processes
Article
Electric Vehicle Charging Load Prediction Model Considering
Traffic Conditions and Temperature
Jiangpeng Feng 1 , Xiqiang Chang 1,2, *, Yanfang Fan 1 and Weixiang Luo 1
Citation: Feng, J.; Chang, X.; Fan, Y.;
Luo, W. Electric Vehicle Charging
Load Prediction Model Considering
Traffic Conditions and Temperature.
Processes 2023, 11, 2256. https://
doi.org/10.3390/pr11082256
Academic Editors: Hongyu Wu and
Bo Liu
Received: 13 June 2023
Revised: 23 July 2023
Accepted: 25 July 2023
Published: 26 July 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Processes 2023, 11, 2256. https://doi.org/10.3390/pr11082256
1 College of Electrical Engineering, Xinjiang University, Urumqi 830047, China;
xjufengjiangpeng@163.com (J.F.); 18733535871@163.com (Y.F.); lwx18839160071@163.com (W.L.)
2 State Grid Xinjiang Electric Power Supply Company, Urumqi 830011, China
* Correspondence: 107552104061@stu.xju.edu.cn
Abstract: The paper presents a novel charging load prediction model for electric vehicles that takes
into account traffic conditions and ambient temperature, which are often overlooked in conventional
EV load prediction models. Additionally, the paper investigates the impact of disordered charging
on distribution networks. Firstly, the paper creates a traffic road network topology and speed-flow
model to accurately simulate the driving status of EVs on real road networks. Next, we calculate the
electric vehicle power consumption per unit kilometer by considering the effects of temperature and
vehicle speed on electricity consumption. Then, we combine the vehicle’s main parameters to create
a single electric vehicle charging model, use the Monte Carlo method to simulate electric vehicle
travel behavior and charging, and obtain the spatial and temporal distribution of total charging load.
Finally, the actual traffic road network and typical distribution network in northern China are used
to analyze charging load forecast estimates for each typical functional area under real vehicle–road
circumstances. The results show that the charging load demand in different areas has obvious spatial
and temporal distribution characteristics and differences, and traffic conditions and temperature
factors have a significant impact on electric vehicle charging load.
Keywords: electric vehicles; traffic conditions; Monte Carlo method; spatio-temporal distribution;
load forecasting
1. Introduction
In recent years, with the increasingly stringent environmental requirements in coun-
tries, the electrification of transportation has been seen as an effective measure to achieve
energy savings, emission reduction, and carbon reduction [1]. Electric vehicles are an
important vehicle for promoting the transformation and upgrading of vehicle electrification
in transportation systems [2]. By the end of 2022, the ownership of pure electric vehicles in
China reached 10.45 million [3]. With the large-scale promotion and application in China, a
large number of electric vehicles are linked to the power grid for charging, which not only
raises the total demand for electricity but also intensifies the load peak-to-valley difference
and has a more adverse effect on the safe and stable operation of the power distribution
network [4]. Moreover, the charging demand of electric vehicles is also randomly influ-
enced by random driving behavior characteristics, which makes the charging requirements
of electric vehicles different from other loads, and the traditional load forecasting model
is no longer applicable to the charging requirements of the electric vehicle [5]. The load
forecasting model is not only the basis for assessing the ability of distribution networks
to accept the electric vehicle but also a necessary prerequisite for promoting research on
charging station planning and vehicle–road network integration and interaction, and it is
also of great significance for the operation and construction of future power systems [6].
Based on the above reasons, it is necessary to establish a model that can precisely describe
the spatio-temporal distribution characteristics of the electric vehicle charging demand.
https://www.mdpi.com/journal/processes
Processes 2023, 11, 2256 2 of 19

On the basis of the survey data of the National Household Vehicle Travel Survey
(NHTS) released by the United States [7], assuming that electric cars have the same travel
pattern as fuel cars, the fitted curves of departure time, charging time, and daily driving
distance are obtained. In [8,9], the Monte Carlo random sampling method is used to
simulate the charging load distribution of electric vehicles. In [10,11], the spatial and
temporal distribution characteristics of charging loads of different types of EVs in different
charging areas and on different typical days are studied according to the scale of EVs
and the development level of charging facilities in China and taking into account the
randomness of EV movements. In addition, the impact of EV user behavior decisions
and real-time charging tariffs on the spatial and temporal distribution of charging load
is considered in [12]. In [13], a coupled system considering the “vehicle-charging station-
distribution network” is constructed, and the spatio-temporal distribution of EV charging
load is simulated by the source–end matrix, taking into account the coupling of EVs with
the transportation network and distribution network. In [14], a prediction model of the
spatial and temporal distribution of EV charging load is proposed based on the law of
gravity, and the spatial and temporal distribution of EV charging load under the fusion
of “vehicle-road-station-grid” multi-source information is simulated. Notwithstanding
the above-mentioned research, which constructs a prediction model of EV charging load,
it ignores the stochastic characteristics of the EV driving process and charging process
and sets the characteristic parameters of EV charging location and charging duration
as fixed values, which cannot truly reflect the specific driving behavior and charging
process of EVs. To solve this problem, [15,16] proposes a spatial–temporal model of electric
vehicles based on stochastic travel chains and Markov decision theory. In [17], the spatio-
temporal distribution of EV charging demand is predicted by considering different dates
and functional areas with the inclusion of random travel chains. Considering multiple
charging behaviors and different destinations, a Monte Carlo approach is introduced to
simulate the driving and charging behavior of electric vehicles in [18]. Despite the fact that
the electric vehicle load forecasting model proposed has a certain effectiveness in the above
study, the impact of traffic jams, ambient temperature, and other factors on the driving
process of electric vehicles is not considered in predicting the charging load of electric
vehicles, and the charging load of electric vehicles cannot be accurately predicted.
According to the analysis above, electric vehicles are closely coupled to the trans-
portation system and the distribution network. For the sake of describing the travel and
charging demand characteristics of the electric vehicle more realistically in the dimensions
of time and space and making the emulation results more precise, the coupling relation-
ship between traffic flow and electric current, as well as the influence of factors such as
traffic jams and ambient temperature, on the driving process and charging process of the
electric vehicle needs to be sufficiently studied. According to the above context, the paper
proposes a spatio-temporal distribution prediction model for electric vehicle charging load
considering road information and environmental temperature.
Compared with past studies, the contributions of this paper include the following:
(1) This study introduces the effects of traffic conditions and ambient temperature on
EV electricity consumption into the field of EV charging load forecasting, which
solves the problem of these two important factors of traffic conditions and ambient
temperature being ignored in the modeling process of load forecasting and further
improves the accuracy of electric vehicle charging load forecasting compared with
previous modeling methods.
(2) Proposed the shortest travel time as the goal of the path planning algorithm, according
to the current driving section of the road class and traffic density dynamic planning
route, so that the user can effectively avoid the shortest distance but the road speed
limit is low and the road conditions for the congested road, more in line with the
actual vehicle driving law.
(3) The spatial–temporal distribution characteristics of electric vehicle charging load and
its impact on the distribution network are analyzed to provide a basis for guiding
Processes 2023, 11, x FOR PEER REVIEW 3 of 20

speed limit is low and the road conditions for the congested road, more in line with
Processes 2023, 11, 2256 the actual vehicle driving law.; 3 of 19
(3) The spatial–temporal distribution characteristics of electric vehicle charging load and
its impact on the distribution network are analyzed to provide a basis for guiding the
orderly charging
the orderly of electric
charging vehicles
of electric and the
vehicles andrational planning
the rational of theofdistribution
planning net-
the distribution
work in the future.
network in the future.

2.2.Electric
ElectricVehicle
VehicleDriving
DrivingCharacteristics
CharacteristicsModeling
Modeling
Electric
Electricvehicles
vehicles have
have the dual characteristics
characteristicsof oftraffic
trafficand
andmobile
mobileload
loadandandthe
thedriving
driv-
ing characteristics
characteristics of of electric
electric vehicles,and
vehicles, andtheir
theircharge–discharge
charge–discharge behavior
behavior links thetheurban
urban
distributionnetwork
distribution networkand and transportation
transportation network
network together;
together; the model
the model of theofcoupled
the coupled
“ve-
“vehicle-road-network”
hicle-road-network” system
system is shown
is shown in in Figure
Figure 1. 1.
The The topologyand
topology andtraffic
trafficsaturation
saturationofof
theurban
the urbanroad
roadnetwork
networkdirectly
directlyaffect
affectthe
thedriving
drivingspeed
speedofofelectric
electricvehicles
vehiclesand
andmotorists’
motorists’
path decisions,
path decisions, which,
which, ininturn,
turn,affect thethe
affect spatial andand
spatial temporal
temporaldistribution of charging
distribution loads
of charging
and the state of the distribution network; therefore, it can be seen that the detailed
loads and the state of the distribution network; therefore, it can be seen that the detailed portrayal
of the tripofcharacteristics
portrayal of electricof
the trip characteristics vehicles
electricisvehicles
the foundation for the establishment
is the foundation of the
for the establish-
“vehicle-road-network” coupling system.
ment of the “vehicle-road-network” coupling system.

z (W): Working Area

time Parking
y (R): Residential Area
time
Driving (C): Commercial Area
time Coupling Relationship
Distribution Network Line
(C) Distribution Network Node
Charging
Vehicle Driving
time
Trajectory track
Layer
(W)

(R)
x
(C)
(W)
Traffic Road
Network
Layer
(R)

Distribution
Network
Layer

Figure 1. “Vehicle-road-network” coupling system.

