A genetic based algorithm for loss reduction

in distribution systems

A. Moussa a, M. El-Gammal a, E.N. Abdallah a and A.I. Attia b

a Elect. Eng. Dept., Faculty of Eng., Alexandria University, Alexandria, Egypt
b Research & Energy Conservation Sector, Alexandria Elect. Distrib. Co., Alexandria, Egypt

The distribution system is considered not only as one of the important parts of the electric
power system but one of the most complicated systems created by the mankind. It
constitutes the link between electricity utilities and consumers. Usually, it suffers from
unbalanced feeder structures and unbalanced loading which affects system power quality
and electricity price. This paper introduces a Genetic based Algorithm (G.A) to determine
the states of the switches for minimum loss configuration. The proposed G.A is applied for
multi-objective programming to solve the reconfiguration problem in distribution systems.
The problem of feeder configuration can be looked upon as an optimization problem, where
the objective function reflects the different goals that the individual utilities may pursue. A
weighed multi-objective function is used in this paper to investigate the utilities interests.
The algorithm can be directed to minimize the losses which are a major sign of better power
quality. Also, the operator has the ability to direct it to minimize the active power loss,
voltage drop, complex power and neutral current of the main transformer. A radial distribu-
tion system is used to demonstrate the capability of the proposed G.A.
 " "#$ %"&  '   !
               
 /" ."( 01  +   %#$  %  +   ,  (&  -   .  .() 
" "  67 8  % 9) .( .  4 5 #  .2()   3 #$  )
%"= @"7 .   += .( +>(    +& +  ?' !7 +( 2#  (8($  / 7 )
."  )) +   +      '  !  -  + +( ?'       )
%"(' A" 8 , 8(0 B  -     4) 2#  =  . #$ - % )  8 )  /  +  
.8(  87 6C  &    .( D    4 ?'  =  ./ .( +  #$ %&

Keywords: Distribution network, Reconfiguration, Multi-objective function, Loss reduction,

Genetic algorithm

1. Introduction During normal operating conditions,

networks are reconfigured for two purposes:
Distribution systems are the networks that (a) loss reduction to reduce overall system
transport the electric energy from bulk power loss.
substations or sources to many services or (b) load balancing to relieve network overloads.
loads. In most cases distribution system is Many techniques have been proposed for
radially structured because it has some solving feeder reconfiguration problem
advantages over meshed network, such that through switching operation. For example,
lower short circuit and simpler switching and Goswami et al. [1] presented a heuristic algo-
protecting equipment. On the other hand, it rithm utilizing the concept of optimal flow pat-
provides lower reliability. Generally, network tern for the minimum loss configuration of
reconfiguration is needed to provide service to distribution feeders. Jin-Cheng et al. [2] pro-
as many consumers as possible following fault posed a solution algorithm, based on a loss re-
condition, or during planned outages for duction formula and a line flow updating for-
maintenance purposes, reduce system losses mula for the network reconfiguration problem.
and balance the loads to avoid overloading of In [3], the developed algorithm is based on
network elements. partitioning the distribution network into
groups of load buses, such that the line sec-

Alexandria Engineering Journal, Vol. 43 (2004), No. 6, 729-735 729

© Faculty of Engineering Alexandria University, Egypt.
 A. Moussa et al. / Loss reduction

tion losses between the groups of nodes are Min.

minimized. M.S. Kandil et al. [4] presented an
approach based on heuristic search strategies m
to determine the switching actions for mini- TS u = ∑ S uj , (1)
j =1
mum loss configuration and/or transformers
load balancing. The authors of [5] proposed a
network reconfiguration algorithm based on where; m is the total number of feeder
branch exchange for load balancing. S.I. segments of the object feeder and S uj is the
Mohamed et al. [6] used artificial neural net- unbalance of the three complex powers of
work (ANN) to reconfigure the feeder that re- phases.
duces the active power losses. Feeder recon-
figuration through switching operation is a
1 2
complicated combinatorial optimization prob- S uj = p
∑ S j −S jo . (2)
lem. Genetic algorithms have recently been 3 p =a,b,c
used to solve many difficult engineering prob-
lems and are particularly effective for com-
binatorial optimization problems with large In which, the S jo stands for an ideal per phase
and complex search spaces. In this paper, a loading , and can be determined through the
G.A is presented for multi-objective pro- following relation:
gramming to solve the reconfiguration prob-
lem. Five objectives are considered in conjunc-
tion with network constraints.
(
S oj = S aj + S bj + S cj / 3. ) (3)
The G.A is basically a stochastic searching
algorithm. It is capable of solving non- and Sjp is the complex power of individual
smooth, non-continuous and non- differenti- loading phase p, p= a, b, c.
able problems for parallel computation to find The S uj = 0 means the complex power of
global or near global optimal solutions. The re-
the j th feeder segment is perfectly balanced.
sults of the case studies demonstrate the
b) Minimize the total line losses:
effectiveness of the solution algorithm and
Min
proved that the G.A is suitable to solve this
kind of problems.
m
TL l = ∑ ∑ (I jp ) 2 .r jp + (I ne 2 ne , (4)
j ) .r j
2. Problem formulation j =1 p =a, b, c

