IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO.
1, FEBRUARY 2004
Power Distribution System Switching Operation
Scheduling for Load Balancing by Using
Colored Petri Nets
Yu-Lung Ke, Member, IEEE, Chao-Shun Chen, Member, IEEE, Meei-Song Kang, Jaw-Shyang Wu, Member, IEEE,
and Tsung-En Lee
Abstract—This study attempts to determine the daily load
patterns of service zones, line switches, distribution feeders, and
main transformers by using customer information in a customer
information system (CIS) and information about distribution
transformers in the outage management information system
(OMIS) in Taiwan Power Company (Taipower). When a power
distribution system is operating under normal conditions, the
reconfiguration of feeders for balancing loads among distribution
feeders is obtained by the colored Petri nets (CPN) inference
mechanism, which improves the operating performance of dis-
tribution systems. A practical Taiwan power distribution system,
with daily load patterns derived by a load survey, is used for a
computer simulation and, thus, determines the effectiveness of the
proposed methodology to improve the balancing of the feeder load
for distribution systems by considering the load characteristics of
the service customers.
Index Terms—Colored Petri nets, feeder reconfiguration, load
balancing, load patterns, switching operations.
I. INTRODUCTION
W ITH the vigorous development of the economy in
Taiwan, Taipower has experienced the continuous
increase of load density over the whole power distribution
system. The distribution feeders may become overloaded due
to load growth and substation planning and it complicates the
distribution system operation in areas with high load density.
With the usage of more and more air conditioners in the
commercial and residential areas, the air conditioning loading
has contributed more than 35% of the summer peak loading of
Taipower. It is estimated that the system power consumption
increases by 600 MW per 1 temperature rises when the
temperature exceeds 28 C. The difference between the system
load at the daily peak and the offpeak in the summer season
has increased from 4577 MW in 1989 to 8440 MW in 1998.
Many main transformers in the substations are heavily loaded
Manuscript received May 26, 2003. This work was supported by the Na-
tional Science Council of the Republic of China under Contract NSC 89-2213-
E168-010.
Y.-L. Ke is with the Department of Electrical Engineering, Kun Shan
University of Technology, Tainan 710, Taiwan, R.O.C. (e-mail: yu-
lungke@mail.ksut.edu.tw).
C.-S. Chen is with the Department of Electrical Engineering, Na-
tional Sun Yat-Sen University, Kaohsiung 804, Taiwan, R.O.C. (e-mail:
cschen@mail.ee.nsysu.edu.tw).
M.-S. Kang is with the Department of Electrical Engineering, Kao Yuan In-
stitute of Technology, Kaohsiung 821, Taiwan, R.O.C.
J.-S. Wu and T.-E. Lee are with the Department of Electrical Engineering,
National Kaohsiung University of Applied Sciences, Kaohsiung 807, Taiwan,
R.O.C.
Digital Object Identifier 10.1109/TPWRS.2003.821433
0885-8950/04$20.00 © 2004 IEEE
629
and the loading factors have deteriorated. Fig. 1 presents
the total number of main transformers in Taipower and the
corresponding loading factor (LF). It is found that the loading
factors of 122 main transformers exceeded 95% and 33 main
transformers were overloaded. To solve the problem, the
mobile transformers have been used for the temporary relief
of peak loading. Besides, the load transfer among distribution
feeders and main transformers are considered to achieve the
load balance by proper reconfiguring the distribution system.
Load imbalance and uneven load distribution among feeders
are important causes of increased system loss. Planning feeder
reconfiguration by switching operations to achieve load balance
among distribution feeders will reduce system losses [3]–[12]
and increase the operation flexibility by evenly distributing the
capacity reserve among the main transformers.
In a large power distribution system, feeder reconfiguration
can change the system topology by changing the open/close
states of the sectionalizing switches and tie switches under both
normal and abnormal operating conditions to reduce the power
loss and solve the overloading problem [3]–[12]. The feeder re-
configuration is a complex combinatorial and constrained opti-
mization problem because of the enormous number of combi-
nations of candidate switches [13]. It becomes more and more
difficult to reconfigure feeders based on the experience of distri-
bution operators. A more systematic operation and computer de-
cision support strategy must be designed to reconfigure feeders
quickly and effectively in the power distribution system.
An integrated distribution automation system (DAS) has been
successfully implemented by Taipower to increase the system
reliability and service quality. The optimal switching operation
