Solving Distribution Feeder Reconfiguration and concurrent DG

Installation problems for Power loss Minimization by Multi

Swarm Cooperative PSO algorithm
Soumitri Jena, Sushil Chauhan

Abstract—This paper presents a novel combinational Position of particle in ℎ iteration.

approach to network reconfiguration and DG installation
problems for power loss minimization in distribution network. A Velocity of particle in ℎ iteration.
multi swarm cooperative population based PSO method is used Best value of the fitness function that has
keeping in view the complexity of maintaining the radial been achieved by particle before
structure and operational constraints. Multi swarm cooperative iteration .
particle swarm optimization (MC-PSO) algorithm is an extension
of PSO algorithm which is based on the cooperative behavior of
Best value of fitness function that has been
different organisms living in an eco-system for surviving. The achieved by any particle.
optimal points for DG installation and DG sizing are carried out Random number between 0 and 1.
simultaneously. Different approaches for loss minimization are
undertaken and compared with each other. The effectiveness of Acceleration constant.
the approach is tested with IEEE 33-bus and 69-bus test systems ,
with encouraging results. Best value of fitness function that has been
achieved by any particle from the slave
Keywords—reconfiguration; distribution system; distributed swarm.
generation; power loss; particle swarm optimization.
Total number of elements in th loop.

NOMENCLATURE
I. INTRODUCTION
Total Power loss in the network.
Total number of branches. D ue to high power loss in distribution systems utilities are
Current flowing in the branch . facing severe financial setbacks. Reduction in distribution
losses can compensate these financial as well as power
Status of branch (0=open, 1=close) deficiency problems. Feeder reconfiguration arises from the
Resistance of branch . idea of altering the status (on/off) of tie line switches and
Total number of nodes. sectionalizing switches. Normally these switches are used for
Voltage of the kth node. maintenance and isolation of loads in case of fault. As load is
dynamic in nature, hence there should not be a fixed
, Generation at bus k. configuration for all the time. Network reconfiguration can
Real Power load at bus . help in reduction of real power loss and can also improve the
Reactive Power load at bus . voltage profile. The complexity arises because of the radial
| | Determinant of reduced incidence matrix. structure of distribution system and reliving a particular node
not possible. Previously DG units are used for improvement of
Real Power flowing from bus . voltage profile of the network and to compensate the
Reactive Power flowing from bus . deficiency in available power to current load level.
, Total Power loss in the network after Simultaneous reconfiguration and DG installation can further
reconfiguration. reduce the loss and improve voltage profile. These two
Real Power flowing from bus after techniques are researched individually numerous times in
,
reconfiguration. literature but comparatively fewer studies are done considering
simultaneous effects.
, Reactive Power flowing from bus after
reconfiguration. Reconfiguration is a combinational multi objective integer
, Voltage of the kth node after reconfiguration. complex problem. Many approaches have been developed in
Total Power loss in the network after DG this regard in recent past. Merlin and Back [1] were the first to
,
propose reconfiguration for loss reduction. They proposed a
installation.
heuristic programming to determine the optimum network.
Physical length of the point of DG installed They first closed all the switches to form meshed network then
(at bus ) from the source of feeder. successively opened switches to maintain radial structure.
Total length of the feeder. Several other approaches such as branch exchange methods
Real power injection from DG. were proposed by Baran and Wu [2] and Civnlar et al. [3].
Power demand at bus .

