Electrical Power and Energy Systems 54 (2014) 123–133

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Optimum network reconfiguration based on maximization of system

loadability using continuation power flow theorem
M.M. Aman a,⇑, G.B. Jasmon a, A.H.A. Bakar b, H. Mokhlis a,b
a
Department of Electrical Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysia
b
UM Power Energy Dedicated Advanced Centre (UMPEDAC) Level 4, Wisma R&D, University of Malaya, Jalan Pantai Baharu, 59990 Kuala Lumpur, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a new algorithm for network reconfiguration based on maximization of system loa-
Received 9 July 2012 dability. Bifurcation theorem known as Continuation Power Flow (CPF) theorem and radial distribution
Received in revised form 20 June 2013 load flow analysis are used to find the maximum loadability point. Network reconfiguration results are
Accepted 29 June 2013
also compared with existing technique proposed in literature. In the proposed method, to find the opti-
mum tie-switch position, a Discrete Artificial Bee Colony (DABC) approach is applied. Graph theory is
used to ensure the radiality of the system. The proposed algorithm is tested on 33-bus and 69-bus radial
Keywords:
distribution networks, each having 5-tie switches. The result shows that using the proposed method the
Continuation Power Flow (CPF)
Discrete Artificial Bee Colony (DABC)
kVA margin to maximum loading (KMML) increases, overall voltage profile also improved and the distri-
Graph theory bution system can handle more connected load (kVA) without violating the voltage and line current con-
Loadability straints. Results further show that the voltage limit is an important factor than the line current
constraints in adding further load to the buses.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
With the ever increasing load demand and the lack of capital
investment in the transmission system, the importance of enhanc-
ing the existing distribution system capacity has increased. The
system capacity is usually limited by two factors, namely thermal
limits and voltage limits. Thermal limit or thermal capacity is the
ampacity or maximum current carrying capacity limit of the con-
ductor. The current carrying capacity is limited by the conductor’s
maximum design temperature [1]. However the voltage limit is the
allowable minimum–maximum voltage variation for safe opera-
tion of power system and connected load. The study in [2] has con-
cluded that the maximum loadability of the distribution system is
limited by the voltage limit rather than the thermal limit.
In literature, minimization of losses in power system was a
major concern for power system researchers. Among different
techniques for power loss reduction including Distributed Genera-
tion (DG) placement and shunt capacitor placement, network
reconfiguration is also utilized. Network reconfiguration is defined
as altering the topological structures of the distribution feeders, by
changing the position of tie and sectionalizing switches. Network
reconfiguration is a key tool in operation of medium voltage distri-
bution system and in improving the reliability of the system. The
structure of medium voltage distribution network is designed
⇑ Corresponding author. Tel.: +60 3 7967 5348; fax: +60 3 7954 7551.
E-mail address: mohsinaman@gmail.com (M.M. Aman).
mesh, provided with tie and sectionalizing switches. However un-
der normal operation, medium voltage distribution networks oper-
ates in radial manner [3,4].
Most of the researchers have considered power losses as an
objective function in network reconfiguration. Analytical and
heuristic search based optimization techniques have been utilized
by different researchers to perform the network reconfiguration.
Authors have formulated the problem in different manners. In
[3], Civanlar has analyzed the problem of network reconfiguration
considering minimization of losses, load flow based approach was
utilized. In [5], Baran and Wu extended the work of [3] and in-
cluded a new load balancing index in addition to minimization of
losses, simplified load flow approach is used. In [6], Shirmoham-
madi utilized the heuristic approach to find the minimum losses
of the network. The radial system is converted to mesh, later-on
optimum power flow pattern is made and the branch carrying
the minimal current is removed.
Later on, several optimization based algorithms have been
developed considering minimization of losses as an objective
function. Zhu utilized the refined genetic algorithm [7], Venkatesh
et al. utilized the Fuzzy adaptation of Evolutionary Programming
technique [8], Zhu et al. performed the network reconfiguration
based on modified heuristic solution and experience system oper-
ation rules [9]. Srinivasa et al. in [10] used the Harmony Search
Algorithm, Sathish in [11] solved the problem using Bacterial
Foraging Algorithm, Yuan-Kang et al. [12] utilized the Ant Colony
Algorithm. de Oliveria et al. [13] mixed the problem of network

