Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India

(Paper ID: 190)

POWER NETWORK RECONFIGURATION FOR

CONGESTION MANAGEMENT AND LOSS MINIMIZATION
USING GENETIC ALGORITHM

Sananda Pal*, Sawan Sen**, Samarjit Sengupta*

*
University of Calcutta, Kolkata, India
** Kalyani Govt. Engineering College, Kalyani, India
1
[email protected], [email protected], [email protected]

Keywords: Congestion Management, Distribution Loss, profile, manage load congestion and enhance system
Genetic Algorithm, Network Reconfiguration reliability. Thus the aim of distribution network
reconfiguration is to find a radial operating configuration that
optimizes certain objectives while satisfying all the
Abstract operational constraints without islanding of any bus. The
greater the number of switches, the greater is the possibilities
for reconfiguration and the better are the effects. However,
Network reconfiguration is referred as operational schemes to
reconfiguration may cause congestion in certain areas, which
alter the network topology by closing and opening the
is a redundant situation and needs proper management.
sectionalizing and tie switches of power distribution system
and thus it allows to control the power flow from substation to
The concept of reconfiguring a network for loss reduction was
power consumers with additional benefits such as load
first introduced by Merlin and Back [1] proposing a branch
balancing, real power loss reducing, optimizing the load
and bound method to search for minimum loss condition of a
sharing between parallel circuits by directing power flow
power network. Reconfiguration itself is a non-linear
along contractual paths. Restructuring of the power network,
optimization problem. Different classical methods like Gauss-
however, may increase the tendency of overloading and thus
Siedel method, Newton Raphson method, Lambda iteration
congestion in certain areas. This congestion may lead to
method are used to optimize linear, continuous, and
violation of system voltage or transmission capacity limits and
differentiable objective functions. To optimize non-linear
threaten the power network security and reliability. On this
multi-objective functions, different evolutionary algorithms
view point, this paper presents a Genetic Algorithm (GA)
came into existence.
based new reconfiguration algorithm of the network which
will able to identify the most congested area of power network
A new method using Artificial Bee Colony algorithm (ABC)
and fabricate the least loss condition after alleviating overload
has been proposed for determining the sectionalizing switch to
and overvoltage as well as ensuring efficient network
be operated in order to solve the distribution system loss
operation. To establish the efficacy of the proposed algorithm,
minimization problem in [2]. Again, as an alternative
case studies have been carried out on modified IEEE-14 bus
approach, loss reduction by both capacitor placement and
and IEEE-30 bus power network and the results are found to
reconfiguration has been presented by L.Yang and Z. Guo [3].
be encouraging.
Literature review reveals that, the evolutionary algorithms are
random, stochastic, and robust which are useful for
optimization of non-linear problems. The application of
1 Introduction Genetic Algorithm for optimizing a reconfiguration problem
was proposed by K. Nara, A. Shiose, M. Kitagawa, and T.
Distribution networks are generally structured in a mesh but Ishihara [4]. Other metaheuristics methods like Particle
operated in radial configuration for effective co-ordination of Swarm Optimization, Simulated Annealing, Tabu Search,
their protective schemes and to reduce the fault level. The Plant growth Simulation have been applied to solve different
reconfiguration of a distribution system is a process that alters optimization problems [5-8].
the feeder topological structure by managing the open/close
status of sectionalizing switches, which are normally closed, Congestion management is also a complicated non-linear
and tie-switches, which are normally open. Reconfiguration of problem especially in the deregulated power system. An
radial distribution system is a very effective and efficient efficient method to alleviate overload was presented in [9].
means to reduce distribution network losses, improve voltage Other significant congestion management schemes suitable for