Distribution feeders contain a number of where:

switches that are normally closed and others
that are normally open. Under normal operat- I p and r p are the current and resistance of
j j
ing conditions, distribution engineers periodi-
cally reconfigure distribution feeders by phase p of the j th feeder segment, respec-
opening and closing of switches in order to tively.
increase networks reliability and/or reduce The last term is included to count for the
line losses. In this section, the feeder recon- line losses in the neutral wire of the j th
figuration problem is formulated as a multi- branch segment.
objective optimization problem, which can be c) Minimize the average voltage drop:
solved efficiently using G.A [7]. Five different Min.
objective functions are considered. One or
more of these objectives are considered by the 1 n
distribution systems planning engineers.
AV d = ∑VDk , (5)
n k =1

2.1. Objective functions

in which, n is the total number of load points
a) Minimize the total complex power unbal- of the object feeder, and
ance:

730 Alexandria Engineering Journal, Vol. 43, No. 6, November 2004

 A. Moussa et al. / Loss reduction

Vnomi −V kp and negative-sequence voltage unbalance fac-

1
VD k = ∑ ×100% , (6) tor at load point k, respectively.
3 p =a ,b ,c V nomi
3. Multi-objective optimization problem
in which Vnomi represents the nominal phase
voltage, Vkp denotes the magnitude of the Combining eqs. (1) through (10), a multi-
phase voltage of phase p at load point k, and objective function is obtained. In this paper a
VDk denotes the average of three phases weighed multi-objective function is used by
voltage drops at load point k. including weighting factor for each term of the
d) Minimize the neutral current of the main objective function. The multi-objective opti-
transformer: mization problem will be as follows:
Min. Min.

F=w1 TSu +w2 TLl +w3 AVd +w4 ItN+ w5 Tdt , (11)
I tN = ∑I tP . (7)
P =a ,b ,c subject to:

Where; Itp represents the current of phase p of VD k < VD s k = 1,...., n 

the main transformer that feeds the object 
feeder, and ItN is the neutral current of the s 
d o,k < d o k = 1,...., n 
main transformer. 
e) Minimize the total voltage unbalance factor: 
s , (12)
Min.
d 2,k < d 2 k = 1,...., n 

Tdt = Td0 +Td2. (8) 

N
It < It N s 
Where:

1 n and
Td0 = ∑ ( d 0 ,k ) 2 , (9)
n k =1
w1+w2+w3+w4+w5=1. (13)
and In which VDS, d0s and d2s are the specified
values for voltage drop, zero-and negative-
Vk( 0 ) sequence voltage unbalance factors for all load
d0 ,k = .
Vk( 1) points, respectively, I tNS is the specified neu-
tral current for the main transformer at the
distribution substation and wk denotes a
1 n
Td 2 = ∑ ( d 2 ,k ) 2 , weighting factor that can be adjusted by dis-
n k =1 tribution engineers for their request.

and 4. Solution algorithm for feeder

reconfiguration
V k( 2 )
d 2 ,k = . (10) The selection of an optimum configuration
V k( 1) among discrete numerous switching options
requires solution of a complicated combinato-
Vk(0), Vk(1), and Vk(2) denote the zero-, positive- rial optimization problem. G.A has recently
and negative- sequence voltage at load point k, proved as an effective tool for solving this type
respectively, and do, k and d2,k denote the zero- of problems with large and complex search

Alexandria Engineering Journal, Vol. 43, No. 6, November 2004 731

 A. Moussa et al. / Loss reduction

spaces. The search of any G.A starts with a Each string is divided into a number of sub
random generation of a population of strings. strings equals the number of the problem
variables. Each sub string consists of a
number of genes to present one of the
variables in a certain coding system. Fig. 1
depicts the flow chart of the proposed G.A
approach.

