2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

Optimal Switch Placement in Distribution Power

System Using Linear Fragmented Particle Swarm
Optimization Algorithm Preprocessed by GA
S.Golestani #1, M.Tadayon*2,
#
Iran Power Plant Project Management (Mapna group)
P.O. Box 19395-6448, Tehran, Iran
1
[email protected]
*
Behrad consulting engineers
Esfahan, Iran
2
[email protected]

Abstract—In recent years, Distributed Generation (DG) has Kennedy and Eberhart [17] introduced the idea of
become an efficient alternative to traditional distribution continuous particle swarm optimizer in 1995 and many
systems, therefore new algorithms should be able to solve researchers have been developing and modifying different
problems in presence of DGs. Reliability improvement and cost versions of the PSO in different disciplines.
reduction are two important goals of utilities which are usually in The original PSO described above has been developed for
opposite of each other. The Islanding operation scheme is
solving continuous optimization problems. However, lots of
dynamically formed when fault occurs, which is based on the
location of the fault, location of switches and the actual status of practical engineering problems are formulated as discrete
distribution network operation before fault occurs. Optimal combinational optimization problems.
allocation of switches in distribution power systems can improve The discrete binary version of PSO for discrete
reliability of a power system by reducing the total time of fault optimization problems has been introduced in 2001. In
detection, isolation and restoration. In this paper, a three-state addition, researchers frequently model continuous domain
approach inspired from the discrete version of a powerful problems in binary terms, and they solve those problems in
Particle swarm optimization (PSO) algorithm is developed and discrete high dimensional spaces, featuring qualitative
presented to determine the optimum number and locations of distinctions. In recent years, PSO with fast convergence has
two types of switches (sectionalizing and breaker switches) in
opened up a big opportunity to be employed in power systems
radial distribution systems. The novelty of the proposed
algorithm is a new linear method for fragmentation particles and [15]-[16].
probability calculation of switch type offer for each candidate In [2] PSO algorithm is used for Determine the optimum
locations. Genetic preprocessing algorithm is used in proposed number and location of switches in distribution system and
algorithm for generating suitable first population. It enhances results on RBTS BUS4 test system shows this method is
the PSO algorithm and decreases time of optimization. RBTS- comparable to genetic algorithm for solving this problem.
BUS4 test system indicates advantages of the algorithm. In this paper, a three-state approach inspired from the
discrete version of a powerful PSO algorithm is developed
Keywords—Distribution power system, Particle swarm and presented to determine the optimum number and locations
optimization, Genetic algorithm, Reliability, switch placement, of two types of switches such as [2] but phasor is not used in
Distributed generation.
the proposed algorithm and a new linear method for velocities
I. INTRODUCTION and states limitation introduced.
In proposed scheme, Genetic preprocessing algorithm is
The reliability index of each service zone defined by the used for generate suitable first population in PSO and
boundary switches is derived to solve the expected energy not summarize search space for enhance main PSO algorithm
served due to fault contingency. A sectionalizing switch is a operation to escape from local minimums and decrease time
device that isolates a faulted part from the system so that the of optimization.
healthy part can still be electrically supplied and the In the proposed method, MATLAB and DIgSILENT have
interruption duration is minimized and the customer combined to produce fast and complete algorithm. In this
interruption cost is then determined according to the customer method designed network in DIgSILENT software and data
type and power consumption within the service zone. use in optimization algorithm. Relation between MATLAB
Therefore, Switch placement plays an important role in and DIgSILENT has been introduced in fig1.
automated distribution network, where the sectionalizing This method is same as [1] but in the proposed method
switches can be remotely activated. genetic algorithm is used for preprocessing and particle swarm
Achieving high-distribution reliability levels and optimization algorithm for main optimization.
concurrently minimizing capital costs can be considered as the
main issues in distribution power system optimization. To II. GENETIC ALGORITHM PREPROCESSOR
improve system reliability for radial distribution systems The simple discrete PSO is not used for switch allocation in
under fault conditions, switch placement schemes are distribution system because in the real distribution networks
proposed by various algorithms such as Genetic Algorithm reliability optimization exist a huge search space and this
[1], Simulated Annealing [7], Immune Algorithm [5], QEA algorithm is not able to reach global optimization. In the
based algorithm [9] and other heuristic methods. proposed method for clustering search space and increasing
the ability of simple PSO, Genetic algorithm by original and

978-1-61284-286-8/11/$26.00 ©2011 IEEE 537

Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.
 2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

new operators selected for preprocessing and generating progress convergence procedure. With this modification in
several desirable population to use in first PSO population. operators, generating useless and malformed chromosomes
The stop conditions intended in this method is determined will be decreased. For example if at start mutation operator
number of repeat iterations. change location of one switch on the point 5 to point 30, when
procedure find a solution near with optimal solution, dynamic
operator utmost change location of switch 2 or one point.
C. Perturbation and mutation operator
If a solution obtained by some search techniques is very
sensitive to small perturbations of its parameter values, it may
not be good to use this solution in certain situations [9]. A
similar situation is observed in switch placement in presence
of DG where displacement of one switch, changes planning
cost greatly. This paper provides the basic concept of a new
technique, which extends the application of GA to domains
that require identification of robust solutions. Small change in
one or several switches locations will be performed in
perturbation operator. It should be noted that the rate of
modifications reduces in the direction of progress
convergence procedure [1]. In usual mutation operator is used
for escape from local minimum but if this operator adapted by
progress convergence procedure, GA optimization time
decrease.