291
 Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India
(Paper ID: 190)
various electricity market structures have been reported in buses 10-14) and IEEE-30 bus system have been used as test
literature [10], while congestion management ensuring voltage bed of the proposed algorithm.
stability was reported in [11]. Highly optimized numerical
algorithms have been developed using soft computing
2.2 Main Objective
techniques to provide a better solution to the congestion
management problem by substation reconfiguration in [12].
But literature survey [13] shows that very modest of these The principle objective of distribution network reconfiguration
techniques do consider the network losses and hence network is to find a radial operating structure that minimizes the
efficiency while dealing with congestion management system real power loss while alleviating congestion. If x
problems. Thus, reconfiguration for loss reduction along with denotes the variables, that is, operating state of the switches of
proper management of congestions has the scope of plenty of a distribution system and c represents the present voltage
research and improvement. magnitude, angle and other states of the system, then f(x, c)
may be considered as the objective function which gives a
Reconfiguration along with congestion management is a non-
measure of the real power loss of operating the system in the
linear complex combinatorial optimization problem with
multiple operational constraints, which must not be violated current state c. The corresponding state c must be consistent
while finding an optimal or near optimal solution. It also has with Kirchhoff’s current and voltage laws to satisfy the
an additional dimension of social welfare. The problem of electrical constraints and other operational constraints of the
finding the optimal topological configuration of a power system by not exceeding the physical limitations of any of the
transmission system was considered with the aim of providing system components.
system operators with a tool suited for congestion
management in [14], whereas congestion management by Thus the objective is to find a network configuration which
reconfiguration with the help of micro-DG placement was minimizes f(x, c) while satisfying all the constraints. This
presented in [15]. problem can be given in a very general formulation as a single
objective function with inequality constraints as:
In this paper, a GA based approach has been proposed as a
powerful tool for achieving global optimum results with all
operational constraints. The operating conditions that allow Minimize f(x, c) (1)
alleviating overloads and overvoltage with optimum system [ȯ6
parameters were determined with smart definition of
chromosomes. The proposed algorithm has been successfully such that: G(xu) d0 (2)
applied for finding optimal topological configuration of power
network with least real power loss as well as identification of where S is the set of all possible configurations and any
congestion and removal of the same. The objective function solution x satisfying non-linear constraints represented by (2) is
and constraints were dealt separately and the proposed
one of the best possible configuration. If the total number of tie
algorithm has been validated on modified IEEE-14 bus and
and sectionalizing switches in the system is ns, then the present
IEEE-30 bus system.
operating state of the switches is represented as a state vector x
= [x1, x2,…, xns] where individual switch states xi ȯ^`”L
2 Distribution Network Reconfiguration with ”Qs,(where xi = 1 indicates that switch i is closed, and xi = 0
Assessment of Loss and Identification of indicates that it is open). In the present problem f(x, c) is the
Congestion summation of active power losses in each line, then,
n
f ( x, c ) ¦P
i 1
i
i=1, 2….n (3)

2.1 Understanding the problem where, Pi is the real power loss in ith line

The network reconfiguration has been performed by changing The inequality constraints are as follows:
the operating condition of the switches in transmission lines
connecting different buses. The problem has been formulated Vi min ” Vi ” Vimax (4)
by considering the open and closed conditions of the switches
which are represented by zeroes (0s) and ones (1s) Ii min ” Ii ”Iimax (5)
respectively, and that all the lines had a switch. The main
objective of the proposed algorithm is to find the least real Also the fact that [ȯ6 justifies that all the configurations are
power loss condition without any congestion through the not possible solution as one of the most important constraints
process of network reconfiguration. The modified IEEE- 14 is that all bus should receive power and none should be
bus system (adding lines in between buses 1-6 and between islanded.

