
with less charging time. Similar findings were also presented data (both historical and real-time) for traffic flow, renewable
in [28]–[31]. generation and CS operational states. Based on the collected
There are several researches focusing on how an EV data, the system will forecast renewable generations, CS
charging management scheme could be used to shape EV charging demands and traffic flow for the next 1-24 hours.
charging loads to match renewable energy generation. The After that, the price of the CSs near renewable generation
feasibility of using EV charging scheduling to match the wind will be established according to demand and generation. EV
power was investigated in [32], while [33] proposed an EV drivers who need to recharge their vehicles will subsequently
charging scheduling algorithm for buildings equipped with select CSs considering charging price, distance to the CSs and
photovoltaic (PV) systems. The aforementioned charging the available charging slots. It is expected that the pricing
management/scheduling approaches facilitated the renew- scheme will mitigate the gap between the charging demand
able energy consumption by providing charging recommen- and th renewable generation, and at the same time balance
dations to EV drivers but the willingness of EV drivers the traffic flow during the peak times.
to participate in the scheme was not fully considered. To illustrate the detailed pricing mechanism, this section
In addition, there are few researches on EV charging pric- will introduce the dynamic pricing methodology in part A
ing methodology aiming to facilitate the use of renewable and the operation of the pricing methodology within the road
energy. network in part B. It should be noted that the mathematical

VOLUME 8, 2020 13163

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 1. Schematic diagram of the pricing methodology adopted in traffic network.

models of CSs, traffic network and charging behavior are also When the SoC of an EV is below the threshold value (i.e.
included in part A. 20% in this paper), the EV must immediately go to a CS to
charge. Otherwise the charging probability of the EV is neg-
A. DYNAMIC PRICING METHODOLOGY atively correlated with the charging state. Factors of the EV
1) OPERATIONAL STATE OF CHARGING STATION charging probability are used to evaluate the number of EVs
The operational state of a CS indicates its servicing busyness to be charged in different time slots and can be modelled as :
level, which can be reflected by the ratio of occupied chargers

1, if SoC < 20%
to the total number of chargers at the CS. The operational state A= SoC − 20% ∗
is expressed as follows: exp( τ ), if 20% < SoC < 100%
SoC − 100%
i,t = ni,t /Ni
OSC SC SC (4)
(1)
OFC i,t /Ni
nFC FC
i,t = (2)
4) SELECTION OF CHARGING STATION BY EV DRIVERS
According to the research presented in [19], EV drivers will
2) ROAD TRAFFIC FLOW STATE
be influenced by traveling time and charging cost when
The road traffic flow state is an indication of the busyness
they select CS to charge their vehicles. Certain researches
level of a road, which is calculated using current traffic flow
(e.g. [34], [35]) established a probability model for the selec-
and maximum road capacity. It is also called vehicle density.
tion of CSs using linear models based on the price model
Eq. (3) shows the expression of traffic flow rate.
presented in [36]. Several economists further improved the
Rm,t = nRm,t /Nmmax (3) model to a non-linear one [37]. The non-linear model is
considered as more suitable to represent the consumer behav-
3) EV CHARGING PROBABILITY ior in real-world (e.g. people will not easily switch to an
EV drivers are expected to charge their EVs once the State of alternative product with a small price difference unless a
Charge (SoC) is low and the model of EV charging probabil- threshold value is exceeded). In this paper, the willingness
ity is established below based on the model proposed in [33]. of the jth EV driver to choose the ith CS considering price

13164 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

5) DYNAMIC PRICING MODEL

The dynamic pricing methodology proposed in this paper
aims to facilitate renewable energy consumption near
the CSs.
A number of rules are preset to make the pricing method-
ology more realistic:
1) If generation from renewable resources exceeds the
forecasted charging demand and the charging demand
is within the maximum capacity of the CS, the charging
price will decrease.
2) The reduced price will not be lower than the Feed-in-
Tariff (FIT) of renewable energy to ensure the motiva-
tions of CSs to participate into the pricing mechanism.
3) If the renewable generation cannot satisfy the predicted
FIGURE 2. Probability Factor of choosing alternative charging stations
charging demand, the charging price will be maintained
with certain distance and charging price. the same as CSs that have no renewable generation
nearby or onsite.
4) In extreme circumstances where the charging demand
difference is modelled by (5)-(6). forecast is beyond the maximum capacity of the CS,
the charging price will also be kept consistent with that
1
wci,j,t = lj ∗ Si,j − + oc (5) adopted at other CSs.
e−f (pi,t ,p¯t )+α0
pi,t − p¯t To enhance users’ motivations, a surplus-vs-demand based
f (pi,t , p¯t ) = α1 ∗ (6) pricing model ensuring a lower price when compared with the
p¯t
market price is proposed in this paper.
where f (pi,t , p¯t ) is a function of the charging price per kWh
of the CS. λi,t = pt − q(t, γi,t ) (11)
The willingness to select the ith CS considering the distance
s.t.
is modelled by (7)-(8).
1 λi,t = pt , ifq(t, γi,t ) < 0 (12)
wdi,j,t = lj ∗ Si,j − + od (7)
e−f (di,j,t )+α2 λi,t = λFIT , if pt − q(t, γi,t ) ≤ λFIT (13)
f (di,j,t ) = α3 ∗ (1 − (α4 ∗ di,j,t + α5 )−1 ) (8) pGi,t
γi,t = D (14)
pi,t
For EV drivers who need to charge their EVs, the willing-
ness function (which depends on the price and distance) to q(t, γi,t ) = ai,t · γi,t
2
+ bi,t · γi,t + ci,t (15)
select a specific CS is expressed in (9).
In Eq. (11), q(t, γi,t ) denotes the price reduction part
wi,j,t = ω1 ∗ wci,j,t + ω2 ∗ wdi,j,t (9) related to the surplus-vs-demand ratio. Eq. (12) and (13) are
used to ensure that the dynamic price will not be lower than
From the aforementioned charging willingness model, the renewable FIT and not be higher than the charging price
when EV drivers need to recharge their EVs, they will select a of CSs with no renewable generation. Eq. (15) is the dynamic
CS with the highest willingness value within the willingness price reduction equation.
value set as shown in (10).
B. OPERATION OF THE PRICING METHODOLOGY IN THE
C = max(G(wi,j,t )) (10)
ROAD NETWORK
The willingness distribution of the EV driver versus the 1) ASSUMPTIONS
price difference and cost is shown in Figure2. CS selection behaviors are modelled in the previous section
It can be observed in Figure 2 that most of EV drivers and the following assumptions are made for the operation of
will not choose an alternative CS if the percentage of the the proposed pricing methodology.
price difference is low. For the distance factor, similar to the Assumption 1: It is assumed that the EV drivers can sub-
price factor, EV owners will usually not choose to charge scribe for charging service to obtain real-time information
their EV if the CS is far away from them. The situation will of the EV CSs and the EV CSs will be able to receive the
dramatically change when the discounted price reaches the charging booking requests from the EV drivers. This com-
threshold value (e.g. 0.5 $/kWh) or the distance is shorter munication can be realized by the information management
than 2km. The willingness of EV owners will be improved system and is realistic considering the rapid development of
significantly. the charging service and the navigation apps.

