1
Coordinated Charging of Multiple Plug-In Hybrid
Electric Vehicles in Residential Distribution Grids
Kristien Clement, Edwin Haesen, Student Member, IEEE and Johan Driesen, Member, IEEE
Abstract—Alternative vehicles based on internal combustion
engines (ICE), such as the hybrid electric vehicle (HEV), the plug-
in hybrid electric vehicle (PHEV) and the fuel-cell electric vehicle
(FCEV), are becoming increasingly popular. HEVs are currently
commercially available and PHEVs will be the next phase in the
evolution of hybrid and electric vehicles. The batteries of the
PHEVs are designed to be charged at home, from a standard
outlet in the garage, or on a corporate car park. The electrical
consumption for charging PHEVs may take up to 5% of the
total electrical consumption in Belgium by 2030. These extra
electrical loads have an impact on the distribution grid which is
analyzed in terms of power losses and voltage deviations. Firstly,
the uncoordinated charging is described where the vehicles are
charged immediately when they are plugged in or after a fixed
start delay. This uncoordinated power consumption on a local
scale can lead to grid problems. Therefore coordinated charging
is proposed to minimize the power losses and to maximize the
main grid load factor. The optimal charge profile of the PHEVs is
computed by minimizing the power losses. The exact forecasting
of household loads is not possible, so stochastic programming is
introduced.
Index Terms—battery charging, distribution grid, load flow
analysis, plug-in hybrid electric vehicles, stochastic programming.
I. I NTRODUCTION
ECAUSE of the large dependency of most countries
B on imported fossil fuels and the soaring oil prices,
it is essential to look for alternatives. Other major drives
for the development of hybrid electric vehicles (HEVs) are
an increasing concern for energy efficiency and limiting
greenhouse gases such as CO2 to reduce global warming
and to meet Kyoto restrictions. Downsizing the internal
combustion engine (ICE) is possible without losing
performance as the electric motor can generate a power
boost. The HEV always operates near optimum efficiency
and therefore consumes less energy to deliver the same
performance as a conventional vehicle (CV).
Plug-in hybrid electric vehicles (PHEVs) will be the
next phase in the evolution of hybrid electric vehicles. Their
batteries will be charged by plugging into electric outlets
or on-board electricity generation. These vehicles can drive
full power in electric-only mode, over a limited range. As
such PHEVs offer valuable fuel flexibility [1]. PHEVs may
have a larger battery and a more powerful electric motor
compared to a HEV, but their range is still very limited [2].
The capacity of the batteries is in the range of 10 kWh [3].
Kristien Clement, Edwin Haesen and Johan Driesen are with the
Department of Electrical Engineering, Katholieke Universiteit Leuven,
Kasteelpark Arenberg 10 box 2445, 3001 Heverlee, Belgium, e-mail:
Kristien.Clement@esat.kuleuven.be.
978-1-4244-3811-2/09/$25.00 ©2009 IEEE
There are several barriers to overcome for a successful
commercialization of PHEV. The storage of electrical energy
in batteries represents the most important limit. A PHEV
will have an additional battery capacity. Because of these
extra batteries, the ICE can be better downsized. Contrary to
fuel-cell electric vehicles, the basic charging infrastructure
for PHEVs already exists, although grid enforcements may
be necessary.
There are two main places where the batteries of PHEVs
can be recharged: either on a car park, corporate or public,
or at home. The focus here lies on the latter. From the
PHEV owner point of view, the batteries of the PHEV have
to be charged overnight so the driver can drive off in the
morning with a fully-charged battery. From the distribution
grid operator point of view, the power losses during charging
have to be minimized and transformer and feeder overload
have to be avoided. Not only power losses, but also power
quality (e.g. voltage dips, unbalance, harmonics, etc . . . ) is
essential to the distribution grid operator as well as to grid
customers. Overnight recharging can also increase the loading
of base-load power plants and smoothen their daily cycle [3]
or avoid additional generator start-ups which would enhance
the general efficiency.
TREMOVE [4] quantified the Belgian passenger vehicles for
the period 2005-2030 based on a simulation of consumer
behavior within a business as usual scenario. For new
technologies, such as HEVs, it takes some time to penetrate
the market, but they will have an important share after some
years. HEVs will take over around 7% of the market by
2010 and around 30% by 2030 in this scenario. PHEVs are
an option within the HEVs. The electrical consumption for
charging PHEVs will take at maximum about 5% of the total
electrical energy consumption in Belgium [5] assuming all
HEVs are PHEVs and depending on the scenarios of the
PRIMES model [6]. This energy would be taken from the
low voltage grid. The impact of charging the PHEVs batteries
on the distribution grid is not yet understood, but the authors
believe that it could cause a high risk of overload in case of
uncoordinated or passive charging.
In [7], the uncoordinated and coordinated charging of
the batteries of the PHEVs are discussed and deterministic
household profiles are used, assuming that there is a perfect
knowledge of the future data. In this paper, the analysis is
elaborated to stochastic data. These stochastic aspects reflect
an error in the forecasting of the daily load profiles.
 2