Figure 1. “Vehicle-road-network” coupling system.

2.1.
2.1.Initial
InitialTravel
TravelTime
Time
The
Theinitial
initialtravel
traveltime
timeof ofprivate
privatecars
carsininaaday
dayisisclosely
closelyrelated
relatedtotothe
thespatio-temporal
spatio-temporal
distribution
distribution of charging demand for the EV. Applying the statistical resultsof
of charging demand for the EV. Applying the statistical results ofthe
the2017
2017
NHTS,
NHTS,and andusing
usingdata
datafitting
fittingmethods
methodsfromfromstudy
study[19],
[19],the
thestarting
startingtime
timeofofprivate
privatecar
car
driving
drivingbehavior
behaviorin inaaday
dayisisextracted
extractedtotostudy
studythe
thetravel
travelcharacteristics
characteristicsofofelectric
electricvehicles.
vehicles.
The
Theprobability
probabilitydensity
densityfunction
functionfor forthe
theinitial
initialtravel
traveltime
timeof ofelectric
electricprivate
privatecarscarscan
canbebe
calculatedusing
calculated usingreal
realdata.
data.This
Thisprobability
probabilitydistribution
distributionisisthen
thenfitted,
fitted,and
andititisisdiscovered
discovered
thatthe
that theprobability
probabilitydensity
densityfunction
functionof ofthe
theconsumer’s
consumer’sinitial
initialtravel
traveltime
timeisisrepresented
representedby by
Equation
Equation (1): (1): " #
1 ( x1 − µ1 )2
f ( x1 ) = √ exp − (1)
x1 σ1 2π 2σ1 2

In Equation (1), µ1 is the expected value of the initial trip time, and σ1 is the initial
travel time’s standard deviation fitted with µ1 = 7.89 and σ1 = 1.96.
Processes 2023, 11, 2256 4 of 19

2.2. Driving Distance

The charging frequency and duration required for each charge for EV customers are
closely related to driving distance, and there are obvious variances in driving distance
across EV users due to various car habits. Furthermore, the driving distances of various
types of electric vehicles vary greatly. Private automobile travel areas are primarily centered
in cities, where cars are primarily used for commuting and amusement, and driving lengths
are quite short and regular. As a result, the primary research object for this study is electric
private cars. According to the 2017 NHTS statistics, the probability density function for the
distance driven by electric private cars roughly fits a lognormal distribution, as shown in
Equation (2): " #
1 (ln x2 − µ2 )2
f ( x2 ) = √ exp − (2)
x2 σ2 2π 2σ2 2

In Equation (2), µ2 is the driving distance’s expected value, and σ2 is the driving
distance’s standard deviation fitted with µ2 = 3.68 and σ2 = 0.88.

2.3. Parking Time

The duration of EV consumers’ stays at a destination location is highly tied to the
function of the area in which the destination is located. When a user travels to a work
area, the user’s stay is longer due to the consumer’s work needs; when the user travels
to a commercial area for entertainment and consumption, the consumer’s stay is shorter;
and when the consumer travels to a residential area, the consumer’s stay is also longer
due to the consumer’s home life needs. By fitting the statistical results of the 2017 NHTS,
the dwell time t of private cars in different areas is discovered to obey distinct types of
generalized extreme value distributions. The generalized extreme value distribution is
used to mathematically model the probability density of parking time in working areas, as
represented in Equation (3):
t−438.455
(
z1 = 164.506 h i (3)
1
f (z1 ) = 164.506 exp −(1 − 0.234z1 )4.27 × (1 − 0.234z1 )3.27

The Weibull distribution is used to mathematically model the probability density of

parking time in residential areas, as represented in Equation (4):
(
z2 = t
1.153
h
z2 1.153
i
z2 0.153 (4)
f ( z2 ) = 195.787 exp −( 195.787 ) × ( 195.787 )

The generalized extreme value distribution is used to mathematically model the

probability density of parking time in commercial areas, as represented in Equation (5):
t−68.520
(
z3 = 41.761 h i (5)
1
f (z3 ) = 41.761 exp −(1 + 0.657z3 )−1.52 × (1 + 0.657z3 )−2.52

2.4. Initial State of Charge

At present, the range of electric vehicle batteries in China can meet the consumer’s
urban travel needs for more than two days, so the consumer’s battery is not fully charged
when he sets off initially, and the initial probability density function of SOC (state of charge)
is as shown in Equation (6):

4.352S0 3.352 , 0 < S0 ≤ 1


f ( S0 ) = (6)
0, S0 = 0

In Equation (6), S0 is the initial battery SOC of the electric vehicle.

 when he sets off initially, and the initial probability density function of SOC (state of
charge) is as shown in Equation (6):

4.352S0 3.352 ,0  S0  1
f (S0 ) =  (6)
0, S0 = 0
Processes 2023, 11, 2256 5 of 19
In Equation (6), S0 is the initial battery SOC of the electric vehicle.

2.5. Travel Chain Model

2.5. Travel Chain Model
Travel chains are used to describe the space movement process of electric vehicles
Travel chains are used to describe the space movement process of electric vehicles
within a day, including travel origins and destinations, because electric vehicles have spa-
within a day, including travel origins and destinations, because electric vehicles have
tial movement characteristics, and the spatial and temporal distribution of charging loads
spatial movement characteristics, and the spatial and temporal distribution of charging
is loads
affected by their uncertain movement process. Electric private automobiles travel to
is affected by their uncertain movement process. Electric private automobiles travel to
destinations
destinations with
witha high degree
a high degreeof regularity, and
of regularity, travel
and chains
travel cancan
chains bebe
used to characterize
used to characterize
this process. According to the functional differences of the destinations,
this process. According to the functional differences of the destinations, they can they canbebe di-
divided
vided into three categories: the residential area (R), the working area (W), and
into three categories: the residential area (R), the working area (W), and the commercial the com-
mercial areaThese
area (C). (C). These different
different types
types of of also
areas areas also provide
provide electricelectric
vehiclevehicle charging
charging ser-
services. The
vices. The three basic travel chain types discussed in this paper are R–C–R,
three basic travel chain types discussed in this paper are R–C–R, R–W–R, and R–W–C–R, R–W–R, and
R–W–C–R,
as shownas inshown
Figurein 2. Figure 2.

(R) (C) (R) (W)

(R) (W) (C)

Figure
Figure 2. Different
2. Different forms
forms of travel
of travel chain
chain structures.
structures.

The
The shareofofvarious
share varioustravel
travel chains
chains for
for daily
dailyactivities
activitiesusing
usingprivate vehicles,
private as reported
vehicles, as re-
by the 2017 NHTS, is displayed in Table 1.
ported by the 2017 NHTS, is displayed in Table 1.

Table
Table 1. Percentage
1. Percentage of of different
different forms
forms of of travel
travel chain.
chain.