VOLUME 8, 2020 13165

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

Step 3: If the analysis of Step 2 indicates that the predicted

charging demand is lower than the renewable generation,
the charging price will be changed by the pricing mechanism.
Step 4: The updated charging price information will be
broadcasted.
Step 5: The CS will wait for charging requests from EV
users.
Step 6: If there is an available charging space at the CS
upon the receipt of a charging request, the space will be
booked for that EV user. Otherwise, if no charging space is
free, the charging request will be declined. Note that the CS
will keep waiting for the charging requests if it is not fully
occupied.

III. HIGHWAY TRAFFIC FLOW AND RENEWABLE

ENERGY FORECAST
FIGURE 3. Workflow of the proposed pricing methodology.
This section presents how highway traffic flow and renewable
energy generation can be forecasted. The highly accurate
forecasting methodology can assist the implementation of the
Assumption 2: It is assumed that historical information proposed pricing methodology for EV CSs.
about the renewable energy and traffic flows can be obtained
from the Met Office and Transportation Bureau, respectively. A. HIGHWAY TRAFFIC FLOW FORECAST
The data will be processed and used for prediction in an To predict hourly traffic flow, an artificial neural network
higher level data aggregator which is implemented using (ANN) model is built using the historical traffic flow data
cloud computer servers. from January to April in 2019 collected from the Open Data
Assumption 3: It is assumed that an EV driver will choose Portal of Transport Infrastructure Ireland.
the CS with high willingness value. Since the traffic conditions differ on weekdays and week-
Assumption 4: It is assumed that the EV driver will go to ends, the historical traffic data is split into two parts, consist-
the selected CS to charge their cars immediately after the ing of 65 weekdays and 21 weekends. The ANN model is
charging requests are accepted by the CS. a three-layer feed backward neural network with one hidden
Assumption 5: It is assumed that the CS near the renewable layer containing 60 neurons. The activation function of the
generation has the priority to use the associated renewable rectified linear units is applied in the network and the L-BFGS
energy. algorithm is selected as the solver for the weight optimization
Assumption 6: It is assumed that the CSs in areas without because it has great performance when dealing with small
renewable generation will not change their charging price, datasets.
which is consistent with the regional electricity price. That The network model uses the square error loss function L to
is, the TOU charging price is adopted and the charging price evaluate the distance difference between the target and pre-
differs in peak hours and off peak hours. dicted values, which allows candidate solutions of the training
procedure to be ranked and compared. Eq. (16) presents the
2) OPERATION OF THE PRICING METHODOLOGY loss function L.
The comprehensive workflow of the proposed pricing scheme 1 β
is presented in Figure 3. The number of pricing time slots is L(n, n̂, W ) = kn − n̂ k22 + k W k22 (16)
2 2
defined as 24 or 48 for each day, and the CS will periodically
broadcast messages containing information about the formu- where β k W k22 is an L2-regularization term that penalizes
lated price and the availability before and during that time. complex models.
In each time slot, the price will be unchanged. Steps of the Simulations are performed to determine the optimal num-
workflow are described and explained below. ber of neurons in input and output layers for training the ANN
Step 1: When the pricing scheme starts operating, the CS model. The neurons of input and output layers range from
will first collect the latest data of traffic flow, renewable 1 to 24. The coefficient of determination R2 is selected as the
generation and charging price without adjustment attributed evaluation score for each training.
to renewable energy. PN −1
(yi − ŷi )2
2
R y, ŷ = 1 − Pi=0


Step 2: Using the trained forecasting models and collected N −1 2
(17)
data, the CS will predict charging demand and renewable i=0 (yi − ȳi )
generation for the next time slot. In this way, the gap between The performance of the ANN model with different input
the required charging capacity and CS’s maximum charging and output neurons on weekdays and weekends is presented
capacity can be identified. in Figure 4. The y-axis represents the number of neurons in

13166 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 4. Performance of ANN model with different input and output

neurons: (a) weekdays; (b) weekends.