0142-0615/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2013.06.026
124 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133

reconfiguration with capacitor placement considering minimiza- In this paper, the parameter that is used to characterize the loa-
tion of total energy loss as an objective function. Zin et al. in [14] dability of the system is the maximum loadability margin kmax, as
has used the heuristic search algorithm to find the minimum shown in Fig. 1. The loadability factor k is used to modify con-
power losses as an objective function, introduced by the author nected load as given by Eq. (1) [21]:
in [5]. The update principle of heuristic search is made by finding 
PL1 ¼ PL0 þ k0 P D
the branch having minimal current, till the optimum solution is ð1Þ
made. In the latest research, Rao et al. in [15] has combined the Q L1 ¼ Q L0 þ k0 Q D
network reconfiguration with distributed generation placement, where PL0 and QL0 are the initial active and reactive power loads
considering minimization of power losses as fitness function. Most respectively.
of the existing research is based on finding optimum tie switch PG1, PL1 and QL1 are the modified active and reactive power loads
combinations based on minimization of power losses. respectively.
In the present study, a new strategy, based on voltage stability k0 is the initial loadability factor that multiplies variable powers
criteria is proposed for solving the reconfiguration problem. The PD and QD, also called power directions.
objective is to maximize the system loading margin or kVA margin When power directions PD and QD are in the vector direction de-
to the point of maximum loadability of the system. Loadability is fined by PL0 and QL0, Eq. (1) changes to Eq. (2):
defined as the capacity of the system with which the maximum 
load could be connected without going into the voltage instability PL1 ¼ ð1 þ k0 ÞPL0 ¼ kPL0
0 ð2Þ
region. Voltage stability has been used in solving other power sys- Q L1 ¼ ð1 þ k ÞQ L0 ¼ kQ L0
tem problems including optimal power flow solution [16] and opti-
where k = 1 + k0 is the loadability factor.
mum DG placement [17,18]. The major research considering
The CPF method utilizes predictor–corrector approach for the
maximum loadability in network reconfiguration is found in [8].
complete solution of PV nose curve. Predictor step is based on
In [8], the author has developed the Maximum Loadability Index
the computation of the tangent vector and a corrector step is ob-
(MLI) based on the voltage stability criteria and used as a fitness
tained either by means of a local parameterization or a perpen-
function to solve the network reconfiguration problem. Fuzzy-EP
dicular intersection [19,20,22]. For mathematical modelling of
approach was used by the author to find the optimum tie switch
CPF, consider a curve AC, shown in Fig. 2, given by the following
combinations.
model
In the proposed method, Continuation Power Flow (CPF) is used
for finding the maximum loading margin of the system. With CPF, gðy; kÞ ¼ 0 ð3Þ
the maximum loadability margin or maximum loadability index where y represents the state variables and k represents system
kmax is calculated which corresponds to saddle node bifurcation parameter used to drive the system from one state of loading to
(SNB) point or point of voltage collapse. The CPF results will also another.
be compared with the radial distribution load flow methods. Net-
work reconfiguration differs from other optimization problems 2.1. Predictor step
(e.g. DG sizing, shunt capacitor sizing, unit commitment and oth-
ers), the tie switch positions always occur in a discrete manner To find the predictor vector, consider an equilibrium point ‘P’ in
(e.g. 7, 13, 19, 25, 28) thus a Discrete Artificial Bee Colony (DABC) Fig. 2. At equilibrium point ‘P’, the following relation applies:
approach is proposed in finding the optimum tie switch combina-
tions. To ensure the radiality of the system for different switch gðyp ; kp Þ ¼ 0 ð4Þ
  
combinations, Graph Theory (GT) approach is applied. dg   dy
  þ dg 

 ¼ 0 ¼ r g ð5Þ
dk p dkp

The paper is organized as follows: In Section 2, the overview of y p
dk p
Continuation Power Flow (CPF) is presented and the mathematical
model is given. Section 3 presents the graph theory approach in And the tangent vector can be approximated by:
checking the radiality of the system for different tie switch combi- 
dy Dyp
nations. Section 4 presents Discrete Artificial Bee Colony (DABC) sp ¼  ð6Þ
dkp Dkp
for finding the optimum switch configuration. Section 5 provides
the definition of Voltage Deviation Index (VDI) for measuring
power quality and kVA Margin to Maximum Loadability (KMML).
In Section 6, the proposed algorithm is presented and in the last
section 7, the proposed algorithm is applied on 33-bus and 69- Predictor
bus radial distribution test systems. The results are shown and also
A B
discussed in detail. Predictor
Corrector
Voltage Magnitude V

C D
2. Continuation Power Flow (CPF) theorem Exact Solution
E
CPF theorem is commonly used in power system to solve the
load flow problem. Most of the theorems get diverges after reach- F
Critical Point or G
ing the critical point or point of voltage collapse as in the case of
Limit Induced Bifurcation or Saddle-Node Bifurcation (SNB), and Point of collapse Corrector
thus the power flow equations have no solution at SNB point. In
comparison to other power flow theorems, CPF has an advantage
in terms of complete solution of nose curve even after reaching
the SNB point. CPF theorem is based on bifurcation model. Bifurca-
tions occur when the system stability changes due to a change of λmax
Loading parameter λ
system parameters [19,20]. In CPF, predictor–corrector approach
is used to solve the PV or kV nose curve, as shown in Fig. 1. Fig. 1. Continuation power flow theorem – solving approach.
 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133 125

Fig. 2. Predictor vector computation.