TripTrip Chain
Chain Form
Form Specific
Specific Gravity/%
Gravity/%
R−C−R
R-C-R 23.123.1
R−W−R
R-W-R 52.852.8
R−W−C−R 24.1
R-W-C-R 24.1

3. Traffic
3. Traffic Network
Network and
and Distribution
Distribution Network
Network Modeling
Modeling
3.1.3.1. Traffic Network Topology
Traffic Network Topology
In order to quantitatively explain the road structure in this paper, a graph theory
In order to quantitatively explain the road structure in this paper, a graph theory
approach is employed to model the two-way traffic network. The topology of the traffic
approach is employed to model the two-way traffic network. The topology of the traffic
network is shown by E(G). A(i,j) is the set of road weights of the traffic network in E(G)
used to describe the connection relationship between the nodes in E(G); the relationship
between E(G) and A(i,j) is as shown in Equation (7):

E( G ) = [ A(i, j)] (7)

The specific values of the traffic network road weights A(i,j) are as shown in Equation (8):

 lij , (i, j) ∈ E( G )
A(i, j) = 0, i = j (8)
inf, (i, j) ∈ / E( G )

 (8):

lij ,(i , j)  E(G)


A(i , j) = 0, i = j (8)
inf,(i , j)  E(G)
Processes 2023, 11, 2256  6 of 19

In Equation (8), i and j both denote the nodes of E(G) in the traffic road network; lij
denotes the distance between nodes i and j in E(G); and inf denotes that the two road
networkIn Equation
nodes (8), i and
concerned j both
are not denote
adjacent totheeach nodes of E(G) in the traffic road network; lij
other.
denotes the distance between nodes i and
As an example, taking the traffic network depicted in j in E(G); and inf denotes
Figure that the twoofroad
3, the intersections
network nodes concerned are not adjacent to each other.
the six road sections constitute a traffic road network set whose corresponding matrix
As an example, taking the traffic network depicted in Figure 3, the intersections of the
E(G) is as shown in Equation (9):
six road sections constitute a traffic road network set whose corresponding matrix E(G) is
as shown in Equation (9):  0 l12 inf l14 inf 
 
 l21 0 0 l l23 infinf l l25 inf  
E(G) = inf l32
12
0 inf l35  
14
(9)
  l21 0 l23 inf l25 
 l41
E( G ) = inf l32 0 inf 45 l35 
inf inf 0 l  (9)
inf
 l41l infl infl 0 0 l45 
 52 53 54 
inf l52 l53 l54 0

l12 l23
1 2 3

l14 l25
l35

4 5
l45
Figure 3. Road traffic network topology.
Figure 3. Road traffic network topology.
3.2. Improvement of Road Traffic Impedance Model
3.2. 3.2.1.
Improvement of Road
Improved TrafficImpedance
Roadway Impedance Model
Model
3.2.1. Improved Roadway
The travel time of Impedance Model
electric vehicles on the road can be calculated according to the BPR
The travel
(Bureau time ofRoads)
of Public electricmodel
vehicles onThe
[20]. theroad
roadcan can be
be classified
calculatedasaccording
smooth (0to<the S ≤BPR1.0) or
congested
(Bureau of Public < S < 2.0)
(1.0 Roads) by the
model road
[20]. The section
road cansaturation S. Since
be classified as the road(0
smooth traffic
< S ≤network
1.0) or is
dynamic,
congested (1.0S <varies over
S < 2.0) bytime. The equation
the road of road saturation
section saturation S isroad
S. Since the as shown
trafficinnetwork
Equationis(10):
dynamic, S varies over time. The equation of road saturation S is as shown in Equation
(10): Q
S= (10)
C
Q
In Equation (10), Q is the section of Sroad’s = traffic flow, which varies dynamically (10)with
C
time; and C is the section of road’s maximum capacity.
According
In Equation to Q
(10), theissection of road’s
the section traffic
of road’s condition,
traffic the road
flow, which section
varies impedance
dynamically model
with
corresponding
time; to different
and C is the section roadmaximum
of road’s capacity.S is obtained. The function can reflect
section saturation
the change intoroad
According the resistance
section of with road
road’s congestion.
traffic condition,Considering the dynamic
the road section flow and
impedance
speed
model limit of the road,
corresponding the improved
to different equation
road section is obtained
saturation S is [21], as shown
obtained. The in Equation
function can(11):
(
t0 (1 + α(S) β ), 0 ≤ S ≤ 1.0
Tr = (11)
t0 (1 + α(2 − S) β ), 1.0 < S < 2.0

In Equation (11), α and β are the impedance impact factors, usually set to 0.15 and 4,
respectively; Tr is the actual time required to go through this section; t0 is the zero-flow
travel time required to go through this section, which can be obtained from Equation (12):

Lij
t0 = (12)
V0

In Equation (12), Lij is the path length from road node i to node j; V 0 is the zero-flow
velocity of the indicated road, and there are differences in the zero-flow velocity of different
levels of roads.
Processes 2023, 11, 2256 7 of 19

3.2.2. Improved Node Impedance Model

Most current studies set the road parameters as static parameters, which makes it
difficult to characterize the impact of road intersections on the whole process of vehicle
traffic, so this paper uses a modified Webster model to calibrate the nodal impedance of
road intersections [22], which can be expressed as shown in Equation (13):
" #
c (1 − λ )2 S2
Tc = 0.441 + + 6.996 (13)
2(1 − λS) 2q(1 − S)

In Equation (13), c is the signal period; λ is the ratio of the effective green duration of
a signal phase to the total duration of the cycle; q is the vehicle arrival rate of the driveway;
S is the road saturation.
Considering that the time consumed in the process of vehicle travel is not only im-
pacted by the road congestion situation but also by the control of road intersection signals,
the road traffic impedance should include two parts, road section impedance and road
node impedance, and then the improved road traffic model can be expressed as shown in
Equation (14):
Th = Tr + Tc (14)

3.3. Path Selection Method

In graph theory, a common method for solving path planning problems is Dijkstra’s
algorithm. When the total length h of the path is represented by Lh , the goal of the
Dijkstra algorithm is to find the path that minimizes Lh among all possible h values, which
ignores the impact of actual road traffic factors on the vehicle movement process [23].
In the actual driving process, users prefer to choose the path without congestion; if the
Dijkstra algorithm’s objective is the minimum value of Th , where Th indicates the road
traffic impedance, i.e., the total time consumed through path h, planning for the dynamic
shortest journey time can be achieved, and electric vehicle users can combine road traffic
information when they reach the road intersection, and, according to the current traffic
congestion, the shortest travel time path can be replanned when the EV user reaches the
road intersection.

3.4.
Processes 2023, 11, x FOR PEER REVIEW Regional Distribution Network Model 8 of 20
On the basis of establishing a dynamic transportation network node model, the IEEE
33 standard node distribution network model is used to connect distribution network
nodes and transportation
transportation network
network nodes.
nodes. Figure
Figure 44 displays
displays the topology
topology diagram
diagram of the
distribution network.
distribution network.

Main Grid
Distribution Network Node
23 24 25
Distribution Network Line

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

26 27 28 29 30 31 32 33
19 20 21 22

Figure 4. Power distribution network topologic model.

Figure 4. Power distribution network topologic model.

4.
4. Electric
Electric Vehicle Charging Load
Vehicle Charging Load Model
Model
The spatio-temporal distribution
The spatio-temporal distributionofofelectric
electricvehicle
vehiclecharging
charging load
load in in
thethe transporta-
transportation
tion network is associated with the distribution of traffic travel. Electric vehicles
network is associated with the distribution of traffic travel. Electric vehicles are impactedare im-
by
pacted
externalby external environmental
environmental factors,road
factors, including including
speedroad speed
limits, limits,
traffic traffic congestion,
congestion, road grade,
road grade, and ambient temperature during normal travel, which, in turn, affect the spa-
tio-temporal distribution of electric vehicle charging time in terms of duration and charg-
ing location in the transportation network. By coupling the impact of external environ-
mental factors on the electricity expenditure per unit kilometer to the process of electric
vehicle travel, a more refined model of electric vehicle energy expenditure per unit kilo-
Processes 2023, 11, 2256 8 of 19

and ambient temperature during normal travel, which, in turn, affect the spatio-temporal
distribution of electric vehicle charging time in terms of duration and charging location
in the transportation network. By coupling the impact of external environmental factors
on the electricity expenditure per unit kilometer to the process of electric vehicle travel, a
more refined model of electric vehicle energy expenditure per unit kilometer considering
traffic conditions and ambient temperature is established.