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

Goswami and Basu [4] approached network reconfiguration by
optimum flow pattern algorithm. They proposed a two stage
algorithm by calculating sensitivities in first stage followed by
a branch exchange method.
Apart from the classical and heuristic techniques several AI
based techniques have been proposed in this regard. Nara et al.
[5] proposed an algorithm based on Genetic Algorithm (GA) to
determine the minimum loss configuration. Su et al. [6]
presented algorithm based on ant colony search (ACS)
algorithm. Das [7] presented fuzzy multi objective approach
incorporating heuristic rule base. Rao et al. [8] proposed a
harmonic search algorithm (HSA) for finding out the optimum
configuration.
With the advances in green and small scale energy
technologies DG penetration level is increasing day by day. So Fig. 1. Distribution system with DG installed.
it is highly desirable to find the optimum size and location for
DG installation. Griffin et al. [9] suggested that inappropriate B. Power loss reduction due to reconfiguration
DG size and location can lead to higher loss than the system Due to network reconfiguration, changed topology leads to
without DG. Sensitivity analysis for DG installations can be different line flows in the system. Hence total power loss after
done in a similar way as is done for capacitor allocation for network reconfiguration can be written as:
voltage profile improvement [10]. Row and wan [11] presented
a method for finding out the optimum DG location by =∑ . , ,
(2)
,
repetitive load flow algorithm. Kim et al. [12] presented GA ,
based algorithms for finding out optimal DG installation Where, , , , , , are the sending end real power, reactive
points. However large number of load flow solutions make the
power and voltage. Percentage power loss reduction can be
algorithm computationally slow.
calculated as:
In this paper a multi-swarm cooperative PSO based ,
approach is used to combine both of these problems. PSO is a % = ∗ 100 (3)
metaheuristic population based optimization technique used to
solve many real world problems. Though PSO is used for C. Power loss reduction due to DG installation.
solving continuous optimization problems, some modifications Real power injection to the system also leads to different
make it efficient to be applied to discrete optimization power flows in the network. DG installation also relives the
problems also. A hybrid technique using SPSO [13] and PSO overloaded lines. Installation of DG units at candid locations
is used to combine the two problems to reach optimum leads to optimum loss. Power loss due to a single DG installed
solution. While determining the optimum location and size [14] in the system can be given by the empirical expression:
for DG units repeated load flow is avoided. The algorithm is
tested with IEEE 33-bus and IEEE 69-bus test distribution , = ( + )+ , + , −2 , −2 .
systems.
(4)
II. PROBLEM FORMULATION
Percentage power loss reduction can be calculated as
A. Objective Function ,
The objective of the approach is to minimize the real power % = ∗ 100 (5)
loss across the network. Due to high / ratio in distribution It may be notated that inappropriate DG installations may lead
system losses are mainly resistive in nature. to higher losses.
=∑ | | (1) D. Sensitivity analysis for DG installation
Subjected to ≤ | |≤ Sensitivity analysis is the process of selecting most
sensitive buses with respect to certain objectives. The objective
, ≤| |≤ , here is to minimize real power loss.
∑ , = ∑ . +∑ Fig. 2 shows the 3D curve of power loss versus DG size for
Radial constraint: | |= ±1 the IEEE 33-bus test distribution system [2]. The noticeable
point here is that for any bus with the increase of DG size the
Most of the distribution utilities work in radial mode due to power losses are reduced up to a particular minimum value
operational constraints. According to Kirchhoff “The spanning (optimum point) then it starts increasing. High DG size does
trees of a connected graph are singular (n − 1) × (n − 1) sub not necessarily guarantee minimum losses. Hence optimum
matrices of the reduced incidence matrix, and determinant of sizing of DGs at candid locations are required. However
these matrices are all±1” [15]. previously sensitivity analysis required lot of repetitive load

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

flow computations. Acharya et al. [14] suggested an analytical = + ∑ ( − ) (8)
,
method based on “exact loss” formula [16]. This method is
computationally fast and does not involve repetitive load flow
calculations.
Where,
=( − ), Hence real power injection from DG at bus
= + + ∑ , ( − ) (7)
250

After finding out the optimum size at each bus, the

approximate loss can be calculated by (6). The buses are
P o w e r L o s s (K W )

