Professional Documents
Culture Documents
Ijepes D 22 04124
Ijepes D 22 04124
Agustín Doña
Abstract: One way to encourage the use of electric vehicles (EVs) is to reduce the connection
times by using higher power charging points at homes. Nonetheless, a massive
adoption of this charging technology can potentially stress the grid above the limits due
to the additional power requirement. So, it is important to assess how much the
aggregate power will increase in comparison with the slow-charging process and how
this increase could be minimized. Hence, in this work, an EV charging coordination
model is developed to compare the impact on system demand of residential slow and
fast charging plans by minimizing the maximum aggregate demand of the system,
where “slow and fast” correspond to the minimum and maximum charging speed for
AC 1-phase in mode 3 of the IEC 61851-1 standard. To create an accurate model, we
used realistic UK data to mimic load profiles behaviour and EV profiles requirements,
also including constraints to model the time intervals at which EVs are not at home.
Additionally, given the stochasticity of customers' behaviour, we solved the
optimization problem for thousands of different groups of households and EVs in order
to get valid conclusions. The results show that due to the greater flexibility provided by
fast-charging events (i.e., shorter connection times), if no coordination is considered,
the aggregate effects only increase by 15% in comparison with the slow-charging case.
On the other hand, if the process is coordinated, then almost the same maximum
aggregate demand is found regardless of whether the charging process is fast or slow.
Eduardo Martinez
The University of Manchester
alex.martinezcesena@manchester.ac.uk
Opposed Reviewers:
Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Cover Letter
Dear Editor of the International Journal of Electrical Power and Energy Systems,
We are submitting a manuscript for consideration of publication in the International Journal of Electrical
Power and Energy Systems. The manuscript is entitled “Is single-phase residential EV fast-charging a good
idea? Evidence from optimal charging plans based on realistic data”. It is important to remark that this
work has not been published nor submitted simultaneously elsewhere.
Given the climate crisis that we live in, and the need to reduce global CO2 emissions, it is imperative to
electrify the transport sector. A way to encourage the adoption of electric vehicles is by using higher power
charging points at homes. In this work, we developed an EV charging coordination model to compare the
impact on system demand of residential slow and fast charging plans for thousands of different groups.
We considered “slow” and “fast” residential charging as the minimum and maximum charging speed for
AC 1-phase allowed in the UK in mode 3 of the IEC 61851-1 standard. Furthermore, to create an accurate
model we included realistic load profiles, EV charging profiles and statistics of mobility patterns.
The results show that incorporating residential fast-charging is effectively a good idea. In fact, when
coordination is not considered, the aggregate effects only increase by 15% in comparison with the slow-
charging case (less vehicles are likely to be charged at the same time due to the shorter connection time).
Furthermore, if residential fast-charging is coordinated, the maximum aggregate demand would reach
levels similar to those of coordinated slow-charging. This finding could allow policymakers to encourage
the adoption of higher power charging points at homes, which in turn encourages users to adopt electric
vehicles thanks to the shorter connection times.
Highlights
• Higher power residential EV charging brings more flexibility to users and the grid
• UK data is used to model load profiles, EV requirements and vehicle user behaviour
• When coordination is included, fast and slow charging reach similar peak demand
Abstract
One way to encourage the use of electric vehicles (EVs) is to reduce the con-
nection times by using higher power charging points at homes. Nonetheless,
a massive adoption of this charging technology can potentially stress the grid
above the limits due to the additional power requirement. So, it is impor-
tant to assess how much the aggregate power will increase in comparison with
the slow-charging process and how this increase could be minimized. Hence,
in this work, an EV charging coordination model is developed to compare
the impact on system demand of residential slow and fast charging plans by
minimizing the maximum aggregate demand of the system, where “slow and
fast” correspond to the minimum and maximum charging speed for AC 1-
phase in mode 3 of the IEC 61851-1 standard. To create an accurate model,
we used realistic UK data to mimic load profiles behaviour and EV profiles
requirements, also including constraints to model the time intervals at which
EVs are not at home. Additionally, given the stochasticity of customers’
behaviour, we solved the optimization problem for thousands of different
groups of households and EVs in order to get valid conclusions. The results
show that due to the greater flexibility provided by fast-charging events (i.e.,
shorter connection times), if no coordination is considered, the aggregate ef-
fects only increase by 15% in comparison with the slow-charging case. On
the other hand, if the process is coordinated, then almost the same maximum
aggregate demand is found regardless of whether the charging process is fast
or slow.
