Beruflich Dokumente
Kultur Dokumente
CAPACITOR PLACEMENT IN
DISTRIBUTION SYSTEMS, A NEW
FORMULATION
M. H. Shwehdi, A. Mantawi ,S. Selim, A. AkShehri, KFUPM and G. K. A!-Bassam, Saudi Aramco
I&
Term-Capacitor Placement, Capacitor Compensated
Distribution Lines, Power System Economics, Genetic
Algorithm,
1. NOMENCLATURE
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
electrical utility?
B. General Solution Algorithm
C. Differences in Methods
Implementation of the solution algorithm varies in the
following forms:
lRepresentation
D. Simplifying Assumptions
In their effort to simplify the solution to the CPP, early pioneers
of the CPP have either ignored some of the terms in the
objective function or made assumptions that simplified the
formulation of the problem and the solution thereof. For
instance, early papers assumed a uniformly distributed load
along radial feeders. This configuration greatly simplified the
problem where it was converted to a calculus problem with
closed form solutions developed based on the gradient of the
objective function. However, error has resulted when this
assumption was applied to nonqadial feeders with non-uniform
f61.
(1)
(2)
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
(3)
5d
Scale Bus Data based on load level
3.2-Crossover
LOCATION
BINARY
VALUE
STATUS
Set iteration =1
E, Optimization
A program was written to implement the solution of the CPP
based on use of GA as an optimization technique (see Fig-I for
the flow chart). Two strings were used, one for the location of
the capacitor and the other for the sizes of the capacitors. In
addition, 4 decision-levels were used to execute mutation on
the binary coded genes.
e,
Iteratlo"=
Iteration+l
LOCATION
BINARY
VALUE
SIZE(KVAR)
I1
I
I ON
1
o
12
13
In
I1
I OFF
I ON
2
n
3
i
c
&Find best
solution
f
. I . . . . . . . "
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
2) Crossover:
The crossover is not restricted to a whole chromosome.
Instead, a chromsome could be cut in half to allow for the
reproduction to occur.
3) Mutation:
Mutation is performed based on four random decisions. First, a
random number is generated to indicate whether the gene is to
be processed for mutation or not. This number is checked
against a predefined probability of mutation. After ruling that
the gene is to be mutated, another random number is generated
that represents the total number of chromosome bits that are
candidates for mutation in this gene. These hits are not
necessaily in the same gene chromosome or in sequence. We
have yet to decide which exact bit locations are to be mutated
in the gene. For this, another string of random numbers is
generated that represents the exact locations of the bits to be
mutated. Now, havng decided which gene is to be mutated,
total number of mutations per gene, locations of the bits to be
mutated, the final random decision is to check the probability
of mutating that particular bit. For this, another random number
is generated. With all the above probabilities stored in a look
up table, the full gene bit locations are scanned and mutated
by switching ones to zeros and zeros to ones.
4) Convergence:
There is no rule for convergence. Basically, the GA algorithm is
repeated for a predetermined number of maximum iterations. A
possibility that was considered and abandoned is the
comparison between the previous and current best solutions
where the program is to be terminated when the difference is
equal to, or less than, a certain limit. However, this criterion
could lead to premature convergence to a non-optimal solution
in a flat or semi-flat portion of the search space. Hence,
decision was made to proceed with the GA iterations
regardless of the relative differences between values.
5) Solution Update:
The population was updated at the end of each iteration with
results from reproduction and mutation. In order to preserve
high fitness genes, certain percentage of the population best
solutions are rolled to the next generation without
reproduction, crossover, or mutation.
V. TEST SYSTEM, CASES & RESULTS
A. Test System
An anempt was made to test the program and fitness function
2) Comprehensive Revenue:
Fitness -3 was applied to the Test System. Complete gene
crossover was selected for the GA algorithm. The capacitor
locations were initialized to the same locations in [I] and [J
and were modified in the CA algorithm using the
LOCATIONSTRING gene. Mutalion was applied using the
four decision levels to both the SCILUTION gene as well as the
LOCATIONSTRING gene.
3) Hybrid:
In this case, Fitness-2 was used in the CA algorithm but the
revenue was calculated based ori Fitness-3. This is similar to
CASBl but with the decision to install the capacitors made
based on the revenue calculated by Fitness-3. Results.
4) Location Control:
Using Fitness2 for the GA algoritlun, the LOCATIONSTRMG
process of gene location manipulation has been deactivated
(CASWa). The Location String process was then activated
(CASWb).
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
C. Results:
It is clear that solving the CPP using Fitness-2 as the objective
function results tangible loss reduction, however, the
calculated revenue is not encouraging.
Moreover, using Fitness-3 independently to solve the CPP will
give a false indication of high revenue since losses in the
system increased rather than decreased. This is due to the
marginal contribution of the cost of real power loss reduction
in the comprehensive revenue calculations (proposed in
Fitness-3) compared to the benefits from reducing the reactive
power, which, hy virtue of addition of capacitors, have
increased considerably. This result should not lead the
reviewer to conclude that Fitness-3 incorrectly calculates
revenues. Rather, the conclusion should be Fitness-3 is
suitable for reflection of the true revenue from capacitor
placement but should not be used for the selection of
capacitor.
TABLE I
VIII W CONCLUSION
It has been demonstrated that use of loss reduction alone to
calculate revenue from CP will result much smaller values
compared to using a comprehensive benefits-cost analysis.
Furthermore, the results presented demonstrated that use of a
LOCATIONSTRING has no significant impact on loss
reduction achieved from CP, although it leads to different
capacitor arrangement among the permissible locations.
The research also arrived at the fact that the CP has marginal
negative impact on some buses with respect to symmetrical
three-phase fault current. However, CP has positive impact on
the system resonance in odd harmonics 1 to 1 5 .
IX ACKNOWLEDGMENT
The authors would like to acknowledge King Fahd
University of Petroleum and Minerals and Saudi Aramco Oil
for the technical and logistics SupPO*.
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
X. REFERENCES
[I]
[6] Richardson
Handbook,
Richardson
BIOGRAPHIES
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.
Figure 4a
CP IMPACT ON HARMONIC RESONANCE
0.01200
0.00800
HI
0.00000
Elm
Figure 4b
CP IMPACT ON HARMONIC RESONANCE
is
0.02000
I
%'E
0.01500
zs:
:a =
5:
a 1
0.01000
0.00500
0.0w00
Figure 4c
CP IMPACT ON HARMONIC RESONANCE
15g
85 E
as:
0.02000
0.01500
8; 2
0.01wo
Z
IB
0.00500
az
HI
0.00000
B"*
Authorized licensed use limited to: P R Engineering College. Downloaded on May 28, 2009 at 09:58 from IEEE Xplore. Restrictions apply.