Fig 1, MATLAB and DIgSILENT combination

A. Chromosome definition
Two type of switch (sectionalizing and breakers)
considered in proposed algorithm therefore each gene consists
of three statuses. Fig 2 shows appropriate definition of
chromosomes.

Fig 3. Flowchart GA preprocessor [1]

Fig 2. Definition of chromosomes

Best convergence has happened where population size is 40
and iteration number in preprocessing is 80 and in main
In proposed chromosome definition: processing, is less than 50. With little size of population, time
1 : Suggests sectionalizing switch of the iteration is very smaller than large population.
2`: Suggests breaker switch
0 : Non switch III. PARTICLE SWARM OPTIMIZATION
B. Dynamic operators Main process in proposed algorithm performed by PSO
algorithm on clustered search space by GA.
In the immune optimization algorithm rate of solution
change has been controlled by difference between cost of each A. Continuous Version of PSO
solution and best cost in previous iterations. In the proposed The particle swarm optimization consists of a swarm of
method rate of modifications reduces in the direction of particle flying through the search space. Each particle i in the

538
Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.
 2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

swarm contains parameters for position (Xti) and velocity

(Vti), D is the dimension of the search space, t is the number
of current iteration. The position of each particle represents a
potential solution to the optimization problem. At each
iteration step, each particle adjusts its velocity based on its
momentum, the influence of its best position (Pti) tracked, and
the best position found by the neighbor (PtG) so far, then
computes a new position to examine. The PSO formulas are
given as follows [15]

Vt V t  1  c . rand .(P t  1  X t  1 )
i i 1 1 i i (1)
t  1
 c 2 . rand 2 .(PG  X i )t  1 Fig 4, Sigmoid function

X it X it 1  Vit (2)
C. Trinary Particle Swarm Optimizer
Where
In the proposed algorithm two type of switch optimized
c1 and c2 are constants known as the cognitive and social
acceleration coefficients. therefore three status 0, 1, 2 used for determine existence and
type of switch in each location. Sigmoid decision function is
not appropriate for this problem. In the proposed algorithm
B. Binary Particle Swarm Optimizer probability of each decision controlled by a decision function
Kennedy and Eberhart developed a discrete binary version such as curve shown in fig 5. for example if preprocessing
of PSO for binary problems [18]. They proposed a model state determine that 30% of candidate locations selected for
where in the probability that a particle will decide ”yes” or switch installation we can use suitable function that
”no”, or ”true” or ”false”, or will make some other binary probability of 0 is about 70%. By this method without straiten
decision, is a function of personal and social factors as search space increase the probability of proper agent
follows: generation. Fig 6 shows a sample of preprocessing result
indicating decision function.
Prob(x ik 1) f(x ik , v ik , p ik , p g
k) (3)

Prob(x ik 0) 1  Prob(x ik 1) (4)

Where
xik : Position of ith particle in kth iteration
vik : Velocity of ith particle in kth iteration
pik : tracked best position of ith particle in kth iteration
p gk : Best position found by the neighbour
Prob(x ik 1) : Probability of (x ik 1)
Fig 5, proposed decision function
The sigmoidal function shown in Equation 5, and in Figure 4,
maps the interval of to a range of [0, 1].
1
S(Vik ) (5)
1  exp(  Vik )

Where
vik : Velocity of ith particle in kth iteration

Fig 6, probability of each status based on preprocessing results

539
Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.
 2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