4.1. Model of Electric Vehicle Electricity Expenditure per Unit Kilometer Taking into Account
Vehicle Speed and Temperature
4.1.1. Model Considering Road Conditions
Driving speed and battery energy expenditure of electric vehicles are linked with
different road categories and under varying traffic jam scenarios. The electricity expenditure
per unit kilometer for electric vehicles Ei (i = 1, 2, 3, 4) can be calculated under real-time
congestion conditions by calculating the measured data. Ei (i = 1, 2, 3, 4) can be expressed
by Equation (15) as follows:

= 1.52 −5
V − 0.004V +2.992 × 10 V + 0.247

 E1
= 0.004V + 5.492

E2 V − 0.179

(15)
 E3
 = −0.001V + 1.531
V + 0.21
= −0.002V + 1.553

E4 V + 0.208

In Equation (15), E1 , E2 , E3 , and E4 are the electricity expenditure per unit kilometer
(kWh/km) of the vehicle traveling on the highway, main trunk, secondary main trunk, and
branch way, respectively; V indicates the real-time driving speed of the electric vehicle
(km/h), which can be obtained from Equation (16):

Lij
V= (16)
Th

4.1.2. Model of On-Board Air Conditioner Considering Temperature Effects

As the largest single auxiliary load on electric vehicles, one of the influences of tem-
perature on the on-board electric battery capacity is mainly embodied by the electricity
expenditure of the air conditioner in summer and winter. When it is a high temperature
inside the cabin in summer, owners may choose to open the on-board air conditioner to
cool the vehicle, which increases the battery’s electricity consumption. When temperatures
inside the cabin are low in the winter, owners may choose to turn on the air conditioning to
raise the temperature, resulting in increased battery electricity consumption. The above
analysis shows that the operating state of the on-board air conditioner affects the battery
electricity expenditure and thus the spatio-temporal distribution of the electric vehicle
charging load.
The air conditioning electricity expenditure of the electric vehicle at distinct tempera-
tures after the air conditioning is opened is as shown in Equation (17):

L 1.324
 0.03347 V ( Tset − Tamb )
 , Tamb < Tset
Eh = 0, Tamb = Tset (17)
 0.03669 L ( T
 1.084
V amb − Tset ) , Tamb > Tset

In Equation (17), Eh is the electricity expenditure of air conditioning when the ambient
temperature is Tamb and the vehicle travels L km at speed V. The real-time vehicle speed V
can be obtained from Equation (16); Tamb is the current ambient temperature; and Tset is
the vehicle cockpit set temperature, which is set to 22 degrees Celsius.
Processes 2023, 11, 2256 9 of 19

4.1.3. Model of Battery Losses in Electric Vehicles Considering Different

Ambient Temperatures
In summer, when temperatures are high, the high external temperature causes the self-
discharge of the vehicle battery to become more pronounced and the rate of battery aging
to accelerate and causes problems such as restricted power and overheating protection,
resulting in a certain degree of battery capacity loss. When the temperature is low in
winter, the internal resistance of the battery increases, and the internal electrochemical
process slows down, affecting the battery capacity and hindering charging and discharging
performance. The above analysis shows that ambient temperature affects the battery
performance of electric vehicles and therefore the spatio-temporal distribution of the
charging load.
The loss per unit kilometer Et of the on-board battery of the electric vehicle without
considering the air conditioning load as a function of the ambient temperature T is as
shown in Equation (18):

 −3.2547 × 10−3 T − 6.3657 × 10−5 T 2 + 6.5058 × 10−6 T 3 + 2.4575 × 10−7 T 4



Et = −8.4631 × 10−9 T 5 − 1.9318 × 10−10 T 6 + 0.21077, Tamb < Tset (18)

1.011 × 10−2 T + 2.6278 × 10−4 T 2 + 0.6777, Tamb ≥ Tset


In Equation (18), Et is the power loss of the on-board battery due to temperature when
the ambient temperature is Tamb (not considering the air conditioning load).

4.1.4. Model of Electric Vehicle Electricity Expenditure Taking into Account Vehicle Speed
and Temperature
From the above analysis, it can be concluded that vehicle speed and temperature
jointly affect the electricity consumption per unit kilometer. Consequently, a model is
established to count electricity consumption per unit kilometer considering vehicle speed
and temperature, as shown in Equation (19):

ET = Eh + Et + Ei (i = 1, 2, 3, 4) (19)

In Equation (19), ET represents the real-time electricity consumption per unit kilometer
of the electric vehicle when the vehicle speed is V and the ambient temperature is T, unit:
kWh/km.

4.2. Electric Vehicle Charging Method

The fast charging power is selected as 60 kW, and the slow charging power is selected
as 7 kW. Considering the charging cost and battery loss of consumers, it is assumed that
consumers prioritize the slow charging method. However, if the SOC does not reach the
preset expectation during the parking time period, i.e., there is not enough power for the
next leg of the electric vehicle’s route, as shown in Equation (20), then the consumer will
choose the quick charging method.

Pslow Tpi i
+ socinit < socexp (20)
CEV

In Equation (19), Pslow denotes the slow charging power; and Tpi denotes the parking
time of the vehicle at node i.
Since the current quick charging technology can guarantee that the battery SOC of an
electric vehicle with a capacity of 60 kWh can be charged from 0 to 80% in 20 min [24], the
situation in which the SOC of the battery after quick charging is less than the SOC required
for the next trip is ignored.
Processes 2023, 11, 2256 10 of 19

4.3. Electric Vehicle Charging Load Calculation

The spatio-temporal information of each distribution network node is counted based
on the coupling relationship between transportation network nodes and distribution net-
work nodes and then on statistics data on the charging load for the entire region.
According to the characteristic quantity of electric vehicles, the total charging load
at node i of the distribution grid at time t is calculated. The entire load of the nodes
in different functional areas is then calculated by adding the grid node’s base load and
the whole EV charging load, finally obtaining the spatio-temporal distribution of the EV
charging load. The prediction model predicts the first 24 h at an interval of 1 min, as shown
in Equation (21):
N
Pai = Paib + ∑ Pj,t
i
(2 ≤ i ≤ 33, t, i ∈ Z ) (21)
i =1

In Equation (21), Pai denotes the total load at the distribution network’s node i of area
a; Paib
denotes the base load at the distribution network’s node i of area a; N denotes the
number of electric vehicles performing charging activities near the distribution network’s
node i at moment t; Pj,ti denotes the charging power of the electric vehicle with serial

number j at the distribution network’s node i at moment t.

4.4. Load Forecasting Model Solving Process

The Monte Carlo method is utilized to simulate the spatio-temporal distribution of
charging load on the basis of the created model of movement characteristics and charge
Processes 2023, 11, x FOR PEER REVIEW 11 of 20
characteristics of electric vehicles. Figure 5 depicts the flow chart for the prediction solution
of the spatio-temporal distribution of the charging load of electric vehicles.

Start
Is there an emergency charging
requirement for the next leg of the trip?
Y N
Input data: total number of electric vehicles N,
battery capacity C, road class, temperature, etc. Fast charging at the current Slow charging at the current
location, record charging and location, record charging and
location information location information

According to the probability density function, the main

state parameters of electric vehicles are extracted, n=1.
Record the Spatio-temporal distribution of
electric vehicle charging load demand
Extract the destination ( j ) based on the first distance
traveled and the current position

Is the next destination the

j=j+1 N
endpoint?
Based on temperature and road traffic conditions, obtain
the total power consumption per unit mileage, calculate
travel time, arrival time, and battery charge state Y

n<N? N n=n+1

According to the probability density function, generate Y

the stopping time and the next travel distance

End

Figure 5. Flow chart of charging load spatio-temporal distribution prediction.

Figure 5. Flow chart of charging load spatio-temporal distribution prediction.
5.
5. Example
Example Analysis
Analysis
To verify the validity
To verify the validityofofthe
theproposed
proposedmodel,
model,anan actual
actual urban
urban road
road network
network in ainregion
a re-
gion of Hebei Province in northern China is selected for simulation analysis.
of Hebei Province in northern China is selected for simulation analysis.

5.1. Simulation Parameter Setting

The actual area indicated in Figure 6 is used as an example; the selected area’s acreage
is approximately 50.4 km2, and the perimeter is approximately 30.6 km. The spatio-tem-
 Figure 5. Flow chart of charging load spatio-temporal distribution prediction.

5. Example Analysis
Processes 2023, 11, 2256 11 of 19
To verify the validity of the proposed model, an actual urban road network in a re-
gion of Hebei Province in northern China is selected for simulation analysis.

5.1.
5.1.Simulation
SimulationParameter
ParameterSetting
Setting
The
The actual area indicatedininFigure
actual area indicated Figure6 is used
6 is asas
used anan
example;
example;thethe
selected area’s
selected acreage
area’s is
acreage
approximately 50.4 km 2 , and the perimeter is approximately 30.6 km. The spatio-temporal
is approximately 50.4 km , and the perimeter is approximately 30.6 km. The spatio-tem-
2
distribution of electric
poral distribution vehiclevehicle
of electric charging load in load
charging various functional
in various areas isareas
functional calculated over a
is calculated
24 h period.
over a 24 h period.

W R C

Figure6.6.The
Figure Thestreamlined
streamlinedtransportation
transportationroad
roadnetwork
networkininthe
theurban
urbanarea.
area.