TABLE 1. Numerical summary of optimal number of input and output

neurons.

FIGURE 5. Schematic diagram of data processing based on ANN.

the output layer. The x-axis represents the number of neurons

in the input layer. The z-axis represents the R2 score for each
training. The color of the bar represents the interval of the
score (blue, R2 < 0.8; green, 0.8 ≤ R2 ≤ 0.85; yellow,
0.85 ≤ R2 < 0.9; red, 0.9 ≤ R2 < 0.95; purple, 0.95 ≤ R2 ). FIGURE 6. Prediction of hourly traffic flow on weekdays: (a) single
weekday; (b) five consecutive weekdays; (c) all weekdays.
The simulation results of training the ANN model on
weekdays and weekends demonstrate that the R2 score is
enhanced with more input neurons and fewer output neurons.
The negative value of R2 represents that it is practical to
replace the prediction with the mean of target value for a
better model. Table 1 provides a summary of the optimal
number of input and output neurons with the highest R2 score
for training the ANN model.
FIGURE 7. Prediction of hourly traffic flow on weekends: (a) single day of
From Table 1, it can be found that the optimal numbers of a weekend; (b) one weekend; (c) all weekends.
neurons for the input and output layer are 24 and 1, regardless
of weekday or weekend. The R2 score reaches over 0.99 when
TABLE 2. Accuracy of traffic flow predictions.
the best score is specified as 1, which brings perfect training
for the model.
Considering the optimal number of the input/output neu-
rons, the historical traffic data is split into a training set and
a testing set by a ratio of 7:3 for training the ANN model.
Figure 5 shows a schematic diagram of data processing.
Figure 6(a) gives a prediction of hourly traffic flow on The average R2 scores of the testing set on weekdays and
a single weekday. Figure 6(b) presents a prediction of traf- weekends are calculated separately.
fic flow on five consecutive weekdays. Figure 6(c) shows
a prediction of all weekdays. The red curve represents the B. RENEWABLE ENERGY FORECAST
prediction results, and the blue curve denotes the actual value The proposed data processing method of finding the optimal
of the target. number of input/output neurons based on ANN is also veri-
Figure 7(a) gives a prediction of hourly traffic flow on fied to be practical and precise in renewable energy (solar and
a single weekend day. Figure 7(b) presents a prediction of wind power) prediction. The historical data of solar and wind
traffic flow on a weekend. Figure 7(c) shows a prediction of generation are collected from an industrial park in Jiangsu,
all weekends. China.
To comprehensively evaluate the performance of the ANN Figure 8 gives the prediction results of solar power for
model, a numerical summary of the R2 score of weekday a single day and a week, respectively. Figure 9 shows the
and weekend predictions is conducted as shown in Table 2. performance of the ANN model on wind power prediction.

VOLUME 8, 2020 13167

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

TABLE 4. Parameters of charging stations.

FIGURE 8. Prediction of solar power: (a) single day; (b) one week.

TABLE 5. Time-of-use (TOU) tariff.

FIGURE 9. Prediction of wind power: (a) single day; (b) one week.

TABLE 3. Accuracy of SOLAR and wind power prediction.

TABLE 6. Types of renewable energy generation and their rated power.

The red cure represents the prediction and the blue curve
denotes the actual value of the target.
The average R2 scores of the testing set are calculated sep-
arately for the solar and wind power prediction. Table 3 gives
the statistical summary of R2 scores for the solar and wind
power prediction.

IV. CASE STUDY

A. ROAD NETWORK TESTBED ESTABLISHMENT
To evaluate whether the proposed pricing methodology is
effective for facilitating renewable energy and improving the TABLE 7. Feed-in-tariff (Fit) minimum values of renewable energy types.
usage of CSs, a testbed based on the real traffic network of
Dublin’s metropolitan circle is established. The traffic data
is collected from the Dublin traffic agency, and the locations
of CSs refer to the actual refueling stations in the area. The
Dublin road network used in this paper is modelled as shown
in Figure 10.
It is assumed that a total of 13 CSs are located in the
modelled road network. The number of fast-charging and
slow-charging facilities of the 13 CSs and their available For renewable generation, the data are collected separately
slots are given in Table 4. Aiming to evaluate whether the from a wind power project and a solar project in Ireland.
proposed pricing methodology can facilitate the consumption The installed capacity of wind farms and PV power stations
of renewable energy, we assume that C4 and C13 are located nearby and their typical hour-by-hour generation data are
in a wind-rich area (near an off-shore wind farm), C7 is within shown in Table 6 and Figure 11.
a wind-rich area (near an on-shore wind farm), and C6, C7, It should be noted that the lowest charging price follows the
and C13 are in a solar-rich area (near a solar farm). Other FIT of renewable energy. The FITs for on-shore wind power,
CSs are considered as neutral, and are not near renewable off-shore wind power, and solar power are given in Table 7.
resources, which means the charging price follows the TOU According to the forecast traffic flow information on the
tariff. The TOU tariff used in this paper is shown in Table 5. seven highways, we assume that 1% of the vehicles on the

13168 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 10. Traffic network testbed based on Dublin area.

FIGURE 12. Number of EVs on seven highways to be charged

hour-by-hour.

be found that the charging peak time periods for EVs are
07:00-9:00 and 16:00-18:00.