From Eqs. (5) and (6):

 Fig. 4. Corrector vector computation using local parameterization.
1 dg 
sp ¼  r y gjp ð7Þ
dkp
Dyp ¼ sp Dkp ð8Þ of the variable to be fixed depends on the bifurcation manifold of g,
as shown in Fig. 4. The value of q in case of local parameterization
A step size control ‘k’ has to be chosen for the determinant: is given by
k qðy; kÞ ¼ yci  ðypi þ Dypi Þ ð14Þ
Dkp ffi ð9Þ
jjsp jj
k sp Or
Dyp ffi ð10Þ
jjsp jj qðy; kÞ ¼ kc  ðkp þ Dkp Þ ð15Þ
where k = ±1 and sign determines the increase or the decrease of k. In the present study, CPF theorem is used to find the fitness
function (maximum loadability – kmax) of proposed algorithm.
2.2. Corrector step To compare the CPF results, the maximum loadability will also
be computed using radial load flow analysis. Voltage stability
In the corrector step, a set of n + 1 equations are solved i.e. and optimization tool will be used to compute the maximum
loading of the system [23]. Here it is noted that different tie
gðy; kÞ ¼ 0 ð11Þ
switch combinations can results in non-radiality of the system,
qðy; kÞ ¼ 0 ð12Þ thus the graph theory will be used to maintain the radiality of
the system. Section 3 will present a brief overview of graph theory
where the solution of g must be in the bifurcation manifold and q is
approach used in network reconfiguration to ensure the radiality
an additional equation to guarantee a non-singular set at the bifur-
of the system.
cation point. The perpendicular intersection and the local parame-
terization is used to solve the equation for q.
The illustration of perpendicular intersection corrector method 3. Graph theory for maintaining the radility of the system
is shown in Fig. 3, the expression of q is given by Eq. (13):
 T   Different tie switch positions in case of solving reconfiguration
Dyp yc  ðyp þ Dyp Þ problem may results in a mesh circuit. Thus a graph theory will be
qðy; kÞ ¼ ¼0 ð13Þ
Dkp kc  ðkp þ Dkp Þ utilized to ensure the system is always operating in radial manner
and all buses are connected with source node. In general, the graph
For the local parameterization, one of the parameter i.e. k or a
G (V, E) is a pair of sets mapped by the pair of vertices or nodes (V)
variable y is forced to be a fixed to find the value of q. The choice
through the edges (E). The vertices V are usually labelled as
V = {v1, v2,    , vn}.The vertices are said to be adjacent to each other
if they are inter-connected through an edge [24,25]. The example
of nodes and edges are shown in Fig. 5.

(a)

Fig. 5. (a) A 6-bus radial distribution system, (b) Simplified form of 6-bus system
Fig. 3. Corrector vector computation using perpendicular intersection. with nodes (V) and edges (E).
126 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133

