International Journal of Smart Grid and Clean Energy

Optimization of pricing policy of electric vehicle charging

station based on big data
Xu Tiana, Yuan Xub, Xiangcheng Zhanga, Xue Maa, Su Zhangb
a
Economic & Technological Research Institute of State Grid Qinghai Power Co, 810000 XiNing, China
b
State Key Laboratory of Advanced Electromagnetic Engineering and Technology, Huazhong University of Science and Technology,
430074 WuHan, China

Abstract
The guideless charging behaviour of electric vehicles(EVs) will lead to load fluctuation and increase the difficulty of
power system control. By formulating a reasonable pricing strategy, the orderly charging behaviour of EVs can be
fully encouraged, which is conducive to improving the stability of the system and the economic benefits of charging
stations. Compared with the current pricing strategy optimization method, the pricing strategy based on big data will
provide a research methodology to capture the relationship between charging load and pricing strategy. Firstly, by
analysing EV user’s behaviour with big data, a charging load predicting model is proposed. Secondly, a causality
model is built to study the relationship between pricing strategy and EV load. Thirdly, the optimization model about
pricing strategy of charging station is established under three angles of minimum loss in power system, highest user
satisfaction and maximum benefit of charging station. Then, the optimization problem solved by using particle swarm
optimization(PSO). Finally, by taking the IEEE9 as an example, the validity of the EV charging load predicting
model and the feasibility of the charging station pricing strategy optimization model are analysed and verified.

Keywords: Electric vehicle, load forecasting, big data, charging station pricing policy

1. Introduction

Electric Vehicle (EV) has incomparable advantages over traditional vehicle in energy-saving and
emission-reduction. Therefore, the popularity of EVs is an inevitable trend. The disorderly charging
behaviour of EVs might have negative impacts to power grid such as overload and power quality
degradation. Thanks to the energy storage characteristic, a proper charging strategy of EVs can not only
depress or even eliminate the negative effects to power grid, but also improve the economic benefits of
charging stations. As the most important controllable factor, pricing policy is the key to guide the orderly
charging behaviour of EVs. To reduce the negative impacts and take full advantage of positive effects, it
is essential to study the relationship between time-space distribution of EV load and charging station
pricing policy.
In the study of pricing policy of charging station, Reference [1] considers the secure operation of
power grid and the individualized demand of EV users, a multi-objective charging optimization model
based on Time-of-Use(TOU) pricing policy is proposed, which considering the maximization economic
benefits of power grid and charging stations. Reference [2] based on the charging/discharging
characteristics of EV and demand theory, On the basis of analysing charging/discharging cost and benefit
of EVs, a peak-valley TOU pricing policy model is established, which aims at minimizing load
fluctuation. On the basis of considering both the benefit of power grid and the satisfaction of EV’s user, a
response model of EV’s charging power to electricity price changes is established in Reference [3], and
the optimal peak-valley TOU pricing policy is designed.

* Manuscript received June 6, 2019; revised January 17, 2020.

Corresponding author. E-mail address: [email protected].
doi: 10.12720/sgce.9.3.596-606
 Xu Tian et al.: Optimization of pricing policy of electric vehicle charging station based on big data 597
In previous studies, the optimization model of charging station pricing policy is based on optimizing
the system load, model not only ignores the individual differences of pricing policy in different charging
station but also neglects the individual differences of corresponding degree for each EV user. In this study,
the accurate extend of the impact of pricing policy can be determined by data-mining techniques which
based on massive dataset. Based on the results, the EV load might be regulated by devising different TOU
pricing policy for each charging station.
The remainder of the paper is organized as follows: Section 2 analyses the sources and characteristics
of big data of EV, then proposes a data mining-based model to generate the big data of EVs. Section 3
uses support vector regression(SVR) to model the relationship between the EV load and pricing policy,
which is the instant when the price of electricity changes. Section 4 establishes optimization model of
charging station pricing policy, the pricing policy of charging station is optimized by using particle
swarm optimization (PSO) algorithm. Section 5 presents our conclusions.