200

ranked according to the losses for minimum DG size, in

150
ascending order. It has been observed that losses given by the
exact loss formula (4) follows similar pattern to that of the
actual losses. Hence large load flow calculations can be
100
2
1.8
avoided. This method avoids the error in bus ranking due to
1.6
1.4
1.2 25
30 high initial slope but lesser power loss reduction [14].
1 20
0.8
0.6 15
0.4
0.2
0
5
10
III. OVERVIEW OF MULTI SWARM COOPERATIVE PSO
DG Size (MW) Bus No.
A.Fundamental PSO
Particle swarm optimization is inspired from the social and
Fig. 2 Effect of DG size on Power losses. organizational behavior of organisms like birds and fishes. It
The real power loss for a system given by “exact loss” was developed by Kennedy and Eberhart [17]. When the birds
formula [16]. move from one place to another in search for food their
movement (position and velocity) depends on their own
knowledge and knowledge of their neighbors ( and ).
= + + ( − ) In each step the velocity and position are updated by the
equation
(6) = + − + −

Where = cos( − ) and = sin( − ) and = + (8), (9)

= + is the ℎ element of [Zbus]. is the inertia weight [18]. Inertia Weight governs how
much of the previous velocity should be retained for the current
230 iteration. Previously in the literature it has been found that for
the best performance inertia weight varies from 0.9( ) to
220
0.4( ). So the inertia weight is set according the
210
following equation
200
= − × (10)
POWER LOSS (KW)

190

180 These settings allow the algorithm to search a large area

170
initially then as the iteration progresses it refines the search for
a smaller area.
160

150
B. Multi Swarm Cooperative PSO
140
Multi swarm cooperative PSO was proposed by B.Niu et
0 0.2 0.4 0.6 0.8 1
DG SIZE (MW)
1.2 1.4 1.6 1.8 2
al. [19]. Basic PSO reflects the cooperative relationship
among individuals within a group. However there are some
Fig. 3 Loss variations due to DG size at bus no. 18 organisms in ecosystem who depend on other organisms for
Power loss at an arbitrary bus is parabolic with respect to DG their survival. Some parasites live on dead or infected cells of
size. At minimum loss point the rate of change of losses with living organisms. This phenomenon is called symbiosis.
DG size becomes zero. MCPSO is based on master-slave model. The model consists
of one master swarm associated with several slave swarms.
= 2∑ ( − )=0 (7) Master swarm evolves from its own knowledge and
,
knowledge of the slave swarms.
It can now be written as, Each slave swarm executes a PSO or variant of PSO which
− + ∑ ( − )= 0 includes finding out the local best, velocity and position
,
update of each particle and creating new population. The
particles in master swarm updates from its own previous

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

experience and experiences of the slave swarm. The idea is Care should be taken so that no two loops contain the same
implemented by introducing a new term in the velocity update switch. The switches which don’t take part in any loop
formula (8). The resulting equations are formation are excluded from the search space. Hence a particle
in the master swarm is defined as follows:
= + − + − +
, 1 2 3 4 5
( − ) (9)

= + (10)

Fig. 4 Master Slave model.

The process flow for MCPSO algorithm can be summarized as

Step 1. Initialize a population of master and slave swarms.

Step 2. Initialize swarm parameters.
for iteration ≤ max. iteration
Step 3. Evaluate fitness of each particle in the
population.
Parallel to the process
Step 4. Parallel to the process run the PSO
for slave swarm and find the local
best.
Step 5. Update position and velocity of
each particle in slave swarm.
Wait for the processes to over.
Step 6. Select the fittest global solution for master
and slave swarm.
Step 7. Evolve new master swarm by updating the Fig 5. IEEE 33-bus radial distribution system
position and velocity.
1= [ 2, 3, 4, 5, 6, 7, 18, 19, 20]
end
2= [ 8, 9, 10, 11, 21, 35]
IV. APPLICATION OF MC-PSO TO CURRENT PROBLEM
3=[ 12, 13, 14, 34]
Both reconfiguration and DG installation problems are
complex combinatorial integer based problems. Hence solution 4=[ 15, 16, 17, 29, 30, 31, 32, 36]
vector for the optimization problem is also combinatorial. Part
5=[ 22, 23, 24, 25, 26, 27, 28, 37]
of the solution vector is concerned with reconfiguration and the
other part is concerned with DG installation. The particles are updated by the following equations [13] :
For network reconfiguration in order to define a radial = (11)
( )
structure only the open switches are to be decided [8] and the
number of switches to be opened is equal to the number of , if <1
closed loop formed by closing all the tie line switches , if <2
(normally open) of the network. In the literature of graph = , if <3 (12)
theory it is well known that if a graph of spanning tree …………………………..
configuration is having number of nodes number of ,if <
branches then the total number of closed loops formed would
be ( − + 1) . For selection of the candid buses i.e. the buses which are
more sensitive to real power injection, the method described in
The dotted lines in Fig. 5 shows the available tie lines. To section II.D is used. The number of DGs to be installed,
maintain the radial structure one switch needed to be opened depends on the availability and feasibility of energy sources in
from each loop. Hence the search space is defined as the a practical distribution system. More percentage loss reduction
switches belonging to each loop. can be achieved by a single DG of appropriate size than