(OSW) program [14]–[19] is to solve the feeder reconfiguration
to achieve load balance among distribution feeders and main
transformers so that the system loss can be minimized and
system overloading due to the use of air conditioners during the
summer can be prevented. Furthermore, the quality of power
supply has become increasingly important to customers as power
distribution systems operate under more constrained operation
conditions. Accordingly, power distribution systems must ben-
efit from more efficiently made operation decisions in response
to handle the increasingly harsh operating environment, and the
reliability of the power supply must increase.
To support the decision-making process of optimal inter-
feeder switching operation, the CPN inference mechanism
approach with operation rules is performed to derive the ap-
propriate combination of switching operations for load transfer
among distribution feeders under system normal operation
630
Fig. 1. Loading factor of main transformers in Taipower system.
conditions or overload contingency conditions by considering
the daily load patterns of customers. The daily load profile of
each distribution feeder in the Taipower distribution system is
determined by considering the typical load patterns of various
customers. The power consumption information of customers
within each service area is derived by the customer information
system (CIS) and outage management information system
(OMIS) in Taipower. The CPN inference mechanism, com-
bined with the utility operation rules, can therefore efficiently
determine the effective combination of switching operations to
increase the system operation performance by balancing the
load among distribution feeders and main transformers.
II. DISTRIBUTION SYSTEM OPERATION
An optimally load-balanced power distribution system is re-
garded as in an ideal state, such that all feeders and main trans-
formers are operated at the same load level. To determine the
load balance among distribution feeders after load transfer, an
equivalent distribution feeder current over a day is defined as
in (1) [14]–[19], where represents the current loading of the
feeder during hour .
(1)
The ideal feeder load level , and the ideal transformer
load level are defined by (2) and (3) [14]–[19], respec-
tively. The terms and represent the current loading of
feeders and transformers, respectively; and represent
the feeders’ and the transformers’ rated capacities; and
are the total number of feeders and transformers, respectively
(2)
The feeders’ load balance index and the
(3)
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 1, FEBRUARY 2004
transformers’ load balance index are defined by the least
square formulas in (4) and (5) [14]–[19] respectively, where
and represent the percentage current loading of the
feeders and transformers, respectively. The load balance index
is zero for an ideal load-balanced power distribution system. A
smaller load balance index implies better system performance
(4)
(5)
To support the CPN inference mechanism for switching
operational decision-making, the object-oriented programming
language C++ is used to generate an object-oriented data-
base of loads associated with customers, feeders, and main
transformers. Given the difference between the typical load
patterns of various types of customer, load transfer among
distribution feeders by interfeeder switching will be effective
to improve load balancing among distribution feeders and main
transformers.
III. COLORED PETRI NETS AND INFERENCE MECHANISM
A. Colored Petri Nets (CPN) [14]–[19], [21]–[27]
Colored Petri nets form a bipartite directed weighting multi-
graph, which is formalized using bag theory. A colored Petri
net structure is a sextuple with ,
where
finite nonempty set of place nodes;
finite nonempty set of transition nodes.
and are disjoint; that is,
and ;
: finite set of types, called color sets,
which determine the operations and
functions that can be used in the net
inscription;
: input token color transfer function,
which communicates between the
colors of the place node and each color
of the transition node;
: output token color transfer function
that communicates between the colors
of the place node and each color of the
transition node;
: initial marking.
B. Enabled Transition Node
A place node is an input location of a transition node
and is represented by ; is an output place node of a
transition node and is represented by . In a col-
ored Petri net, each token color can carry complex information
or a database. Transition node is enabled with respect to color
for the marking M, if and only if the number of tokens in all
of the input place nodes of transition node is not less than
the number of colors by the input function of the directed arc
that connects place node to transition node . Equation (6)
KE et al.: POWER DISTRIBUTION SYSTEM SWITCHING OPERATION SCHEDULING FOR LOAD BALANCING 631