rected graph it is easy to switch of the connection between 4–6 or

(b)
6–4 (a46 = a64 = 0). However in the case of directed graph, it will be
essential to determine either flow of power is 4-to-6 or 6-to-4, then
the corresponding link can be switched off.
In order to measure the connectivity of the system, two search-
ing methods are commonly used, Breadth First Search (BFS) and
Depth First Search (DFS) methods [26]. In current application of
network reconfiguration, DFS method is utilized. DFS method
returns the order of the bus in the system, which is later used to
rearrange the line-data according to the network flow that is
mapped out.
The steps in solving the network reconfiguration problem are
briefly presented here:
Fig. 5 (continued)
 Step 1: Initialize the bus system.
 Step 2: Generation of sparse and Adjacency Matrix (A).
 Step 3: Generation of random tie switch positions and updating
In the present case, the node or vertices (V) represents the the Adjacency matrix (A).
system buses and the edge (E) represents the distribution lines  Step 4: Checking Radiality of the system. In this step the graph is
(connecting two buses). In graph theory, two different types of traversed in DFS order, starting from source node1 to nth node
graphs are normally used i.e. directed graph and undirected in the system [command: graphtraverse ()] and the path is
graph. A directed graph is formed by vertices connected with di- mapped out [command: graphpred2path()]. If the length of the
rected edges. In directed graph, there is always a certain direction graphpred2path() is found equal to 2, then the circuit is
or flow in the graph between the nodes. While in undirected believed to be a radial system (status = 1), otherwise non-radial.
graph, the nodes are connected to each other in a bidirectional In case of non-radiality in the system (status = 0), the program
manner. exits and Step 3 is repeated.
In the present problem of checking radiality using graph theory,  Step 5: Checking connectivity among nodes. If status==1, the
it is important to create an adjacency matrix which will contain all DFS order of the radial graph is obtained and stored in variable
the information among connectivity of different nodes or buses. To radial_pathmapped. For example in Fig. 5, the following
create the adjacency matrix, a sparse formation will be done using radial_pathmapped is obtained, given by Eq. (18). The length
the linedata. Eq. (16) represents the adjacency matrix for the direc- of radial_pathmapped must be equal to the system total nodes
ted graph according to Fig. 5. Here it can be seen that each element (nbus(traverse) = nbus(system)).
in adjacency matrix aij = 1 if (i, k) 2 E and aij = 0 if (i, k) R E, where E
is a set of directed edges {(1, 2),(2, 3), (3, 4),(2, 5),(5, 6)}. 0 1
1
0 1 B C
0 1 0 0 0 0 B2C
B C B C
B C B C
B0 0 1 0 1 0C B C
B C B3C
B C B C
radial pathmapped ¼ B C ð18Þ
B C B4C
B0 0 0 1 0 0C B C
B C B C
A¼B C ð16Þ B C
B C B5C
B0 0 0 0 0 0C @ A
B C
B C
B0 0 0 0 0 1C 6
B C
@ A
 Step 6 Carried out load flow analysis and compute the maxi-
0 0 0 0 0 0 mum loadability of the system (kmax).
Different tie-switch combinations will result in different maxi-
Eq. (17) represents the adjacency matrix for the undirected mum loadability, thus the optimum tie switch configuration will
graph according to Fig. 5. Here it can be seen that each element be found out by using the discrete form of Artificial Bee Colony
in adjacency matrix aij = 1 if (i, k) 2 E and aij = 0 if (i, k) R E, where (ABC) algorithm.
E is a set of undirected edges {(1, 2),(2, 1),(2, 3),(2, 5),(3, 2),(3, 4),
(4, 3),(5, 2),(5, 6),(6, 5)}. For undirected graph, the adjacency matrix 4. Discrete Artificial Bee Colony (DABC)
is symmetric since the edges are bi-directional i.e. aij = aji.
0 1 Artificial bee colony (ABC) is one of the most recently pro-
0 1 0 0 0 0 posed algorithm in optimization, introduced by Karaboga and
B C
B C Basturk [27]. ABC is inspired from the foraging behavior of
B1 0 1 0C
1 0
B C honeybees. The bee colony consists of three different groups
B C
B C of bees including employed bees, onlookers and scouts. Initially
B0 1 0 1 0 0C
B C employed bees and onlookers are divided into halves. One
A¼B C ð17Þ
B0 0 1 0 0 0C employed bee is employed for every single food source.
B C
B C Employed bee from their memory determined a food source
B C
B0 1 0 0 0 1C within the neighborhood of the food source. Employed bee
B C
@ A exchange information with onlookers within the hive and then
0 0 0 0 1 0 the onlookers select one of the food sources. When employed
bee has an abandoned of food source, it becomes a scout
In the present case of solving network reconfiguration problem, bee and starts to search a new food source randomly. Thus
undirected graph is used. Here it can be observed that using undi- a scout bee is an x-employed bee.
 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133 127

Table 1
Proposed Discrete Artificial Bee Colony (DABC) method for optimum tie switch positions.

Parent solution a1 a2 a3 a4 a5
The randomization process is started
First run
First tie position is randomly replaced xi1 a2 a3 a4 a5
Second tie position is randomly replaced a1 xi2 a3 a4 a5
Third tie position is randomly replaced a1 a2 xi3 a4 a5
Fourth tie position is randomly replaced a1 a2 a3 xi4 a5
Fifth tie position is randomly replaced a1 a2 a3 a4 xi5
The randomization process is again repeated
Second run
First tie position is randomly replaced yi1 a2 a3 a4 a5
Second tie position is randomly replaced a1 yi2 a3 a4 a5
Third tie position is randomly replaced a1 a2 yi3 a4 a5
Fourth tie position is randomly replaced a1 a2 a3 yi4 a5
Fifth tie position is randomly replaced a1 a2 a3 a4 yi5