2 Data Mining-Based Model

Data Mining (DM) techniques are adopted to study the relationship between the EV load and various
factors. DM techniques always need a great quantity of input dataset. Since the big data in this case are
unavailable in the real world, the first step in modeling is to analyse the factors related to EV charging
load. The second step is to generate great quantity of simulation data. With this model, a massive
simulation dataset can be obtained.

2.1. Factors of EV load

Factors of EV load are divided into three aspects as following:

2.1.1. Factors of EV
In terms of a single EV, its charging load is mainly determined by the nature characteristics of battery,
the user’s habits and the travel demand. Mainly depends on the following factors:
1) Battery characteristics: Including the battery size, charge/discharge rate, self-discharge rate and
battery type, which determine the charging/discharging power and charging load curve.
2) User's habits: Including preferences in charging location, expected state of charge(SOC) and
charging mode. The difference in charging habits will disperse EV load in time and space.
3) Travel demand: Including travel time, distance, frequency and destination. These factors will affect
the charging time, charging power, charging location and so on.
4) Penetration of EV: It is mainly related to the price of EV and the subsidy policy. It determines the
proportion of EV charging load to system load.

2.1.2. Factors of EV charging station

EV charging load is also affected by the charging mode, location and pricing policy of EV charging
stations. Mainly depends on the following factors:
1) Charging characteristics: Including charging power and efficiency of charging station, which affect
charging duration and queuing waiting time;
2) Location of charging station: Influenced by the traffic flow in the area where the charging station is
located, which affect the spatial distribution of EV load.
3) Pricing policy of charging station: A controllable factor that will encourages EV owners to change
their charging behaviour, which affect the temporal distribution of EV load.

2.1.3. Factors of distribution network

Distribution network provides information such as structure of distribution network and real-time
electricity price for EV users and charging station operators.
Considering the value, timeliness and reliability of the above factors, the following factors closely
related to EV charging load are selected. The usage of factors are shown in Table 1.
598 International Journal of Smart Grid and Clean Energy, vol. 9 , no. 3, May 2020

Table 1. Factors used to simulation the charging schedules

Side Factors Symbol Data usage
Battery size E0 Calculate benchmark data
State of charge at Calculate the charging
SOCt
time t demand
Battery charging Calculate the charging
Rp
rate duration
Current energy
Electric Vehicle Et Analyze charging location
stored in the batter
Distance of trip j
li,j Analyze whether to charge
for EVi
The upper limit Calculate the charging
SOCm
value of SOC duration
The lower limit Calculate the charging
SOCa
value of SOC duration
Instant when the
tup Analyze whether to charge
price increase
Instant when the
Charging Station tdown Analyze whether to charge
price decrease
Efficiency of Calculate the charging
η
charging station efficiency
Distribution Electricity price of
Pg Device the pricing policy
Network power grid
EV number i EV number
Other Trip number j Trip number
Line number k Power line number

2.2. Simulation of the charging schedules

In order to predict the EV charging load, the charging schedules should be known. It is reasonable to
assume that traditional vehicles are just being replaced by EVs in the future. Consequently, charging
schedule model is established based on the historical travel information of traditional vehicles. The
information can be obtained from the National Household Travel Survey(NHTS), which has adopted real-
world travel dataset of 150,147 households.
The following assumptions are made. First, the study considers a simple TOU pricing policy with only
two price gradients in one day. tup and tdown are the instants when the prices increases and decreases.
Second, the charging power and efficiency are constant. Finally, the power consumption per kilometer is
fixed. The process of charging schedules model is shown as Fig. 1.
Start

Input datas; i=1,j=1

N
End i
n
Y
N
j
Si i=i+1
Y
E>Li/Kpe
N
Y
E=E-Li/Kpe

The destination has

j=j+1
a charging station?
N
Y

N The owner is
willing to charge?

