1236 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO.

2, APRIL 2008

Optimal Distribution Voltage Control and

Coordination With Distributed Generation
Tomonobu Senjyu, Senior Member, IEEE, Yoshitaka Miyazato, Student Member, IEEE,
Atsushi Yona, Student Member, IEEE, Naomitsu Urasaki, Member, IEEE, and
Toshihisa Funabashi, Senior Member, IEEE

Abstract—In recent years, distributed generation, as clean nat- from an upstream to a downstream system in the conventional
ural energy generation and cogeneration system of high thermal electric power system. Thus, voltage control of an electric power
efficiency, has increased due to the problems of global warming system postulates the condition of the flow of electric power as
and exhaustion of fossil fuels. Many of the distributed generations
are set up in the vicinity of the customer, with the advantage that uniflow. When distributed generation increases, by the conven-
this decreases transmission losses. However, output power gener- tional control technique, it is expected that voltage variation of
ated from natural energy, such as wind power, photovoltaics, etc., each node becomes a problem. In order for voltage variation to
which is distributed generation, is influenced by meteorological not damage the equipment and other devices of the user, con-
conditions. Therefore, when the distributed generation increases sumer voltage should be maintained within the range defined by
by conventional control techniques, it is expected that the voltage
change of each node becomes a problem. Proposed in this paper is the electric law in Japan. Shunt capacitor (SC) determines the
the optimal control of distribution voltage with coordination of dis- on–off control using the moving average voltage value, from
tributed installations, such as the load ratio control transformer, self-end information [1]; thus, frequent operation is avoided and
step voltage regulator (SVR), shunt capacitor, shunt reactor, and voltage variation control is thus attained. However, in order to
static var compensator. In this research, SVR is assumed to be a opt for on–off control only for self-end information, the question
model with tap changing where the signal is received from a cen-
tral control unit. Moreover, the communication infrastructure in of optimal control of voltage for the whole system needs to be ad-
the supply of a distribution system is assumed to be widespread. dressed. With centralized control, voltage control that overlooks
The genetic algorithm is used to determine the operation of this the entire power system is possible. Methods of online control SC
control. In order to confirm the validity of the proposed method, of a power distribution system have been reported [2]–[4]. Other
simulations are carried out for a distribution network model with suggested methods include coordinated control, such as SC and
distributed generation (photovoltaic generation).
load ratio control transformer (LRT) [5], [8], [9], and a scheme
Index Terms—Centralized control, coordinated control, distri- that sends voltage of a substation (tap of LRT) and step voltage
bution system, genetic algorithm (GA), voltage/reactive power regulator (SVR) [6]. Others include cooperation control using a
control.
controllable load in addition to LRT or SVR [7], and a combined
Heuristic-algorithmic approach for reactive power optimization
I. INTRODUCTION [10]. In the power distribution system with photovoltaics equip-
ment, the FACTS devices’ inverter minimum capacity aimed at
HE twin problems of global warming and exhaustion of
T fossil fuels have seen an increase in the usage of distributed
generation, such as clean natural energy generation, cogen-
voltage variation within a restricted condition can be determined
[11]. However, although all controllers (LRT, SVR, SC, ShR,
SVC) exist in the real power distribution system, the voltage
eration system of high thermal efficiency, and many others in optimal control of an actual power distribution system is not fully
recent years. Many of the distributed generations are set up in considered. Moreover, the cooperation control in consideration
the vicinity of the customer, and there is an advantage that this of all the controllers is not reported.
decreases transmission loss, has a short period of construction, This paper proposes the optimal control of distribution voltage
and low investment risk. However, the natural energy power gen- with coordination of distribution installations, such as LRT,
eration of wind power, photovoltaics, etc., which is distributed SVR, SC, ShR, and SVC. In the research, SVR is assumed to be
generation, is influenced by meteorological conditions. There- a model with tap changing where the signal is received from a
fore, the distributed generation connected to the downstream central control unit. Further, the communication infrastructure
power system may produce backward flow which is not assumed in the supply of distribution system is assumed to be widespread.
in conventional systems. The flow of electric power is uniflow, The objective is to try and control the voltage of each node to the
optimal value by determining the sending voltage of the substa-
Manuscript received December 27, 2006; revised June 1, 2007. Paper no. tion in a certain time section, the tap location of SVR, on–off of
TPWRD-00842-2006. SC and ShR, and capacity of SVC. The controller determines
T. Senjyu, Y. Miyazato, A. Yona, and N. Urasaki are with the Department of
Electrical and Electronics Engineering, Faculty of Engineering, University of
the amount of operation based on a GA, which is an excellent
the Ryukyus, Okinawa 903-0213, Japan (e-mail: [email protected]). technique in the high-speed approximation of a solution, and
T. Funabashi is with the Meidensha Corporation, Tokyo 141-8565, Japan a feasible area is discretely set in a large-scale optimization
(e-mail: [email protected]). problem. In order to confirm the validity of the proposed method,
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. simulations are carried out for a distribution network model
Digital Object Identifier 10.1109/TPWRD.2007.908816 with distributed (photovoltaic) generation.
0885-8977/$25.00 © 2007 IEEE