II. A SSUMPTIONS AND MODELING

In section II, some assumptions for the analysis of the
problem are made such as the time and the duration of the
charging and the different scenarios for the daily load profiles
that are used. A standard grid is chosen for representing a
radial distribution network.
GRID
A. Charging periods
In this paper, the batteries of the vehicles are assumed to
be charged at home. Fig. 1 shows the percentage of trips by
vehicle as function of time, meaning that these vehicles are not
available for charging. Based on Fig. 1, four essential charging Figure 2. Residential grid topology with 34 nodes.
periods are defined. The first charging period is during the
evening and night. Most of the vehicles are at home from
power is given on a 15 minute time base. Fig. 3 represents a
21h00 up to 06h00 in the morning. The next charging period
random household load for an arbitrary day during winter.
is during the night between 00h00 and 06h00. Some PHEVs
are immediately plugged in on return from work for being
ready to use throughout the evening. The number of vehicles 800
that will be charged in the evening will be smaller. These
vehicles are charged between 18h00 and 21h00. The last case 700
is charging during the day, from 10h00 till 16h00. Because
the focus in this paper lies on charging at home, the number 600
of vehicles that will be charged during the day will be lower.
Power [W]

500

12
400

10
300
Number of trips [%]

8
200
0 2 4 6 8 10 12 14 16 18 20 22 24
Time [h]
6

Figure 3. Household load during winter.

4

2
III. U NCOORDINATED CHARGING
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Uncoordinated charging means that the batteries of the
Time [h] vehicles are starting to charge immediately when they are
plugged in or after a user-adjustable fixed start delay. The
Figure 1. Percentage of trips by vehicle at each hour. vehicle owners do not have the incentive nor the essential
information to schedule the charging profile of the batteries to
optimize the grid utilization. The fixed start delay is introduced
B. Grid topology to give the vehicle owner the possibility to start charging using
The radial network used for this analysis is the IEEE 34 off-peak electricity tariffs.
node test feeder shown in Fig 2. This network is downscaled
from 24.9 kV to 230 V so this grid topology represents a A. Power flow analysis
residential radial network. The line impedances are adapted to
A load flow analysis is performed to assess the voltage
achieve tolerable voltage deviations and power losses. Each
deviations and the power losses in the selected distribution
node is a connection with a residential load and some of
grid. This analysis is based on the backward-forward sweep
the connections which are randomly chosen, have PHEVs
method to calculate the node-currents, line-currents and node-
recharging.
voltages [8]. At the initialization step, a flat profile of 230 V
is taken for the node-voltages. A constant power load model is
C. Load scenarios used at all connections. In the backward step, the currents are
From an available set of residential load measurements, two computed based on the voltages of the preceding iteration.
large groups of daily winter and summer load profiles are se- In the forward step, the voltages are computed based on
lected. The load profiles cover 24 hours and the instantaneous the voltage at the root node and the voltage drops of the
 3