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

combination of number of DGs. Previously it is observed [8] within 2 MW. The performance of the network for three
that the percentage loss reduction is not significant when the different load levels: 0.5 (light), 1(nominal), 1.6 (heavy) is
DG numbers is increased more than three. As it is observed determined.
from the Fig 3 most of the DGs produce the optimum loss point
within 2 MW. So, the DG size is kept between 0-2 MW. Hence
the slave particle here can be defined as follows.
Size 1 Size 2 Size 3
candid bus.1 candid bus. 2 candid bus. 3

It is to be noted that, DG sizes taken here is considered as pure

real power injections to the network at unity power factor. The
candid bus locations are determined as the configuration based
on the master particle. Hence the fitness value (power loss) is
defined as the solution taken from the particle from master
swarm and the fittest particle from the slave swarm. Hence a
solution vector is as like.
s1 s2 s3 s4 s5 Size Size Size
1 2 3

based on reconfiguration sizes of DGs

at candid buses
For selection of the fittest combination of DG sizes at candid
buses basic PSO method is used and the position and velocities
are updated by equations (9) and (10).
V. RESULTS
The proposed method is tested with IEEE 33-bus and Fig 6. Flow chart for proposed method.
IEEE 69-bus radial distribution systems. The data for the As seen from Table.1 for light load condition the total
systems can be found from [2] and [20]. The following five power loss for base case is 47.07 KW. The loss has been
cases are studied for each of the above systems. reduced to 33.20 , 24.31, 24.02, and 14.87 KW for case 2,case
Case 1: The base case i.e. neither reconfigured nor 3, case 4, and case 5 respectively. The percentage loss
DG installed. reduction has been improved for case 2 to case 5 as
Case 2: Only reconfiguration applied to the base 29.47,48.35, 48.97 and 68.40. Similarly for nominal loading
case. conditions total losses has been reduced from 202.68 to 138.39,
Case 3 : Only DG units installed to base case 100.97, 99.25, 61.24 for case 2 to case 5 respectively. The
Case 4 : DG units are installed after reconfiguration percentage loss reduction has been improved for case 2 to case
of the network. 5 as 31.45, 50.18, 51.03 and 69.78. For heavy load conditions
Case 5: Reconfiguration is done in concurrent with the losses have been reduced from 575.36 to 377.34, 272.7,
DG installation. 264.82 and 152.72 for case 2 to case 5 respectively. The
The algorithm is coded with MATLAB and tested with the percentage loss reduction is improved from 34.42 to 52.54,
two systems on a computer with Dual core, 1.73 GHz and 53.97 and 73.45 for case 2 to case 5.
2GB RAM. 600
Power loss (KW)

A. 33-bus test distribution system 400

This test system is a radial distribution system with 33
buses, 32 sectionalizing switches and 5 tie branches. While
200
referring to Fig 5 the solid lines are the sectionalizing switches
which are normally closed and dotted lines indicate tie line
switches which are normally open. The total real power load 0 Case 1 Case 2 Case 3 Case 4 Case 5
on the system is 3715 KW and total reactive power load is Light Nominal Heavy
2300 KVAR. Fig. 7 Loss reduction for different cases in 33-bus test system.
The candidate buses for DG installations are determined by
the algorithm suggested by [14]. The top three ranked buses
are selected for DG installation and the size of DGs is kept

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

Table 1.
Performance of 33-bus test system.