TABLE II
PROPOSED SWITCHING OPERATION BY COMPARISON OF CPN AND BIP
APPROACHES

Fig. 2. Switching operation model by colored Petri nets.

TABLE I
RELATIONSHIP BETWEEN INPUT PLACES AND OUTPUT PLACES
FOR EACH TRANSITION

TABLE III
SWITCHING OPERATIONS SCHEDULING FOR LOAD BALANCING

[14]–[19], [21]–[27] specifies the enabling rule of transition for

colored Petri nets
(6)

C. Firing an Enabled Transition Node

An enabled transition node , with respect to a color , can
be fired. The new marking is obtained after transition node
is fired with respect to the color . This new marking
is obtained from the current marking by the following rela-
tionship [14]–[19], [21]–[27]:

D. Inference in CPNs
The CPN views a system by describing the state transition
of systems using place nodes to represent conditions, and tran-
sition nodes to represent events or activities. If all of the place node is activated. The activated transition node can be fired and
nodes that enter a transition node receive token colors, then the the token colors changed between the entering place nodes and
transition node is enabled. If the guard function of the enabled the outgoing place nodes. When the place node stands for com-
transition node is evaluated to take action, then the transition pleted work, the inference is then determined. In CPN, several
632
Fig. 3. Practical 18-feeder distribution system.
TABLE IV
FEEDER EQUIVALENT CURRENT COMPARISON
tokens may be present and several transition nodes may be acti-
vated simultaneously. Consequently, passing tokens can be pro-
cessed along many paths simultaneously at the same time to
achieve the parallel-like inference.
In practical distribution system modeling, CPN is used to
model the practical switching in a distribution system, the
on/off states of line switches and energized/de-energized/ter-
minal states of service zones, which are represented by place
nodes, specify the variable distribution system states during
the inference process. The service section is defined as the
terminal zone (T) when it is the final section in a feeder; every
service section of a feeder except the terminal zone is regarded
as a nonterminal zone (NT). Fig. 2 presents the switching
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 1, FEBRUARY 2004
operation model using CPN. The numbers of place nodes,
transition nodes, and directed arcs in a CPN are calculated from
(7), (8), and (9) [14]–[19], where , , , and
represent the numbers of service zones, sectionalizing switches,
feeders, and faulted zones, respectively. Table I presents the
relationships between the input place nodes and output place
nodes of each switching operation of switch , according to the
switching operation model obtained using Colored Petri nets.
The relevant expressions are given by (10)–(13) [14]–[19]
(7)
(8)
(9)
(10)
(11)
(12)
(13)
In Fig. 2, switch is fired to open to release the loading of
z1 when all of the states of three input place nodes of the tran-
sition node are met: s(C) implies switch is closed; z2(E&NT)
implies z2 is an energized, nonterminal zone; z1(E&T) implies
that the objective zone z1 is energized and terminal. The fired
transition node will open switch to generate three new output
place nodes: s(O) is with switch is opened; z1(DE&NT) im-
plies z1 is a de-energized and nonterminal zone, while z2(E&T)
implies that z2 is an energized and terminal zone according to
the “if then ” rule as expressed in the notation
in (10). Similarly, in (11) is to release the loading of
z2 by opening switch .
KE et al.: POWER DISTRIBUTION SYSTEM SWITCHING OPERATION SCHEDULING FOR LOAD BALANCING
Fig. 4. Daily load profiles of distribution feeders.
To demonstrate the load balancing of distribution systems by
feeder reconfiguration, a sample distribution system is selected
for computer simulation. All of the switches are initially opened
to maximize the freedom of load balancing switching operation
decisions and are closed one by one, to pick up the adequate
loading of service zones by considering the load distribution
in each zone such that the redistribution of loading in distribu-
tion feeders may be divided equally as possible, subject to the
following constraints 1) The number of switching operations is
633
Fig. 5. Load balance index comparison.
minimized; 2) the radial structure of distribution system is main-
tained; and 3) additional overload contingency associated with
load balancing switching operations is not permitted.
IV. COMPUTER SIMULATIONS
In this paper, a distribution system with 18 distribution
feeders, which serve the mixed loads of various types of cus-
tomers in Taipower is used for case study. The current ratings
of each feeder and main transformer are 450 A and 1350 A,
respectively. Fig. 3 shows a one-line diagram and Fig. 4 plots
the daily load profiles of distribution feeders.
The original feeder load balance index of the original system
at 3 P.M. is 80.3; the feeder load balance index was markedly
reduced to 62.1 and 57.3 by CPN and BIP [20] approaches,
respectively. However, the proposed switching operations in-
volve six and eight switched pairs, obtained by CPN and BIP ap-
proaches, respectively, as shown in Table II. The proposed CPN
approach has lower operation costs because it involves fewer
pairs of switching operation than BIP approach, but the load
balances among distribution feeders by CPN and BIP approach
are almost equal.
Without performing the switching operation, the loading dis-
tributions among distribution feeders are uneven and the system
loss is large. Table III lists the daily scheduled switching load
balancing operations derived by the proposed colored Petri nets
inference mechanism. Table IV compares the equivalent current
in distribution feeders during switching operations (SW). The
feeder load balance index at 3 P.M. and 7 P.M. have been reduced
from 64.1 to 50.3, 68.2 to 49.2, and that of main transformers
has also been improved from 19.7 to 14.2, 22.8 to 12.6, respec-
tively, by execution of the proposed switching operations. The
equivalent feeder current of feeder #3, #17, #10, and #6 have
been adjusted from 362 to 314 A, 144 to 170 A, 359 to 304
A, and 178 to 235 A, respectively, by the proposed switching
operations. The load distribution among distribution feeders is
divided even more effectively better by the balance index of
feeders and main transformers as listed in Fig. 5 so that the
system losses are reduced and the system operation performance
is improved. Fig. 6 presents the resultant daily load profiles of
634 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 1, FEBRUARY 2004

