Electrical Engineering in Japan, Vol. 189, No.

4, 2014
Translated from Denki Gakkai Ronbunshi, Vol. 133-B, No. 2, September 2012, pp. 157–166

A Study on Compensating Voltage Drop in Distribution Systems due to Nighttime

Simultaneous Charging of Electric Vehicles Utilizing Charging Power Adjustment
and Reactive Power Injection

YUKI MITSUKURI,1 RYOICHI HARA,1 HIROYUKI KITA,1 EIJI KAMIYA,2 SHOJI TAKI,2
NAOYA HIRAIWA,2 and EIJI KOGURE2
1 GraduateSchool of Information Science and Technology, Hokkaido University, Kita 14, Nishi 9,
Kita-ku, Sapporo 060-0814, Japan
2 The Tokyo Electric Power Company, Inc, 1-1-3 Uchisaiwai-cho, Chiyoda-ku, Tokyo 100-8560, Japan

SUMMARY ing concerns about energy security, sources of renewable

energy are attracting increasing attention. It is expected
To reduce the environment load, electric vehicles that the use of solar generation in power distribution sys-
(EV) have been attracting attention and expectations. since tems will grow at a fast pace. One problem related to the
they can run with at good fuel efficiency and without ex- wider introduction of distributed power sources (including
haust gases. It is expected that EV will become more and solar generators) in distribution systems is voltage increase
more popular. Generally, EVs would be used for driving caused by reverse power flow from distributed generators.
during daytime and be charged during the nighttime, con- Voltage maintenance in power distribution systems at the
sidering lifestyle and time-of-use patterns. If most EV own- consumer site, which is the end system connection, is an
ers adopt this use pattern, feeder voltage profiles could be important part of system operation and includes measures
greatly affected, causing severe voltage drops. The battery against the well-known problem [1] of overvoltage caused
charger for EVs described in this papers consists of a self- by nonopening of capacitors installed for improvement of
commutated inverter, which in principle can control not the customers’ power quality at times of light load. Sev-
only active power but also reactive power. We propose a eral methods for solving this problem have been proposed
basic feeder voltage regulation algorithm, “PQ control”, [2–6].
using EVs which involves adjustment of the EV charging Electric vehicles (EV) have also been attracting at-
schedule and reactive power injection. In addition, we pro- tention as a way of reducing the environmental impact
pose “prediction control”, which is based on “PQ control” of high fuel costs and exhaust gas emissions in the near
and is takes account of prediction of the receiving volt- future. A publication by the Ministry of Economy, Trade
age. Advanced “communication control” regulation algo- and Industry of Japan, “Next-Generation Vehicle Strategy
rithms, utilizing communication among EVs and the dis- 2010” [7] provides target figures for next-generation motor
tribution system operator, are also proposed. The purpose vehicles (including EVs) and indicates that that their num-
of this paper is to confirm the effectiveness of the three ber is likely to increase. The electric energy required for EV
proposed algorithms by simulations. C⃝ 2014 Wiley Peri- operation will probably come from such charging sources
odicals, Inc. Electr Eng Jpn, 189(4): 9–21, 2014; Published as households, commercial facilities, and public facilities.
online in Wiley Online Library (wileyonlinelibrary.com). This implies a considerable increase in the load on power
DOI 10.1002/eej.22558 distribution systems. Considering today’s lifestyle and the
variable electricity tariffs dependent on the time of day,
Key words: electric vehicles; voltage drop; distribu- the most likely pattern of EV use will be driving during
tion system; dispersed voltage control. the day and charging during the night. If many EV owners
decide to use their vehicles according to this pattern, the
1. Introduction profile of distribution system use may change considerably
[8, 9]. Taking account of the effect of these changes on the
Against the background of such social issues as re- voltage profile of distribution systems, it appears important
duction of environmental impact of fossil fuels and grow- to analyze existing methods of voltage control.

C⃝ 2014 Wiley Periodicals, Inc.