B. NUMERICAL RESULTS OF PRICING METHODOLOGY

PERFORMANCE ON ESTABLISHED TESTBED
The testbed of the Dublin traffic network is established
based on the data and parameters presented in part A above.
FIGURE 11. Forecast output of renewable-rich charging stations: (a) in
wind-rich areas; (b) in solar-rich areas. The prices of CSs in renewable-rich areas can be simulated
based on the workflow and pricing methodology described
road are EVs that need to be charged, and their positions in section II, and the prices of CSs are given in Figure 13.
are generated randomly. Figure 12 presents the typical hour- It shows the prices of C4, C10 and C12 in wind-rich areas
by-hour traffic flow data of the seven highways. It can and the prices of C6, C7 and C13 in solar-rich areas.

VOLUME 8, 2020 13169

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

The traffic flow peak period from 07:00 to 09:00 par-

tially overlaps with the electricity peak period, when CSs
C6, C7 and C13 have the greatest effect on the traffic flow.
Figure 17 depicts the traffic flow data between 08:00 and
09:00.
In addition to the morning peak period, the traffic flow will
also enter a peak time from 16:00 to 18:00 in the afternoon.
During this period, the traffic flow is dominantly influenced
by CSs C4, C6, C7 and C13, as indicated by the traffic flow
data between 16:00 and 17:00 in Figure 18.
During the electricity peak period at night from 17:00 to
21:00, only CS C4 obviously affects the traffic flow. Figure 19
demonstrates the traffic flow data between 20:00 and 21:00.
The electricity off-peak period at night lasts from 21:00 to
00:00 and CS C10 affects the traffic flow in addition to C4.
The example traffic flow data between 21:00 and 22:00 is
shown in Figure 20.
In the above typical time slots, different CSs play the main
role in diverting traffic flow, which is related to different
levels of renewable generation, traffic condition and charging
prices. The CSs that obviously change EV drivers’ selection
in typical time slots are listed in Table 8.
FIGURE 13. Charging prices of renewable-rich charging stations: (a) in
wind-rich areas; (b) in solar-rich areas. C. ANALYSIS OF PRICING METHODOLOGY
IN WIND-RICH AREAS
According to Figure 13(a), the CSs in wind-rich areas (C4,
As illustrated in [20], the results indicate that EV owners C10 and C12) have competitive prices during three time peri-
will select the nearest CS to charge their EV if there is no ods 00:00-07:00, 12:00-17:00 and 19:00-00:00. As surplus
obvious change in the charging price. To explore whether wind energy usually arises at night, the CSs can facilitate
the proposed pricing methodology will affect the usage of wind energy consumption during 19:00-00:00. Figure 15(a)
charging facilities in renewable-rich areas, the outcomes of shows that C4 and C10 can increase the wind energy con-
CS selection with and without the pricing methodology are sumption rate by up to 82.97% and 11.72%, respectively. The
compared. reason why C12 cannot significantly improve wind energy
Figures 14 illustrate the change in utilization of six consumption is because the wind generation capacity located
renewable-rich CSs for each individual time slot. Looking at near C12 is relatively small.
the purple line, it is clear that the utilization rate of CSs C4, Furthermore, as 00:00-07:00 is the electricity valley time
C6 and C7 significantly grows whilst a decrease in utilization period of TOU, and the traffic flow on the roads is at a low
rate is observed for C12. Besides, C10 and C13 experience level, the price advantages of C4, C10 and C12 in wind-rich
both an increase and decrease in the utilization rate during areas are not obvious.
specific time slots. During 12:00-17:00, C6, C7 and C13 in solar-rich areas
The amount of renewable generation consumed in each also have price advantages, thus C4, C10 and C12 become
time slot by the aforementioned six CSs is reported by bar less attractive to EV drivers.
charts in Figure 15 whilst the increase rate of renewable In the period from 19:00 to 21:00, solar panels cannot gen-
energy consumption resulting from the application of the erate any power and the price advantages of C4 and C10 are
proposed pricing methodology is represented by the lines. improved accordingly. As seen in Figure 19, C4 attracts
Figure 15(a) shows the wind energy consumption of C4, a number of EVs from the M2, M3, M4 and M7 roads.
C10 and C12 whilst Figure15(b) shows that of C6, C7 and In normal conditions, EV drivers on their original roads will
C13. select the nearest CS to charge. After the pricing methodology
Information for typical time slots is selected in order to is adopted, the EVs will move to C4 to charge due to the
analyze the impact of the proposed pricing methodology on cost benefit. C10 is located on the M50 road, which is a
the consumption of wind and solar energy as well as the traffic motorway with low level traffic flow and is relatively far
flow of EVs waiting for recharging. away from the road junction. The long distance has a larger
In the electricity valley period from 0:00 to 08:00, negative impact on drivers’ willingness and C10 thus cannot
renewable-rich CSs negligibly affect the traffic flow, which effectively influence the traffic flow (i.e. attract considerable
can be reflected by the traffic flow data between 02:00 and amount of EVs from other roads) even though it has a price
03:00, as shown in Figure 16. advantage.

13170 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 14. Usage of six renewable-rich charging stations without/with proposed pricing methodology: (a) C4; (b) C10; (c) C12;
(d) C6; (e) C7; (f) C13.

VOLUME 8, 2020 13171

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 14. (Continued.) Usage of six renewable-rich charging stations without/with proposed pricing methodology: (a) C4; (b) C10;
(c) C12; (d) C6; (e) C7; (f) C13.

TABLE 8. Charging stations that obviously work in typical time slots.