Table I
lines between the nodes. The backward and forward sweep R ATIO OF POWER LOSSES TO TOTAL ENERGY [%] FOR THE 4 K W
is formulated as a matrix multiplication. The currents and CHARGER .
voltages are updated iteratively until the stopping criterion
Charging period Season 0% 10% 20% 30%
based on node-voltages is reached.
21h00-06h00 Summer 1.6 2.0 2.5 2.9
Winter 3.1 3.5 4.0 4.4
00h00-06h00 Summer 1.4 1.9 2.6 3.3
B. Methodology Winter 2.8 3.3 4.0 4.7
10h00-16h00 Summer 0.9 1.5 2.3 3.0
At the start of a 24-hour cycle, a daily profile is randomly Winter 1.3 1.9 2.7 3.3
selected and placed at each node. In the first case, no PHEVs 18h00-21h00 Summer 1.0 1.9 3.4 4.6
are assumed so this is taken as a reference case. The next Winter 1.7 2.7 4.2 5.4
three cases have a PHEV penetration of respectively 10, 20
Table II
and 30%. The PHEVs are randomly placed. For example, for M AXIMUM VOLTAGE DEVIATIONS [%] FOR THE 4 K W CHARGER .
the 30%-penetration case, almost one third of the nodes will
have a PHEV-load randomly placed. Charging period Season 0% 10% 20% 30%
The profile for charging the PHEVs is kept straightforward. 21h00-06h00 Summer 4.3 4.8 5.7 6.4
Winter 7.3 8.0 9.0 9.8
The batteries are charged at a constant power of 4 kW. The 00h00-06h00 Summer 3.4 3.9 5.2 6.4
charger of 4 kW is chosen because the maximum power output Winter 6.8 7.0 7.9 9.1
of a standard outlet is 4.6 kW. So this is the largest charger that 10h00-16h00 Summer 1.7 3.0 4.6 5.8
can be used for a standard outlet at home. The starting point Winter 2.6 3.9 5.5 6.8
18h00-21h00 Summer 1.8 3.2 5.2 6.8
of the PHEVs charging is randomly chosen for each individual Winter 3.1 4.6 6.7 8.4
vehicle within a specific period such that the vehicles still are
fully charged at the end of the charging period. It is assumed
that the batteries of the vehicles are empty at the first time Fig. 4 depicts the voltage profile of the last node of
step. the distribution grid for a penetration degree of 0% and
For every quarter of an hour, the backward-forward power 30% during the winter. This figure shows only one charging
flow analysis is repeated to compute the voltage at each node example and is not the average of several samples. There
until convergence is obtained. The results are discussed in the is clearly a decrease of the voltage in the presence of the
next paragraph. PHEVs during the charging period which is between 00h00
and 06h00. Between 02h00 and 04h00, all the vehicles are
charging and the voltage drop during these hours is the largest.
C. Results
The power needed for charging these vehicles is significantly
The impact of uncoordinated charging on the distribution higher compared to the household loads during the night. The
grid is illustrated by computing the power losses and the small difference in voltage deviations during the rest of the
maximum voltage deviation for the different charging periods. day is caused by the different selected load profiles for both
The power losses are normally distributed over 1000 samples cases.
and the number of samples is large enough to have an accurate
average. The results of the 4 kW charger are showed in Table
230
I and II.
Table I depicts the ratio of the power losses to the total 225

load. This load includes the daily household load and the 220
charging of the PHEVs if present. The power losses are always 215
Voltage [V]

higher in the winter season compared to the summer season

in all cases. The increase of the number of PHEVs indicates a 210

significant increase in the ratio of power losses. Power losses 205

are important for the operator of the distribution grid. The
200
operator will compensate higher losses by increasing its grid 0% PHEV
tariffs. The regulator, who has to approve these tariffs, can in 195 30% PHEV
turn force the operator to take appropriate measures to cope 190
0 2 4 6 8 10 12 14 16 18 20 22 24
with PHEVs. Time [h]
Not only the power losses, but also the voltage deviations of
the grid voltage (230 V) which are represented in Table II, are Figure 4. Voltage profile with 30% PHEVs compared to the voltage profile
with 0% PHEV.
important for the operator of the distribution grid. An increase
in the number of PHEVs indicates a significant increase in
the voltage deviations. Voltage deviations up to 10% in low
voltage grid are acceptable according to the EN50160 norm IV. C OORDINATED CHARGING
which is mandatory in Belgium. For a penetration of 30%, In the previous section, the charging of the batteries of the
some of the voltage deviations are close to 10%, especially PHEVs starts randomly, meaning immediately when they are
during the winter season. plugged in or after a fixed start delay. The idea of this section is
 4

to achieve optimal charging by minimizing the power losses

to optimize to grid utilization. A direct coordination of the
charging will be done by smart metering and by sending
signals to the individual vehicles.