Load level
Case Description
Light (0.5) Nominal (1.0) Heavy (1.6)

Base case (case 1) Switches opened 33,34,35,36,37 33,34,35,36,37 33,34,35,36,37

Power loss (KW) 47.07 202.68 575.36
Minimum Voltage (p.u) 0.9583 0.9131 0.8527

Only reconfiguration (case 2) Switches opened 7,9,14,32,37 7,9,14,32,37 7,9,14,32,37

Power loss (KW) 33.20 138.93 377.34
Minimum Voltage (p.u) 0.9712 0.9423 0.9045
% Loss reduction 29.47 31.45 34.42

Only DG installation (case 3) Switches opened 33,34,35,36,37 33,34,35,36,37 33,34,35,36,37

DG size in MW 0.2297 (6) 0.4926 (6) 1.8700 (6)
(Candid. bus) 0.4672 (26) 0.9742 (26) 1.4250 (26)
0.5487 (7) 1.1208 (7) 1.2310 (7)
Power loss (KW) 24.31 100.97 272.7
Minimum Voltage (p.u) 0.9764 0.9530 0.9260
% Loss reduction 48.35 50.18 52.54

DG installation after reconfiguration (case 4) Switches opened 7,9,14,32,37 7,9,14,32,37 7,9,14,32,37

DG size in MW 0.3501 (6) 0.7172 (6) 1.1744 (6)
(candid. bus) 0.0311 (26) 0.1230 (26) 0.1031 (26)
0.4096 (27) 0.7793 (27) 1.3923 (27)
Power loss (KW) 24.02 99.25 264.82
Minimum Voltage (p.u) 0.9747 0.9482 0.9145
% Loss reduction 48.97 51.03 53.97

Reconfiguration with concurrent DG installation (case 5) Switches opened 7,11,14,26,32 6,9,12,27,32 7,12,32,35,37
DG size in MW 0.3265 (29) 0.9353 (29) 1.7372 (29)
(candid bus) 0.5135 (25) 0.6187 (25) 1.4157 (25)
0.3445 (15) 0.9640 (15) 1.4470 (15)
Power loss (KW) 14.87 61.24 152.72
Minimum Voltage 0.9865 0.9736 0.9620
% Loss reduction 68.40 69.78 73.45

Fig. 7 shows that case 5 has the superiority over the other The comparison of results for nominal loading obtained by
cases and also case 3 (only DG installation) and case 4 (DG the proposed method with GA, RGA and HSA [21] is
installation after reconfiguration) yields almost the same presented in Table 2.
result. The optimal configuration for nominal load case (case Table 2. Comparison of results for 33-bus test system.
5) is presented in Fig.8. As the loading increases from light to Description MC-PSO HSA[21] GA[21] RGA[21]
heavy percentage loss reduction is almost the same. Switches 6,9,12,27,32 7,10,14,32,28 7,10,28,32,34 7,9,12,31,27
The voltage profile of the system is also seen improved. opened
DG size in 0.9353(29) 0.5258(32) 1.9633 1.774
The minimum voltage for the system is improved from case 1 MW 0.6187(25) 0.5586(31)
to case 5. As the system loading increases there is a fall in (Candidate
bus)
0.9640(15) 0.5840(33)

system voltage. It is also observed that the overall voltage Power loss 61.24 73.05 75.13 74.32
(KW)
profile of the network is improved by simultaneous Minimum 0.9736 0.9700 0.9766 0.9691
reconfiguration and DG installation (case 5).The shape of Voltage
%Loss 69.78 63.95 62.92 63.33
voltage curve almost remains same for all the three loading reduction
conditions.
As we move from light loading to nominal loading and B. 69-bus test distribution system
then over to heavy loading conditions there is an increase in
IEEE 69-bus radial distribution system is large
size of DG required.
system with 68 sectionalizing branches and 5 tie lines. The
total real power load on the network is 3801.9 KW and 2694.1
KVAR of reactive power load. The initial parameters for MC-