292
 Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India
(Paper ID: 190)

3 Problem Formulation with Genetic Algorithm best among all congestion-free states was considered as the
optimal solution. Figure 1 shows the flow diagram of the
proposed program. The over-current or overvoltage
3.1 Application of Genetic Algorithm constraints could not be directly used as constraint equations
in the problem as the population are bit-string type, that is,
GA is a global adaptive search technique based on the variables can be 0 or 1 only. Bit-sting type population ignores
mechanics of natural genetics. It is used to optimize various all the constraint equations.
problems by using biological evolution based methods like
crossover, mutation, and selection for reproduction and theory
of survival of the fittest. To solve a specific problem with GA,
a function called objective function is to be constructed which
allows different possible solutions to be evaluated. The
algorithm will then take those solutions, which seem to
estimate the best-fit solution. Problem encoding is done first
by defining the control variables. Each control variable is
called a gene, while all control variables integrated into one
vector is called a chromosome. The GA always deals with a
set of chromosomes called a population. A next generation or
new population set is formed by altering chromosomes of the
existing population. It needs three genetic operators: selection,
crossover, and mutation for this purpose.

In addition to these operators there are some parameters of the

GA, population size being an important one. The population
size consists of a number of chromosomes. If there are not a
lot of chromosomes, the GA only has a few possibilities to
perform crossover. Thus only a small part of the search space
is explored; and if there are many chromosomes in a
population set, the GA will slow down. It has been proven that
after some limit by increasing the population size, the GA will
not find the optimal solution faster. The population size
depends on the type of encoding and the problem that needs to
be solved. Figure 1. Flow diagram of GA based proposed algorithm

Figure 2 shows the modified IEEE-14 bus used to test the

In the present problem, the operating conditions of the
program along with the switching arrangements. It shows a
switches are considered as control variables. The values of
switch in each line in between the buses.
these variables are either zero or one. A set of initial
population has been generated with population size twenty.
Each chromosome was evaluated through power flow by
Newton-Raphson method to generate the distribution system
real power loss.

First, GA was applied to the problem to find out the condition

of the switches, which gives the minimum real power loss
provided all the buses receive power. The set of variables,
which creates this least loss condition, is considered one of the
best-fit states of the switches. This particular condition was
taken as an apparently best condition. A checking was then
done to find out whether any bus had an overvoltage or any
line had crossed its current limit. If there was no such
condition, the apparently best condition was taken as one of
the most eligible condition. Or else GA was again called
ignoring the apparent best, to get the congestion free least real
power loss condition. After all such corrections, the GA was
again called to find next best condition without any Figure 2. Modified IEEE-14 bus system with switching
congestion. After the end of the number of set iterations, the arrangement

293
 Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India
(Paper ID: 190)
Each chromosome had to go through load flow which was Best: 9.98752 Mean: 9.99717
20
done by N-R method to yield real power loss due to power Best fitness

flow. The power flow equations from [16] thus used were: Mean fitness

Fitness value
15
n
Pi ¦V
k 1
i V j yik cos(G k  G i  J ik ) fip (6) 10

5
0 20 40 60 80 100 120 140 160 180 200
Generation
n Current Best Individual

¦ Vi V j yik sin(G k  G i  J ik )
1
Qi fiq (7)

Current best individual

k 1
0.5

The line flows are calculated by

0
0 5 10 15 20
Number of variables (22)

I ij I s  I pi (Vˆi  Vˆj ) ys  Vˆi y pi (8)