Y
Charge

Fig. 1. Process of charging schedules model

 Xu Tian et al.: Optimization of pricing policy of electric vehicle charging station based on big data 599
First, the input data are the variables in Table 1, which include E0, SOCt, Rp, Et, li,j, SOCm, SOCa, tup,
tdown, η for each EV. Second, the daily travel information for trip j of EV i is read. If the remaining energy
will enable the distance of trip j, EV won’t charge. Then, the model judges whether a charging
infrastructure is available at the destination of trip j. The probability that charging infrastructure offers a
service is denoted as  , and its value varies from 0 to 1. Third, when a charging service is available, the
model judges whether the owner is willing to charge based on the owner’s charging habits. Finally, the
lowest cost is pursued under the condition that the energy will enable to continue the next trip j+1.

3. SVM Predicting Model

A large amount of valuable data is hidden in the big data related to EV load. Through the data-mining
technology, the charging load of each EV can be analysed. Support Vector Machine (SVM) is a learning
system using a linear function in high-dimensional feature space, by some nonlinear function mapping the
input vector to a high-dimensional feature space. In this paper, support vector machine (SVM) is used to
predict the influence degree of pricing policy on EV load.
In this paper, the SVM model constructs the loss function based on structural risk minimization and
determines the prediction function by minimizing the objective function shown in Formula (1).

 
 

n
min  1 wT w  C     * 
  2
i 1
i i 


 
 s.t. f xi - yi    i

(1)

 i  
 y  f x   *
i i
 *
 i  0,i  0, i  1, 2...n

Where C is the punishment coefficient,  i and  i are the relaxation factor, and  is the insensitive loss
*

coefficient.
Formula (1) can be transformed into the dual problem of Formula (2) by Lagrange multiplier.

   
 n
 max   yi i* i  i* i 
 
 i 1 

 1 n n

*
  *
 2   i i   j  j K xi , x j
 i 1 j 1
 
(2)

 
 n
 s.t .  i* i 0
 i 1
 *
 0  i ,i C ,i 1,2...n

Where,  i and  i are the Lagrange multiplier; K ( xi , x j ) is a kernel function that satisfies Mercer's
*

condition. In this paper, Gaussian kernel function shown in equation (3) is selected as the kernel function.
 x ,x 
  
K xi , x j  exp  
 2
i j 


(3)
 

Where,  is the broadband parameter of Gaussian kernel function.

2

The optimal predicting function can be obtained by solving equation (2), as shown in equation (4).

  
n
f  x    i i* K xi , x j  b (4)
i 1
600 International Journal of Smart Grid and Clean Energy, vol. 9 , no. 3, May 2020

Equation (5) can be obtained from KKT condition.

 
b  yi     ai*  ai K  xi , x j 
n
(5)
i 1

In our study, yi is the load of EV i and x is up down . The function is shown as equation (6).
(t , t , t )

2
n  ( tup
i i
,t )  ( tup , tdown , t )
P(tup , tdown , t )   (i  i* )e
,tdown
b (6)
i 1

By solving the SVM predicting model, the causality model is obtained, which can quantify the impact
of TOU pricing policy on EV charging load.

4. Pricing Policy Model of Charging Station

The charging station's price signal or charging service plan can mobilize users' charging behaviour or
exchange users' charging control right. Users' response to pricing policy can change the temporal and
spatial distribution of load. Therefore, the pricing policy model is established under the following three
angles of maximizing the benefit of charging station, improving user’s satisfaction and reducing power
system loss.