allel-like inference mechanism with operation rules in models

of the colored Petri nets. One of the Taipower distribution sys-
tems, which consists of 18 distribution feeders and serves the
mixed loads of various customer types, was selected for com-
puter simulation to demonstrate the effectiveness of the pro-
posed methodology. By executing the optimal switching opera-
tion derived by the colored Petri nets inference mechanism, the
load distribution among the distribution feeders by load transfer
switching operations was effectively improved. According to
the computer simulation, the proposed methodology provides
an effective tool with which distribution system engineers can
solve the system reconfiguration for loss reduction and load bal-
ancing among distribution feeders and main transformers.

Fig. 6. Resultant load profiles of distribution feeders after scheduled load
balancing switching operations.
the distribution feeders by execution of proposed daily load bal-
ancing switching operations.
V. CONCLUSION
In this paper, the typical load patterns of various customer
types and the energy consumption of all customers served by
each distribution feeder are used to calculate the hourly current
flows of line switches and the loads in each service zone. Op-
timal switching operations for load balancing among distribu-
tion feeders were derived by applying the colored Petri nets par-
REFERENCES
[1] C. S. Chen, J. C. Hwang, and C. W. Huang, “Application of load survey
systems to proper tariff design,” IEEE Trans. Power Syst., vol. 12, pp.
1746–1751, Nov. 1997.
[2] , “Determination of customer load characteristics by load survey
system at taipower,” IEEE Trans. Power Delivery, vol. 11, pp.
1430–1435, July 1996.
[3] M. E. Baran and F. F. Wu, “Network reconfiguration in distribution
system for loss reduction and load balancing,” IEEE Trans. Power De-
livery, vol. 4, pp. 1401–1407, Apr. 1989.
[4] C.-C. Liu, S. J. Lee, and K. Vu, “Loss minimization of distribution
feeders: optimality and algorithms,” IEEE Trans. Power Delivery, vol.
4, pp. 1281–1289, Apr. 1989.
[5] H.-D. Chiang and R. Jean-Jumeau, “Optimal network reconfigurations
in distribution systems: part 1: a new formulation and a solution method-
ology,” IEEE Trans. Power Delivery, vol. 5, pp. 1902–1907, Nov. 1990.
[6] , “Optimal network reconfigurations in distribution systems: part 2:
solution algorithms and numerical results,” IEEE Trans. Power Delivery,
vol. 5, pp. 1568–1574, July 1990.
[7] D. Shirmohammadi, “Service restoration in distribution network recon-
figuration,” IEEE Trans. Power Delivery, vol. 7, pp. 952–958, Apr. 1992.
[8] K. Aoki, T. Ichimori, and M. Kanezashi, “An efficient algorithm for
load balancing of transformers and feeders by switch operation in large
scale distribution systems,” IEEE Trans. Power Delivery, vol. 3, pp.
1865–1872, Oct. 1988.
[9] J. S. Wu, K. L. Tomsovic, and C. S. Chen, “A heuristic search approach
to feeder switching operations for overload, faults, unbalanced flow and
maintenance,” IEEE Trans. Power Delivery, vol. 6, pp. 1579–1585, Oct.
1991.
[10] T. Taylor and D. Lubkeman, “Implementation of heuristic search strate-