The operation of Discrete Artificial Bee Colony (DABC) is nearly

similar to the original ABC except in terms of searching new food
searching procedure (mutation). In the DABC, the Employed and
Onlooker Bees will search the new food source at the neighborhood
location that exists in their memory in greedy manner. The muta-
tion process of DABC in case of solving network reconfiguration is
shown in Table 1.
Here it can be observed that in DABC certain level of heuristic
knowledge is retained based upon last feasible solution to ensure
fast convergence in the optimization process. In mutation process
of DABC, each parent solution [a1, a2, a3, a4, a5] is mutated in diag-
onal manner to obtain the next generation. A single line switch
(e.g. a1) is randomly changes at each ith run (a1 ? xi1). Here xi rep-
resents the new tie switch for the first run and yi represents the
new tie switch for the second run that is randomly generated.
The continuation power flow is run for each successful switch com-
bination (radial system only) and maximum loadability margin is
measured. The process of searching by employed, onlooker and
scout bees are continue until it reaches the maximum number of
iteration. Fig. 6. Effect of network reconfiguration on kVA margin to maximum loadability.

5. Voltage Deviation Index (VDI) and kva margin to maximum

loadability (KMML) From Fig. 6, it can be observed that the curve A (with optimum
network reconfiguration) will have a better voltage profile than
In [8], the author has developed the Voltage Deviation Index curve B at each loading. Further it can also be noted that using
(VDI) to quantify the power quality of the distribution network. the optimum network reconfiguration, the operating point of the
For an N-bus radial network, VDI is defined in: system (kV) can be increased from O1 to O2 (within the allowable
branch current limits (ILimit) and voltage limits VLimits).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PNVB 2
i¼1 ðV i  V LimitðiÞ Þ 6. Proposed algorithm for optimum network reconfiguration
VDI ¼ ð19Þ
N
In this section, a new algorithm is proposed for optimum net-
where NVB are the number of buses that violate the voltage limits. work reconfiguration on the basis of maximization of system loa-
VLimit is the upper or lower limit of the ith bus voltage magni- dability. CPF will be used to find the maximum loading margin.
tude when the respective upper or lower limit violation occurs, is The CPF results will also be compared with radial load flow analy-
given by sis [29]. To find the maximum loadability or loading factor (kmax) of
the system, the active and reactive load is increased on all buses
V min  V Limit  V max ð20Þ (using Eq. (22)) with equal loading factor of 0.01, till the divergence
is observed.
In the present case,

Pnew ¼ P0  Loading FactorðkÞ
ð22Þ
0:95  V Limit  1:05 ð21Þ Q new ¼ Q 0  Loading FactorðkÞ

kVA Margin to Maximum Loadability (KMML) is defined in [28] to where k is a loading factor, Po and Qo is initial active and reactive
represent the additional load from the operating point ‘O’ to the power load, connected with ith bus, and Pnew and Qnew is final active
point of voltage collapse, as shown in Fig. 6. and reactive power load, connected with ith bus.
128 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133

Maximization of system loadability is considered as fitness The additional constraints include, there should be no looping
function, given by Eq. (23). in the system and all nodes must be connected with the source
node.
f ¼ Maxfkmax g ð23Þ

status ¼¼ 1 ðradial systemÞ
where f is a fitness function and kmax is the maximum system load- ð25Þ
ing factor or loadability. status ¼¼ 0 ðnon-radial systemÞ
Subjected to following line constraints,
Different tie switch positions will result in radial as well non-radial
Iimin  IiLimit  Iimax ð24Þ system, graph theory will select only radial distribution system. The
DABC algorithm is applied on radial distribution test systems to find
where ILimit is a ith branch current limit. the optimum tie switch positions, which will result in maximum

Fig. 7. Flow chart of proposed algorithm.

 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133 129

Fig. 8. Radial distribution test systems with 5 tie switches.

Table 2
Proposed method results for optimum network reconfiguration for 33-bus test system.

Test case Base case Method [28] Proposed method using radial load flow Proposed method using CPF
Open tie-switches 33,34,35,36,37 06,14,09,32,37 10,14,32,28,06 10,14,32,28,06
System loading margin kMax 3.41 5.1 5.52 5.50
kV
0.5500 0.8700 0.9100 0.90
Power loss (MW) 0.21099 0.12812 0.12952 0.12900
Vmax 1 1 1 1
Vmin 0.9038 0.9416 0.9445 0.9445
KMML 10530.14 17914.34 19749.46 19662.08
NBVV 21 4 3 3
VDI 0.0245 0.0021 0.0013 0.0013
130 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133

loadability of the system. The complete flow chart of the proposed method [28]. In [28], the author has implemented the MLI based
algorithm is shown in Fig. 7. algorithm on 33-bus radial distribution network.
From Table 2, following points are observed:

7. Results and discussion
To test the proposed algorithm 3-phase, 12.66 kV standard 33-
bus [28] and 69-bus [30] radial distribution systems are used. The
base load in 33-bus system is 4369.35 kVA and in 69-bus system is
4659.67 kVA. The other system details including bus and line data
is given in Appendix A. The single line diagram of the system is
shown in Fig. 8.
When the proposed algorithm is tested on 33-bus radial distri-
bution system, results presented in Table 2 are obtained. The ob-
served results are also compared with existing Venkatesh
1. Improving the maximum system loadability has improved the
KMML from 10530.14 kVA to 19749.46 kVA (which is higher
than the KMML of existing method [28] by 10.24%).
2. The loading margin factor (k) is also calculated from 0 up to
the voltage limit (kV), satisfying line constraints also. It was
found that the base case can carry only 2403.14 kVA
(kV = 0.55) without violating the voltage and line constraints.
However the proposed method can carry up to 3976.10 kVA
(kV = 0.91), and the method [28] can sustain only
3801.33 kVA (kV = 0.87). Thus the proposed method can carry
1572.96 kVA more power than then base case and 175 kVA
more power than the method proposed in [28]. These results
are also summarized in Fig. 9.
3. The proposed method was also found better than the method
[28] in terms of better voltage profile (shown in Fig. 10) and
number of buses violating voltage limits (NBVV). The VDI in
case of method [28] has improved 91.4%, however using the
proposed method VDI has improved 94.7% in comparison with
base case.

Table 3
Proposed method results for optimum network reconfiguration for 69-bus test
system.

Test case Base case Proposed method

(optimum results*)
Open tie-switches 69, 70, 71, 72, 73 69, 14, 70, 55, 61
System loading margin kMax 3.22 5.5
kV
0.5800 1
Power loss (MW) 0.22495 0.09859
Vmax 1 1
Vmin 0.9092 0.9495 ( 0.95)
KMML 10344.47 20968.52
NBVV 9 0
VDI 0.0119 0
*
The optimum results with minimum power losses are presented.
Fig. 9. System loading curve for 33-bus radial distribution system.

Fig. 10. Voltage profile curve of 33-bus test system. Fig. 11. System loading curve for 69-bus radial distribution system.
 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133
The proposed algorithm is also applied on 69-bus radial distri-
bution test system and the following results are obtained, tabu-
lated in Table 3.
Fig. 12. Voltage profile curve of 69-bus test system.
Fig. 13. Bus voltages and available line capacity limit curve at point of allowable voltage limits.
131
From Table 3, following points can be concluded:
1. Improving the system loadability has improved the KMML up to
100% from 10530.14 kVA to 19749.46 kVA in comparison with
base case.
2. The loading margin factor (k) is also calculated from 0 up to the
voltage limit (kV). It was found that the base case can carry only
2702.61 kVA (kV = 0.58) without violating the voltage and line
constraints. However the proposed method has improved the
system loading up to 42% (4659.67 kVA corresponding to
kV = 1) without violating the voltage and line constraints. These
results are also summarized in Fig. 11.
3. The overall system voltage profile (shown in Fig. 12) and
voltage quality has been improved from VDI equal to
0.0119 to 0. The minimum bus voltage in base case was
found to be 0.9092 V at bus 65, however using the proposed
method the minimum voltage is found equal to 0.95 V at
bus 61.
4. Fig. 13 shows the relationship between bus voltages and avail-
able line capacity (ILimit  Iflowing).
From Fig. 13, it can be observed that although all lines can carry
more current, however voltage constraints is limiting the system
operating point at 2702.61 kVA (kV = 0.58) in case of base case
and 4659.67 kVA (kV = 1) in optimum case respectively. This also
shows that voltage limits come earlier than the line capacity of
the system.
132 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133