Keywords: Electric Vehicles, Charging Coordination, Distribution
Network, Load Management
PACS: 0000, 1111
1. Introduction
According to the main findings reported by Working Group I (WGI) to
the IPCC Sixth Assessment Report (AR6), it is irrefutable to conclude that
human activity has warmed our planet [1]. The magnitude of this impact is
such that atmospheric CO2 concentrations reached in 2019 are the highest in
at least 2 million years. For these reasons, it is imperative to take immediate
action to reduce CO2 emissions.
According to Global EV Outlook [5], despite the pandemic and the global
economic slowdown, the EV market continues to grow. Thanks to the devel-
opment of regulatory framework in recent years, the drop in battery costs,
and the addition of incentives to protect EV sales from economic recession,
in 2021 EV sales nearly doubled to 6.6 million compared to 2020[5].
2
At the distribution network level, the integration of EV increases system
demand, which can cause feeder overload, increases in power losses, decreases
in voltage profiles, and potentially increases in expansion requirements on the
network infrastructure [8]. These problems could eventually increase when
more home chargers of higher power capacity are integrated, leading to even
higher requirements on grid infrastructure. In consequence, the adoption of
more EVs will bring new challenges for the distribution network operators
since the incorporation of these new demands will imply more efforts in net-
work planning.
This last problem takes special relevance considering that EVs load tends
to coincide at the points of maximum demand of the system [9], which may
also increase the problems of overloading in conductors and transformers as
increase voltage drops. The management of EV charging gives a potential
alternative to the increase in network capacity due to EV integration. Shift-
ing the EV charging load to the non-peak load periods will avoid overloading
problems [10], and therefore smart EV charging will reduce investments in
network capacity by decreasing the overall demand on the system [11].
3
tion model is based on real UK population data using EV charging profiles
[28], along with vehicle user behavior statistics [15] and representative pro-
files of UK residential consumption [16].
2. State of Art
During the last years, numerous investigations have demonstrated the im-
pact of uncoordinated EV charging on distribution networks, covering vari-
ous scenarios of EV penetration, charging paradigms, rate plans, and others.
[17] studies control architecture and market framework designed to handle
the challenges involved with EV charging in distribution networks. The re-
sults showed that if a centralized, advanced EV charging control strategy is
implemented, the system can adopt a larger number of EVs without any grid
reinforcements. By analyzing low voltage networks, [18] and [19] confirmed
that a high number of EVs without coordination implies increases in loading
levels at feeders and transformers, increases in voltage deviations of residen-
tial feeders, and increases in the infrastructure expansion requirements.
4
the EV load by dividing the problem into two stages, where the first of these
solves the minimization of the maximum demand of the system, while in the
second the demand fluctuation is minimized by using the maximum demand
value found in the first stage. More complex systems, such as the one solved
in [24] encompass optimization in a hierarchical way through three levels
of optimization: provincial, municipal, and charging station level, combining
minimizations between the maximum peak demands and cost minimizations.
When studying the effect of users charging their EVs with higher-power
household chargers in an uncoordinated scenario versus a coordinated one, we
found that little research delves into these issues. [25] shows that higher levels
of EV adoption, including residential fast-charging at higher power, stress the
network even more than the same level of slow-charging adoption. [26] stud-
ies the coordination of large-scale PEVs, considering different types of PEVs
charging within the network (1.96 to 7.2[kW ]). This study proposes coordi-
nating the high penetration load of EVs by using non-cooperative game the-
ory. The load management of EVs is modeled considering a pricing system,
designed to minimize the cost of each EV customer while satisfying network
load constraints. [27] shows that in scenarios with 3.3[kW ] EV slow-charging
and 7.2[kW ] EV fast-charging, if the EVs’ charging is not coordinated, the
transformers and the network may experience technical problems.
The limitations of the previous studies [25]-[27], are that they created
the smart EV charging plans without including drivers behavior and without
using actual EV charging data, which affects the feasibility of being able to
shift the EV charging outside peak hours. This is because EV owners may
not be available to charge when the coordination model requires, or they
may have energy requirements that lead them to charge during peak hours.