(1) Considering
(1) Considering the main
main sections
sections of the
the region,
region, there
there are
are 64
64 traffic
traffic nodes
nodes and
and98
98traffic
traffic
sections;
sections;
(2) The actual area is simply divided into three functional categories, namely, residential,
working, and commercial areas;
(3) Combined with the practical road traffic situation in China, the roads are divided
into four types, as shown in Figure 6: the ring highway that is the green road with
a set zero-flow speed of 100 km/h; the main trunk that is the yellow road with a set
zero-flow speed of 60 km/h; the secondary main trunk that is the red road with a set
zero-flow speed of 50 km/h; the branch way that is the blue road with a set zero-flow
speed of 40 km/h;
(4) It is assumed that there are a total of 2500 electric private cars in this area whose initial
locations are evenly distributed around the residential areas.

5.2. Charging Load Forecast Results

Figure 7 shows the total charging load of each functional area in the region. Subse-
quently, the typical nodes in each functional area are selected separately, and the charac-
teristics of EV charging demand in different functional areas are analyzed by taking each
typical node as an example.
Figure 8a shows what happens when electric vehicle charging load is linked to a
typical node in a residential area. The residential area load increases comprehensively
during all hours, showing an obvious double-peak phenomenon. Furthermore, the peak
electric vehicle charging load is stacked on the peak residential basic load, causing the
undesirable effect of adding a peak to a peak, resulting in a further increase in the load
peak-to-valley difference, with the maximum peak-to-valley difference reaching 649 kW.
Figure 8b shows that the charging load in the working area begins to climb at 6:00 a.m.
The charging load in the working area peaks after the morning peak commute to work,
i.e., the peak load occurs at about 12:00 p.m.
 5.2. Charging Load Forecast Results
Figure 7 shows the total charging load of each functional area in the region. Subse-
quently, the typical nodes in each functional area are selected separately, and the charac-
Processes 2023, 11, 2256 teristics of EV charging demand in different functional areas are analyzed by taking12each
of 19
typical node as an example.

Processes 2023, 11, x FOR PEER REVIEW 13 of 20

Spatial–temporaldistribution
Figure7.7.Spatial–temporal
Figure distributionofoftotal
totalcharging
chargingload
loadininvarious
variousfunctional
functionalarea.
area.

2000 2200
Superimposed load
1800
Superimposed load
Basic load
Figure 8a shows what happens when electric vehicle2000charging
Superimposed load
Basic load
load is linked to a typ-
Basic load
1800
1600 ical node in a residential area. The residential area load increases 1800 comprehensively during

Charging demand/kW
Charging demand/kW
Charging demand/kW

1600 1600
1400 all hours, showing an obvious double-peak phenomenon. 1400
Furthermore, the peak electric
1200
vehicle charging load is stacked on the peak residential basic
1400
1200 load, causing the undesirable
1000
effect of adding1200 a peak to a peak, resulting in a further increase 1000 in the load peak-to-valley
800
800
difference, with1000 the maximum peak-to-valley difference reaching 600
649 kW.
600
Figure 8b shows800 that the charging load in the working 400 area begins to climb at 6:00
a.m. The charging 600 load in the working area peaks after the morning peak commute to
200
400 0
0 2 4 6 8 10 12 14
Time/h
16
work, i.e., the peak load occurs at
18 20 22 24 0 2 4 6 8 10 about 12:00 p.m.
12
Time/h
14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time/h

(a) Figure 8c depicts the charging (b) load distribution in the commercial (c) area, with EV
charging loads beginning to grow significantly at 12:00 p.m. and 20:00 p.m., the peak busi-
Figure
ness 8. (a)
hours inTypical
commercialnode load in thewhich
regions, residentialis inarea;
line (b) typical node loadcharacteristics
in the working ofarea; (c)
Figure 8. (a) Typical node load in the residential area;with the business
(b) typical node load in the working the
area;
typical node loads in the commercial area.
commercial area.
(c) typical node loads in the commercial area.
A comprehensive
Figure 8c depicts comparison
the chargingofload Figure 8a–c shows
distribution inthat
the charging
commercial loadarea,
is tightly
withcou-
EV
pled to travel
charging loadsdestination
beginning and travel
to grow time. The charging
significantly at 12:00 load
p.m. peaks in thep.m.,
and 20:00 working area
the peak
after the hours
business morning rush hour and
in commercial in the
regions, residential
which area
is in line after
with thethe evening
business rush hour. The
characteristics of
peak charging
the commercial area. load in the commercial area is concentrated during the midday meal and
dinner times. The charging
A comprehensive load in the
comparison of residential,
Figure 8a–cthe commercial,
shows and theload
that charging working area
is tightly
increases to varying degrees after the EV charging load is linked,
coupled to travel destination and travel time. The charging load peaks in the working causing a significant
change
area afterinthe
themorning
load curve,
rushwhich
hour has
andainsignificant effectarea
the residential on the regional
after distribution
the evening net-
rush hour.
work.
The peak charging load in the commercial area is concentrated during the midday meal
Based times.
and dinner on the The
emulation
chargingresults,
load itincan
the be seen that the
residential, charging load prediction
commercial, can pro-
and the working
vide an important reference for distribution network planning and safe
area increases to varying degrees after the EV charging load is linked, causing a significant and stable opera-
tion. in the load curve, which has a significant effect on the regional distribution network.
change
Based on the emulation results, it can be seen that charging load prediction can provide
5.3.important
an Path Selection Method
reference for distribution network planning and safe and stable operation.
The path planning experiments designed in this paper do not take into account the
5.3.
costsPath Selection
incurred byMethod
electric vehicles in the process of finding charging stations and in the
The path
charging planning
process; experiments
they only comparedesigned
and verify in the
this search
paper do notand
effect takethe
into account
path the
selection
costs
method incurred by electric
presented in this vehicles
paper and in the
the process of finding
traditional shortestcharging stations
path selection and in The
method. the
charging process;
path selection theyare
results only compare
shown and verify
in Figure 9, andthethe
search effect and
comparison the path
results selection
are shown in
method
Table 2. presented in this paper and the traditional shortest path selection method. The
path selection results are shown in Figure 9, and the comparison results are shown in
Table 2.
Processes2023,
Processes 2023,11,
11,2256
x FOR PEER REVIEW 14
13 of 19
20

Figure 9.
Figure 9. Comparison
Comparison of
of path
path selection
selection results.
results.

Table2.2.Comparison
Table Comparisonof
ofdifferent
differenttypes
typesof
ofpath
pathselection.
selection.

Path
Path Selection Type
Selection Type Path Length/km
Path Length/km Total Time/h
Total Time/h
Shortest distance (open)
Shortest distance (open) 35
35 0.764
0.764
Shortest time(open)
Shortest time (open) 47.3
47.3 0.544
0.544
Shortest time
Shortest time (congestion
(congestionoccurs)
occurs) 47.3
47.3 0.679
0.679
Shortest time (change of path)
Shortest time (change of path) 46
46 0.646
0.646

As can
As can bebe seen
seen inin Figure
Figure 9,9, the
the shortest
shortest timetime pathpath selection
selection methodmethod based
based onon traffic
traffic
information combines the
information combines theroad
roadspeed
speed limit
limit and and real-time
real-time traffic
traffic situation
situation in theinpath
the plan-
path
planning
ning process process to dynamically
to dynamically planplanthethe driving
driving path.When
path. Whenthe theroadroadisisclear,
clear, the shortest
shortest
path
path selection
selection method
method tendstends toto select
select thethe road
road with
with aa higher
higher speed
speed limit,limit, as
as shown
shown by by the
the
yellow
yellowarrow
arrowininFigureFigure9, 9,
andandthethe
selected
selectedroadroadsections are all
sections aredistributed
all distributedon the onring
theroad.
ring
The
road.traditional shortest
The traditional path selection
shortest method
path selection ignoresignores
method the influence of roadofspeed
the influence road limits
speed
and chooses
limits the road
and chooses shown
the by the black
road shown by thearrow
blackinarrow Figurein9,Figure
and the 9, selected
and the road sections
selected road
are mostly
sections aredistributed on secondary
mostly distributed arterial roads
on secondary arterial with lower
roads with speed
lowerlimits,
speedwhich
limits,are not
which
given priority in the actual driving process for travel time. In addition,
are not given priority in the actual driving process for travel time. In addition, when con- when congestion
occurs
gestioninoccurs
a certain
in aroad section,
certain the shortest
road section, time path
the shortest selection
time methodmethod
path selection presented in this
presented
paper modifies the path. For example, when congestion occurs
in this paper modifies the path. For example, when congestion occurs in a certain section in a certain section for the
electric
for the vehicle,
electric the shortest
vehicle, thepath planning
shortest pathmethod
planning modifies
method the modifies
original paththe shown
originalbypaththe
yellow
shown arrow
by the to the new
yellow arrowpathtoshown
the new bypath
the blue
shown arrow to blue
by the avoidarrowcongestion.
to avoid The different
congestion.
path passage times
The different and pathtimes
path passage lengthsandare shown
path lengths in Table 2.
are shown in Table 2.
Analysis
Analysis of Table 2 shows that the path length planned by
of Table 2 shows that the path length planned by the
the shortest
shortest route
route selection
selection
method
method is 35.14% longer than that planned by the traditional shortest route selection
is 35.14% longer than that planned by the traditional shortest route selection
method,
method, but but the
the shortest
shortest route
route selection
selection method
method is is 28.79%
28.79% less
less than
than the the traditional
traditional shortest
shortest
route selection method in terms of travel time. Additionally, when
route selection method in terms of travel time. Additionally, when a section of the planned a section of the planned
route
route isis congested,
congested, the the replanned
replanned routeroute consumes
consumes 5.11% 5.11% less
less time
time than
than the
the original
original route
route
which was previously congested, verifying that the shortest route selection method may
which was previously congested, verifying that the shortest route selection method may
make users effectively avoid paths with the shortest distance but low road speed limits and
make users effectively avoid paths with the shortest distance but low road speed limits
congested road conditions.
and congested road conditions.
Processes 2023,11,
Processes2023, 11,2256
x FOR PEER REVIEW 14 of
15 of 19
20