8. Conclusion Table A2
69-Bus test system data.

This paper has presented a new algorithm for network reconfig- From bus To bus Ri,i + 1 (X) Xi,i + 1 (X) P (kW) Q (kVAR)
uration based on maximization of system loadability. Discrete 1 2 0.0005 0.0012 0 0
Artificial Bee Colony (DABC) is used here to find the optimum 2 3 0.0005 0.0012 0 0
tie-switch combinations and the graph theory is used to maintain 3 4 0.0015 0.0036 0 0
4 5 0.0251 0.0294 0 0
the radiality of the system. The proposed algorithm is tested on a
5 6 0.366 0.1863 2.6 2.2
33-bus and 69-bus radial distribution test systems. The proposed 6 7 0.381 0.1941 40.4 30
algorithm has also been compared with existing method which 7 8 0.0922 0.047 75 54
was based on a voltage stability index. Results have shown that 8 9 0.0493 0.0251 30 22
using the proposed method kVA margin to maximum loadability 9 10 0.819 0.2707 28 19
10 11 0.1872 0.0619 145 104
(KMML) has been improved by 100% (in comparison with base
11 12 0.7114 0.2351 145 104
case), overall system voltage profile has been improved, numbers 12 13 1.03 0.34 8 5
of buses violating voltage limits have been reduced significantly, 13 14 1.044 0.345 8 5.5
system operating point has been improved by 40% and thus the 14 15 1.058 0.3496 0 0
15 16 0.1966 0.065 45.5 30
system is able to handle more power system load (kVA) without
16 17 0.3744 0.1238 60 35
violating the bus voltages and line current constraints. 17 18 0.0047 0.0016 60 35
18 19 0.3276 0.1083 0 0
19 20 0.2106 0.069 1 0.6
Acknowledgments 20 21 0.3416 0.1129 114 81
21 22 0.014 0.0046 5 3.5
This work was supported by the Bright Spark Programme of 22 23 0.1591 0.0526 0 0
23 24 0.3463 0.1145 28 20
University of Malaya and HIR/MOHE research grant (Grant Code:
24 25 0.7488 0.2475 0 0
D000004-16001). 25 26 0.3089 0.1021 14 10
26 27 0.1732 0.0572 14 10
3 28 0.0044 0.0108 26 18.6
Appendix A 28 29 0.064 0.1565 26 18.6
29 30 0.3978 0.1315 0 0
30 31 0.0702 0.0232 0 0
See Tables A1 and A2. 31 32 0.351 0.116 0 0
32 33 0.839 0.2816 14 10
Table A1 33 34 1.708 0.5646 19.5 14
33-Bus test system data. 34 35 1.474 0.4873 6 4
3 36 0.0044 0.0108 26 18.55
From bus To bus Ri,i + 1 (X) Xi,i + 1(X) P (kW) Q (kVAR)
36 37 0.064 0.1565 26 18.55
1 2 0.0922 0.0477 100 60 37 38 0.1053 0.123 0 0
2 3 0.493 0.2511 90 40 38 39 0.0304 0.0355 24 17
3 4 0.366 0.1864 120 80 39 40 0.0018 0.0021 24 17
4 5 0.3811 0.1941 60 30 40 41 0.7283 0.8509 1.2 1
5 6 0.819 0.707 60 20 41 42 0.31 0.3623 0 0
6 7 0.1872 0.6188 200 100 42 43 0.041 0.0478 6 4.3
7 8 1.7114 1.2351 200 100 43 44 0.0092 0.0116 0 0
8 9 1.03 0.74 60 20 44 45 0.1089 0.1373 39.22 26.3
9 10 1.04 0.74 60 20 45 46 0.0009 0.0012 39.22 26.3
10 11 0.1966 0.065 45 30 4 47 0.0034 0.0084 0 0
11 12 0.3744 0.1238 60 35 47 48 0.0851 0.2083 79 56.4
12 13 1.468 1.155 60 35 48 49 0.2898 0.7091 384.7 274.5
13 14 0.5416 0.7129 120 80 49 50 0.0822 0.2011 384.7 274.5
14 15 0.591 0.526 60 10 8 51 0.0928 0.0473 40.5 28.3
15 16 0.7463 0.545 60 20 51 52 0.3319 0.1114 3.6 2.7
16 17 1.289 1.721 60 20 9 53 0.174 0.0886 4.35 3.5
17 18 0.732 0.574 90 40 53 54 0.203 0.1034 26.4 19
2 19 0.164 0.1565 90 40 54 55 0.2842 0.1447 24 17.2
19 20 1.5042 1.3554 90 40 55 56 0.2813 0.1433 0 0
20 21 0.4095 0.4784 90 40 56 57 1.59 0.5337 0 0
21 22 0.7089 0.9373 90 40 57 58 0.7837 0.263 0 0
3 23 0.4512 0.3083 90 50 58 59 0.3042 0.1006 100 72
23 24 0.898 0.7091 420 200 59 60 0.3861 0.1172 0 0
24 25 0.896 0.7011 420 200 60 61 0.5075 0.2585 1244 888
6 26 0.203 0.1034 60 25 61 62 0.0974 0.0496 32 23
26 27 0.2842 0.1447 60 25 62 63 0.145 0.0738 0 0
27 28 1.059 0.9337 60 20 63 64 0.7105 0.3619 227 162
28 29 0.8042 0.7006 120 70 64 65 1.041 0.5302 59 42
29 30 0.5075 0.2585 200 600 11 66 0.2012 0.0611 18 13
30 31 0.9744 0.963 150 70 66 67 0.0047 0.0014 18 13
31 32 0.3105 0.3619 210 100 12 68 0.7394 0.2444 28 20
32 33 0.341 0.5302 60 40 68 69 0.0047 0.0016 28 20
21 8 0 2 – – 11 43 0.5 0.5
9 14 0 2 – – 13 21 0.5 0.5
12 22 0 2 – – 15 46 1 0.5
18 33 0 0.5 – – 50 59 2 1
25 29 0 0.5 – – 27 65 1 0.5