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

PSO is same as of 33-bus system. The number of DGs and DG size is also kept same as of the previous system.
(b)
1
0.99

0.97

0.95

Voltage (p.u)
0.93

0.91

Case 1
0.89
Case 2
Case 3
0.87
Case 4
Case 5
0.85
5 10 15 20 25 30
Bus number

(c)

Fig. 9(a)-(b)-(c) Voltage profile for 33-bus test system under light(0.5)-
nominal(1) and heavy(1.6) loading conditions.
Similar trends as for 33-bus test system are also observed
in 69-bus test system. Under light loading conditions the
system losses has been found reduced from 51.60 KW for base
case to 23.69, 20.26, 12.70,11.76 for case 2 to case 5. The
percentage loss reduction achieved has been found 54.20,
60.74, 75.24 and 77.20 respectively for case 2, case 3, case 4
and case 5.
For the system under nominal rating the power loss has
been reduced from 224.95 KW to 101.32, 83.05, 53.35 and
Fig 8. Optimal configuration for IEEE 33-bus test system 40.91 for case 2 to case 5. The percentage loss reduction
achieved is 54.96, 63.08, 76.28 and 81.81 for case 2 to case 5
respectively. Under heavy loading conditions the system
1
losses for base case 652.38 has been found reduced to
0.995 272.53,219.50,151.06 and 104.96 for case 2 to case 5 . The
0.99 percentage loss reduction achieved is 58.23, 66.35, 76.84 and
0.985 83.91 for case 2 to case 5 respectively.
Unlike 33-bus system ,here significant loss reduction has
Voltage (p.u)

0.98

0.975
been achieved from case 3 to case 4.However as the system
loading increases there percentage loss reduction has not been
0.97
Case 1
significant.
0.965 Case 2
Case 3 As observed from the results of both the systems,
0.96 Case
Case
4
5
reconfiguration with concurrent DG installation case is found
0.955
5 10 15 20 25 30
to be the superior one over the other cases for power loss
Bus number reduction and overall voltage profile improvement.
(a)
800
1

0.99
600
0.98

0.97
400
Voltage (p.u)

0.96

0.95 200
0.94
Case 1
0.93 Case 2 0
Case 3
0.92 Case 4
Light Nominal Heavy
Case 5 case 1 case 2 case 3 case 4 case 5
0.91
5 10 15 20 25 30
Bus number
Fig. 10 Loss reduction for different cases in 69-bus test system.

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

Table 3.
Performance of 69-bus test system

Load level
Case Description
Light (0.5) Nominal (1.0) Heavy (1.6)

Base case (case 1) Switches opened 69,70,71,72,73 69,70,71,72,73 69,70,71,72,73

Power loss (KW) 51.60 224.95 652.38
Minimum Voltage (p.u) 0.9567 0.9092 0.8445

Only reconfiguration (case 2) Switches opened 13,57,61,69,70 12,18,58,61,69 13,18,55,61,69

Power loss (KW) 23.63 101.32 272.53
Minimum Voltage (p.u) 0.9754 0.9494 0.9045
% Loss reduction 54.20 54.96 58.23

Only DG installation (case 3) Switches opened 69,70,71,72,73 69,70,71,72,73 69,70,71,72,73

DG size in MW 0.7623 (61) 1.5529 (61) 1.9337 (61)
(Candid. bus) 0.0177 (62) 0.0323 (62) 0.6575 (62)
0.1426 (63) 0.2872 (63) 0.4598 (63)
Power loss (KW) 20.26 83.05 219.50
Minimum Voltage (p.u) 0.9844 0.9683 0.9485
% Loss reduction 60.74 63.08 66.35