5.4.
5.4.Consideration
Considerationofofthe
theImpact
ImpactofofTraffic
Trafficand
andTemperature
TemperatureFactors
Factorsononthe
theCharging
ChargingLoad
Load
Taking
Takinginto
intoaccount
accountthe theinfluence
influenceofoftraffic
trafficand
andtemperature
temperaturefactors,
factors,the
thecomparative
comparative
results
resultsof
ofEV
EV charging
charging load before and
load before and after
afterthe
theintroduction
introductionofoftraffic
trafficand
andtemperature
temperature in
in Figure 10 are obtained, and the proportion of energy consumption accounted
Figure 10 are obtained, and the proportion of energy consumption accounted for by dif- for by
different factors
ferent factors is obtained
is obtained to to compare
compare andand analyze
analyze thethe effect
effect of traffic
of traffic andand temperature
temperature fac-
factors on charging
tors on charging load.load.

25 Consider road conditions and temperature

Consider road conditions and air-conditioning
Consider road conditions only
20
Base load without road conditions and temperature
Charging load/MW

15

10

0
0 2 4 6 8 10 12 14 16 18 20 22 24
Time/h
Figure10.
Figure 10.The
Theeffect
effectofoftraffic
trafficand
andtemperature
temperaturefactors
factorson
oncharging
chargingload.
load.

From
FromFigure
Figure10, 10,ititcan
canbebeseen
seenthat
thattraffic
trafficconditions
conditionsand andambient
ambienttemperature
temperaturefactors
factors
have
haveaagreater
greaterimpact
impacton oncharging
chargingload. load. Taking
Takingthe thehigh-temperature
high-temperatureenvironmentenvironmentin in
summer
summeras as an example,
example,in inthe
thecase
caseofof considering
considering onlyonly
thethe traffic
traffic roadroad conditions,
conditions, the
the peak
peak
hours hours of vehicle
of vehicle travel travel are 7:00–10:00
are 7:00–10:00 a.m.a.m. and 18:00–20:00
and 18:00–20:00 p.m.,p.m., corresponding
corresponding to the tohigh
the
high saturation
saturation of traffic
of the the traffic
roadroad sections
sections at this
at this time. time.
TheThe traffic
traffic mileage
mileage consumption
consumption of
of ve-
vehicles in the process of driving increases, and the charging load starts
hicles in the process of driving increases, and the charging load starts to surge at this time. to surge at this time.
After
After22:00
22:00p.m.,
p.m.,thethesaturation
saturationof oftraffic
trafficroad
roadsections
sectionsisislow, low,owing
owingto tothe
thecomparatively
comparatively
small
small number of vehicles on the road, and the energy consumption of trafficmiles
number of vehicles on the road, and the energy consumption of traffic milescaused
caused
by
bytraffic
trafficroad
roadconditions
conditionsstarts startstotodecrease.
decrease. InInthethecase
caseof of
considering
considering only thethe
only traffic factor
traffic fac-
on
torthe
on charging
the charging load, thethe
load, charging
charging loadloadcaused
caused byby thethetraffic road
traffic roadcondition
conditionduring
duringthe the
peak travel period increases by 30.63% compared with the case
peak travel period increases by 30.63% compared with the case that does not consider the that does not consider the
traffic
trafficroad
roadcondition,
condition,obtaining
obtainingan anoverall
overallincrease
increaseofof20.97%.
20.97%.In Inthe
thecase
casewhere
whereonlyonlythethe
ambient temperature causes the air conditioning to turn on,
ambient temperature causes the air conditioning to turn on, as the ambient temperature as the ambient temperature
gradually
graduallyincreases
increasesover overtime,
time,thetheelectricity
electricityexpenditure
expenditureof ofthe
theon-board
on-boardair airconditioning
conditioning
caused by the ambient temperature begins to rise continuously
caused by the ambient temperature begins to rise continuously during the high-tempera- during the high-temperature
hours of 10:00
ture hours a.m.–16:00
of 10:00 p.m., p.m.,
a.m.–16:00 increasing the charging
increasing the charging load load
by 12.90%
by 12.90%compared
comparedto the to
charging load without the on-board air conditioning, with an overall increase of 6.52%.
the charging load without the on-board air conditioning, with an overall increase of 6.52%.
Considering only the battery loss due to ambient temperature, the charging load during
Considering only the battery loss due to ambient temperature, the charging load during
the high-temperature period increases by 17.63% compared to that which does not consider
the high-temperature period increases by 17.63% compared to that which does not con-
the battery loss due to ambient temperature, obtaining an overall increase of 9.74%. The
sider the battery loss due to ambient temperature, obtaining an overall increase of 9.74%.
overall charging load increases by 38.28% compared to charging loads that do not take
The overall charging load increases by 38.28% compared to charging loads that do not
these two factors into account, verifying the validity of the model.
take these two factors into account, verifying the validity of the model.
From Figure 11, the effect of different temperatures on the charging load can be seen;
From Figure 11, the effect of different temperatures on the charging load can be seen;
due to the difference in the cooling and heating principles of the on-board air conditioning
due to the difference in the cooling and heating principles of the on-board air conditioning
of electric vehicles, the heating power of the on-board air conditioner in the winter low-
of electric vehicles, the heating power of the on-board air conditioner in the winter low-
temperature environment is higher than the cooling power of the on-board air conditioning
temperature environment is higher than the cooling power of the on-board air condition-
in the summer high-temperature environment, and battery capacity loss due to temperature
ing in is
factors the summer
higher high-temperature
in cold winter environments environment, and battery capacity
than in high-summer loss due
environments, to tem-
resulting
Processes 2023, 11, x FOR PEER REVIEW 16 of 20

Processes 2023, 11, 2256 15 of 19

perature factors is higher in cold winter environments than in high-summer environ-

ments, resulting in the winter peak charging load and the total charging demand in winter
in the higher
being winter than
peakin charging
summer, load
andand
thethe total charging
charging demand
load caused in winter
by the winterlow-tempera-
being higher
than in summer, and
ture environment the charging
is 20.35% higherload
thancaused by theby
that caused winter low-temperature
the high environment
summer temperature. The
is 20.35% higher than that caused by the high summer temperature. The analysis
analysis shows that the low-temperature environment is particularly detrimental to the shows
that thevehicle
electric low-temperature environment
range, validating is particularly detrimental to the electric vehicle
the model.
range, validating the model.

