Professional Documents
Culture Documents
Abstract: A genetic algorithm (GA) is employed to search for the optimal supply restoration
strategy in distribution networks. An ‘integer permutation’ encoding scheme is adopted in which
each chromosome is a list of indices of switches. The status of each of these switches is decided
according to graph theory subject to the radiality constraint of the distribution networks. Each
chromosome then maps to a feasible network topology. A special gene ‘0’ is also introduced into
the chromosome. Instead of representing a switch, t h s is a flag that keeps some parts of the
network disconnected enabling the GA to find the optimal load shedding strategy where necessary.
The proposed algorithm has been tested on a practical system and shown to find an optimal
postfault supply restoration strategy, and also the optimal load shedding point when total demand
cannot be supplied.
148 IEE Proc -Gener Trmm Dutrib Vol 149, No 2, March 2002
~
9
I d
I
-
@ lnfeed
I *s44
Load
f Transformer
+ Load
Closed Switct
Closed Switcl
OpenSwltch
0 Open Switch
0
0
a,
40-
-
about 3.5min of computer time for the GA to reach the a
20
optimal supply restoration solution. L
O , . J , . . , I
1 51 101 151 201 251 301
5.2 GA for optimal load shedding iteration no.
Suppose now that the system is operated in a stressed
condition at 120% base load level. With the same GA Fig. 5 GA evolution process for cptimal load shedding, using ‘0’
parameters as in Section 5.1, the minimum objective gene
function value achieved is 13.166, of which 3.26 is due to -OBF
load shedding and 0.905 is for network loss, and 9.0 is for
switch operations. The resulting network is shown in Fig. 4.
The GA evolution process is shown in Fig. 5. Since the system is in a very si.ressed condition and cannot
It can be seen from Fig. 5 that, at the beginning of GA supply all loads without violaling security constraints, one
evolution, the objective function is high due to the or more selected loads have to be shed. The optimal load
unreasonable network topologies generated by the (initially shedding point is indicated by the position of the ‘0’gene in
randomly positioned) ‘0’ gene. However, through evolution, the final chromosome. Its position is ‘tuned’ by the GA
the ‘0’ gene moves towards the end of the chromosome. evolution to minimise the GA objective function value. The
load is shed if it can only be connected to the network via
switches located behnd the ‘0‘gene in the final chromo-
some. In this case, as can be seen in Fig. 4, switch S49 is
I I
tripped off to shed the appropriate load.
5.3 Results
From the test results, it can be seen that at the base load
level where full supply restoration can be achieved, the
proposed algorithm gives an optimal reconfigured network
and supply restoration strategy.
When the system is in a stressed condition, and the
network can not supply all of its loads, the position of the
‘0’gene and the genes following it in the GA chromosome
determine the optimal load shedding points, so that the
network can supply all other loads.
The radiality constraint is respected by the graph theory-
based switch setting algorithm. The security limit con-
straints are unlikely to be violated due to the heavy penalty
weightings selected for them in the objective function of the
GA. It can be seen from the resulting objective function
values that in both examples, no overload exists in any line,
and no voltage excursion occiirs at any bus. In the case of
0 lnfeed
violations of these security lirrut constraints, the GA would
LI Transformer
prefer to shed some loads to correct the problem, since the
-+ Load
penalty weighting value for the unsupplied load term is
1 Closed Switch
OpenSwitch 1 smaller.
In both cases, the optimum network configurations are
acheved after about 340 generations of evolution. It takes
approximately 3.5min of computer time for the system to
read the data from the database, run the GA for 340
generations and display the suggested supply restoration
result and optimal load shedding point (if applicable) to the
user. This level of perfonnance would allow online
application in the distribution control centre.
It can be clearly seen that, by omitting the switch
Fig. 4 Network confguration after optimal load shedding in operation cost term and lost load term in the objective
stressed condition (at 1200?base load level) function formulae, the proposed algorithm could also
chromosome length = 64 (‘0’gene, pool size = 50) function as a network loss minimiser for distribution
crossover rate = 0.6 network operation and planning. The ‘0’ gene will not be
mutation rate = 0.08125 required in this application.