9
 The EV by itself may be regarded as a sort of
rechargeable battery. Naturally, each EV will be charged
in accordance with decisions made by the owner, but
this opens the possibility of designing control algorithms
such that EV charging (including discharging from the EV
to the distribution system) will contribute to more effi- Fig. 1. Power system model.
cient operation of the distribution system. For example,
Ref. [10] proposes a charging algorithm intended to im-
prove the load curve of the entire distribution system. Refs.
[11] and [12] investigate the shifting of EV charging times
for the purpose of minimizing distribution system losses.
When a self-commutated inverter is used for EV
charging, it is possible to functionally control not only
active power but also reactive power. Since the use of
active power in the system can strongly influence the
convenience of EV owners, it must be considered in
system design, but reactive power can be used for voltage Fig. 2. Distribution system model with branches.
control within limits that do not seriously affect the
convenience of EV owners [13, 14]. Ref. [15] proposes where Vs is the voltage at the sending terminal; Vr is the
an algorithm for compensating the voltage drop caused voltage at the receiving terminal; P is the active power
by simultaneous nighttime charging of EVs by injecting passing through the receiving terminal; Q is the lagging
reactive power from EVs into the distribution system. In reactive power passing through the receiving terminal; R
addition compensating the voltage drop by using EVs, it is the resistance of the transmission line; and X is the
is technically possible to limit voltage surges caused by reactance of the transmission line.
reverse power flow and to improve power quality. Therefore, the system voltage can be regulated if it is
In this paper, we propose a basic algorithm for volt- possible to control either P or Q.
age regulation (“PQ control”), which consists of adjusting
the EV charging schedule and injecting reactive power from
EVs. Based on the practical application of the PQ control, 2.2 Voltage control in the distribution system
we also propose “prediction control”, which is a voltage model
control algorithm based on prediction of the receiving volt-
age, and “communication control”, an algorithm consisting As can be seen in Fig. 2, an actual distribution system
of exchange of information on voltage violations between consists of a primary feeder and secondary feeders (for
EVs and the system operator. The purpose of this investi- simplicity, in Fig. 2 the resistances of the feeders and the
gation was to confirm the effectiveness of these algorithms. impedances of the pole transformers are not shown). Here,
In contrast to Ref. [15], in which the output of we consider regulation of the voltage at node #As by means
reactive power was examined for only one time period, of the reactive power of nodes #As and #Bs shown in the
in this study we investigate the adjustment of charging figure. Nodes #As and #Bs are low-voltage nodes on the
schedules covering several time periods. The method of load side of the pole transformers. In accordance with Eq.
reactive power control proposed in Ref. [16] uses the (1), the voltage drop VS − V As from #S to #As can be
inverter available capacity, and therefore the inverters must calculated as
be treated as distributed power sources; but in the method
proposed in this paper, EV charging is treated as a load. VS − V As = (VS − V A ) + (V A − V As )
XpQA Xs Q A
2. Voltage Control in the Distribution System using ≅ + , (2)
VA V As
Active and Reactive Power
where X p and X s are the reactances of the primary and
2.1 Voltage control in a two-node system secondary systems; Q A and Q As are the amounts of lagging
reactive power flowing into #A and #As; and V A and V As
The voltage drop Vs − Vr from the sending terminal are the voltages at nodes #A and #As.
to the receiving terminal of the power system, as shown in When the reactive power at #As is changed by an
Fig. 1, can be approximately represented by amount ΔQ, this change ΔQ is supplied along the path
#S ∼ #A ∼ #As, thus affecting the terms in Q A and Q As
PR + QX in Eq. (2). Therefore, the voltage change ΔV As As at #As
Vs − Vr ≅ , (1)
Vr due to the reactive power adjustment ΔQ at #As can be

10
calculated as follows if changes in the reactive power loss
are disregarded:
X p ΔQ X s ΔQ
ΔV As As ≅ + . (3)
VA V As
When the reactive power at #Bs is changed by ΔQ, this
change ΔQ is supplied along the path #S ∼ #B ∼ #Bs, thus
affecting only the term in Q A but not the term in Q As in Eq.
(2). Therefore, the voltage change ΔV As Bs at #Bs due to
the reactive power adjustment ΔQ at #As can be calculated
as follows:
X p ΔQ
ΔV As Bs ≅ . (4)
VA Fig. 4. PQ control diagram.
Comparison of Eqs. (3) and (4) shows that the value of
ΔV As As is clearly greater than the value of ΔV As Bs . EV based on use of the idle capacity of the EV inverter
Since the impedance of the secondary feeder is greater connected to the system. Also, since discharge of active
than that of the primary feeder, especially in the pu system power into the power system strongly affects customers’
(the square of the voltage transformation ratio), ΔV As As convenience, in this paper we do not consider this function.
is much greater than ΔV As Bs . This means that voltage
regulation by reactive power near the node at which volt-
age regulation is performed is more effective than voltage 3.2 Proposed method 1: PQ control
regulation at a remote point.
Figure 4 shows the relationship between the active
power (P) and the reactive power (Q) in the EV battery.
3. Proposed Algorithm of Voltage Regulation Using The EV can independently control both active power and
EV reactive power inside a circle whose radius is equal to
the inverter capacity. When EV is connected to the power
3.1 EV model system via the inverter, the inverter charges the battery to
the maximum capacity (marked in the diagram as “normal
We assume that in addition to the battery to store the charge”) in the shortest possible time. It is expected that
energy necessary to drive the vehicle, the EV analyzed in with a growing number of EVs in operation and with an
this paper is also equipped with a self-commutated inverter increase in demand for EV charging, the drops in the system
for AC/DC conversion. Although currently used EVs do voltage will be greater. However, when there is time before
not have this feature, in principle it is possible to freely the next trip, charging can be kept within the “reduce charg-
exchange active and reactive power between an EV and the ing power” zone in the diagram, which makes it possible not
power system by appropriately controlling the inverter and only to reduce the voltage drop but also to control reactive
the battery. With this in mind, in this paper we consider power by using the idle capacity of the inverter (marked by
the use of an EV as a load that can be used for voltage “reactive power control” in the diagram). In this way the
regulation in the distribution system by not only using the charging power (active power) can be reduced to reasonable
active power for battery charging, but also controlling the limits and the idle capacity of the inverter can be used for
flow of reactive power in both directions as shown in Fig. 3. reactive power control. We call this method “PQ control”.
This process should be implemented in such a way as not The instructions for PQ control are as follows: First,
to block charging for subsequent trips. Here, we examine normal charging is performed in accordance with Eqs. (5)
the regulation of the distribution system voltage by the and (6). These equations state that until the battery is fully
charged, only active power is used for charging, at the full
capacity of the inverter.
{
E (t = st)
L e (t) = ∑ c (5)
E c − t−ΔT
k=st P(k)ΔT (t > st)
{
S (L e (t) > 0)
P(t) = (6)
0 (L e (t) = 0).
Here, L e (t) is the amount of energy required for EV charg-
Fig. 3. EV model. ing at time t (kWh); E c is the amount of energy necessary