5. Application

To show the validity, and efficiency of the

proposed G.A, it is tested on the distribution
system shown in fig. 2. This system includes
one substation, four feeders, 32 branches, 5
tie lines, 32 buses and 37 switches [8]. The
system data are illustrated in tables 1 and 2.
The solution space contains 237 possible
combinations.
The proposed method can obtain the
optimal solution rapidly and accurately.
We consider two different three-phase
unbalanced load patterns to demonstrate the
ability and efficiency of the proposed solution
algorithm. The first load pattern in phases A,
B and C is 0.4, 0.4 and 0.2 p.u respectively,
while the second load pattern is 0.1, 0.2 and
0.7 respectively. For each load pattern five
cases are considered according to the selecting
weighting factors.
Case 1: w2 =1, which means that, the solution
algorithm is directed only to minimize the total
active power losses.
Case 2: all weighting factors are set to 0.2.
Case 3: w2 = 0.6, while the remainder weight-
ing factors are set to 0.1.
Case 4: w1, w2 and w3 are all set to 0.1 while
w4 and w5 are set to 0.35.

Fig. 1. The flow chart of the proposed G.A. approach.

Fig. 2. Network structure of the test system.

732 Alexandria Engineering Journal, Vol. 43, No. 6, November 2004

 A. Moussa et al. / Loss reduction

Case 5: w1, w4 and w5 are all set to 0.3, while Table 2

Line data
w2 and w3 are set to 0.05.
The optimal decisions to reconfigure the
Line From To R(Ohm) X(Ohm)
distribution system under study for the first
and second load pattern are given in tables 3 1 0 1 0.0922 0.0470
and 4, respectively. 2 1 2 0.4930 0.2511
From the results of tables 3 and 4 for the 3 2 3 0.3660 0.1864
suggested load patterns, it can be noted that 4 3 4 0.3811 .1941
the G.A is a very effective tool when studying 5 4 5 0.8190 0.7070
Table 1 6 5 6 0.1872 0.6188
Load data 7 6 7 0.7114 0.2351
8 7 8 1.0300 0.7400
Bus P (kw) Q(kvar)
9 8 9 1.0440 0.7400
1 -100 -60
10 9 10 0.1966 0.0650
2 -90 -40
11 10 11 0.3744 0.1238
3 -120 -80
12 11 12 1.4680 1.0550
4 -60 -30
13 12 13 0.5416 0.7129
5 -60 -20
14 13 14 0.5910 0.5260
6 -200 -100
15 14 15 0.7463 0.5450
7 -200 -100
16 15 16 1.2890 1.7210
8 -60 -20
17 16 17 0.7320 0.5740
9 -60 -20
18 1 18 0.1640 0.1565
10 -45 -30
19 18 19 1.5042 1.3554
11 -60 -35
20 19 20 0.4095 0.4784
12 -60 -35
21 20 21 0.7089 0.9373
13 -120 -80
22 2 22 0.4512 0.3083
14 -60 -10
23 22 23 0.8980 0.7091
15 -60 -20
24 23 24 0.8960 0.7011
16 -60 -20
25 5 25 0.2030 0.1034
17 -90 -40
26 25 26 0.2842 0.1447
18 -90 -40
27 26 27 1.0590 0.9337
19 -90 -40
28 27 28 0.8042 0.7006
20 -90 -40
29 28 29 0.5075 0.2585
21 -90 -40
30 29 30 0.9744 0.9630
22 -90 -40
31 30 31 0.3105 0.3619
23 -420 -200
32 31 32 0.3410 0.5320
24 -420 -200
33* 7 20 2.0000 2.0000
25 -60 -25
34* 8 14 2.0000 2.0000
26 -60 -25
35* 11 21 2.0000 0.5000
27 -60 -20
36* 17 32 0.5000 0.5000
28 -120 -70
37* 24 28 0.5000 0.5000
29 -200 -600 *Normally open branches
30 -150 -70
31 -210 -100 multi-objective functions. The results in table
32 -60 -40 4 demonstrate the flexibility and capability of
the G.A in solving the feeder reconfiguration
problem.

Alexandria Engineering Journal, Vol. 43, No. 6, November 2004 733

 A. Moussa et al. / Loss reduction

Table 3
Genetic algorithm results with load pattern 0.4, 0.4, and 0.2

Case no. Case 1 Case 2 Case 3 Case 4 Case 5

Items
Switch off [7 9 14 28 32] [7 9 14 28 32] [7 9 14 28 32] [7 9 14 28 32] [7 10 14 28 32]
Switch on [33 34 35 36 37] [33 34 35 36 37] [33 34 35 36 37] [33 34 35 36 37] [33 34 35 36 37]
TSu (kva) 2252.682 2252.862 2252.862 2252.862 2247.555
TLL ( kw) 158.121 158.121 158.121 158.121 158.856
AVd (%) 2.761 2.761 2.761 2.761 2.765
INt (Amp) 126.607 126.607 126.607 126.607 126.61
Tdt (%) 2.011 2.011 2.011 2.011 2.014
Fitness* 0.6711 0.6837 0.6774 0.7029 0.725
(p.u)