A. Optimization
By minimizing the power losses, the owners of PHEVs will
no longer be able to control the charging profile. The only
degree of freedom left is for the owners to indicate the point in
time when the batteries must be fully charged. For the sake of
convenience, the end of the indicated charging period is taken
as point in time when the vehicles must be fully charged. The
power of the charger varies between zero and maximum and is
no longer constant. The coordinated charging is analyzed for
the same four charging periods. The range of the penetration
degree remains the same and the vehicles are also randomly
placed. The starting time of charging is not arbitrary selected
any more, but determined by the optimization program. The
same IEEE 34 node test grid is used. For each charging
period and season, the power losses and voltage deviations
are calculated and compared with the values of uncoordinated
charging.

B. Methodology
The objective function is to minimize the power losses
which are treated as a reformulation of the non-linear power
flow equations. This non-linear minimization problem can be
tackled as a sequential quadratic optimization [9]. The power
of the charging obtained by the quadratic programming cannot
be larger than the maximum power of the charger Pmax . The
batteries must be fully charged at the end of cycle, so the
energy which flows to the batteries must equal the capacity of
the batteries Cmax . xn is zero if there is no PHEV placed and
is one if there is a PHEV at node n. The goal is to minimize
power losses while taking into account these constraints. The
quadratic programming (QP) uses the formulas of (1) and (2).
t
max lines

min 2
Rl · Il,t (1)
t=1 l=1
 ∀t, ∀n  {nodes} : 0 ≤ P ≤ P
 max n,t max
s.t. ∀n  {nodes} : tt=1 Pn,t · t · xn = Cmax (2)
xn  {0, 1}
Fig. 5 represents the outline of the algorithm of coordinated
charging. The vehicles are randomly placed after the selection
of a daily load profile and the number of PHEVs. A flat
voltage profile is assumed and the node-voltages are computed
with the backward-forward sweep method assuming that there
are no PHEVs. The quadratic optimization is performed to
determine the optimal charging profile and the node-voltages
are computed again. This process is repeated until the power
losses based stopping criterion is reached.
Figure 5. Algorithm of coordinated charging.
C. Results
This paragraph describes the results of the coordinated
charging to illustrate the impact on the distribution grid. Table
III and IV represent respectively the power losses and the
maximum voltage deviations for the coordinated charging
during the different charging periods. These results must be
compared with Table I and II.
The voltage deviations are in accordance to the EN50160
norm and the maximum voltage deviations for a penetration
degree of 30% is now well below 10%. For a larger number of
vehicles, the coordinated charging would not be able to reduce
the maximum voltage deviations well below 10%. To comply
with this norm, the distribution grid has to be reinforced to
maintain voltage deviations below 10%. The voltage deviation
during the evening peak is larger than the voltage deviation
caused by the extra load of charging vehicles. Probably, the
vehicles will not be charged at full power rate during this
peak to obtain the objective to minimize the power losses.
For a vehicle penetration of 30%, the number of vehicles is
increased and the charging is more distributed. This increases
the voltage deviation to a level which is higher compared to
the evening peak level.
For all charging periods and seasons, the power losses are
 5