DG installation after reconfiguration (case 4) Switches opened 13,57,61,69,70 12,18,58,61,69 13,18,55,61,69

DG size in MW 0.3200 (61) 1.2464 (61) 1.9972 (61)
(candid. bus) 0.3016 (60) 0.0547 (60) 0.0221 (60)
0.1243 (59) 0.2516 (59) 0.4196 (59)
Power loss (KW) 12.70 53.35 151.06
Minimum Voltage (p.u) 0.9827 0.9596 0.9383
% Loss reduction 75.24 76.28 76.84

Reconfiguration with concurrent DG installation (case 5) Switches opened 13,55,63,69,70 14,56,69,70,73 13,57,63,69,70
DG size in MW 0.3265 (62) 1.0871 (62) 1.8372 (62)
(candid bus) 0.0433 (63) 0.3559 (63) 0.4957 (63)
0.2071 (60) 0.4341 (60) 0.6374 (59)
Power loss (KW) 11.76 40.91 104.96
Minimum Voltage 0.9839 0.9770 0.9565
% Loss reduction 77.20 81.81 83.91

Table 4. Comparison of results for 69-bus test system.

Description MC-PSO HSA[21] GA[21] RGA[21]
parallel with solving reconfiguration problems. The problem
Switches 14,56,69,70,73 13,17,58,61,69 10,15,45,55,62 10,14,16,55,62 of repeated load flows while calculating the sensitivity factors
opened
DG size in 1.0871(62) 1.0666(61) 2.0292 2.0654 has been avoided by using the exact loss formula. The
MW
(Candidate
0.3559(63)
0.4341(59)
0.3525(60)
0.4527(62)
proposed method is tested with different system loading
bus) conditions light, nominal and heavy for both 33-bus and 69-
Power loss 40.91 40.30 46.50 44.23
(KW) bus test distribution systems. The comparison with existing
Minimum
Voltage
0.9770 0.9736 0.9727 0.9742
methods like GA, RGA and HSA shows the effectiveness of
%Loss 81.81 82.08 73.38 80.32 the method. The results suggest that reconfiguration with
reduction
concurrent DG installation yields in minimum loss and overall
voltage profile improvement. The main advantage of this
VI. CONCLUSION algorithm is simplicity in optimum reconfiguration and DG
installation. PSO, which is a meta-heuristic algorithm is
In this paper a new approach for simultaneous network
applied in the same problem for both continuous and discrete
reconfiguration and DG installation is presented. Proposed
optimization problem. The algorithm can be easily applied to
method maintains the radial structure and operational
large networks to provide an optimum or a near optimum
constraints while solving the problem. This method
solution.
determines the optimal DG installation points and DG size in