11
for EV charging (kWh); st is the charging start time; P(t)
is the EV charging power (kW) from time t to time t + 1
(below, time period t), which is assumed to be constant
during this time period; ΔT is the duration of the time
period (h); and S is the capacity of the EV inverter (kVA).
We next determine whether there are any violations of
the lower voltage limit due to the charging load. In case of a
violation, the command value of P is determined according
to Eq. (7). This equation is based on the consideration that
charging can be completed at any time before the next trip
and the charging power is restricted to the value obtained
by dividing L e (t) by the time remaining to the next trip.
Fig. 5. Proposed prediction control diagram.
L e (t)
P(t) = (L e (t)) > 0, V (t) < Vmin ), (7)
et − t
where et is the time of the beginning of the next trip, V (t) is
the voltage at time t (V), and Vmin is the minimum voltage
level (V).
The Q command is determined by Eq. (8). This equa-
tion is based on the idea that the unused inverter capacity
created by P control can be used for Q control:
√
Q(t) = − S 2 − P(t)2 (L e (t) > 0, V (t) < Vmin ), (8)
where Q(t) is the reactive power consumed by EV at time
t (kvar).
In order to control the generation of reactive power
for the purpose of compensating the voltage drop, in this
paper the leading reactive power has a negative sign.

Fig. 6. Proposed algorithm for EV.

3.3 Proposed method 2: Prediction control

In the abovementioned PQ control, the PQ control the entire time period in which the charging is possible.
commands are chosen depending on the presence or ab- The magnitude of the adjustment can be found by solving
sence of voltage violations in the node under control at any the optimization problem for the purpose of averaging the
time. Therefore, if attention is focused on a certain time voltage over the entire time at which charging is possible.
period, it is impossible to respond to voltage violations that (Objective function)
may occur after it. There is a possibility that no optimal et−ΔT
∑
instruction for the entire time can be found. f = (Vtarget − Vpre (t))2 (9)
Consequently, in this study we assumed that under t=st
certain conditions based on past actual data, it is possible to
et−ΔT
∑
accurately predict the voltage at the location of the EV plug- V P Q (t)
in at any time, during normal charging or during charging Vtarget = (10)
t=st
(et − st)∕ΔT
under PQ control, and to use this information for voltage
control by the EV (below, this algorithm is referred to as δV δV
“prediction control”). V pr e (t) ≅ V P Q (t) + (t)ΔP(t) + (t)ΔQ(t)
δP δQ
Figure 5 shows a diagram of the proposed prediction
(11)
control. The time is plotted on the horizontal axis and the
voltage at a certain node is plotted on the vertical axis. (Constraint conditions)
The circles in this diagram correspond to predicted voltage
values in normal charging and under PQ control. Since, in {P(t) + ΔP(t)}2 + {Q(t) + ΔQ(t)}2 ≤ S 2 (12)
the example shown in Fig. 5, a violation of the minimum et−ΔT
voltage level at time point t2 is predicted, the purpose of ∑
{P(t)} + ΔP(t)ΔT ≥ E c , (13)
control is to avoid voltage violation at point t2 by adjusting t=st
only the active and reactive power ΔP(t) and ΔQ(t) over