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
SENJYU et al.: OPTIMAL DISTRIBUTION VOLTAGE CONTROL AND COORDINATION 1237

and secondary voltage with current , is expressed in the

formula below

(1)

The input-and-output voltage–current expression of relations of

the equivalent 4-terminal circuit of Fig. 2(b) is given by the fol-
lowing formula:
Fig. 1. Model of SVC. (a) Configuration. (b) Equivalent model.
(2)

When each element of the coefficient procession of (1) and (2)

is set to be mutually equal, the following relations apply:

(3)

Therefore, the impedance of LRT is expressed in the equivalent

4-terminal circuit of Fig. 2(b), and the value of each element is
calculated from (3).
Fig. 2. Model of LRT. (a) Equivalent model of LRT. (b) Four-terminal network.
D. Genetic Algorithm
A genetic algorithm (GA) is a learning algorithm that imi-
II. SVC/LRT/SVR
tates the evolution of organisms [12]. In the natural evolution
A. SVC process of organisms, the reproduction of a set of individuals
that forms a certain generation (i.e., the population) is such that
SVC is equipment that generates reactive power at high speed
those individuals with fitness to environmental adaptation sur-
by controlling the current of a capacitor by the thyristor. Fig. 1
vive with high probability. Reproduction that is based on the de-
shows the fundamental composition and the equivalent circuit
gree of conformity of an individual in a GA as an artificial model
of SVC in this research. For simplicity, SVC is made to be the
that imitates the evolution process of such a organism is per-
source of reactive power as shown in Fig. 1(b).
formed, and the next-generation population is generated through
B. Problem of SVR and LRT crossover and mutation. The process is carried out by repeating
such genetic operations, and if the individual of the last gener-
The present voltage control is a system with LRT, SVR, and ation that fulfills end conditions can be found, the semioptimal
tap location from which the voltage/current measurement value solution in question may be determined. The flow of the basic
is calculated, which is self-end information [6]. The load central operation of the GA shown in Fig. 3 is explained below.
point voltage of a target value is defined beforehand, from which 1) Initialization: individuals with random chromosomes are
the voltage/current value of a self-end and the voltage of the generated that set up the initial population .
load central point is estimated. The transformer tap location is 2) Reproduction: the degree of conformity of each object is
controlled by SVR so that the voltage value is settled at the calculated and an individual is reproduced under a fixed
desired value. However, the following problems arise. rule depending on the degree of conformity. Here, some in-
• Voltage distribution of each feeder gets complicated from dividuals with a low degree of conformity will be screened,
the reverse power flow by dispersed power system and while individuals with a high degree of conformity will in-
nonuniformity of load distribution. Consequently, even if crease.
it fits in the operating standard voltage at the load central 3) Crossover: new individuals are generated by the method of
point, it may deviate from operating standard voltage at intersection that has been set up.
other points such as end portion, etc. 4) Mutation: this is performed by an operation determined by
• LRT/SVR is coordinated by the time period and sensitivity the installed mutation probability or mutation, and a new
voltage but is not exactly optimal operation for the entire individual is generated.
power system. 5) End judging: if end conditions are fulfilled, the best indi-
Therefore, this research considers that SVR is equal to LRT vidual thus obtained is the semioptimal solution in ques-
from which tap changes are by centralized control. tion. Otherwise, return to 2).
C. Model of LRT
The GA parameters used are shown in Table I.
The equivalent circuit of LRT is shown in Fig. 2(a). This is
changed into an equivalent 4-terminal circuit and is shown in III. SIMULATION
Fig. 2(b). Suppose that the number-of-turns ratio of the winding
of a transformer has now changed times from a standard A. Method of Determination for Operations of Each Device
transformation ratio. Then, the relationship between the pri- In this research, the amount of operation of each device is
mary voltage displayed by the unit method with current , determined by solving the objective function and voltage opti-

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
1238 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 2, APRIL 2008

Fig. 4. Residential distribution network.