Figure 11. The effect of different temperatures on charging load are compared.
Figure 11. The effect of different temperatures on charging load are compared.
Comprehensive analysis of the simulation results shows that the dynamic energy
Comprehensive
consumption analysis of the
model established simulation
in this paper canresults shows obtain
accurately that thethe dynamic energy
actual energy
consumption model
consumption established
of electric vehiclesininthis paper can
different accurately
ambient obtain the
temperatures andactual energy
different con-
traffic
sumption ofwhich
conditions, electric vehicles
further in different
improves ambient temperatures
the accuracy of the electricand different
vehicle load traffic con-
prediction
ditions,and
results which further improves
is expected to promote thethe
accuracy
sharingof of
theinformation
electric vehicle
andload prediction
win–win results
cooperation
and is expected to promote the sharing of information and win–win
between electric vehicle users, the traffic network, and the distribution network to improvecooperation between
electric
the vehicle users,
comprehensive the traffic
benefits network,
to society and the adistribution
and provide reasonable basisnetwork to improve
of planning the
for the
comprehensive
future layout of benefits to society
the distribution and provide a reasonable basis of planning for the fu-
network.
ture layout of the distribution network.
5.5. Analysis of the Impact of Uncontrolled Charging on the Distribution Network
5.5. Analysis of the Impact
In accordance with theof spatial–temporal
Uncontrolled Charging on the Distribution
distribution Networkloads obtained,
of the EV charging
an effort is made to assess
In accordance with thethespatial–temporal
impact of macroscale electric vehicle
distribution of the EV disorderly
charging charging
loads ob-on
the distribution
tained, an effortnetwork.
is made to assess the impact of macroscale electric vehicle disorderly
Fromon
charging Figures 12 and 13,network.
the distribution it is obvious that, with access to the EV charging load, the
successive increase in load at
From Figures 12 and 13, it is the individual nodeswith
obvious that, in the different
access to thefunctional
EV charging areasload,
leadstheto
asuccessive
drop in the nodal voltage at each node.
increase in load at the individual nodes in the different functional areas leads
As seen
to a drop in Figure
in the 14, node
nodal voltage at18 is also
each the most affected as it is far away from node 1,
node.
the power
As seen in Figure 14, node 18 is also theismost
node, and, as the electric vehicle connected
affected to as
node
it is18,
farthe
awaynode voltage
from node at1,
this node drops
the power node,from
and,0.9631
as theto 0.9447,vehicle
electric which is
exceeds
connectedthe lower
to nodelimit
18,ofthe
thenode
nodevoltage
voltage.at
A comprehensive
this node analysis
drops from 0.9631 to of Figures
0.9447, 12–14
which shows the
exceeds that,lower
during theof
limit peaktheelectricity load
node voltage.
period, with the connection of electric vehicle charging load, the node voltage of each node
has a relatively large reduction which seriously affects the voltage quality for consumers.
 Processes2023,
Processes 2023,11,
11,xxFOR
FORPEER
PEERREVIEW
REVIEW 17 ofof 20
17 20
2023, 11, x FOR PEER REVIEW 17 of 20
Processes 2023, 11, 2256 16 of 19

Figure12.
Figure 12.Nodal
Nodalvoltage
voltageatatdifferent
differentmoments
momentsafter
afterthethecharging
chargingloadloadisislinked
linkedto tothe
thedistribution
distribution
network.
Figure 12. Nodalnetwork.
Figure at Nodal
voltage12. voltage
different at different
moments aftermoments after the
the charging load charging load
is linked toisthe
linked to the distribution network.
distribution
network.
Lowerlimit
Lower limitof
ofvoltage
voltagedeviation
deviation
Lower limit of voltage deviation
1.00
1.00
1.00
0.99
0.99
0.99
voltage/.p.u
Node voltage/.p.u

0.98
0.98
Node voltage/.p.u

0.98
0.97
0.97
0.97
Node

0.96
0.96
0.96
0.95
0.95
0.95
0.94
0.94
00 33 66 99 12 12 15 15 18 18 21 21 24 24 27 27 30 30 33 33
0.94 Node number
0 3 6 9 12 15 18 21 Node24 number
27 30 33
Node number
Figure13.
Figure
Figure 13.Voltage
13. Voltageof
Voltage ofeach
of eachnode
each nodeafter
node aftercharging
after chargingload
charging loadis
load isislinked
linkedto
linked tothe
to thedistribution
the distributionnetwork.
distribution network.
network.
Figure 13. Voltage of each node after charging load is linked to the distribution network.
0.97
0.97 Lowerlimit
Lower limitofofvoltage
voltagedeviation
deviation
0.97 Lower limit of voltage deviation
voltage/.p.u
Node voltage/.p.u

0.96
0.96
Node voltage/.p.u

0.96
Node

0.95
0.95
0.95

0.94
0.94
00 22 44 66 88 10 10 1212 1414 16
16 18 18 20
20 22
22 24
24
0.94
0 2 4 6 8 Time/h
Time/h
10 12 14 16 18 20 22 24
Time/h
Figure 14.
Figure14.
Figure Node
14.Node voltage
Nodevoltage at
voltageat each
ateach moment
eachmoment
momentofof node
ofnode 18.
node18.
18.
Figure 14. Node voltage at each moment of node 18.
Processes 2023, 11, x FOR PEER REVIEW 18 of 20

A comprehensive analysis of Figures 12–14 shows that, during the peak electricity
Processes 2023, 11, 2256 load period, with the connection of electric vehicle charging load, the node voltage of17each of 19
node has a relatively large reduction which seriously affects the voltage quality for con-
sumers.
Figure 15
Figure 15 clearly
clearly shows
shows that,
that, around
around 12:00
12:00 p.m.
p.m. and
and 20:00
20:00 p.m.,
p.m., the
the overall
overall network
network
loss of the distribution network is larger because there are more EV charging
loss of the distribution network is larger because there are more EV charging loads at loads at each
each
functional area node during this time period. Under the base load, the
functional area node during this time period. Under the base load, the peak losses arepeak losses are
0.199 MW and 0.172 MW, respectively, and when the charging load is
0.199 MW and 0.172 MW, respectively, and when the charging load is linked, the peak linked, the peak
losses are
losses are 0.220
0.220 MW
MW and
and 0.193
0.193 MW, respectively, increasing
MW, respectively, increasing by
by 10.6%
10.6% and
and 12.5%.
12.5%. It
It can
can be
be
seen that,
seen that, along
along with
with the
the charging
charging load
load being
being connected
connected toto the
the distribution
distribution network,
network, the the
distribution network’s network loss increases considerably, posing a serious threat
distribution network’s network loss increases considerably, posing a serious threat to the to the
distribution network’s safe and economical functioning.
distribution network’s safe and economical functioning.

Superimposed load
250
Basic load

200
Network loss/kW

150

100

50

0
0 2 4 6 8 10 12 14 16 18 20 22 24
Time/h
Figure 15. Distribution network loss.

6. Conclusions
6. Conclusions
In this paper, considering the effects of road congestion and ambient temperature on
In this paper, considering the effects of road congestion and ambient temperature on
the electricity consumption of electric vehicles per unit kilometer, an electric vehicle charg-
the electricity consumption of electric vehicles per unit kilometer, an electric vehicle
ing load demand prediction model containing multiple stochastic processes is developed
charging load demand prediction model containing multiple stochastic processes is de-
and solved using Monte Carlo simulation, which is used for predicting the spatio-temporal
veloped and solved using Monte Carlo simulation, which is used for predicting the spatio-
distribution of electric vehicle charging loads under the influence of road conditions and
temporal distribution of electric vehicle charging loads under the influence of road condi-
ambient temperatures, as well as investigating the impact of disordered charging on the
tions and ambient temperatures, as well as investigating the impact of disordered charg-
distribution network. Through path planning experiments and spatio-temporal prediction
ing on the distribution network. Through path planning experiments and spatio-temporal
of electric vehicle charging demand in different functional areas under actual road network
prediction of electric vehicle charging demand in different functional areas under actual
distribution network conditions, the results verify the validity of the proposed model and
road network distribution network conditions, the results verify the validity of the pro-
method, and the conclusions and main contributions are as follows:
posed model and method, and the conclusions and main contributions are as follows:
(1) The charging loads of different electric vehicles are obviously affected by the charac-
(1) The charging
teristics loads of areas,
of functional different
andelectric vehicles
there are are differences
obvious obviously affected
betweenby thecharging
the charac-
teristics of functional
characteristics areas,functional
of different and there areas.
are obvious differences
The total chargingbetween the charging
load of electric vehi-
characteristics
cles has obvious “double-peak” characteristics, and, if not reasonably guided,vehi-
of different functional areas. The total charging load of electric it is
cles has obvious
superimposed on“double-peak”
the peak value characteristics,
of the base load,and, if notinreasonably
resulting an unfavorableguided, it is
impact;
(2) superimposed
Electric vehicle onusers
the peak
tendvalue of the base
to choose roadsload,
withresulting in an levels
high traffic unfavorable
and flatimpact;
road
(2) Electric vehicle users tend to choose roads with high traffic levels and
conditions for route planning. In order to simulate user behavior more realistically, flat road con-
ditions for uses
this paper routeanplanning.
improvedInshortest
order to simulate
path userwith
algorithm behavior more realistically,
road weights this
as the objective
paper uses
function toan improved
plan shortest
the driving paths path algorithm
of electric with road
vehicles, whichweights
reducesas the
the traveling
objective
function to plancompared
time by 28.79% the driving paths
with of electric vehicles,
the traditional shortest path which reduces the traveling
algorithm;
time by 28.79% compared with the traditional shortest path algorithm;
(3) Ambient temperature and traffic conditions have an impact on electric vehicle electric-
(3) Ambient temperature
ity consumption. and to
In order traffic conditions
analyze haveofan
the impact impact
these two on electric
factors vehicle
on the elec-
charging
tricity
load ofconsumption. In order
electric vehicles, thisto analyze
paper the impact
introduces of these
traffic two factors
conditions on the charg-
and ambient tem-
perature in the modeling process, which can accurately obtain the actual electricity
consumption of electric vehicles under different environmental conditions and im-
prove the model prediction accuracy by 38.28% compared with the charging load
prediction model which does not consider these two factors and provides a reasonable
planning basis for the layout and capacity allocation of the distribution network.
Processes 2023, 11, 2256 18 of 19