978-1-5090-2157-4/16/$31.00 ©2016 IEEE

 [13] Tamer M. Khalil , Alexander V. Gorpinich “Reconfiguration for loss
REFERENCES reduction of Distribution systems using selective particle swarm
optimization”, International Journal of Multidisciplinary sciences and
Engineering , vol.3 , No.6, June 2012,pp 16-21
[1] A. Merlin and H. Back, “Search for a minimal-loss operating spanning [14] Naresh Acharya, Pukar Mahat, and Nadarajah Mithulananthan. "An
tree configuration in an urban power distribution system,” in Proc. 5th analytical approach for DG allocation in primary distribution
Power System Computation Conf. (PSCC), Cambridge, U.K ,1975, pp.1- network." International Journal of Electrical Power & Energy
18 Systems,Vol.28, no. 10 , 2006, pp 669-678.
[2] M. E. Baran and F. Wu, “Network reconfiguration in distributed system [15] Perrin Wright, “On minimum spanning trees and determinents ”,
for loss reduction and load balancing” IEEE Trans. Power Del., vol.4, Mathematics magazine ,Vol.73 , No.1 ,Feb 2000.
no.2, pp. 1401-1407, April. 1989.
[16] Elgerd IO., Electrical Energy system theory : an introduction., McGraw-
[3] S. Civanlar, J. Grainger, H. Yin, and S. Lee, “Distribution feeder Hill, 1971.
reconfiguration for loss reduction,” IEEE Trans. Power Del., vol.3, no.3,
pp. 1217-1223, Jul 1988. [17] J. Kennedy , R.C. Eberhart,“Particle swram optimization” , Proc. of the
IEEE international conference on Neural Networks, Pearth , Austrilia ,
[4] S. K Goswami and S. K Basu, “A new algorithm for the reconfiguration 1995 , pp. 1942-1948
of distribution feeders for loss minimization,” IEEE Trans. Power Del. ,
vol.7, no.3, pp. 1484-1491, july 1992. [18] Y. Shi, R.C. Eberhart, “A modified particle swarm optimizer”, Proc. Of
IEEE international conference on Evolutionary Computation, Anchorage
[5] K. Nara, A. Shiose, M. Kitagawoa and T. Ishihara, “Implementation of , AK, May 1998, pp- 69-73
genetic algorithm for distribution systems loss minimum reconfiguraion
”, IEEE Trans. Power syst., vol.7 ,no. 3, pp. 1044-1051, Aug.1992. [19] B. Niu , Y. Zhu , X. He , H. Wu “MCPSO- A Multi swarm cooperative
particle swarm optimizer ”, Applied mathematics and computation 185
[6] C.T. Su, C.F. Chang, J.P. Chiou, “Distribution network reconfiguration ,2007 , pp. 1050-1062.
for loss reduction by ant colony search algorithm”, Electr. Power. Syst.
Res., 2005. Pp. 190-199. [20] M. E Baran, F. Wu “ Optimal sizing of capacitors placed on radial
distribution systems”, IEEE Trans. Power Del., vol.4, no. 1, pp. 735-
[7] D. Das, “A fuzzy multi-objective approach for network reconfiguraton 743, 1989.
of distribution systems”, IEEE Trans. Power Del., Vol. 21, no.1 , pp-
202-209, Jan. 2006. .[21] R. Srinivasa Rao,K.Ravindra, K.Satish ,S.V.L. Narasimham, “Power
loss minimization in distribution system using network reconfiguration
[8] R.Srinivasa Rao, S.V.L Narasimham, M.R Raju, and A.Srinivasa in the presence of distributed generation ” ,IEEE Tans. Power syst.,Vol
Rao,”Optimal network reconfiguration of large-scale distribution system 28, No.1, February 2013,pp. 317-325
using harmony search algorithm”,IEEE Trans. Power Syst. ,vol.26 ,no.
3 , pp. 1080-1088, Aug, 2011.
[9] T. Griffin , K Tomosovic, D Secrest, A. Law “Placement of dispersed
generations systems for reduced losses ”, Proceedings of the 33rd Hawaii Soumitri Jena is a post graduate student in the Department of Electrical
international conference on sciences, Hawii, 2011. Engineering at National Institute of Technology, Hamirpur, India.His areas of
[10] K.Prakash and M. Sydulu, “Particle swarm optimization based capacitor intrest include distributed generation and distribution system automation.
placement on radial distribution systems”, Proc. IEEE Power
Engineering Society General Meeting, 2007 , pp. 1-5.
[11] NS Row , Y-H Wan “optimum location of resources in distributed Sushil Chauhan is a professor in the Department of Electrical Engineering at
planning”, IEEE Trans PWRS 1994; 9(4): 2104-20. National Institute of Technology, Hamirpur, India. His areas of intrest include
[12] K-H Kim, Y-J Lee, S-B Rhee, S-K Lee, S-K You,”Dispersed Generator AI applications in power system, Power system analysis and design.
placement using fuzzy –GA in distribution systems”, Proceedings of
2002 IEEE Power Engineering Society Summer Meeting, Chicago, IL,
July 2002; 3:1148-53

978-1-5090-2157-4/16/$31.00 ©2016 IEEE