TABLE II
LINE IMPEDANCES

Fig. 3. GA algorithm flowchart.

TABLE I
GA PARAMETERS

mization of the whole power distribution system. The controller

determines the amount of operation based on a GA, and a fea-
sible area is discretely set in a large-scale optimization problem.
The objective function is set up so that a voltage margin can be
fully secured to a steep voltage variation, and power loss is also
considered. The objective function will be shown

(4)

subject to the constraint

(5) Fig. 5. Load curve of residential area.

(6)
B. Condition in a Residential Area Power Distribution System
where
A residential area power distribution system model is shown
th node voltage target value; in Fig. 4. The distribution line capacity of this system model is
th node voltage; 2500 kVA, and the photovoltaics equipment is introduced into
sending voltage; a node N33. Each line impedance is shown in Table II. The
, min, max value of sending voltage; daily load curve of a residential area power distribution system
ratio of the tap of transformer; is shown in Fig. 5, and a photvoltaics output pattern is shown
in Fig. 6. These data form a power distribution system common
, min, max the ratio of the tap of transformer;
in Japan [11], [13], [14]. The load pattern of the residential area
, weighting values.
power distribution system of Fig. 5 has the feature by which
In order to take into consideration the maximization of the peak load comes at 20:00 unlike an actual commercial region
voltage margin and reduction of power loss with sufficient or an industrial region. In this time, photovoltaics always take a
balance, the weighting coefficients are determined as 1.0 phase factor of 1 for operation. Moreover, the installation part of
and 10.0 by trial and error. SVR is accomplished on the part between N6–N7 that can adjust

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
SENJYU et al.: OPTIMAL DISTRIBUTION VOLTAGE CONTROL AND COORDINATION 1239

Fig. 7. Voltage profile (case 1).

Fig. 6. Capacity factor of photovltaic generation.

TABLE III
CAPACITY OF DEVICES

the voltage of the whole power distribution system by trial and

error. Also, SC and ShR are installed at the place of a large line
drop (N2, N4, N6, N8, N10, N12).

C. Simulation Conditions
The division of three kinds of time zones is undertaken. Sim-
ulations are performed for the case of four controls: tap control
of a transformer, the tap of a transformer and the tap of SC and
ShR, the tap of transformer, as well as the tap of SC and ShR
and SVC control. The capacity of each device used in the sim- Fig. 8. Voltage profile (case 1).
ulations is shown in Table III. It was determined that such de-
vices’ capacity should not deviate from the value of the devices’
capacity installed in an actual system. Moreover, the tap ratios
of LRT and SVR are taken as ten taps of 0.02 units from 0.9
to 1.1.

D. Simulation Results
Voltage distribution on each condition of the time zone at
6:00 of case 1 is shown in Figs. 7–9. Voltage distribution on
each condition of the time zone at 12:00 of case 2 is shown in
Figs. 10–12. Voltage distribution on each condition of the time
zone at 20:00 of case 3 is shown in Figs. 13–15. The control
states of are shown in Table IV, while the con-
trol states of and
are shown in Tables V and VI, respectively.
Moreover, when all of the devices in each case are used, from
Table VII, the minimum value of the objective function solved Fig. 9. Voltage profile (case 1).
using the GA is shown. In the case of each condition, the flow of
the current shown in Fig. 4 is shown in Table VIII. Case 1 shown
in Figs. 7–9 is the value of the minimum load consumed, and it tion of SVC. The consumer load is a middle value and case 2
is photovoltaics at 18% output time. In order for the conditions shown in Figs. 10–12 are an instance of photovoltaics being the
of this case to not be severe compared with other cases, each maximum outputs. In case 2 and from the reverse power flow
node voltage may be considered as the target value by control- from Fig. 10 and Table VIII, the distributed generation can be
ling all taps. Besides, each node voltage hardly changes com- confirmed. Likewise, when there is no control, it can be con-
pared with all types of device control and also without installa- firmed that there is a deviation from the voltage minimum value,

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
1240 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 2, APRIL 2008

Fig. 10. Voltage profile (case 2). Fig. 13. Voltage profile (case 3).

Fig. 14. Voltage profile (case 3).

Fig. 11. Voltage profile (case 2).

Fig. 15. Voltage profile (case 3).