Not including these constraints makes it uncertain if the problems can be
eliminated considering high levels of fast-charging adoption.
5
Figure 1: Representative diagram of the coordination problem to solve, where (a) is a
sample of 10 EV slow-charging profiles, and b) is the optimally ordered charging blocks.
process. This greater versatility also benefits users since the residential fast-
charging would allow them more usage time with their vehicles. Furthermore,
this research will contribute to the realistic analysis of EV charging coordi-
nation by using real EV load profiles and unavailable time windows.
3. Methodology
EV charging coordination seeks to decrease the maximum coincident de-
mand of EV load and thus avoid potential network reinforcements. With
this idea, we solved the management problem by minimizing the maximum
aggregate demand of the system. Figure 1 represents the task done by the
optimization problem (solved for slow and fast household charging), and
consists of arranging energy rectangles of EV charging profiles (blue ones)
to decrease the system demand. As is shown in (a), the peak demand is
at night, then to smooth the curve, the model in (b) moves the EV energy
rectangles to low demand hours and skips the hours where the EVs are not
at home to be charged (orange rectangles). As a result, in the coordinated
case (b), most of the energy rectangles are moved to the early morning hours.
6
Figure 2: Global outline of the proposed methodology.
Energy
4.1 Analysis
Requirements
without
coordination
1. Uncoordinated 3. EVs Charging
fast charging Coordination 4. EVs Charging
profiles construction Model Coordination
Low EV Results
Unc. charging (slowor fast)
4.2 Statistical
2. EVs
Analysis of
unfeasible time
optimal
intervals
charging plans
Residential Methodology steps
Demand Profiles Input variables and data
7
Figure 3: General diagram for the analysis of charging plans.
Uncoordinated charging
Evaluation of objective
Random PEV charging
values and ADMD
1° Selection 2° coordination 3°
n=1 reduction percentages
First launching n = n+1
Until reaching the number of launches 4°
8
4. Fast-charging profiles construction
The proposed EV charging coordination was analyzed through two types
of residential EV charging: fast and slow household charging. Slow-charging
corresponds to the data of 2000 residential daily charging profiles at 3.6[kW ].
These profiles were built using the functions and data presented in [28], based
on My Electric Avenue [14]. On the other hand, fast-charging corresponds to
the use of fast chargers at the residential level and is carried out at 7.2[kW ].
For fast-charging, 2000 profiles were also created, which were built from the
previous slow-charging EV profiles using the following considerations:
• The energy requirements of the EVs built from the measurement cam-
paign remains unchanged; since the fast-charging is completed at higher
power, these requirements can be satisfied in a shorter period, assum-
ing that the use pattern (i.e., daily kilometers traveled per driver) does
not change between slow and fast charging cases.
• In each time interval in which a vehicle is charged at 3.6[kW ], the
transfer to fast charge of 7.2[kW ] takes half the charging time.
• The period in which fast-charging occurs is within the same time win-
dow as slow-charging and lasts half the time.
The period in which fast-charging occurs during the slow-charging time
window is a decision that depends on people’s behavior. Since this cannot be
determined with certainty, we performed three slow-to-fast charging transfer
sensitivities: one in the beginning, one in the middle, and one at the end
of the slow-charging time window by analyzing the aggregate system. Here
the aggregate demand was divided by the number of charges (in this case,
2000) to study the average effect on the individual load that each customer
has when moving from slow-to-fast charging. Aggregate demand divided by
the number of loads will be called ADD (After Diversity Demand).
After building the fast-charging profiles for the three sensitivities pre-
viously defined, we calculated the aggregate demand for each case. The
resulting ADD considering only EVs is shown in Figure 4a, and the ADD
taking into account both EVs and dwellings is shown in Figure 4b. As can
be seen, during peak hours, the aggregated profiles for each sensitivity are
very similar when EVs and dwellings are considered.
9
Figure 4: After Diversity Demand (ADD) comparison. Figures 4a and 4b have the dif-
ference between the fast-charge ADD curve of each parameterization and the slow-charge
ADD curve colored.