Table 4
Genetic algorithm results with load pattern 0.1, 0.2, and 0.7

Case no. Case 1 Case 2 Case 3 Case 4 Case 5

items
Switch off [14 20 21 25 29] [7 9 14 28 32] [10 19 22] [10 16 20 23] [9 15 21 23]
Switch on [33 34 35 36 37] [33 34 35 36 37] [34 35 37] [34 35 36 37] [34 35 36 37]
TS u (kva) 8943.48 6195.37 9862.03 9448.09 9372.76
TLL (kw) 233.927 311.667 212.746 215.558 114.4
AVd (%) 4.522 2.861 4.705 3.489 2.171
INt (Amp) 221.676 374.67 175.001 184.712 199.892
Tdt (%) 6.233 5.945 7.225 9.058 6.428
Fitness* 0.4228 0.6325 0.5296 0.6012 0.6632
(p.u)
*Fitness is the weighed multi-objective function

Case 5 in table 4 leads to lower active distribution feeders and 32 buses distribution
power losses than case1 in which the system showed the efficiency and capability of
weighting factor is directed towards the G.A in solving this type of problems. The
minimization of losses only. So using the G.A algorithm can be directed easily by experience
for solving the multi-objective optimization of the operator to minimize the total active
problem for feeder reconfiguration leads to power losses and at the same time improving
minimizing or improving total active power or minimizing total complex power, average
losses, total complex power, average voltage voltage drop, neutral current of the trans-
drop, neutral current of the main transformer former and total voltage unbalance factor
and total voltage unbalance factor. which are a major sign of better power quality.

6. Conclusions References

A genetic algorithm approach has been [1] S.K. Goswami and S.K. Basu, “A New
presented in this paper to solve the weighed Algorithm for the Reconfiguration of
multi-objective function for distribution Distribution Feeders for Loss
system reconfiguration. Five different objec- Minimization”, IEEE Trans. On Power
tives are considered in conjunction with Delivery, Vol. 7 (3), pp. 1484 -1491 (1992).
network constraints. Numerical results of four

734 Alexandria Engineering Journal, Vol. 43, No. 6, November 2004

 A. Moussa et al. / Loss reduction

[2] Jin Cheng Wang, Hsiao Dong Chiang and [5] M.A. Kashem, V. Ganapathy and G.B.
Gary R. Darling, “An Efficient Algorithm Jasmon, “Network Reconfiguration For
For Real-Time Network Reconfiguration in Load Balancing in Distribution Networks” ,
Large Scale Unbalanced Distribution IEE Proc. Gener. Transm. Distrib., Vol.
Systems”, IEEE Trans. On Power Systems, 146 (6), pp. 563-567 (1999).
Vol. 11 (1), pp. 511-517 (1996). [6] S.L. Mohamed, E.A .Mohamed and A.R.
[3] R.J. Sarfi, M.M.A. Salama and A.Y. Abu El Wafa, “Distribution Network
Chikhani, “Distribution System Reconfiguration For Power Loss Reduction
Reconfiguration For Loss Reduction: An Using Artificial Neural Network”, Seventh
Algorithm Based on Network Partitioning International Middle East Power Systems
Theory”, IEEE Trans. On Power Systems, Conference MEPCON’ 2000, Ain Shams,
Vol. 11 (1), pp. 504-510 (1996). Egypt, pp. 259-264 (2000).
[4] M.S. Kandil, A.M. Riad, A. El-Hosseiny and [7] Tsai Hsiang Chen and Jeng Tyan Cherng,
S.M. Abou Zaid, “Artificial Intelligence “Optimal Phase Arrangement of
Approach For Loss Reduction and Load Distribution Transformers connected to a
Balancing in Distribution Systems”, Fifth Primary Feeder For System Unbalance
International Middle East Power Improvement and Loss Reduction Using a
Conference MEPCON’97, Alex. , Egypt, pp. Genetic Algorithm”, IEEE Trans. On Power
479-486 (1997). Systems, Vol. 15 (3), pp. 994-1000 (2000).
[8] E. Mesut Baran and F. Felix Wu, “Network
Reconfiguration In Distribution Systems
For Loss Reduction and Load Balancing”,
IEEE Trans. On Power Delivery, Vol. 4 (2),
pp.1401-1407 (1989).

Received May 24, 2004

Accepted September 13, 2004

Alexandria Engineering Journal, Vol. 43, No. 6, November 2004 735