TABLE IV
Fig. 12. Voltage profile (case 2).
STATES OF LRT + SVR

and that a large voltage margin can be taken by controlling all

devices cooperatively. In case 3 shown in Figs. 13–15, the con-
sumer load is the maximum value, and the output of the photo-
voltaic generation equipment is 0. It can be inferred that the de- the voltage control and the tap control of SVR. This shows that
viation from the maximum voltage value occurs only by sending the voltage variation range is more than 1, and this exceeds the

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
SENJYU et al.: OPTIMAL DISTRIBUTION VOLTAGE CONTROL AND COORDINATION 1241

TABLE V of all the devices as in the other cases. If all devices are coopera-
STATES OF LRT + SVR + SC + ShR tively controlled according to this simulation result, then a large
voltage margin can be taken. Further, reduction of each control
device’s capacity, such as SC, ShR, and SVC, can be expected
by performing cooperation control.
Optical-fiber use is widespread in Japan. Actually, advanced
distribution automation systems using an optical-fiber network
are developed in [15]. Therefore, the central coordination of
voltage regulating devices is possible in the near future.

IV. CONCLUSION
TABLE VI In this paper, centralized cooperation control of sending
STATES OF LRT + SVR + SC + ShR + SVC voltage, and SVR, SC, ShR, and SVC are proposed. Each control
device is cooperatively controlled for optimal voltage of each
time section for the case of a residential area system where
photovoltaic generation equipment is introduced and assumed to
be a widespread part of the communication infrastructure. The
conventional SVR acquires self-end information, estimates the
voltage drop to the load central point, and determines the tap po-
sition. However, the transformer is considered to be the part that
determines the tap location by centralized control from the opti-
mizing control of the whole power distribution system voltage.
The controller determines the amount of operation, based on the
GA that expedites approximation of solution from the feasible
TABLE VII
FITNESS VALUES
area discretely set as a large-scale optimization problem. It is
confirmed that the proposed technique contributes to voltage
adjustment of the power distribution system into which the large
voltage margin of each node could be taken and the distributed
generation is introduced by performing centralized cooperation
TABLE VIII control of the power distribution equipment. These proposed
CHANGE OF POWER FLOW IN EACH CASE methods require a reliable communication infrastructure. If the
communication system fails, the voltage regulation devices with
a broken communication system detect a failure and change the
control methods to control self-information. Since the informa-
tion on locating the fault cannot be acquired, the control based on
empirical rules (a time, a season) can be assumed to be carried
out that can be put into control conditions, and the system can
respond by carrying out concentrated control of the other voltage
regulation devices with the unbroken communication system.
Optimizing voltage for the whole power distribution system,
including the output cooperation of distributed generation, will
be the subject of a future work.

REFERENCES
[1] K. Kabemura, K. Yonekura, T. Tsukamoto, K. Hashimoto, and M.
Hara, “Application of dispersed autonomous voltage control system to
a real high voltage distribution network,” Elect. Eng. Jpn., vol. 146, no.
1, pp. 27–36, 2004.
[2] N. I. Santoso and O. T. Tan, “Neural-net based real-time control of
capacitors installed on distribution systems,” IEEE Trans. Power Del.,
vol. 5, no. 1, pp. 266–273, Jan. 1990.
[3] J. C. Wang, H. D. Chiang, K. N. Miu, and G. Darling, “Capacitor place-
ment and real time control in large-scale unbalanced distribution sys-
tems: Loss reduction formula, problem formulation, solution method-
ology and mathematical justification,” IEEE Trans. Power Del., vol. 12,
no. 2, pp. 953–958, Apr. 1997.
[4] J. C. Wang, H. D. Chiang, K. N. Miu, and G. Darling, “Capacitor place-
ment and real time control in large-scale unbalanced distribution sys-
tems: Numerical studies,” IEEE Trans. Power Del., vol. 12, no. 2, pp.
maximum limit from a viewpoint of the solution obtained for 959–964, Apr. 1997.
[5] Z. Gu and D. T. Rizy, “Neural networks for combined control of capac-
power transmission. Moreover, it can be confirmed that a large itor banks and voltage regulators in distribution systems,” IEEE Trans.
voltage margin can be taken by carrying out cooperation control Power Del., vol. 11, no. 4, pp. 1921–1928, Oct. 1996.