4.1. The optimization objective of pricing policy

The optimization objectives will be established from the following three perspectives:
Maximizing the benefit of charging station: In the economic market, charging stations maximize their
economic benefits by formulating reasonable electricity prices based on the day-ahead load forecasting,
as shown in Formula (7).
N T
,i i  t t 
max C    Pch t I t t   
(7)
i 1 t 1

is the charging power of EV i at time t ,  t and t are the

Pcht ,i
Where, C is the benefit of charging station,
t
price of power grid and charging station at time t, respectively. I i is the charging state of EV i at time t,
among them, Ii  1 means that the EV i is charging while Ii  0 means the opposite.
t t

Improving user’s satisfaction: The charging station optimize the pricing policy according to the
response degree of users to pricing policy. Users are encouraged to change their charging behaviour then
their consumption capacity is fully tapped，as shown in Formula (8).
T t min
  t Pev,i Ii cos ti
max   1  t 1 (8)
i
cos timax cos timin

 P
Where, i is the satisfaction of EV i owner, ev ,i is the rated power of EV i,
cos timax
and
cos timin
represent the consumption of EV i when charging at the maximum and minimum electricity price,
respectively.
Reducing power system loss: In the case of incentive compensation for the charging station by the
power grid, the charging station actively guides the EV to play a role in reducing the loss of power system
and equipment. Loss refers to the active power loss of lines, transformers and reactive power
compensation equipment, as shown in Formula (9).

 
m R
min P   Pk2 Qk2 k (9)
d
k 1 U k2
 Xu Tian et al.: Optimization of pricing policy of electric vehicle charging station based on big data 601

Where, Pd is the loss of power system, Pk and Qk are the active power and reactive power of line k
respectively, and U k is the voltage of line k.
Establishment of the pricing policy model considering the above three directions, the expression is
shown in Formula (10).
N
max f  1C  2 i  3 Pd (10)
i 1

Where, 1 , 2 and 3 are the weight coefficients that the sum is equal to 1, N is the number of EVs.

4.2. Constraints
In terms of constraints, the constraints related to the unit are consistent with the traditional problems
and are no longer listed. For a single vehicle, the following constraints should be considered:
1) Constraint on the use of EVs:
SOC  SOC  SOC (11)
a t m

Where, SOCt is the state of charge at time t, SOCa and SOCm are the minimum and maximum state of
charge by users, respectively. That is to say, the charging state of electric vehicles in the travel process
always within the user's expectation range.
2) Technical constraints of EVs:

0P P (12)
c,i c,max

Where, Pc ,i is the charging power of EV at time t, and Pc ,max is the maximum allowable charging power
of EV.
3) Energy Balance Constraints of EVs:
t
SOCt 1  SOCt  c , h Pcht I t t 
dch
t
Pdch 
1 I t  (13)

Where, SOCt 1 is the state of charge at time t+1; SOCt is the state of charge at time t; ch is the charging
efficiency of EV; dch is the discharge efficiency of EV; Pdch is the discharge power of electric vehicle at
t

time t.
For the charging station, the following constraints should be considered:
1) Charging power constraint of charging station:

0  Psta.t  Psta.max (14)

Where, Psta.t is the charging power of charging station at time t, and Psta.max is the maximum allowable
charging power of charging station.
2)Number constraint of charging EVs in charging stations:
0  Nsta.t  Nmax (15)

Where, N sta.t is the number of charging EVs in charging station at time t, and N sta.max is the maximum
allowable number of EV charged in charging station at time t.
For the distribution network, the following constraints should be considered:
1) Constraints of power flow balance:
gi.t Vi  t  , Pi  t  , Qi  t  , Pch,i  t   0 (16)
602 International Journal of Smart Grid and Clean Energy, vol. 9 , no. 3, May 2020

V (t ), Pi (t ), Qi (t ), Pch,i (t )
Where, i respectively represents the voltage at node i at time t, the active power
injection excluding the EV charging load, the reactive power injection excluding the EV charging load,
and the charging power of EVs.
2) Constraints of voltage:
Vmin  Vi  t   Vmax (17)

That is, the voltage of node i at time t is always within the allowable range.
Through the above load predicting model of EV, the charging load of charging stations under different
real-time electricity pricing policies can be predicted. Based on the predicted results, pricing policies are
formulated for each charging station. In this paper, particle swarm optimization (PSO) algorithm is used
to optimize the pricing policy model of charging stations.
1)Obtain charging demands of EVs. According to section 2, the EV charging load is predicted.
2)The optimization objective of pricing policy is optimized by particle swarm optimization algorithm
under the constraint of section 3.4.
3)Users independently respond to the pricing policy of charging station. After the charging station has
issued the pricing policy, users can choose the charging mode and time independently.