Table III
R ATIO OF POWER LOSSES TO TOTAL ENERGY [%] FOR THE 4 K W A. Principles of stochastic programming
CHARGER .
The daily load profiles are the essential input variables. The
Charging period Season 0% 10% 20% 30% uncertainties of these variables can be described in terms of
21h00-06h00 Summer 1.6 1.8 2.1 2.4 probability density functions. In that way, the fixed input vari-
Winter 3.1 3.2 3.5 3.8 ables are converted into random input variables with known
00h00-06h00 Summer 1.4 1.7 2.3 2.8 normal distributions. N independent samples of the random
Winter 2.8 3.2 3.7 4.2
10h00-16h00 Summer 0.9 1.3 1.9 2.3 input variable ω j , the daily load profile, are selected.
Winter 1.3 1.7 2.3 2.8 Equation (3) gives
 the estimation
 for the stochastic optimum
18h00-21h00 Summer 1.0 1.8 3.1 4.1 v̂n . The function g x, ω j gives the power losses and x is the
Winter 1.7 2.6 3.9 5.0
power rate of the charger for all the PHEVs and for time steps.
Table IV fˆN is a sample-average approximation to the objective of the
M AXIMUM VOLTAGE DEVIATIONS [%] FOR THE 4 K W CHARGER . stochastic programming problem.
⎧ ⎫
Charging period Season 0% 10% 20% 30% ⎨ 1 N
  ⎬
21h00-06h00 Summer 4.3 4.3 4.3 4.6 v̂n = min fˆN (x) ≡ g x, ω j (3)
Winter 7.3 7.3 7.3 7.3 ⎩ N ⎭
j=1
00h00-06h00 Summer 3.4 3.4 4.1 4.7
Winter 6.8 6.8 7.0 7.6
10h00-16h00 Summer 1.7 2.2 3.0 3.6
The mean value of the power losses, E (v̂n ), is a lower
Winter 2.6 3.0 3.8 4.4 bound for the real optimal value of the stochastic programming
18h00-21h00 Summer 1.8 2.9 4.6 5.9 problem, v ∗ , as shown in (4).
Winter 3.1 4.3 6.0 7.4

E (v̂n ) ≤ v ∗ (4)
decreasing if the coordinated charging is applied. E (v̂n ) can be estimated by generating M independent
Fig. 6 shows the load profiles of the nodes 2 and 33 with samples ω i,j of the random input variable each of size N. M
a penetration degree of 30% during the charging period from optimization runs are performed in which the non-linear power
00h00 till 06h00. The nodes are chosen at the begin and end flow equations are solved by using the backward-forward
point of the grid feeder. It is clearly that the power output of sweep method. According to (5), v̂nj is the mean optimal value
the charger is not constantly 4 kW but varies. of the problem for each of the M samples. The optimal values
of the M samples constitute a normal distribution.
3000  N

node 2
ˆ j 1   
node 33 v̂nj = min fN (x) := g x, ω i,j
, j = 1...M (5)
2500 N i=1

From equation (6), LN,M is a unbiased estimator of E (v̂n )

Power of the charger [W]

2000
1500
1000
500
0
0 1 2 3
Time [h]
Figure 6. Load profile of the 4 kW charger.
V. S TOCHASTIC PROGRAMMING
The results of the previous sections are based on deter-
ministic or historical data for the daily load profiles. So the
essential input variables are fixed. But for this approach, a
sufficient number of measurement data must be available. Most
of the time, these measurements are not adequate to do a
perfect forecasting of the data. This can be handled by using
a stochastic approach. These aspects represent an error in the
forecasting of the daily load profiles.
which is a statistical lower bound for the true optimal value
[10]. Simulations indicated that in this type of problem the
lower bound converges to the real optimal value when N is
sufficiently high and that most cases converge to a single
solution.
N

j
LN,M = v̂N (6)
j=1
4 5 6
B. Methodology
The impact of the problem of forecasting the daily load
profile for the next 24 hours is the aim of this section. The
daily load profiles of the available set are varied by a normal
distribution function. The standard deviation σ is varied in
such a way that 99.7 % of the samples varies at maximum
25% of the average μ. Positive covariances are added.
C. Results
For 2000 independent samples of the random input variable,
the daily load profile, one optimal charging profile is calcu-
lated. Because there is one optimal charge profile that fits for
the 2000 household loads, it is not possible to give the results