(a) Fast-charging profile ADD by parameteriza- (b) Fast-charging profile ADD by parameteriza-
tion considering only EVs tion considering EVs and Dwellings
2.5 2.5
Slow charging Slow charging
Fast charging, beginning Fast charging beginning
2.0 Fast charging, middle 2.0 Fast charging middle
Fast charging, end Fast charging end
1.5 1.5
[kW]
[kW]
1.0 1.0
0.5 0.5
0.0 0.0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
00:1
01:5
03:3
05:1
06:5
08:3
10:1
11:5
13:3
15:1
16:5
18:3
20:1
21:5
23:3
00:1
01:5
03:3
05:1
06:5
08:3
10:1
11:5
13:3
15:1
16:5
18:3
20:1
21:5
23:3
Time interval Time interval
Table 1: Percentage change of ADMD when moving from slow-to-fast charging by fast-
charging parameterization and type of demand studied (EVs or EVs+Dwellings)
10
the case to analyze throughout this work. For this case, the differences are not
significant among slow-to-fast parametrization; therefore, the optimization
problem and the analyses were made using just the sensitivity in which the
middle of the slow-charging time window coincides with the middle of fast-
charging.
The resulting NTS database after applying the filters and selecting a
random group of 2000 profiles is represented in Figure 5, which shows the
number of trips in progress throughout the selected typical weekday with
a resolution of 10 minutes. The 2000 profiles extracted kept the same pat-
tern compared with the original data (most of the trips happen between the
period 40 and 120 of the time interval, with peaks around the periods 50
and 110). Therefore, the simplified database adequately characterizes the
daily commute distribution of vehicles in the UK. After filtering, to provide
11
6000 Original distribution
Distribution after filters
Distribution after selection of 2000 profiles
5000
4000
N° of trips
3000
2000
1000
0
0 20 40 60 80 100 120 140
24 hours - 10 minutes resolution
Figure 6: Number of EV charging and
Figure 5: Distribution of trips in travelling
progress of the NTS database.
X
A[p, c] = 1, (∀c) (1b)
ρ
X
A[p, c] = 1, (∀p) (1c)
c
12
hoursX
outside
A[p, c] × c(i) ≤ 0, (∀p, ∀c) (1d)
i
13
and increasing the potential number of EVs that could be incorporated into
the system. Coordination will make possible to observe the positive effects
of reduced connection times for fast-charging, facilitating the coordination
process (i.e., more flexibility due to the shorter connection periods). The
mathematical formulation used for the home charging coordination of EVs
is presented below:
( n−1 )
X
(P ) min max q · pi [t] + Dres [t] (2a)
t
i=0
T
X
ηi · q · pi [t] = Eidem ∀i ∈ I (2b)
t=1
E dem
Tmin = min Tmin set , i ∀i ∈ I (2e)
q
14
Equations 2a-2k represent the optimization problem. The decision vari-
able is pi [t], a binary variable that indicates if a vehicle i is charging at the
instant t. The equation 2a corresponds to the minimization of the objec-
tive function given by the maximum consumption of dwellings plus EV loads
throughout the day. Constraint 2b states that the energy demands of users
must be satisfied, considering a feasible charging power of q ∈ {3.6, 7.2}[kW ]
for fast and slow home charging, respectively. Additionally, 2c corresponds to
the constraint of unfeasible time intervals, where it is not possible to charge
the EVs (since they are not at home). It is because of this constraint (2c)
that the use of realistic mobility patterns becomes relevant. The set Θif f
precisely define the unfeasible time intervals for each vehicle i. Charging ef-
ficiency per vehicle (ηi ) can be considered as 1 in the model, since the actual
charging profiles measured in My Electric Avenue project [14] correspond to
the power directly withdrawn from the distribution network.
15
Specifically, n corresponds to the total number of EVs, and T is the number
of intervals where the problem is solved; for 24 hours and a resolution of 10
minutes, T is 144. We modeled a resolution of 10 and 5 minutes for the slow
and fast case, respectively. Doubling the time samples for the fast case solves
a numerical problem of the constraint 2b to satisfy the EV energy for cases
where the charging time windows are half the time of odd time windows of
its slow-charging version.
This base case studies the ADMD index for different numbers of EVs
and dwellings (group size), and for each of these groups, we performed 1000
16
Figure 7: ADMD boxplot for uncoordinated fast and slow-charging based on EV group
size between 1 and 1000
(a) ADMD by group of EVs, fast charge (b) ADMD by group of EVs, slow charge
$ '