5. Case Analysis

In the future development, user's travel process is only to replace traditional vehicles with EVs, so it is
reasonable to assume that the travel process of EV is consistent with that of the traditional vehicle.
Take the IEEE9 system shown in Fig.2 as an example, and the appendix for system parameters are
shown in the appendix. According to the National Household Travel Survey (NHTS), the factors related
to EV charging load are selected as shown in Table 2. The travel behaviour of EVs is analysed to obtain
the charging load of each electric vehicle, and the charging load at each moment is calculated as shown in
Fig.3. Based on the model proposed in this paper, the calculation example designs nine price policies as
shown in Table 3. It is assumed that the price policies of charging stations connected to the same node are
the same, and the price of the power grid and charging stations is shown in Table 4.
Transform T1 Transform T2
Gen 1 1 100MVA 230kV 230kV 230kV 100MVA 5 Gen 2

Line 1 Line 3
16.5kV 18kV
2 3 4

Line 2 Line 5

6 230kV 230kV 7

Line 4 Line 6

230kV 8

Transform T3
100MVA

9 13.8kV

Gen 3

Fig.2. IEEE9 system

 Xu Tian et al.: Optimization of pricing policy of electric vehicle charging station based on big data 603

Fig.3 EV’s charging load in one day

Table 2. The definition and range of variables

Variables Definition Value
E0(kwh) Rated Capacity of EV Battery 14~38
Rp(kw) Charging Rate of Charging Station 1.4~7.7
Rh(kw) Charging Rate in Residential Areas 1.4~7.7
SOCm(%) Maximum state of charge 50~100
SOCa(%) Minimum state of charge 0~50
  % Permeability of Charging Station 0~100
Ef Charging Efficiency of Electric Vehicles 0.8~0.95
Instant when the price of electricity
tup 4~6 h
increases
Instant when the price of electricity
tdown 22~24 h
decreases
tch Charging time of electric vehicles 1~24 h
SOC0 Initial state of charge of electric vehicles 0~100
N Number of electric vehicles 10000
ev Penetration of electric vehicles 10%

Table 3. Electricity pricing policy

Time 1 2 3
Peak-periods 4:00-22:00 4:00-23:00 4:00-24:00
Off-Peak-periods 22:00-4:00 23:00-4:00 24:00-4:00
Time 4 5 6
Peak-periods 5:00-22:00 5:00-23:00 5:00-24:00
Off-Peak-periods 22:00-5:00 23:00-5:00 24:00-5:00
Time 7 8 9
Peak-periods 6:00-22:00 6:00-23:00 6:00-24:00
Off-Peak-periods 22:00-6:00 23:00-6:00 24:00-6:00
Table 4. Electricity price

Electricity price of charging

Electricity price of grid/($/kWh)
station/($/kWh)
Peak-periods 0.071 Peak-periods 0.077
Off-Peak-periods 0.053 Off-Peak-periods 0.057

5.1. EV charging load predicting

The support vector machine(SVM) predicting model was established for the pricing policy of charging
stations, and the charging load of 10,000 EVs was analyzed. According to the statistics, the predicted
604 International Journal of Smart Grid and Clean Energy, vol. 9 , no. 3, May 2020

charging load of EV in a day is shown in Fig 4.

Fig. 4. Charging load forecast of SVM predicting model

The error of SVM predicting model is within 10%, indicating the accuracy of the prediction.