as a ratio of the power losses to the total energy. Therefore,
the absolute value of the power losses is used. The optimal
charging profile is used to determine the power losses for the
2000 individual load profiles. This is the stochastic optimum.
For each of these 2000 load profiles, the optimal charging
profile and the corresponding power losses are also computed,
which is the deterministic optimum. The power losses are also
normal distributed, although the power flow equations have a
non-linear nature.
The power losses of the deterministic optimum are sub-
tracted from the power losses of the stochastic optimum which
is shown in Fig. 7. The value of this difference is always
positive. The forecasting of the daily load profiles introduces
an efficiency loss because the charge profiles of the PHEVs
are not optimal for this specific daily load profile.
Since the power losses are depending on the total load, the
ratio of the power losses to the total energy for the difference
between the stochastic and the deterministic optimum is shown
in Fig. 8. If the standard deviation of the normal distribution is
reduced, the 2000 charge profiles of the deterministic optimum
will converge to the optimal charge profile and the efficiency
loss will reduces.
250
200
150
100
50
0
0 50 100 150
The power losses [Wh]
Fig. 7. Histogram of the difference between the stochastic and the deterministic
optimum of an arbitrary day during winter.
250
200
150
100
50
0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
The ratio of power losses to the total load [%]
Fig. 8. Histogram of the difference between the stochastic and the deterministic
optimum of an arbitrary day during winter.
6
VI. C ONCLUSION
In general, the coordinated charging of PHEVs can improve
power losses and voltage deviations by flattening out peak
power, although in some cases grid enhancement will be
necessary. In this paper, the charging periods are arbitrary
chosen, but the impact of the penetration level is large. At the
first stage of the analysis, historical data is used so there is a
perfect knowledge of the load profiles for every 15 minutes of
the day on a 24-hour cycle. In the second stage, stochastic
programming is introduced. Not being able to forecast the
daily load profiles exactly, increases the power losses. In future
research, the stochastic programming technique will be further
developed.
R EFERENCES
[1] A. Raskin, S. Shah, The emerge of hybrid electric vehicles, Alliance-
Bernstein, 2006.
[2] M. Anderman., The challenge to fulfil electrical power requirements of
advanced vehicles, Journal of Power Sources, 127(2004).
[3] P. Denholm, W. Short, An Evaluation of Utility System Impacts and
Benefits of Optimally Dispatched Plug-In Hybrid Electric Vehicles,
Technical Report, October 2006.
[4] S. Logghe, B. Van Herbruggen, B. Van Zeebroeck, Emissions of road
traffic in Belgium, Transport & Mobility Leuven, January 2006.
[5] Clement K., Van Reusel K., Driesen J., The consumption of electrical
energy of plug-in hybrid electric vehicles in Belgium, EET-European
Ele-Drive Transportation conference, Brussels, Belgium, 30th May - 2nd
June, 2007.
[6] Commission Energy 2030, Belgium’s energy challenges towards 2030,
preliminary report, November 2006.
[7] Clement K., Haesen E., Driesen J. The impact of uncontrolled and
controlled charging of plug-in hybrid electric vehicles on the distribution
grid, EET-2008 3rd European Ele-Drive Transportation Conference,
Geneva, Switzerland, March 11-13, 2008.
[8] Kersting W., Distribution System Modeling and Analysis, Boca Raton,
2002.
[9] Haesen E., Driesen J., Belmans R. Robust planning methodology for
integration of stochastic generators in distribution grids, IET Journal on
Renewable Power Generation, vol.1, no.1, March, 2007.
[10] J. Linderoth, A. Shapiro and S. Wright, The em-pirical behavior of
sampling methods for stochastic programming, Annals of Operations
Research, Vol. 142, No. 1, February 2006.
200
Kristien Clement received the M.S. degree in
electro-mechanical engineering in 2004 with spe-
cialization Energy. Currently, she is working to-
ward her Ph.D in electrotechnical engineering at
the K.U.Leuven division ELECTA. Her research
interests include hybrid and electric vehicles.
0.016
 7

Edwin Haesen (S’05) received his M.Sc. degree in Johan Driesen (S’93-M’97) was born in Belgium in
electrical engineering at the KU Leuven in 2004. 1973. He received the M.Sc. and Ph.D. degrees in
For his Master Thesis on ”Technical Aspects of electrical engineering from Katholieke Universiteit
Congestion Management” he received the European Leuven (K.U. Leuven), Leuven, Belgium, in 1996
Talent Award for Innovative Energy Systems 2005. and 2000 on the finite element solution of coupled
He is currently pursuing a Ph.D. at the KU Leuven as thermal-electromagnetic problems and related appli-
a research assistant at the division ESAT-ELECTA. cations in electrical machines and drives, microsys-
His research interests are in the domain of power tems, and power-quality issues. Currently, he is an
system analysis and distributed generation. Associate Professor with K.U. Leuven and teaches
power electronics and drives. In 2000-2001, he was
a Visiting Researcher with the Imperial College of
Science, Technology and Medicine, London, U.K. In 2002, he was with
the University of California, Berkeley. Currently, he conducts research on
distributed generation, including renewable energy systems, power electronics
and its applications (e.g., in drives and power quality).