With reference to Figure 20, C10 is able to affect the traffic trend is consistent with the PV generation curve. Figure15(b)
flow by attracting more EVs from 21:00 to 00:00. This is shows that solar-rich CSs also facilitate solar energy con-
because C10 has a greater price advantage due to higher sumption during 7:00-17:00 and the maximum increase rate
wind generation and reduced road traffic. The attractiveness is identified around 13:00. C6, C7 and C13 can increase
of C4 during this electricity valley period is reduced com- the solar energy consumption by up to 59.50%, 13.19% and
pared to 19:00-21:00 since the electricity price is the cheapest 14.19%, respectively.
according to the TOU tariff scheme and the competitiveness It is worth noting that C6, C7 and C13 have more compet-
of C4 on price is reduced. itive prices than other CSs during 07:00-13:00. More specifi-
cally, Figure 17 indicates that C6 attracts EVs from M2, M3,
D. ANALYSIS OF PRICING METHODOLOGY IN M7, N81 and M50 roads and these EVs are supposed to go
SOLAR-RICH AREAS to other CSs if the pricing methodology is not implemented.
According to Figure 13(b), the price advantages of C6, C7 and EV drivers on M4 road planning to go to C11 in normal
C13 are noticeable in the period of 07:00-17:00 and such condition are attracted to C7 whilst those on the M7 and

13172 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 15. Renewable consumption and increase ratios for charging stations: (a) in wind-rich areas and (b) in solar-rich areas.

FIGURE 16. Effect of proposed methodology on traffic flow at 02:00-03:00.

N81 roads are attracted to C13. In this period, C6 is dom- generation is similar. However, C7 is located on the M4 road
inant in attracting EVs. Compared with C13, C6 is located carrying the densest traffic flow, which means C7 needs to
at a transportation hub and its distance has less negative satisfy massive charging demand. This inflates the charg-
impact on drivers’ willingness. Therefore, C6 can attract ing price of C7 and reduce its price advantage, compared
more EVs than C13. For both C6 and C7, the amount of solar to C6.

VOLUME 8, 2020 13173

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 17. Effect of proposed methodology on traffic flow at 08:00-09:00.

FIGURE 18. Effect of proposed methodology on traffic flow at 16:00-17:00.

In the period 13:00-17:00, CS C4, C10 and C12 are able E. INFLUENCE OF PRICING METHODOLOGY ON
to compete with C6, C7 and C13 from the perspective of TRAFFIC DIVERSION
charging price. Even though both wind-rich and solar-rich During the traffic flow peak time in the morning between
CSs have the largest price advantages during this period, C6, 07:00-09:00, wind generation is at the lowest level, which
C7 and C13 are more competitive due to the lower FIT of neutralizes the price advantages of C4, C10 and C12. In such
renewable energy. According to Figure 9, C4 attracts a portion circumstances, C6 and C13 fully utilize their capability to
of EVs from the M1 road whilst C6, C7 and C13 attract divert the traffic flow where a certain amount of EVs waiting
certain amount of EVs from the M2, M3, M4, M7, N81 and for recharging are transferred from the congested M4 and
M50 roads. M7 roads to the less busy M3 and N81 roads. Among solar-

13174 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FIGURE 19. Effect of proposed methodology on traffic flow at 20:00-21:00.

FIGURE 20. Effect of proposed methodology on traffic flow at 21:00-22:00.

rich CSs, C7 has the least influence for traffic flow diversion generation almost ceases due to the sunset and C6 no longer
because it is located in the most congested road. has a price advantage. Therefore, its capability to divert the
During the traffic flow peak time in the afternoon from traffic flow is dramatically weakened.
16:00 to 17:00, C6 plays a crucial role in traffic flow diversion During other traffic flow off-peak periods from 9:00 to
where a certain amount of EVs waiting for recharging are 15:00 and 15:00 to 16:00, traffic diversion is mainly accom-
transferred from the congested M4 and M7 roads to the less plished by the solar-rich CS C6. Between 19:00 and 00:00,
busy M3 road. Although there are price advantages at C4, CSs C4 and C10 make the largest contribution to divert-
C10 and C12, the distance to these CSs noticeably frus- ing the traffic flow. More specifically, C4 transfers a por-
trates the willingness of EV drivers, reducing their capability tion of EVs waiting for recharging from the M2 and
to divert the traffic flow. In the period 17:00-18:00, solar M3 roads to the less congested M1 whilst C10 reduces the