Figure 3. Fitness function representing real power loss and
switching conditions for modified IEEE-14 bus system.
Line current Iij at bus i is defined positive in the direction iÆj. The analysis of the above case has been shown in Table 1. The
Sij and Sji being the complex powers into the line at bus i and developed program is used to find the best achievable
j, respectively. condition with least real power loss. The voltage profile of the
buses and the line currents are shown in Table 2 and Table 3
respectively.
Vˆi Iˆij* Vˆi (Vˆi *  Vˆj* ) ys*  Vi y*pi
2
Sij Pij  jQij (9)
Table 1. Case Study for Modified IEEE-14 Bus System
2
S ji Pji  jQiji Vˆj Iˆ*ji Vˆj (Vˆj*  Vˆi * ) ys*  V j y*pi (10) Case
Open From To Loss in
Remarks
Switch Bus Bus MW
Initial network
No
with all
None - - 10.108 overvoltage or
The power loss in line (i--j) is the algebraic sum of power switches
over-current
closed
flows determined from (9) and (10). Optimum
sw17 9 10
network after No
Therefore, SLij = Sij + Sji (11) application of 9.9875 overvoltage or
the developed sw18 9 14 over-current
program
4 Simulation and Results
Table 2. Voltage Profile of Modified IEEE-14 Bus System
The program was tested with modified IEEE-14 bus as well as after Reconfiguration
IEEE-30 bus system. The results obtained are given as separate
cases.
Voltage Voltage
Case I: Verification of proposed algorithm with modified Bus Bus
no. no.
IEEE 14 bus system
Magnitude Angle Magnitude Angle
(pu) (degree) (pu) (degree)
As the foremost step, the active power loss was calculated
with all the closed switches of the network. The program with 1 1.0600 0.0000 8 1.0900 -9.3165
GA was then called with randomly generated population or
generation set. The population size was taken as twenty and 2 1.0450 -3.5369 9 1.0395 -10.3958

two-point crossover and uniform mutation was done. Figure 3 3 1.0100 -10.4252 10 0.9927 -7.3105
show the condition with least loss and with no congestion due 4 1.0205 -7.2722 11 1.0166 -6.1700
to overvoltage or over-current. 5 1.0242 -5.5759 12 1.0323 -5.8263
6 1.0500 -4.8092 13 1.0240 -5.9811
7 1.0549 -9.3165 14 0.9903 -7.6628

294
 Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India
(Paper ID: 190)
Table 3. Line Currents in Modified IEEE-14 Bus System after
Reconfiguration
Line Current Line Current
From To To
From
Bus Bus Angle Bus Angle
pu Bus pu
(degree) (degree)

1 2 1.0708 0.0588 5 6 0.1246 1.9741

1 6 0.6286 0.0331 6 11 0.1880 -0.5871

1 5 0.4684 -0.1499 6 12 0.0902 -0.4092
2 3 0.6306 -0.1646 6 13 0.2297 -0.5110
2 4 0.3859 -0.1255 7 8 0.1994 1.4082
2 5 0.2310 -0.2767 7 9 0.2273 -0.8340

3 4 0.3094 -3.1076 10 11 0.1493 2.5533

4 5 0.6904 -3.0706 10 14 0.0438 -0.3098

4 7 0.2471 0.6041 12 13 0.0294 -0.5154

4 9 0.1100 0.1727 13 14 0.1156 -0.5132
Figure 5. IEEE-30 bus system with switching arragement.

Case II: Verification of proposed algorithm with IEEE 30 The analysis of the above conditions are shown in Table 4.
bus system The voltage profile of the buses and the line currents are
shown in Table 5 and Table 6 respectively.
Initially, the condition with one of the switches closed was
considered and loss calculated. It was found that over-current Table 4. Case Study for IEEE-30 Bus System
was present between buses 1-2. The program with GA was
then called with randomly generated population. The same Open From To Loss in
Case Remarks
options for crossover and mutation were set as in the previous Switches Bus Bus MW
case. Figure 4 show the condition with least real power loss Initial
Over-current
between
and after removal of the over-current congestion. Though the network sw33 24 25 17.4129
Bus 1-2 amounting
loss amount had increased from the first condition, the over- considered
to 1.3429 per unit
current is removed and it may be thus called one of the best Optimum
sw29 21 23
solutions. Figure 5 show the IEEE-30 bus system used along network after
No overvoltage or
with its switching arrangements. application of 17.5420
over-current
the developed sw32 23 24
Best: 17.542 Mean: 18.3146
30 program
Best fitness
Mean fitness
Table 5. Voltage Profile of IEEE-30 Bus System after
Fitness value