12
where Vtarget is the target voltage for prediction control (V);
Vpre (t) is the predicted voltage after output adjustment (V);
V P Q (t) is the predicted voltage in normal charging or under
PQ control (V); ∂V ∕∂P ⋅ (t) is the voltage sensitivity at the
node under control as a function of the active power at the
node under control at time t (V/kW); ∂V ∕∂Q ⋅ (t) is the
voltage sensitivity at the node under control as a function
of the reactive power at the node under control at time t
(V/kvar); ΔP(t) is the adjusted part of the EV active power
(kW); and ΔQ(t) is the adjusted part of the EV reactive
power (kvar).
In Eq. (11) the voltage at the node where an EV is
Fig. 8. Proposed communication control diagram.
plugged in and the sensitivities of the active and reactive
power (∂V ∕∂P ⋅ (t) and ∂V ∕∂Q ⋅ (t)) can be estimated in
advance, for example, by the method proposed in Ref. [17].
is implemented, a voltage violation signal is sent to the
DSO. When the DSO receives information about voltage
3.4 Proposed method 3: Communication
violations, an instruction is sent for PQ control support to
control
the EVs plugged into the same secondary feeder as the EV
from which a signal similar to the “SF control” shown in
In the PQ control and prediction control schemes,
Fig. 7 was sent (below, this control is referred to as “SF
only the EV at the node where voltage violations exist is
control”).
operated in an autonomous and decentralized manner. In
If SF control fails to resolve the issue of voltage vio-
this section, we consider a case in which it is possible to
lations, the EV where these voltage violations occur sends
establish communication between EVs and the distribution
a voltage violation signal to the DSO again. At that time,
system operator (DSO) for the purpose of achieving more
the DSO sends a signal for PQ control support to all EVs
efficient voltage regulation, and we propose an algorithm
plugged into the same secondary feeder (as indicated in
for coordinated voltage regulation using this ability to com-
Fig. 7 by “PF control”) and all EVs that receive this signal
municate. According to the proposed method, the EV is
switch to PQ control (below this control is referred to as
controlled autonomously, but if the issue of voltage viola-
“PF control”). The reason for applying control in two steps
tions is not resolved, the system offers support for the EV.
is that by first operating EVs having higher efficiency, as
A detailed algorithm of the proposed method is
described in Section 2.2, the number of EVs to be charged
shown by the flow charts in Figs. 6 and 7. Each EV is first
can be decreased. The proposed communication control is
charged according to the normal charging pattern (or the
shown graphically in Fig. 8. The dotted lines in the figure
standby pattern in the case of a fully charged EV). Next,
represent the signal flow.
when voltage violations at the nodes are detected or, in the
case of a “PQ control support” signal from the DSO, PQ
control is implemented. If voltage violations at the node
under control cannot be eliminated even when PQ control 4. Simulation

4.1 Simulation conditions

In order to confirm the effectiveness of the proposed

method, we performed a simulation using the distribution
system model shown in Fig. 9. As shown in Fig. 9(a),
the model system had six high voltage nodes (#S and #A
through #E). Node #S represents the distribution substation
and, considering that the voltage sent from this substation
is high during the high load period and low during the low
load period, it is assumed to be 6.60 kV from 8:00 to 22:00
and 6.30 kV from 22:00 to 8:00. As shown in Fig. 9(b), it
is assumed that 30 pole transformers are connected to each
high voltage node (#A through #E) and that the length of
the secondary feeders between nodes #A and #D is 100 m.
Fig. 7. Proposed algorithm for distribution system Node #E, at which the voltage drop becomes relatively
operator. large, is located at a distance of 300 m. The loads connected

13
 Fig. 9. Distribution system model used in simulation.

to the pole transformers are arranged as shown in Fig. 9(c)

and include 12 residential loads. The nodes that are in the
same position relative to the impedance are called nodes #a
through #f, and a number 1 or 2 is assigned to denote the
load number. Thus, this system model, from the distribution
substation to the farthest secondary node #f of the primary
node #E (referred to below as “E f”), has a structure in
which voltage violations can easily occur. The impedance
values provided in the same figure were selected on the
basis of Refs. [17–19].
The load curve of one residential load according to
this distribution system model (not including the active Fig. 10. Load curve of one residential load without EV.
and reactive power of the EV) is shown in Fig. 10. In [Color figure can be viewed in the online issue, which is
this diagram, the consumption of lagging reactive power available at wileyonlinelibrary.com.]
is plotted in the positive direction. Figure 11 shows the
voltage fluctuations with time at each node with the EV
not connected (that is, the power consumption only by the
residential load shown in Fig. 10). Figure 11 shows the as the first load plugged in at the low voltage node #f of
voltage at primary node #A a, which is the closest to branch Tr1. The 361st EV, shown in parentheses on the
the power source and has the highest voltage, and at sec- same line, is connected as the second load at the same
ondary node #E f, which is farthest from the power source location where the load of the first EV is plugged in. Since
and has a lower voltage. The voltages at the other nodes are we assume that cases of EV introduction will be centered
between these values. Below we discuss mainly the voltage around new residential areas and testing grounds, in this
observed at node #E f, having the largest voltage drop. paper we consider primarily cases of EV connections at
The following sections describe simulations carried nodes #D and #E. The total number of such residences
out with the number of EVs as a parameter. For this pur- as well as the maximum number of EVs introduced was
pose, we assume that the number of EVs plugged in for assumed to be 720. The other conditions of the simulation
charging is as given in Table 1. The first EV is connected are listed in Table 2.