gies for distribution feeder reconfiguration,” IEEE Trans. Power De-
livery, vol. 5, pp. 239–245, Jan. 1990.
[11] Q. Zhou, D. Shirmohammadi, and W.-H. E.W.-H. Edwin Liu, “Distribu-
tion feeder reconfiguration for service restoration and load balancing,”
IEEE Trans. Power Syst., vol. 12, pp. 724–729, May 1997.
[12] Y. Y. Hsu, Y. J. Hwu, S. S. Liu, Y. W. Chen, H. C. Feng, and Y. M.
Lee, “Transformer and feeder load balancing using a heuristic search
approach,” IEEE Trans. Power Syst., vol. 8, pp. 184–190, Feb. 1993.
[13] Y.-C. Huang, “Switch reconfiguration of electrical distribution systems
using improved genetic algorithm,” J. Comput., vol. 13, no. 3, pp. 47–54,
Sept. 2001.
[14] C.-S. Chen, Y.-L. Ke, and J.-S. Wu, “Switching operation decision
making by using Petri-net approach for power distribution systems,” J.
Chinese Inst. Eng., vol. 25, no. 1, pp. 17–26, Jan. 2002.
[15] , “Colored Petri nets approach for solving distribution system con-
tingency by considering customer load patterns,” Proc. Inst. Elect. Eng.,
Gen. Transm. Dist., vol. 148, no. 5, pp. 463–470, Sept. 2001.
[16] C.-S. Chen, Y.-L. Ke, J.-S. Wu, and M.-S. Kang, “Application of Petri
nets to solve distribution system contingency by considering customer
load patterns,” IEEE Trans. Power Syst., vol. 17, pp. 417–423, May
2002.
[17] J.-S. Wu, C.-C. Liu, K.-L. Liou, and R. F. Chu, “A Petri net algorithm
for scheduling of generic restoration actions,” IEEE Trans. Power Syst.,
vol. 12, pp. 69–75, Feb. 1997.
[18] J.-S. Wu, “A Petri-net algorithm for multiple contingencies of dis-
tribution system operation,” IEEE Trans. Power Syst., vol. 13, pp.
1164–1171, Aug. 1998.
KE et al.: POWER DISTRIBUTION SYSTEM SWITCHING OPERATION SCHEDULING FOR LOAD BALANCING
[19] Y.-L. Ke, “A study on switching operation decision making by using
petri nets for power distribution systems,” Ph.D. dissertation, National
Sun Yat-Sen Univ., Kaohsiung, Taiwan R.O.C., June 2001.
[20] C. S. Chen, M. S. Kang, J. C. Hwang, and C. W. Huang, “Tempera-
ture adaptive switching operation for distribution systems,” IEEE Trans.
Power Delivery, vol. 16, pp. 694–699, Oct. 2001.
[21] J. L. Peterson, Petri Net Theory and the Modeling of Sys-
tems. Englewood Cliffs, NJ: Prentice-Hall, 1981.
[22] K. Jensen and G. Rozenberg, High-Level Petri Nets Theory and Appli-
cation. New York: Springer-Verlag, 1991.
[23] R. David and H.Hassane Alla, Petri Nets and Crafcet. Englewood
Cliffs, NJ: Prentice-Hall, 1992.
[24] K. Lautenbach, Lectures on Invariant Analysis Methods Definitions,
Theoretical Results, Examples. Taipei, Taiwan R.O.C.: National
Taiwan Univ., 1999.
[25] W. Reisig and G. Rozenberg, Lectures on Petri Nets I: Basic
Models. Berlin, Germany: Springer-Verlag, 1998.
[26] K. Jensen, Colored Petri Nets Basic Concepts, Analysis Methods and
Practical Use. Berlin, Germany: Springer-Verlag, 1997.
[27] E. Best, Raymond Devillers and Maciej Koutny, Petri Net Al-
gebra. Berlin, Germany: Springer-Verlag, 2001.
[28] Operation Manual of Outage Management Information System, Taiwan
Power Company, Business Department, Taipei City, Taiwan, R.O.C.
Yu-Lung Ke (M’98) was born in Kaohsiung, Taiwan, R.O.C., on July 13, 1963.