V. MATHEMATICAL FORMULATION
The objective function of proposed method is to minimize
the sum of interruption and investment costs for distribution
feeders. For compare results of this paper with other
researches expected outage cost (ECOST) is used as an
interruption cost reliability index that should be minimized
using a proposed algorithm. ECOST is used as the objective
function as it responds to the effects of system topology,
interruption duration, load variations, and component random
failures. It also recognizes various customer types and their
nonlinear customer damage function [5] such as[2]
NC NIL
ECOST ¦ ¦ L kj . C jk (r j ). λ j ($ /yr) (8)
j 1k 1
Where
NIL number of isolated load points due to contingency j;
NC number of contingencies;
Lkj curtailed load at load point due to contingency j ;
rj average outage time due to contingency j;
λj average failure rate of contingency j;
Fig 7, Flowchart of proposed algorithm
outage cost in ($/KWh) of load point k due to outage
Cjk(rj)
j with outage duration of rj.
IV. CONSIDERING DG Minimize [Total. Cost ECOST(x1, x 2 x 3,...,xn,
Reliability assessment in this paper performed by (9)
y1, y 2, y 3,...,ym)  nu SEC mu BRC]
DIgSILENT software therefore with this method proposed
algorithm is easy to use on other networks schemed in this Where
software and fast and reliable computation has been
performed. For accurate model Islanding operation of DGs ECOST expected interruption cost;
Proposed algorithm in [12] is used for reliability assessment
Xi ith location where a switch is installed;
considering DG in distribution power system in DIgSILENT
but only one DG was considered such as fig 8. yi ith location where a breaker is installed;
All steps of this algorithm with a little change in
n number of sectionalizing switch;
formulation of calculating indices to consider only one DG in
network. In the proposed algorithm in [12] for each event of a m number of breakers;
contingency on any section,
SEC cost associated with a sectionalizing switch;
1) Both circuit breakers at its ends should be opened
immediately. BRC cost associated with a breaker.
2) If the load demand in the island is beyond the DG
capacity, DG is disconnected from the island instead of VI. METHOD APPLICATION
shedding some loads. The proposed method is applied to a typical urban type
3) ENS is calculated in these conditions by usual methods. configuration connected to bus 4 of the RBTS [19] shown in
fig 9 on two conditions.
A. First conditions
The first condition is given from [2] therefore all customers
and network data are given from [19] and the investment costs
of a sectionalizing and a breaker-type switch are U.S. $4700
and U.S. $11800, respectively [7]. The annual maintenance
cost is 2% of the annual investment cost [2]. Such as [2] the
Fig 8. Islanding phenomenon
monetary values associated with customer damage cost have
been taken from [7] and the life period of the switches is
assumed to be 20 years with an interest rate of 8% [7].

540
Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.
 2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

Best locations for installing breaker-type switches are same as

previous conditions (A).

TABLE IV
RESULTS OF PROPOSED ALGORITHM IN PRESENCE OF DG

Total cost 159058 ($/yr)

Number of
12
Sectionalizing Switches
SEC locations 15,29,22,54,57,4,14,26,32,40,46,58

VII. CONCLUSION
Optimal allocation of switches in distribution power system
has an important effect on reliability but cost of installation
and maintenance has a conflict to that. The switch placement
problem is a combinatorial optimization problem with a
nonlinear and non-differential objective function. In this paper
a new method based on combination of genetic algorithm and
particle swarm optimization algorithm has been proposed.
Results of this methods display the ability of proposed method
in compare of previous researches.
Another ability of this method is DG effect consideration in
Fig 9. RBTS connected to BUS4
distribution power system with usual models and islanding
operation.
Result of proposed algorithm compare with previous Using of genetic algorithm for preprocessing with several
researches in same conditions has been shown in table 1 to 3. aspects including summarizing search space, decision function
estimation and producing good initial population improves the
TABLE I PSO operation.
RESULTS OF PROPOSED ALGORITHM -SECTIONALIZING SWITCHES
ACKNOWLEDGMENT
Total cost 161522 ($/yr)
Number of switch 12
This paper is sponsored by Iran Power Plant Projects
Selected sections 15,29,22,55,57,4,14,26,32,40,46,58 Management, MAPNA Group Co.
MAPNA Group is a conglomeration of a parent enterprise
TABLE II and its 33 subsidiaries engaged in development and
RESULTS OF [2] - SECTIONALIZING SWITCHES
implementation of power, oil & gas, railway transportation
and other industrial projects under EPC & IP schemes as well
Total cost 162028 ($/yr) as manufacturing relative equipment. Since its inception in
Number of switch 12
1992, MAPNA Group has been involved in more than 85
Selected sections 15,29,45,55,57,4,14,26,32,40,46,58
projects valuing over € 17 billion, among them power projects
TABLE III cover more than 52,000 MW in commissioned, under
RESULTS OF [7]
construction & under-negotiation stages, having a share of
86% of Iran’s total grid capacity.
Total cost 189033 ($/yr)
Number of switch 14
Selected sections 15,31,45,53,55,57,10,14,26,36,40,42,48,52 REFERENCES
The results obtained from applying the proposed algorithm [1] M. Tadayon and S. Golestani, “A New Method for Optimal RCS
Placement in Distribution Power System Considering DG Islanding
indicate that the best locations for installing breaker-type Impact on Reliability” IEEE Transmission & Distribution Conference &
switches are the main feeders: 1, 11, 17, 27, 37, 43, and 49, Exposition(Asia and Pacific), 2009, 26-30 Oct. 2009, Seoul
confirming good performance of the algorithm for this system. [2] A. Moradi and M. Fotuhi-Firuzabad, “Optimal Switch Placement in
Distribution Systems Using Trinary Particle Swarm Optimization
B. Second conditions Algorithm” IEEE Transactions on power delivery, VOL. 23, NO. 1,
The second condition is given from [2] but one distributed JANUARY 2008
[3] B. Liu, S. Kun, J. Zou, X. Duan, X. Zheng "Optimal Feeder Switches
generation has been considered at load point that modeled in Location Scheme for High Reliability and Least Costs in Distribution
"PV" with 2MW power generation and 1MVAR reactive System" Proceedings of the 6th World Congress on Intelligent Control
power limit on RBTSbus4 test system. In these conditions and Automation, June 21-23,2006,Dalian,China
results of proposed algorithm for breakers has been not [4] Y. Mao, K.N. Miu "Switch Placement to Improve System Reliability for
changed. In this case study, islanding operation reduces ENS Radial Distribution Systems with Distributed Generation" IEEE
Transactions on Power Systems,Vol.18,No.4,November 2003.
and changes best location of breakers and decreases total cost.