Nevertheless, in the modeling process, this article assumes that there are charging
stations at each node of the road network and that the capacity of the charging stations is
large enough; therefore, the queuing time for charging a car and the process of the user
selecting a charging station when charging is ignored, and further analysis is needed in
the future taking into account the distribution of actual charging stations. In addition, due
to the phenomenon that electric vehicle charging causes peak loads to superimpose on
each other, a more effective, orderly charging strategy will be proposed in the future on the
basis of the vehicle–road network coupling system to guide electric vehicles to charge in an
orderly manner so as to achieve the effect of “peak load shifting”.

Author Contributions: Conceptualization, J.F. and Y.F.; methodology, J.F.; software, J.F.; validation,
J.F., Y.F. and X.C.; formal analysis, W.L.; writing—original draft preparation, J.F.; writing—review
and editing, X.C.; visualization, W.L.; supervision, Y.F.; funding acquisition, X.C. All authors have
read and agreed to the published version of the manuscript.
Funding: This research was funded by the Natural Science Foundation of Xinjiang Uygur Au-
tonomous Region (2022D01C365), 2022 Tianshan Talent Cultivation Programme (2022TSYCLJ0019).
Data Availability Statement: The data presented in this study are available on request from the
corresponding author. The data are not publicly available due to confidentiality agreements.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Tamba, M.; Krause, J.; Weitzel, M.; Lona, R.; Duboz, L.; Grosso, M.; Vandyck, T. Economy-wide impacts of road transport
electrification in the EU. Technol. Forecast. Soc. Chang. 2022, 182, 121803. [CrossRef] [PubMed]
2. Morfeldt, J.; Johansson, D.J.A. Impacts of shared mobility on vehicle lifetimes and on the carbon footprint of electric vehicles. Nat.
Commun. 2022, 13, 6400. [CrossRef] [PubMed]
3. General Office of the State Council of the People’s Republic of China. China’s New Energy Vehicle Holdings Reached 13.1 Million
Units, Showing Rapid Growth. 2023. Available online: https://www.gov.cn/xinwen/2023-01/11/content_5736281.htm (accessed
on 25 May 2023).
4. Song, Y.; Lin, S. Spatial-temporal Distribution Prediction of Charging Load for Electric Vehicle based on Dynamic Traffic Flow.
In Proceedings of the Joint 2019 International Conference on Ubiquitous Power Internet of Things (UPIOT 2019) & 2019 3rd
International Symposium on Green Energy and Smart Grid (SGESG 2019), Chongqing, China, 21–23 August 2019.
5. Yang, X.; Cai, Y.; Zhang, M.; Yang, X. Analysis of charging demand of household electric vehicles based on travel characteristics.
In Proceedings of the 2019 IEEE Innovative Smart Grid Technologies-Asia (ISGT Asia), Chengdu, China, 21–24 May 2019.
6. Jenn, A.; Highleyman, J. Distribution grid impacts of electric vehicles: A California case study. iScience 2021, 25, 103686. [CrossRef]
[PubMed]
7. U.S. Department of Transportation Federal Highway Administration. 2017 National Household Travel Survey; U.S. Department of
Transportation Federal Highway Administration: Washington, DC, USA, 2017.
8. Su, S.; Hui, Y.; Ning, D.; Li, P. Spatial-Temporal Distribution Model of Electric Vehicle Charging Demand Based on a Dynamic
Evolution Process. In Proceedings of the 2018 2nd IEEE Conference on Energy Internet and Energy System Integration (EI2),
Beijing, China, 20–22 October 2018.
9. Qian, T. Calculation of Electric Vehicle Charging Power Based on Spatial-Temporal Activity Model. In Proceedings of the 2018
China International Conference on Electricity Distribution (CICED), Tianjin, China, 17–19 September 2018.
10. Zhang, J.; Yan, J.; Liu, Y.; Zhang, H.; Lv, G. Daily electric vehicle charging load profiles considering demographics of vehicle users.
Appl. Energy 2020, 274, 115063. [CrossRef]
11. Yi, T.; Zhang, C.; Lin, T.; Liu, J. Research on the spatial-temporal distribution of electric vehicle charging load demand: A case
study in China. J. Clean. Prod. 2020, 242, 118457. [CrossRef]
12. Zhang, M.; Cai, Y.; Yang, X. Charging demand distribution analysis method of household electric vehicles considering users’
charging difference. Electr. Power Autom. Equip. 2020, 40, 154–163.
13. Shao, Y.; Mu, Y.; Yu, X.; Dong, X.; Jia, H.; Wu, J.; Zeng, Y. A spatial-temporal charging load forecast and impact analysis method
for distribution network using EVs-traffic-distribution model. Proc. CSEE Smart Grid 2017, 37, 5207–5219.
14. Chen, L.; Zhang, Y.; Figueiredo, A. Spatio-Temporal Model for Evaluating Demand Response Potential of Electric Vehicles in
Power-Traffic Network. Energies 2019, 12, 1981. [CrossRef]
15. Liu, Z.; Zhang, Q.; Zhu, Y. Spatial-temporal distribution prediction of charging loads for electric vehicles considering vehicle-
road-station-grid integration. Autom. Electr. Power Syst. 2022, 46, 36–45.
16. Mostafa, A.; Mohammad, R.; Mohamad, A. A novel spatial-temporal model for charging plug hybrid electrical vehicles based on
traffic-flow analysis and Monte Carlo method. ISA Trans. 2021, 114, 263–276.
Processes 2023, 11, 2256 19 of 19

17. Yan, J.; Zhang, J.; Liu, Y.; Lv, G.; Han, S.; Alfonzo, L.E.G. EV charging load simulation and forecasting considering traffic jam and
weather to support the integration of renewables and EVs. Renew. Energy 2020, 159, 623–641. [CrossRef]
18. Tang, S.; Mu, Y.; Zhou, Y.; Dong, X.; Jia, H. A Spatial-temporal Electric Vehicle Charging Load Forecasting Method Considering the
Coordination among the Multiple Charging Behaviors. In Proceedings of the 2021 Power System and Green Energy Conference
(PSGEC), Shanghai, China, 13–16 May 2021.
19. Wang, H.; Zhang, Y.; Mao, H. Charging load forecasting method based on instantaneous charging probability for electric vehicles.
Electr. Power Autom. Equip. 2019, 39, 207–213.
20. Liu, N.; Zhao, S.C.; He, N. Study of the road resistance function based on BPR function. J. Wuhan Univ. Technol. Transp. Sci. Eng.
2013, 37, 545–548.
21. Yu, T.; Ma, J. A review of the link traffic time estimation of urban traffic. In Proceedings of the 2016 IEEE International Conference
on Intelligent Transportation Engineering (ICITE), Singapore, 20–22 August 2016.
22. Zhang, Y.; Chen, Y.; Qi, S. Model of controlled intersection turning delay based on probabilistic principle. J. Jilin Univ. 2020, 50,
2113–2121.
23. Perger, T.; Auer, H. Energy efficient route planning for electric vehicles with special consideration of the topography and battery
lifetime. Energy Effic. 2020, 13, 1705–1726. [CrossRef]
24. Weiss, M.; Ruess, R.; Kasnatscheew, J.; Levartovsky, Y.; Levy, N.R.; Minnmann, P.; Stolz, L.; Waldmann, T.; Wohlfahrt-Mehrens, M.;
Aurbach, D.; et al. Fast Charging of Lithium-Ion Batteries: A Review of Materials Aspects. Adv. Energy Mater. 2021, 11, 2101126.
[CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.