14
 Table 2. Data for simulations

Parameters Value
Required energy to be charged 18 kWh
Time to start charging 23
Time of departure 7
Capacity of inverter for EV 3 kVA
Time step 0.25 h
Maximum voltage level 107 V
Minimum voltage level 95 V

Fig. 11. Voltage at nodes in absence of EV. [Color figure

can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

Table 1. Order of EV introduction

EVNo. Primary Tr No. Secondary Load No.

1(361) #E Trl #f 1(2)
2(362) #D Trl #f 1(2)
3(363) #E Tr2 #f 1(2)
4(364) #D Tr2 #f 1(2)
... ... ... ... ...
59(419) #E Tr30 #f 1(2) Fig. 12. Voltage at #E f in normal charging. [Color
60(420) #D Tr30 #f 1(2) figure can be viewed in the online issue, which is available
61(421) #E Trl #d 1(2) at wileyonlinelibrary.com.]
62(422) #D Trl #d 1(2)
... ... ... ... ...
119(479) #E TOO #d 1(2)
120(480) #D TOO #d 1(2)
121(481) #E Trl #b 1(2)
... ... ... ... ...
181(541) #E Trl #e 1(2)
... ... ... ... ...
241(601) #E Trl #c 1(2)
... ... ... ... ...
301(661) #E Trl #a 1(2)
... ... ... ... ...
360(720) #D TOO #a 1(2)
Fig. 13. Voltage at #E f with 181 EVs. [Color figure can
be viewed in the online issue, which is available at
wileyonlinelibrary.com.]
4.2 Effectiveness and limitations of PQ control

(1) Effectiveness of PQ control. Fig. 12 shows the and during charging under PQ control. According to this
voltage fluctuations at node #E f during normal charg- diagram, under PQ control, it is possible to avoid voltage
ing. The notations “NoEV”, “60EVs”, “120EVs”, and violations during the period between 23:00 and 24:00. The
“181EVs” in this figure mean “before EV connection”, “60 voltage drop under PQ control between 5:00 and 5:15 is
EVs”, “120 EVs”, and “181 EVs”, respectively. According explained by charging during this time to compensate for
to this diagram, the system voltage becomes lower when charging reduction between 23:00 and 24:00.
vehicles are connected to the system for charging. When Based on Eq. (1), voltage control by reduction of
fewer than 180 EVs are introduced, the voltage remains charging power is effective in the case of high system re-
higher than the minimum level, but when 181 EVs are sistance; in the case of high reactance, voltage control by
connected, a violation of the minimum voltage level occurs injecting reactive power is effective. According to Fig. 13,
between 23:00 and 24:00. PQ control results in a 0.53 V voltage improvement, while
Figure 13 shows the voltage fluctuations when 181 suppression of charging alone results in a voltage improve-
EVs are introduced at node #E f during normal charging ment of 0.30 V; hence, the application of PQ control results

15
 Fig. 15. Voltage at #E f with 241 EVs. [Color figure can
be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

Table 3. PQ control status at each node at 23:00 with 241

EVs

#a #b #c #d #e #f
#D – Normal – Normal Normal Normal
#E – Normal Normal PQ PQ PQ

Fig. 14. Magnitudes of P and Q at #E f with 181 EVs.

[Color figure can be viewed in the online issue, which is mode, “PQ” means PQ control, and “–” means no EVs are
available at wileyonlinelibrary.com.] connected. According to this table, PQ control is imple-
mented at nodes #E d through #E f, where voltage viola-
in a voltage improvement of 57%. Thus, in this simulation tions occur, but since no voltage violations occur at other
system, reduction of charging power is more effective than nodes, PQ control is not implemented. Implementation of
injection of reactive power. We believe that this can be control depending on location can be achieved by using
explained by the fact that the resistive component of the communication control. The results of communication con-
secondary feeders and service wires shown in Figs. 9(b) and trol implementation are discussed in Section 4.4.
(c) is greater than the reactive component.
Figure 14 shows the active power (P) and reactive
power (here Q is positive when lagging occurs) of an EV at 4.3 Effectiveness and limitations of prediction
different times of day. In the case of only normal charging, control
the charging is performed at the maximum inverter capacity
of only 3 kW, while in the case of PQ control, the charging (1) Effectiveness of prediction control. Figure 16
is reduced between 23:00 and 24:00 and reactive power shows the voltage fluctuations at node #E f under PQ con-
control is implemented using the excess inverter capacity. trol and prediction control with 241 EVs plugged in for
As can be seen in the same figure, the charging reduction in charging. Prediction control consists of the application of
the period from 23:00 to 24:00 is compensated by charging
for 15 min starting at 5:00.
(2) Limitations of PQ control (time bias). Figure 15
shows the voltage at #E f with 241 EVs connected for
charging. According to this diagram, voltage violations oc-
cur in the period from 23:00 to 24:00, but at other times
there is some margin before a voltage violation can occur.
Prediction control is a method of improving control effec-
tiveness by averaging the voltages at such times. The effect
of prediction control is explained in Section 4.3.
(3) Limitations of PQ control (location bias). Table 3
gives the conditions of PQ control at each node at 23:00
with 241 EVs plugged in for charging. In this table the Fig. 16. Voltage at #E f with 241 EVs. [Color figure can
lines correspond to primary nodes, the columns correspond be viewed in the online issue, which is available at
to secondary nodes, “normal” means the normal charging wileyonlinelibrary.com.]