VOLUME 8, 2020 13175

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

amount of EVs on the N81 by distributing them to the less c) Other traffic flow periods
dense M50. During the off-peak periods of 09:00-15:00 and
In the period 00:00-07:00 when the excess wind energy 15:00-16:00 in the daytime, traffic flow diversion is
needs to compensate the generation gap caused by the lack of mainly realized by CS C6. In the off-peak period
solar power, the contribution of CSs to traffic flow diversion between 19:00 and 00:00 during the night time, CSs
is not obvious. This is because there is only a small amount C4 and C10 optimize the traffic flow.
of EVs on the road and the charging spaces are distant from
(3) Other suggestions/findings from the numerical
wind-rich C4, C10 and C12 which discourages the intention
simulation
of EV drivers.
It is essential to consider the location and the traffic flow
V. CONCLUSION
before constructing renewable-rich CSs. Restrictive require-
In this paper, a pricing methodology for CSs in renewable- ments already exist for the realization of renewable gener-
rich areas is proposed. The pricing methodology is evaluated ation. For example, off-shore infrastructure and platforms
on a testbed established using real data of the Dublin traffic need to be built for off-shore wind farms whilst large-scale
network. Conclusions are made below from the perspective of solar stations will require open terrain to accommodate a
facilitating renewable consumption and balancing the traffic large number of solar panels. Based on the case study results,
flow. several suggestions are made for the Dublin area.
(1) The performance of the proposed pricing methodology a) Wind-rich charging station
in terms of facilitating renewable consumption For CS C12 discussed in this paper, the total capacity
Based on the case study results, the level of renew- of the associated wind turbines is relatively small
able consumption can be improved by applying the pricing leading to smaller competitive price advantages. It is
methodology. suggested to increase the capacity of the on-shore
a) Improving wind and solar energy consumption wind turbines near C12 to enhance its capability to
In the time periods 15:00-17:00 and 19:00-00:00, facilitate renewable energy consumption and traffic
wind-rich CS C4 can improve the level of wind flow. For CSs C4 and C10 that are near off-shore
energy consumption by up to 82.97% and the aver- wind farms, it is advised to place them as close to the
age improvement is 30.73%. The solar-rich CS road junction as possible. In this way, traffic on con-
C6 can increase the solar energy consumption by gested roads can be diverted to less busy roads such as
up to 59.50% and an average increase of 29.28% is M1 and M50 and off-shore wind energy consumption
achieved. can be improved.
b) Optimal time period for improving wind and solar b) Solar-rich charging station
energy consumption In terms of CSs that are built on congested roads (e.g.
The price advantages of wind-rich CSs C4, C10 and C7), it is suggested to reduce the number of this type
C12 mainly arise when the wind speed is at the peak of stations when planning future infrastructure con-
value in the afternoon and evening. Considering the struction. CSs that are close to renewable generation
daytime when the price advantages of solar-rich CSs and located on less congested roads (e.g. C6) should
are also competitive, the actual periods when the traf- be promoted because they can easily meet the target
fic is diverted to facilitate wind energy consumption of optimizing the traffic flow and consuming larger
are 15:00-17:00 and 19:00-00:00. amount of renewable energy.
In terms of solar-rich CSs C6, C7 and C13, price c) Complementarity between wind and solar
advantages are obtained during the daytime with a consumption
strong level of irradiation. In order to better facilitate renewable energy con-
(2) The improvement of traffic flow during peak hours and sumption and balance the traffic flow using the pro-
off-peak hours posed pricing methodology, it is advised to take
The proposed pricing methodology can assist the optimiza- into account the complementarity between wind and
tion of traffic flow during peak and off-peak hours and can solar consumption when planning the construction
reflect the complementarity between wind energy and solar of renewable-rich CSs. Both wind-rich and solar-rich
energy consumption. CSs should be built near the roads with convenient
a) Morning traffic flow peak hours interconnections and appropriate traffic flow.
In the peak period between 07:00 and 09:00, CSs
C6 and C13 mainly divert the traffic flow. REFERENCES
b) Evening traffic flow peak hours [1] J. Meckling and J. Nahm, ‘‘The politics of technology bans: Industrial
During the first evening peak period from 16:00 to policy competition and green goals for the auto industry,’’ Energy Policy,
vol. 126, pp. 470–479, Mar. 2019.
17:00, CS C6 diverts the traffic flow by transferring
[2] A. Adepetu and S. Keshav, ‘‘The relative importance of price and driving
EVs waiting for recharging from the congested range on electric vehicle adoption: Los Angeles case study,’’ Transporta-
M4 and M7 roads to the less dense M3 road. tion, vol. 44, no. 2, pp. 353–373, Mar. 2017.