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.
1242 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 2, APRIL 2008

[6] J. Yoshinaga, T. Hirai, J. Motohashi, T. Takano, M. Wataabe, and Y. Yoshitaka Miyazato (S’06) was born in Okinawa, Japan, in 1984. He received
Kowada, “Development of central voltage control method for distribu- the B.S. degree in electrical engineering from the University of the Ryukyus,
tion systems,” in Proc. 16th Annu. Conf. Power Energy Society. Inst. Okinawa, in 2006, where he is currently pursuing the M.S. degree.
Elect. Eng. Jpn. (in Japanese), 2006, pp. 25-1–25-6. His research interests include power system control.
[7] F. C. Lu and Y. Y. Hsu, “Fuzzy dynamic programming approach to re- Mr. Miyazato is a student member of the IEEJ.
active power/voltage control in a distribution substation,” IEEE Trans.
Power Del., vol. 12, no. 2, pp. 681–688, Apr. 1997.
[8] R. H. Liang and C. K. Cheng, “Dispatch of main transformer ULTC
and capacitors in a distribution system,” IEEE Trans. Power Del., vol.
16, no. 4, pp. 625–630, Oct. 2001.
[9] R. H. Liang and Y. S. Wang, “Fuzzy-based reactive power and voltage
control in a distribution system,” IEEE Trans. Power Del., vol. 18, no. Atsushi Yona (S’06) was born in Okinawa, Japan, in 1982. He received the B.S.
2, pp. 610–618, Apr. 2003. degree in electrical engineering from the University of the Ryukyus, Okinawa,
[10] Y. Deng, X. Ren, C. Zhao, and D. Zhao, “A heuristic and algorithmic in 2006, where he is currently pursing the M.S. degree.
combined approach for reactive power optimization with time-varying His research interests include the wind and photovoltaic energies, and neural
load demand in distribution systems,” IEEE Trans. Power Syst., vol. 17, networks.
no. 4, pp. 1068–1072, Oct. 2002. Mr. Yona is a student member of the IEEJ.
[11] T. Kondo, J. Baba, and A. Yokoyama, “Voltage control of distribu-
tion network with a large penetration of photovoltaic generations using
FACTS devices,” (in Japanese) Inst. Elect. Eng. Jpn. Trans. Power En-
ergy, vol. 126, no. 3, pp. 347–358, 2006.
[12] M. Sakawa and M. Tanaka, Genetic Algorithm (in Japanese). Tokyo,
Japan: Asakura, 1995. Naomitsu Urasaki (M’98) received the B.S., M.S., and Ph.D. degrees from the
[13] “Harmonics obstacle preventive measures of a power distribution University of the Ryukyus, Okinawa, Japan, in 1996, 1998, and 2004, respec-
system,” (in Japanese) Elect. Technol. Res. Assoc., vol. 37, no. 3, 1981.
tively, all in electrical engineering.
[14] “New electric power network system study group report,” Inst. Appl.
Since 1998, he has been with the Department of Electrical and Engineering,
Energy, 2003, (in Japanese).
[15] T. Kasajima, R. Endo, Y. Wada, Y. Kudo, and H. Kanawa, “The de- Faculty of Engineering, University of the Ryukyus, where he is an Associate
velopment of the advanced distribution automation system with optical Professor. His research interests include the areas of motor drives.
fiber network of Tokyo Electric Power Co., Inc.,” in Proc. IEEE Power Dr. Urasaki is a member of the Institute of Electrical Engineers of Japan.
Eng. Soc. General Meeting, 2004, vol. 12, pp. 1441–1445.

Tomonobu Senjyu (A’89–M’02–SM’06) was born in Saga, Japan, in 1963. He

received the B.S. and M.S. degrees in electrical engineering from the University Toshihisa Funabashi (M’90–SM’96) was born in Aichi, Japan, in 1951. He
of the Ryukyus, Okinawa, Japan, and the Ph.D. degree in electrical engineering received the B.S. degree from the Nagoya University, Aichi, in 1975, and the
from the Nagoya University, Nagoya, Japan, in 1986, 1988, and 1994, respec- Ph.D. degree from the Doshisha University, Kyoto, Japan, in 2000, both in elec-
tively. trical engineering.
In 1988, he joined the University of the Ryukyus, where he is now a Pro- He has been with the Power System Engineering Division, Meidensha Cor-
fessor. His research interests include stability of ac machine, advanced control poration, Tokyo, Japan, since 1975, and is now a Senior Engineer, engaged in
of electrical machines, and power electronics. research on power system analysis.
Prof. Senjyu is a senior member of the IEEJ. He is also a Chartered Engineer Dr. Funabashi is a member of the Institute of Electrical Engineers. He is also
in the U.K. a Chartered Engineer in the U.K.

Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:40 UTC from IE Xplore. Restricon aply.