16
 Fig. 17. Magnitudes of P and Q at #E f with 241 EVs. Fig. 19. Magnitudes of P and Q at #E f with 393 EVs.
[Color figure can be viewed in the online issue, which is [Color figure can be viewed in the online issue, which is
available at wileyonlinelibrary.com.] available at wileyonlinelibrary.com.]

by the same amount. In addition, charging is concentrated

in the period from 6:00 to 7:00, when no charging takes
place under PQ control. Thus, the voltage is averaged over
the entire charging period and voltage violations during the
period from 23:00 to 24:00 are avoided.
(2) Limitations of prediction control. When the
number of EVs connected for charging is increased to
392, no violations of the minimum voltage level occur, but
when 393 vehicles are introduced, as shown in Fig. 18,
the voltage drops below the minimum voltage level even
Fig. 18. Voltage at #E f with 393 EVs. [Color figure can
when prediction control is implemented. According to the
be viewed in the online issue, which is available at
same diagram, voltage violations appearing between 23:00
wileyonlinelibrary.com.]
and 24:00 under PQ control can be prevented by predic-
tion control. However, although no voltage violations were
PQ control in cases when voltage violations are expected observed under PQ control between 5:00 and 5:15, under
to remain even after implementation of PQ control (as indi- prediction control the voltage dropped below the minimum
cated by “PQ control” in the diagram) by adjusting the PQ voltage level.
command value in advance so that no violations occur. As The magnitudes of P and Q control in the process
shown by “prediction control” in Fig. 16, voltage violations of voltage adjustment are shown in Fig. 19. The plot of
appearing between 23:00 and 24:00 under PQ control alone active power under prediction control shown in Fig. 19(b)
can be avoided by the application of prediction control. indicates that the PQ reference values can be chosen so that
A diagram of the magnitude of PQ control at the charging is reduced in the period between 23:00 and 24:00,
time of control adjustment is shown in Fig. 17. Analysis where voltage violations exist under PQ control, and more
of voltage violations taking place between 23:00 and 24:00 power is allocated for charging in the period between 5:00
during the implementation of PQ control shows that during and 5:15, where excess power capacity occurs. However,
prediction control the charging power is reduced more than Table 1 indicates that when 361 or more EVs are plugged
under PQ control and the reactive power output is increased in, the EVs are plugged in not only as the first load but

17
 Table 4. Status of PQ control at #E at 23:00 with 241
EVs

#a #b #c #d #e #f
(a) Only PQ control
#D – Normal – Normal Normal Normal
#E - Normal Normal PQ PQ PQ
(b) With SF control
#D – Normal – Normal Normal Normal
#E – SF SF PQ PQ PQ
Fig. 20. Voltage at #E with 241 EVs. [Color figure can
be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

also as the second load at #E f, as shown in Fig. 9(c). Since

the system impedance from the distribution substation side
is the same for these two loads and their load curves are
the same as the curves without EVs shown in Fig. 10,
the voltages at these points are also the same. Therefore,
prediction control is applied to both the first and the second
EVs at node #E f. Since the two loads have common power Fig. 21. Voltage at #D and #E with 361 EVs. [Color
supply paths from the distribution substation other than the figure can be viewed in the online issue, which is available
service lines, voltage control at one load strongly affects the at wileyonlinelibrary.com.]
voltage at the other load. Nevertheless, since no information
about the application of prediction control to any EV is
communicated to other EVs, the magnitude of PQ control Table 5. Status of PQ control at #E at 23:00 with 361
for a specific EV is determined on the basis only of informa- EVs
tion about this specific EV, as shown in Fig. 19. As a result,
from the viewpoint of the entire system, the consumption of (a) Only SF control
active power during the period from 5:00 to 5:15 becomes #a #b #c #d #e #f
too high, causing violations of the low voltage limit. Due
#D Normal Normal Normal Normal Normal Normal
to the autonomous and decentralized nature of prediction
#E SF SF PQ PQ PQ PQ
control, it is impossible to determine the reference value
for PQ control based only on information about a specific (b) With PF control
#a #b #c #d #e #f
node. A shortcoming of prediction control is its inability
#D PF PF PF PF PF PF
to coordinate charging at a specific node with charging at
#E SF SF PQ PQ PQ PQ
other nodes, and therefore it is not an optimal charging
pattern for the entire system.