13176 VOLUME 8, 2020

S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

[3] Z. Yu, S. Chen, and L. Tong, ‘‘An intelligent energy management system [24] E. Azadfar, V. Sreeram, and D. Harries, ‘‘The investigation of the major
for large-scale charging of electric vehicles,’’ CSEE Power Energy Syst., factors influencing plug-in electric vehicle driving patterns and charg-
vol. 2, no. 1, pp. 47–53, Mar. 2016. ing behaviour,’’ Renew. Sustain. Energy Rev., vol. 42, pp. 1065–1076,
[4] B. Nykvist, F. Sprei, and M. Nilsson, ‘‘Assessing the progress toward Feb. 2015.
lower priced long range battery electric vehicles,’’ Energy Policy, vol. 124, [25] D. Chang, ‘‘Financial viability of non-residential electric vehicle
pp. 144–155, Jan. 2019. charging stations,’’ Luskin Center Innov., Los Angeles, CA, USA,
[5] J. Xiong, K. Zhang, Y. Guo, and W. Su, ‘‘Investigate the impacts of PEV Tech. Rep. 90095-1656, 2012.
charging facilities on integrated electric distribution system and electrified [26] X.-H. Sun, T. Yamamoto, and T. Morikawa, ‘‘Fast-charging station choice
transportation system,’’ IEEE Trans. Transport. Electrific., vol. 1, no. 2, behavior among battery electric vehicle users,’’ Transp. Res. D, Transp.
pp. 178–187, Aug. 2015. Environ., vol. 46, pp. 26–39, Jul. 2016.
[6] W. Su, H. Eichi, W. Zeng, and M.-Y. Chow, ‘‘A survey on the electrification [27] Y. Yang, E. Yao, Z. Yang, and R. Zhang, ‘‘Modeling the charging and route
of transportation in a smart grid environment,’’ IEEE Trans. Ind. Informat., choice behavior of BEV drivers,’’ Transp. Res. C, Emerg. Technol., vol. 65,
vol. 8, no. 1, pp. 1–10, Feb. 2012. pp. 190–204, Apr. 2016.
[7] F. Mwasilu, J. J. Justo, E.-K. Kim, T. D. Do, and J.-W. Jung, ‘‘Electric [28] H. Yang, Y. Deng, J. Qiu, M. Li, M. Lai, and Z. Y. Dong, ‘‘Electric vehicle
vehicles and smart grid interaction: A review on vehicle to grid and renew- route selection and charging navigation strategy based on crowd sensing,’’
able energy sources integration,’’ Renew. Sustain. Energy Rev., vol. 34, IEEE Trans. Ind. Informat., vol. 13, no. 5, pp. 2214–2226, Oct. 2017.
pp. 501–516, Jun. 2014. [29] S. Bessler and J. Grønbæk, ‘‘Routing EV users towards an optimal charging
[8] M. Amini and O. Karabasoglu, ‘‘Optimal operation of interdependent plan,’’ World Electr. Vehicle J., vol. 5, no. 3, pp. 688–695, May 2018.
power systems and electrified transportation networks,’’ Energies, vol. 11, [30] F. Malandrino, C. Casetti, C.-F. Chiasserini, and M. Reineri, ‘‘A game-
no. 1, p. 196, Jan. 2018. theory analysis of charging stations selection by EV drivers,’’ Perform.
[9] C. Li, C. Liu, K. Deng, X. Yu, and T. Huang, ‘‘Data-driven charging Eval., vols. 83–84, pp. 16–31, Jan. 2015.
strategy of PEVs under transformer aging risk,’’ IEEE Trans. Control Syst. [31] B. Wang, B. Hu, C. Qiu, P. Chu, and R. Gadh, ‘‘EV charging algorithm
Technol., vol. 26, no. 4, pp. 1386–1399, Jul. 2018. implementation with user price preference,’’ in Proc. IEEE Power Energy
Soc. Innov. Smart Grid Technol. Conf. (ISGT), Feb. 2015.
[10] T. H. Bradley and A. A. Frank, ‘‘Design, demonstrations and sustainability
[32] Q. Huang, Q.-S. Jia, Z. Qiu, X. Guan, and G. Deconinck, ‘‘Matching EV
impact assessments for plug-in hybrid electric vehicles,’’ Renew. Sustain.
charging load with uncertain wind: A simulation-based policy improve-
Energy Rev., vol. 13, no. 1, pp. 115–128, Jan. 2009.
ment approach,’’ IEEE Trans. Smart Grid, vol. 6, no. 3, pp. 1425–1433,
[11] B. Bilgin, P. Magne, P. Malysz, Y. Yang, V. Pantelic, M. Preindl,
May 2015.
A. Korobkine, W. Jiang, M. Lawford, and A. Emadi, ‘‘Making the case
[33] Y.-M. Wi, J.-U. Lee, and S.-K. Joo, ‘‘Electric vehicle charging method for
for electrified transportation,’’ IEEE Trans. Transport. Electrific., vol. 1,
smart homes/buildings with a photovoltaic system,’’ IEEE Trans. Consum.
no. 1, pp. 4–17, Jun. 2015.
Electron., vol. 59, no. 2, pp. 323–328, May 2013.
[12] M. H. Amini, J. Mohammadi, and S. Kar, ‘‘Distributed holistic framework [34] W. Yuan, J. Huang, and Y. J. Zhang, ‘‘Competitive charging station pricing
for smart city infrastructures: Tale of interdependent electrified transporta- for plug-in electric vehicles,’’ IEEE Trans. Smart Grid, vol. 8, no. 2,
tion network and power grid,’’ IEEE Access, vol. 7, pp. 157535–157554, pp. 627–639, Mar. 2017.
2019. [35] F. Kong, X. Liu, and I. Lee, ‘‘Joint rate control and demand balancing
[13] M. H. Amini, M. P. Moghaddam, and O. Karabasoglu, ‘‘Simultaneous for electric vehicle charging,’’ in Proc. IEEE/ACM 3rd Int. Conf. Internet-
allocation of electric vehicles’ parking lots and distributed renewable Things Design Implement. (IoTDI), Apr. 2018.
resources in smart power distribution networks,’’ Sustain. Cities Soc., [36] R. S. Winer, ‘‘A reference price model of brand choice for frequently
vol. 28, pp. 332–342, Jan. 2017. purchased products,’’ J. Consum. Res., vol. 13, no. 2, p. 250, Sep. 1986.
[14] H. N. T. Nguyen, C. Zhang, and J. Zhang, ‘‘Dynamic demand control [37] G. J. Tellis, ‘‘Advertising exposure, loyalty, and brand purchase: A two-
of electric vehicles to support power grid with high penetration level stage model of choice,’’ J. Marketing Res., vol. 25, no. 2, p. 134, May 1988.
of renewable energy,’’ IEEE Trans. Transport. Electrific., vol. 2, no. 1,
pp. 66–75, Mar. 2016.
[15] G. Pan, W. Gu, Z. Wu, Y. Lu, and S. Lu, ‘‘Optimal design and operation
of multi-energy system with load aggregator considering nodal energy
prices,’’ Appl. Energy, vol. 239, pp. 280–295, Apr. 2019. SUYANG ZHOU received the B.Eng. degree from
[16] M. Alizadeh, H.-T. Wai, A. Goldsmith, and A. Scaglione, ‘‘Retail the Huazhong University of Science and Tech-
and wholesale electricity pricing considering electric vehicle mobil- nology, in 2009, and the Ph.D. degree from the
ity,’’ IEEE Trans. Control Netw. Syst., vol. 6, no. 1, pp. 249–260, University of Birmingham, in 2015, all in electrical
Mar. 2019. engineering. He is currently a Lecturer with the
[17] H. Yang, S. Yang, Y. Xu, E. Cao, M. Lai, and Z. Dong, ‘‘Electric vehicle School of Electrical Engineering, Southeast Uni-
route optimization considering time-of-use electricity price by learnable versity, Nanjing. Prior to joining Southeast Uni-
partheno-genetic algorithm,’’ IEEE Trans. Smart Grid, vol. 6, no. 2, versity, he worked as the KTP Associate with the
pp. 657–666, Mar. 2015. University of Leicester and a Research and Devel-
[18] M. Schneider, A. Stenger, and D. Goeke, ‘‘The electric vehicle-routing opment Engineer with Cellcare Technology Ltd.,
problem with time windows and recharging stations,’’ Transp. Sci., vol. 48, from 2015 to 2016. He worked as a Data Scientist with the Power Networks
no. 4, pp. 500–520, Nov. 2014. Demonstration Centre (a joint Research Centre between the University of
[19] E. S. Rigas, S. D. Ramchurn, N. Bassiliades, and G. Koutitas, ‘‘Congestion Strathclyde and U.K. distribution network operators), from 2016 to 2017.
management for urban EV charging systems,’’ in Proc. IEEE Int. Conf. His main research interests include integrated energy systems, artificial
Smart Grid Commun. (SmartGridComm), Oct. 2013.
intelligence in electrical domain, and the demand side management.
[20] M. Gharbaoui, L. Valcarenghi, R. Brunoi, B. Martini, M. Conti, and
P. Castoldi, ‘‘An advanced smart management system for electric vehicle
recharge,’’ in Proc. IEEE Int. Electr. Vehicle Conf., Mar. 2012.
[21] Y. Cao, T. Wang, O. Kaiwartya, G. Min, N. Ahmad, and A. H. Abdullah,
‘‘An EV charging management system concerning drivers’ trip duration YUE QIU received the B.Eng. degree in electrical
and mobility uncertainty,’’ IEEE Trans. Syst., Man, Cybern., Syst., vol. 48, engineering from the Nanjing University of Aero-
no. 4, pp. 596–607, Apr. 2018. nautics and Astronautics, Nanjing, China, in 2018.
[22] H. Qiu, B. Zhao, W. Gu, and R. Bo, ‘‘Bi-level two-stage robust optimal She is currently pursuing the master’s degree with
scheduling for AC/DC hybrid multi-microgrids,’’ IEEE Trans. Smart Grid, the School of Electrical Engineering, Southeast
vol. 9, no. 5, pp. 5455–5466, Sep. 2018. University. Her research interests include the plan-
[23] H. Qiu, W. Gu, Y. Xu, Z. Wu, S. Zhou, and J. Wang, ‘‘Interval-partitioned ning methodologies and operational optimization
uncertainty constrained robust dispatch for AC/DC hybrid microgrids with of EV and integrated energy systems.
uncontrollable renewable generators,’’ IEEE Trans. Smart Grid, vol. 10,
no. 4, pp. 4603–4614, Jul. 2019.