He received the B.S. degree in control engineering from National Chiao-Tung
University, Hsin-Chu, Taiwan, R.O.C., in 1988, the M.S. degree in electrical
engineering from National Taiwan University, Taipei, Taiwan, R.O.C., in 1991,
and the Ph.D. degree in electrical engineering from National Sun Yat-Sen Uni-
versity, Kaohsiung, Taiwan, R.O.C., in 2001.
Currently, he is an Associate Professor with the Department of Electrical
Engineering at Kun Shan University of Technology, Tainan, Taiwan, R.O.C.,
where he has been since 1991. His research interests include distribution au-
tomation, energy management, power quality, and artificial intelligence appli-
cations in power system operation. Since 2002, he has been as a reviewer for
IEEE TRANSACTIONS ON POWER DELIVERY and International Journal of Elec-
trical Power & Energy Systems.
Dr. Ke is a Member of IEEE Power Engineering Society (PES), Industry Ap-
plications Society (IAS), and Systems, Man, and Cybernetics (SMC) Society.
635
Chao-Shun Chen (M’84) received the B.S. degree in electrical engineering
from National Taiwan University, Taipei, Taiwan, R.O.C., in 1976 and the M.S.
and Ph.D. degrees in electrical engineering from the University of Texas, Ar-
lington, in 1981 and 1984, respectively.
Currently, he is a Professor in the Electrical Engineering Department at Na-
tional Sun Yat-Sen University, Kaohsiung, Taiwan, R.O.C. From 1984 to 1994,
he was a Professor with National Sun Yat-Sen University, Kaohsiung, Taiwan,
R.O.C. He was also in charge of electrical and mechanical system planning with
the Kaohsiung Rapid Mass Transport Project, Mass Rapid Transmit Bureau,
Kaohsiung City, Taiwan, R.O.C.
Meei-Song Kang received the M.S. and Ph.D. degrees in electrical engineering
from the National Sun Yat-Sen University, Kaohsiung, Taiwan, R.O.C., in 1993
and 2001, respectively.
Currently, he is an Associate Professor with the Department of Electrical
Engineering at Kao Yuan Institute of Technology, Kaohsiung, Taiwan, R.O.C.,
where he has been since 1993. His research interest is in the area of load survey
and demand subscription service.
Jaw-Shyang Wu (M’95) received the Ph.D. degree in electrical engineering
from the National Sun Yat-Sen University, Kaohsiung, Taiwan, R.O.C., in 1991.
Currently, he is a Professor in the Electrical Engineering Department at Na-
tional Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan, R.O.C.
His research areas are power system operation, expert system, and application
of artificial intelligence.
Tsung-En Lee was born in Tainan Hsien, Taiwan, R.O.C., on March 5, 1958.
He received the B.Sc. degree in electrical engineering from National Taiwan In-
stitute of Technology, Taipei, Taiwan, R.O.C., in 1985, the M.Sc. degree in elec-
trical engineering from National Taiwan University, Taipei, Taiwan, R.O.C., in
1989, and the Ph.D. degree from National Sun Yat-Sen University, Kaohsiung,
Taiwan, R.O.C., in 1994.
Currently, he is an Associate Professor with the Department of Electrical En-
gineering at National Kaohsiung University of Applied Sciences, Kaohsiung,
Taiwan, R.O.C. His research interests include power system operation, and ap-
plication of operation research and artificial intelligence.