Figure 21 shows the voltages at the nodes of the sec-

4.4 Effectiveness of communication control ondary feeders when 361 EVs are connected for charging.
When only SF control is used, voltage violations occur at
Figure 20 shows the voltage at secondary nodes con- node #E f, but under PF control no voltage violations occur.
nected to #E with 241 EVs under PQ control and SF con- Table 5 presents data on the control status at that time. In
trol. Violation of the minimum voltage level at #E f when the table, “PF” denotes PF control. When PQ control was
only PQ control is implemented can be avoided by using applied to all EVs connected to #D, which was not operated
SF control. Table 4 shows the status of the EVs at each under SF control, as noted above, voltage violations at #E f
secondary node, indicating normal charging (“normal”), were avoided due to the implementation of PF control.
charging under PQ control (“PQ”), and charging under SF The voltage behavior at different nodes with an in-
control (“SF”). Since PQ control was applied to the EVs at crease in the number of EVs connected for charging (up
nodes #E b and #E c which, as noted above, did not have to the maximum of 720 vehicles) is shown in Fig. 22. In
voltage violations, violation of the minimum voltage level this case too, the voltage does not drop below the minimum
at #E f could be avoided. level. The status of control of EVs is shown in Table 6.

18
 night hours, and it was demonstrated that compensation by
using the excessive capacity of EV inverters can increase
the voltage and improve power quality in several millisec-
onds. In the future we plan to investigate algorithms of EV
use from this standpoint.

Acknowledgment

This paper was recommended by the chairman of the

Fig. 22. Voltage at #D and #E with 720 EVs. [Color 2011 Joint Committee on Power Technologies and Power
figure can be viewed in the online issue, which is available Systems Technologies.
at wileyonlinelibrary.com.]

REFERENCES
Table 6. Status of PQ control at #E at 23:00 with 720
EVs 1. Electric Technology Research Association: Power fac-
tor problems in power distribution systems and meth-
#a #b #c #d #e #f
ods of their solution. Elect Technol Res 2010;66(1). (in
#D PF PF PF PF PF PQ Japanese)
#E SF SF PQ PQ PQ PQ 2. Hayashi Y. Trends and prospects of voltage control for
distribution systems with distributed generators. IEEJ
Trans PE 2009-4;129(4):491–494. (in Japanese)
5. Conclusions 3. Vovos PN, Kiprakis AE, Wallace AR, Harrison GP.
Centralized and distributed voltage control: Impact on
We have proposed three methods of compensating distributed generation penetration. IEEE Trans Power
the voltage drop caused by charging of EVs by using the Syst 2007;22(1):476–483.
active and reactive power control of vehicles being charged 4. Viawan FA, Karlsson D. Voltage and reactive power
and have confirmed effectiveness of the proposed methods. control in systems with synchronous machine-based
First, we confirmed that by using PQ control it is possible distributed generation. IEEE Trans Power Deliv
to adjust the charging schedule and inject reactive power 2008;23(2):1079–1087.
in an autonomous and decentralized manner. However, in 5. Senjyu T, Miyazato Y, Yona A, Urasaki N, Funahashi
some respects, because of time and location bias, PQ con- T. Optimal distribution voltage control and coordina-
trol is not effective. Consequently, we have proposed pre- tion with distributed generation. IEEE Trans Power
diction control, which eliminates the time bias, and have Deliv 2008;23(2):1236–1242.
demonstrated its effectiveness. However, because of the au- 6. Carvalho PMS, Correia PF, Ferreira LAF. Distributed
tonomous and decentralized nature of the system, the effec- reactive power generation control for voltage rise mit-
tiveness of this algorithm declines when the number of EVs igation in distribution networks. IEEE Trans Power
to be charged is increased. Therefore, we examined the use Syst 2008;23(2):766–772.
of communication control, using a coordinated algorithm 7. Ministry of Economy, Trade and industry (METI):
based on autonomous and decentralized PQ control, which “Next-generation vehicle strategy 2010”: (Online):
makes it possible to eliminate the location bias by sharing http://www.meti.go.jp/English/press/data/20100412
voltage violation information only when necessary. As a 02.html
result, we were able to verify that voltage violations were 8. Qian K, Zhou C, Allan M, Yuan Y. Modeling of
eliminated even with the maximum number of EVs plugged load demand due to ev battery charging in distribu-
in for charging. Since communication control requires a tion systems. IEEE Trans Power Syst 2011;26(2):802–
developed communications infrastructure, it is desirable to 810.
evaluate its cost-effectiveness, but if such a communica- 9. Wu D, Aliprantis DC, Gkriza K. Electric energy
tions infrastructure has already been created for different and power consumption by light-duty plug-in elec-
purposes, control can be successfully implemented by shar- tric vehicles. IEEE Trans Power Syst 2010;26(2):738–
ing information about voltage violations or drops in voltage 746.
below the minimum voltage level. 10. Takagi M, Iwafune Y, Yamamoto H, Yamaji K, Okano
This study was limited to examination of drops in K, Hiwatan R, Ikeya T. Algorithm for bottom charge
voltage caused by simultaneous charging of EVs during the based on load-duration curve of plug-in hybrid electric