25
Reconfiguration
20
Bus Voltage Bus Voltage
no. Magnitude Angle no. Magnitude Angle
15
0 20 40 60 80 100 120 140 160 180 200 (pu) (degree) (pu) (degree)
Generation 1 0.0000 -15.5130
1.0600 16 1.0447
Current Best Individual
1 2 1.0430 -5.3486 17 1.0384 -15.8826
Current best individual

3 1.0217 -7.5430 18 1.0292 -16.4854

4 1.0129 -9.2966 19 1.0258 -16.6917
0.5 5 1.0100 -14.1586 20 1.0294 -16.5122
6 1.0120 -11.0936 21 1.0292 -16.2370
7 1.0034 -12.8785 22 1.0326 -16.1699
0 8 1.0100 -11.8133 23 1.0341 -16.0775
0 5 10 15 20 25 30 35 40
9 1.0505 -14.1454 24 1.0179 -16.6301
Number of variables (41)
10 1.0433 -15.7456 25 1.0166 -16. 2021
11 1.0820 -14.1454 26 0.9989 -16.6224
Figure 4. Fitness function representing real power loss and 12 1.0583 -14.8737 27 1.0243 -15.6752
switching conditions for IEEE-30 bus system. 13 1.0710 -14.8737 28 1.0105 -11.7581
14 1.0441 -15.7378 29 1.0045 -16.9025
15 1.0404 -15.8185 30 0.9931 -17.7834