5.2. Electricity price model of charging station

Matlab was used to solve the optimization of pricing policy model by particle swarm
optimization(PSO) algorithm. In the case, it was assumed that the population number was 30 and the
maximum iteration number was 600. Charging/discharging power and time of EVs are obtained under
different electricity pricing policies. The corresponding optimal solution is obtained according to different
optimization objectives. Since the penetration of EV is 10%, the loss of power system can be ignored.
The model 1 represent the maximizing benefit of charging station model, the model 2 represent the
maximizing user’s satisfaction model, model 3 represent the integrated model by 1  0.5 , 2  0.3 and
3  0.1 . The results are shown in Table 5 to Table 7.
Table 5. The pricing policy of the charging station
Nodes 1 2 3 4 5 6 7 8 9
Model 1 4 8 6 5 3 2 9 7 1
Model 2 2 7 9 3 8 5 6 1 4
Model 3 2 7 4 6 1 9 5 3 8

Table 6. The charging load of charging station

Unit: KW
Nodes 1 2 3 4 5 6 7 8 9
Model 1 253.4 400.8 317.9 404.2 572.3 405.0 322.1 412.8 259.7
Model 2 263.2 301.8 403.0 315.2 303.4 401.2 425.6 400.8 500.7
Model 3 485.8 412.5 428.4 409.5 286.7 308.4 408.7 318.9 260.8

Table 7. Other conclusions of the model

Benefit of the Reducing power

Items user satisfaction
charging station($) system loss(KW)
Model 1 245.55 79% 68.688
Model 2 242.71 89% 68.688
Model 3 243.10 84% 68.688
When considering the maximum benefit of charging station, the satisfaction of users decreases slightly.
In this case, it assumes that the penetration of EV is 10%, and the change of charging load has little effect
on the system. With the development of EVs, when the penetration of EVs is large, the application of
reducing power system loss model can effectively improve the security and economy of the power grid.
By synthesizing various indicators, we can see that the comprehensive model balances the interests of all
parties and achieves a win-win situation among the three parties.
 Xu Tian et al.: Optimization of pricing policy of electric vehicle charging station based on big data 605
6. Conclusion

The formulation of pricing strategy of EV charging stations is of great significance to the operation
and management of charging stations and power grid dispatch. This study based on the big data related to
EV load, a data mining-based model is proposed for modeling the charging schedules of EV, then a
massive simulation dataset about EV load can be obtained. A SVM model is developed to analyze the
relationship between pricing policy and EV load, with which the model can be used to evaluate the
impacts of the pricing policy on the EV load quantitatively. Based on it, an optimization model is
established to devise a TOU pricing policy for charging station. Finally, the IEEE9 is used as an example.
The results show that the electricity pricing policy will affect the user's consumption behavior, and
through the rational formulation of the pricing policy, the optimization objective can be effectively
achieved.

Appendix A: An example appendix

A.1. IEEE9 system bus parameters

Table A1. IEEE9 system bus parameters

Bus number Rated voltage /kV Upper limited voltage /kV Lower limited voltage /kV
1 16.5 18.15 14.85
2 18 19.8 16.2
3 13.8 15.18 12.42
4 230 253 207
5 230 253 207
6 230 253 207
7 230 253 207
8 230 253 207
9 230 253 207

A.2. IEEE9 system AC-line parameters

Table A2. IEEE9 system AC-line parameters

Bus I Bus J R X B
1 4 0 0.0576 0
4 5 0.017 0.092 0.158
5 6 0.039 0.17 0.358
3 6 0 0.0586 0
6 7 0.0119 0.1008 0.209
7 8 0.0085 0.072 0.149
8 2 0 0.0625 0
8 9 0.032 0.161 0.306
9 4 0.01 0.085 0.176

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

Xu Tian, Yuan Xu concuted the research; Xiangcheng Zhang, Xue Ma analyzed the data; Yuan Xu, Su
Zhang wrote the paper; all authors had approved the final version.
606 International Journal of Smart Grid and Clean Energy, vol. 9 , no. 3, May 2020