19
 vehicles. IEEJ Trans PE 2010-8;130(8):727–736. (in 15. Noda T, Kabasawa Y, Fukushima K, Nemoto K, Ue-
Japanese) mura S. A method for compensating customer volt-
11. Fernandez LP, Roman TGS, Gossent F, Domingo CM, age drops due to nighttime simultaneous charging of
Friss P. Assessment of the impact of plug-in electric evs utilizing reactive power injection from battery
vehicles on distribution network. IEEE Trans Power chargers. IEEJ Trans PE 2012-2;132(2):163–170. (in
Syst 2011;26(1):206–213. Japanese)
12. Sortomme E, Hindi MM, MacPherson SDJ, Venkata 16. Tsuji T, Hashiguchi T, Goda T, Shinji T, Tsujita S.
SS. Coordinated charging of plug-in hybrid electric A study of autonomous decentralized voltage pro-
vehicles to minimize distribution system losses. IEEE file control method considering control priority in
Trans Smart Grid 2011;2(1):198–205. future distribution network. IEEJ Trans PE 2009-
13. Mitsukuri Y, Hara R, Kita H, Kamiya E, Hiraiwa N, 12;129(12):1533–1544. (in Japanese)
Kogure E. Advanced operation of distribution system 17. Tanaka S, Suzuki H. Voltage compensation using au-
by using electric vehicles. basic examination of voltage tonomous decentralized control of distributed gen-
drop compensation, The 2011 National Convention of erators. IEEJ Trans PE 2009-7;129(7):869–879. (in
Institute of Electrical Engineers of Japan, 6-057, Os- Japanese)
aka University (2011). (in Japanese) 18. Electric Technology Research Association: ‘State of
14. Mitsukori Y, Hara R, Kita H, Kamiya E, Hiraiwa Power Quality and Technology for Its Control’. Elect
N, Kogure E. Basic study on advanced operation of Technol Res 2005;60(2). (in Japanese)
distribution systems using electric vehicles. Proc. of 19. Home page of Showa Cable System Co., Ltd.
International Conference on Electrical Engineering, http://www.swcc.co.jp/cs/index.html, access date:
(ICEE2011), No. A215 (2011). April 2012. (in Japanese)

AUTHORS (from left to right)

Yuki Mitsukuri (student member) completed the M.S. program in system information science at the Graduate School of
Information Science and Technology of Hokkaido University in 2003. In the same year, he joined Tokyo Electric Power Co.,
Ltd. In 2010 he entered the doctoral program in system information science at the Graduate School of Information Science and
Technology of Hokkaido University. In 2012, he retired from Tokyo Electric Power Co., Ltd. His research work is focused on
power distribution issues.

Ryoichi Hara (member) completed the doctoral program in system information technology at the Graduate School of
Engineering of Hokkaido University in 2003. In the same year, he joined the faculty of the Graduate School of Engineering of
Yokohama National University as a research associate in electrical and electronics engineering and mathematical information
science. In 2006, he became an associate professor in the Department of System Information Science of the Graduate School
of Information Science and Technology, Hokkaido University. He received D.Eng. degree. He is involved in research mainly
on the design, operation, and control of power systems. Member of IEEE and IEIEJ.

20
 AUTHORS (continued) (from left to right)

Hiroyuki Kita (member) completed the M.E. program in electrical engineering at the Graduate School of Information
Science and Technology of Hokkaido University in 1988. In 1989, he became a research associate in electrical engineering at
the Graduate School of Engineering of the same university. In 1995, he was appointed an associate professor of system and
information science at the Graduate School of Information Science and Technology of the same university, and he became a
professor in 2005. He is involved in research principally on the design, operation, and control of power systems. Member of
IEEE, ORSJ, IEIEJ, and the Illuminating Engineering Institute of Japan.

Eiji Kamiya (member) received his M.E. degree in integrated design engineering from the Graduate School of Science
and Technology of Keio University in 2006. In the same year, he joined Tokyo Electric Power Co., Ltd. He is now affiliated
with the Distribution Technology Group at the Power Distribution Headquarters. His research work is focused mainly on power
distribution issues.

Shoji Taki (member) received his M.E. degree in electrical and electronic engineering from the Graduate School of
Engineering of Chiba University in 1997. In the same year, he joined Tokyo Electric Power Co., Ltd. He is now affiliated
with the Distribution Technology Group at the Power Distribution Headquarters. His research work is focused mainly on power
distribution issues. Member of IEIEJ.

Naoya Hiraiwa (member) received a bachelor’s degree in electrical and computer engineering from Nagoya Institute of
Technology in 1994. In the same year, he joined Tokyo Electric Power Co., Ltd. He is now the manager of the power distribution
group of the Tochigi facility. He is involved in research on new electrical materials and the operation of distribution systems.

Eiji Kogure (member) received a bachelor’s degree in electrical engineering from Saitama University in 1987. In the same
year, he joined Tokyo Electric Power Co., Ltd. He is now manager of the Distribution Technology Group at the Technology
Development Research Center. His research work is focused mainly on the development of power distribution equipment and
systems. Member of IEIEJ.

21