295
 Michael Faraday IET International Summit: MFIIS-2015, September 12 – 13, 2015, Kolkata, India
(Paper ID: 190)
Loss Reduction Using Artificial Bee Colony Algorithm”,
International Science Index 21, 2008.
Table 6. Line Currents in IEEE-30 Bus System after [3] Lin Yang and Zhizhong Guo: “Comprehensive
Reconfiguration Optimization for Energy Loss Reduction in Distribution
Networks”,IEEE 2008.
From Line Current From Line Current [4] K. Nara, A. Shiose, M. Kitagawa, and T. Ishihara:
To To
Bus
Bus Angle Bus
Bus Angle “Implementation of genetic algorithm for distribution
pu pu system loss minimum reconfiguration,” IEEE Trans.
(degree) (degree)
Power Syst., vol. 7, no. 3, pp. 1044–1051, Aug. 1992.
1 2 1.1727 0.1042 16 17 0.0441 -0.6238
[5] Wiley, Chichester, UK, 2001. Young-Jae Jeon, Jae-Chul
1 3 0.8300 -0.0707 15 18 0.0674 -0.5711 Kim, Jin-O. Kim, Joong-Rin Shin, and Kwang Y. Lee:
2 4 0.4207 -0.2113 18 19 0.0351 -0.5796 “An Efficient Simulated Annealing Algorithm for
3 4 0.8058 -0.0855 19 20 0.0634 2.4759 Network Reconfiguration in Large-Scale Distribution
2 5 0.7902 -0.1423 10 20 0.0858 -0.6475 Systems”, IEEE Transactions on Power Delivery, Vol. 17,
2 6 0.5793 -0.1172 10 17 0.0609 -1.0124 No. 4, October 2002.
4 6 0.7375 0.0709 10 21 0.2019 -0.8527 [6] Wu-Chang Wu and Men-Shen Tsai: “Feeder
5 7 0.1878 -2.7153 10 22 0.0789 -0.7736 Reconfiguration Using Binary Coding Particle Swarm
6 7 0.3774 -0.1613 15 23 0.0346 -0.7443 Optimization”, International Journal of Control,
6 8 0.2943 -0.0779 22 24 0.0789 -0.7736 Automation, and Systems, vol. 6, no. 4,pp. 488-494,
6 9 0.3296 0.3907 24 25 0.0203 -2.7316 August 2008.
6 10 0.1653 0.1244 25 26 0.0419 -0.8715 [7] Dong Zhang, Zhengcai Fu, Liuchun Zhang: “An
9 11 0.1516 1.3239 25 27 0.0516 2.6580 improved TS algorithm for loss-minimum reconfiguration
9 10 0.2737 -0.5017 28 27 0.1850 -0.0432 in large-scale distribution systems”, Electric Power
4 12 0.4630 0.2111 27 29 0.0626 -0.5369 Systems Research, Volume 77, Issues 5–6, April 2007,
12 13 0.0908 1.3112 27 30 0.0711 -0.5038 Pages 685–694.
12 14 0.0748 -0.5475 29 30 0.0374 -0.4571 [8] P. V. V. Rama Rao and S. Sivanagaraju: “Radial
12 15 0.1704 -0.6011 8 28 0.0052 2.7653 Distribution Network Reconfiguration for Loss Reduction
12 16 0.0816 -0.6800 6 28 0.1899 -0.0513
and Load Balancing using Plant Growth Simulation
Algorithm”, International Journal on Electrical
14 15 0.0136 -0.6392 Engineering and Informatics - Volume 2, Number 4,
2010.
5 Conclusions [9] Sujatha Balaraman, N.Kamaraj: “Congestion
Management in Deregulated Power System using Real
Coded Genetic Algorithm”, International Journal of
In this paper, a GA based algorithm has been proposed to Engineering Science and Technology Vol. 2(11), 2010,
minimize real power loss of a power distribution network 6681-6690.
through suitable reconfiguration along with alleviation of [10] Kumar. A., Srivastava. S. C., Singh. S. N.: “Congestion
congestions. The problem is formulated as a non-linear management in competitive power market: A
Bibliographical survey”, Elect. Power Syst. Res., 2005,
optimization problem with an objective function of 76, pp. 153–164.
minimizing system losses subjected to certain constraints. The [11] Conejo. A. J., Milano. F., Bertrand. R. G.: “Congestion
proposed algorithm has been executed through MATLAB management ensuring voltage stability”, IEEE Trans.
software and the test results have been presented, which shows Power Syst., 2006, 21 (1), pp. 357–364.
that using the GA method; the feeder reconfiguration problem [12] Nasrolahpour, E., Ghasemi, Hassan , Khanabadi, M:
can be solved efficiently for loss reduction as well as “Optimal Transmission Congestion Management by
Means of Substation Reconfiguration”, Electrical
congestion management. Engineering (ICEE), 20th Iranian Conference, pp. 416 –
421, 2012.
Acknowledgement [13] Abhishek Saxena, Seema N. Pandey, Laxmi Srivastava:
“Congestion management in Open Access: A Review”,
The authors acknowledge the usage of the laboratories International Journal of Science, Engineering and
developed under TEQIP-PHASE-I, UCT-CU in the Technology Research (IJSETR) Volume 2, Issue 4, April
Department of Applied Physics, University of Calcutta and 2013
financial assistance in the form of Junior Research Fellowship [14] G. Granelli , M. Montagna, F. Zanellini, P. Bresesti, R.
to the first author from TEQIP-PHASE-II, UCT-CU. Vailati, M. Innorta: “Optimal network reconfiguration for
congestion management by deterministic and genetic
algorithms” Electric Power Systems Research,Volume 76,
References Issues 6–7, April 2006, Pages 549–556.
[15] Shariatkhah, M.H., Haghifam, M.R.: “Using feeder
[1] A.Merlin and H.Back: “Search for a minimum loss reconfiguration for congestion management of smart
operating spanning tree configuration for urban power distribution network with high DG penetration”
distribution System,” in Proc. 5thPower Syst. Integration of Renewables into the Distribution Grid,
Computation Conf. (PSCC), Cambridge, U.K., 1975, CIRED 2012 Workshop , May, 2012.
Paper 1.2/6. [16] Badrul H. Chowdhury: “Load-Flow Analysis in Power
[2] R. Srinivasa Rao, S.V.L. Narasimham, M. Ramalingaraju: Systems”, Sec. 11, Handbook of Electric Power
“Optimization of Distribution Network Configuration for Calculations.

296