VOLUME 8, 2020 13177

 S. Zhou et al.: Dynamic EV Charging Pricing Methodology for Facilitating Renewable Energy With Consideration

FENGHUA ZOU received the B.Eng. degree in XINER LUO is currently an Electrical Engineer with Shenzhen Power Supply
electrical engineering from the School of Elec- Bureau Company, Ltd. Her research interests mainly focus on EV and
trical Engineering, Southeast University, Nanjing, integrated energy systems.
China, in 2018, where he is currently pursuing the
master’s degree. His research interests include the
planning methodologies and operational optimiza-
tion of EV and integrated energy systems. CHENGLIANG WANG is currently an Electrical Engineer with Jiangsu
Fangtian Electric Technology Company, Ltd. His research interests mainly
focus on EV and integrated energy systems.

ZHI WU received the M.Sc. degree in electrical

DI HE received the B.Eng. degree in chemical engineering from the School of Electrical Engi-
engineering from the Dalian University of Tech- neering, Southeast University, Nanjing, China,
nology, Dalian, China, in 2018. He is currently in 2012, and the Ph.D. degree from the Univer-
pursuing the master’s degree with the School of sity of Birmingham, Birmingham, U.K., in 2016.
Cyber Science and Engineering, Southeast Univer- His research interests include renewable energy,
sity. His research interest includes the operational electric market, distribution network planning, and
optimization of integrated energy systems. optimization techniques.

WEI GU (Senior Member, IEEE) received the

B.Eng. and Ph.D. degrees in electrical engineering
PENG YU is currently an Electrical Engineer with Shenzhen Power Supply from Southeast University, China, in 2001 and
Bureau Company, Ltd. His research interests mainly focus on EV and 2006, respectively. From 2009 to 2010, he was a
integrated energy systems. Visiting Scholar with the Department of Electri-
cal Engineering, Arizona State University, Tempe,
AZ, USA. He is currently a Professor with the
School of Electrical Engineering, Southeast Uni-
versity. His research interests include distributed
JINQIAO DU is currently an Electrical Engineer with Shenzhen Power generations and microgrids and active distribution
Supply Bureau Company, Ltd. His research interests mainly focus on EV networks.
and integrated energy systems.

13178 VOLUME 8, 2020