References

[1] Cui JD, Luo WD, Zhou NC. Research on pricing model and strategy of electric vehicle charging and discharging based on
multi view. Proceedings of the CSEE, 2018.
[2] Xiang D, Song YH, Hu ZC, Xu ZW. Research on optimal time of use price for electric vehicle participating V2G. Proceedings
of the CSEE, 2013;(31):15-25.
[3] Gao YJ, Wang C, Lv MK, et al. Opti-mal-time-of-use price model considering satisfaction degree of electric vehicle owners.
Electric Power Automation Equipment, 2014; 34(2):8-12.
[4] Li G, Zhang XP. Modeling of plug-in hybrid electric vehi-cle charging demand in probabilistic power flow calculations. IEEE
Transactions on Smart Grid, 2012; 3(1):492-499.
[5] Bae S, Kwasinski A. Spatial and temporal model of electric vehicle charging demand. IEEE Transactions on Smart Grid, 2012;
3(1):394-403.
[6] Darabi Z, Ferdowsi M. Aggregated impact of plug-in hy-brid electric vehicles on electricity demand profile. IEEE
Transactions on Sustainable Energy, 2011; 2(4):501-508.
[7] Tremblay O, Dessaint LA, Dekkiche AI. A generic battery model for the dynamic simulation of hybrid electric vehicles.
Vehicle Power and Propulsion Conference, 2007. Vppc. IEEE, 2007:284-289.
[8] Letendre S, Watts R. Plug-In Hybrid Electric Vehicles and the Vermont Grid: A Scoping Analysis. 2008.
[9] Weiller C. Plug-in hybrid electric vehicle impacts on hourly electricity demand in the United States. Energy Policy, 2011;
39(6):3766-3778.
[10] Darabi Z, Ferdowsi M. Aggregated impact of plug-in hybrid electric vehicles on electricity demand profile. IEEE Transactions
on Sustainable Energy, 2011; 2(4):501-508.
[11] Tian LT, Shi SL, Jia Z. A Statistical model for charging power demand of electric vehicles. Power System Technology, 2010;
34(11):126-130.
[12] Luo ZW, Hu ZC, Song YH, et al. Study on plug-in electric vehicles charging load calculating. Automation of Electric Power
Systems, 2011; 35(14):36-42.
[13] Shao S, Zhang T, Pipattanasomporn M, et al. Impact of TOU rates on distribution load shapes in a smart grid with PHEV
penetration. Transmission and Distribution Conference and Exposition, 2010 IEEE PES. IEEE, 2010:1-6.
[14] Zhang C, Ding M, Zhang J. A temporal and spatial distribution forecasting of private car charging load based on origin-
destination matrix. transactions of China electrotechnical society, 2017.
[15] Li RQ, Su HY. Optimal allocation of charging facili-ties for electric vehicles based on queuing theory. Automation of Electric
Power Systems, 2011, 35(14):58-61.
[16] Chen L, Nie Y, Zhong Q. A model for electric vehicle charging load forecasting based on trip chains. Transac-tions of China
Electrotechnical Society, 2015, 30(4):216-225.
[17] Wang XF, Shao CC, Wang Xiui, et al. Survey of electric vehicle charging load and dispatch control strategies. Proceedings of
the CSEE, 2013, 33(1):1-10.
[18] Chung KM, Kao WC, Sun T, et al. Decomposition methods for linear support vector machines. IEEE International
Conference on Acoustics, Speech, and Signal Pro-cessing, 2003. Proceedings. IEEE, 2003:IV-868-71 vol.4.
[19] Lou S, Zhang L, Wu Y, et al. Coordination operation of electric vehicles and power system under low-carbon economy.
Transactions of China Electrotechnical Society, 2017; 32(5):176-183.
[20] Liu D, Wang Y, Yuan X. Cooperative dispatch of large-scale electric vehicles with wind-thermal power generating system.
Transactions of China Electrotechnical Society, 2017.
[21] Yuhang H, Yichen P, Cui J, et al. Research on electric vehicle charging station modeling. Power System Pro-tection & Control,
2017; 45(8):107-112.

Copyright © 2020 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC
BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use
is non-commercial and no modifications or adaptations are made.