150 IEE Proc.-Gener. Transm. Distrih., Vol. 149, No. 2,March 2002
6 Conclusions 2 LIU, C.C., LEE, S.J., and VENKATA, S.S.: ‘An expert system
operational aid for restoration and loss reduction of distribution
systems’, IEEE Truns. Power Syst., 1988, 3, (2), pp. 619-626
A GA-based method is proposed to decide the supply 3 SAKAGUCHI, T., and MATSUMOTO, K.: ‘Development of a
restoration and optimal load shedding strategy for distribu- knowledge based system for power system restoration’, ZEEE Trans.,
tion networks. The ‘integer permutation’ encoding scheme Power Appur. and Syst., 1983, PAS-102, (2), pp. 32&329
4 SARFI, R.J., SALAMA, M.M.A., and CHIKHANI, A.Y.: ‘A survey
is adopted in which each chromosome represents the set of of the state of the art in distribution system reconfiguration for system
controllable switches the final states of which the algorithm loss reduction’, Electr. Power Syst. Res., 1994, 31, pp. 61-70
5 NAGATA, T., SASAKI, H., and KITAGAWA, M.: ‘A method of
is to determine. Graph theory guarantees that the chromo- determining target configurations for power system restoration using a
some will be mapped to a unique and valid radial mixed integer programming approach’, Electr. Eng. Jpn, 1995,
distribution network. A ‘0’ gene can be introduced in the 115, (2), pp. 67-77
6 STANKOVIC, A.M., and CALOVIC, M.S.: ‘Graph oriented
chromosome to indicate the optimal load shedding points algorithm for the steady state security enhancement in distribution
when demand is beyond the ability to supply. The algorithm systems’, ZEEE Truns. Power Deliu., 1989, 4, (I), pp. 53%544
has been successfully tested on a practical system. 7 TEO, C.Y.: ‘A computer aided system to automate the restoration of
electrical power supply’, Electr. Power Syst. Res., 1992, 24, pp. 119-
125
7 Acknowledgments 8 TEO, C.Y., and GOOI, H.B.: ‘Restoration of electrical power supply
through an algorithm and knowledge based system’, Electr. Power
Syst. Res., 1994, 29, pp. 171-180
The authors would like to acknowledge the financial 9 GOLDBERG, D.E.: ‘Genetic algorithms in search, optimisation and
support of Northern Electric Distribution, Yorkshire machme learning’, (Addison-Wesley, 1989)
Electricity, and NORWEB Distribution. They also wish 10 NARA, K., SHIOSE, A,, KITAGAWA, M., and ISHIHARA, T.:
‘Implementation of genetic algorithm for distribution systems loss
to thank Mr. Bob Eunson, Mr. Mark Marshall, Mr. Steve minimum re-configuration’, IEEE Trans. Power Syst., 1992, 7, (3),
Cox and Mr. John Westwood of the above companies, for pp. 10441051
11 FUKUYAMA, Y., CHIANG, H.D., and MIU, K.N.: ‘Parallel
their technical contributions to this research. genetic algorithm for service restoration in electric power distribution
systems’, Electr. Power Energy Syst., 1996, 18, (2), pp. I 1 1-1 19
8 References 12 IRVING, M.R., and MACQUEEN, C.N.: ‘Robust algorithm for
load estimation in distribution networks’, IEE Proc.-Gener. Trunsm.
1 CHANG, G., ZRIDA, J., and BIRDWELL, J.D.: ‘Knowledge-based Distrib., 1998, 145, (5), pp. 49%504
distribution system analysis and reconfiguration’,IEEE Trum. Power 13 MICHALEWICZ, Z.: ‘Genetic algorithms + data structures =evolu-
Sysr., 1990, 5, (3), pp. 744749 tion programs’, (Springer, 1996)
ZEE Proc.-Getier. Trunsm. Distrib.. Vol. I49, No. 2, Mnrch 2002 151