541
Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.
 2011 8th International Conference on the European Energy Market (EEM) • 25-27 May 2011 • Zagreb, Croatia

[5] C.H. Lin, C.S. Chen, H.J. Chuang, C.S. Li, M.Y. Huang, C.W. Huang
"Optimal Switching Placement for Customer Interruption Cost
Minimization" IEEE 2006
[6] P.M.S. Carvalho, L.A.F.M. Ferreira, A.J. Cerejo da Silva "A
Decomposition Approach to Optimal Remote Controlled Switch
Allocation in Distribution Systems" IEEE Transactions on Power
Delivery,Vol.20,No.2,April 2005
[7] R. Billinton, S. Jonnavithula "Optimal Switching Device Placement in
Radial Distribution Systems" IEEE Transactions on Power
Delivery,Vol.11,No.3,July 1996.
[8] A. Moradi, M. Fotuhi-Firuzabad "Optimal Switch Placement in
Distribution Systems Using Trinary Particle Swarm Optimization
Algorithm" IEEE Transactions on power delivery, Vol. 23, No. 1,
January 2008
[9] G. Chen "A Novel QEA-based Optimum Switch Placement Method for
Improving Customer Service Reliability" DRPT2008 6-9 April 2008
Nanjing China
[10] R.E. Brown "Reliability Benefits of Distributed Generation On Heavily
Loaded Feeders" IEEE 2007
[11] S. Tsutsui, A. Ghosh "Genetic Algorithms with a Robust Solution
Searching Scheme" IEEE Transactions on evolutionary computation,
Vol.1, No.2 September 1997
[12] S.X. Wang, W. Zhao, Y.Y. Chen "Distribution System Reliability
Evaluation Considering DG Impacts" DRPT2008 6-9 April 2008
Nanjing China
[13] A. Moradi, M. Fotuhi-Firuzabad, M. Rashidi-Nejad "A Reliability
Cost/Worth Approach to Determine Optimum Switching Placement in
Distribution Systems" 2005 IEEE/PES Transmission and Distribution
Conference & Exhibition, Asia and Pacific Dalian, China
[14] L. Jian, X. Jing-qiu, and Ch. Hong-li, “Algorithms on Fast Restoration
of Large Area Breakdown of Distribution Systems under Emergency
States,” Proceedings of the Csee, vol12(24), pp132~139, 2004.
[15] L. Zhiming, W. Cheng, L. Jian “Solving Constrained Optimization via a
Modified Genetic Particle Swarm Optimization” 2008 Workshop on
Knowledge Discovery and Data Mining
[16] J. Sadri and Ch.Y. Suen, “A Genetic Binary Particle Swarm
Optimization Model” 2006 IEEE Congress on Evolutionary
Computation, July 16-21, 2006, Canada
[17] J. Kennedy and R. Eberhart, “Particle swarm optimization,” in Proc.
IEEE Int. Conf. Neural Networks, Perth, Australia, 1995, vol. IV, pp.
1942–1948.
[18] B. E. Wells, C. Patrick, L. Trevino, J. Weir, and J. Steinca, “Applying
particle swarm optimization to a discrete variable problem on an FPGA-
based architecture,” presented at the MAPLD Int. Conf., Washington,
D.C., 2005.
[19] R. N. Billinton and I. Sjarief, “A reliability test system for educational
purposes-basic distribution system data and results,” IEEE Trans. Power
Syst., vol. 6, no. 2, pp. 813–820, May 1991.

542
Authorized licensed use limited to: Sharda University. Downloaded on December 20,2024 at 10:07:07 UTC from IEEE Xplore. Restrictions apply.