Other data: Current carrying capacity of branches 1–5 is 400 A, 6–7 and 25–27 is Other data: Current carrying capacity of branches 1–9 is 400 A, 46–49 and 52–64 is
300 A and for all other branches including ties line is 200 A. 300 A and for all other branches including ties line is 200 A.
 M.M. Aman et al. / Electrical Power and Energy Systems 54 (2014) 123–133 133

References [15] Rao RS, Ravindra K, Satish K, Narasimham SVL. Power loss minimization in
distribution system using network reconfiguration in the presence of
distributed generation. IEEE Trans Power Syst 2013;28:317–25.
[1] Wan H, McCalley JD, Vittal V. Increasing thermal rating by risk analysis. IEEE
[16] El-Dib AA, Youssef HK, El-Metwally M, Osman Z. Maximum loadability of
Trans Power Syst 1999;14:815–28.
power systems using hybrid particle swarm optimization. Electric Power Syst
[2] Prada RB, Souza LJ. Voltage stability and thermal limit: constraints on the
Res 2006;76:485–92.
maximum loading of electrical energy distribution feeders. IEE Proc Gener,
[17] Hedayati H, Nabaviniaki S, Akbarimajd A. A method for placement of DG units
Transm Distrib 1998;145:573–7.
in distribution networks. IEEE Trans Power Deliv 2008;23:1620–8.
[3] Civanlar S, Grainger J, Yin H, Lee S. Distribution feeder reconfiguration for loss
[18] Al Abri RS, El-Saadany EF, Atwa YM. Optimal placement and sizing method to
reduction. IEEE Trans Power Deliv 1988;3:1217–23.
improve the voltage stability margin in a distribution system using distributed
[4] González A, Echavarren F, Rouco L, Gómez T, Cabetas J. Reconfiguration of
generation. IEEE Trans Power Syst 2012:1.
large-scale distribution networks for planning studies. Int J Electr Power
[19] Ajjarapu V, Christy C. The continuation power flow: a tool for steady state
Energy Syst 2012.
voltage stability analysis. IEEE Trans Power Syst 1992;7:416–23.
[5] Baran ME, Wu FF. Network reconfiguration in distribution systems for loss
[20] Milano F. Power system modelling and scripting. Springer Verlag; 2010.
reduction and load balancing. IEEE Trans Power Deliv 1989;4:1401–7.
[21] Rosero MJ, Rios MA. Characterization of the maximum loadability in power
[6] Shirmohammadi D, Hong HW. Reconfiguration of electric distribution
systems due to contingencies in the operative planning scenario. Power Tech,
networks for resistive line losses reduction. IEEE Trans Power Deliv
2007 IEEE Lausanne 2007:1272–7.
2002;4:1492–8.
[22] Hedayati H, Nabaviniaki SA, Akbarimajd A. A method for placement of DG
[7] Zhu JZ. Optimal reconfiguration of electrical distribution network using the
units in distribution networks. IEEE Trans Power Deliv 2008;23(3):1620–8.
refined genetic algorithm. Electric Power Syst Res 2002;62:37–42.
[23] Gözel T, Eminoglu U, Hocaoglu MH. A tool for voltage stability and
[8] Venkatesh B, Ranjan R, Gooi H. Optimal reconfiguration of radial distribution
optimization (VS&OP) in radial distribution systems using matlab graphical
systems to maximize loadability. IEEE Trans Power Syst 2004;19:260–6.
user interface (GUI). Simul Modell Pract Theory 2008;16:505–18.
[9] Zhu J, Xiong X, Zhang J, Shen G, Xu Q, Xue Y. A rule based comprehensive
[24] Harju T. Graph theory; 2011. <http://cs.bme.hu/fcs/graphtheory.pdf>
approach for reconfiguration of electrical distribution network. Electric Power
[accessed 05.06.13].
Syst Res 2009;79:311–5.
[25] Ruohonen K. Graph theory; 2008. <http://math.tut.fi/~ruohonen/
[10] Srinivasa Rao R, Narasimham SVL, Ramalinga Raju M, Srinivasa Rao A. Optimal
GT_English.pdf> [accessed 05.06.13].
network reconfiguration of large-scale distribution system using harmony
[26] Agarwal RC, Aggarwal CC, Prasad VVV. A tree projection algorithm for
search algorithm. IEEE Trans Power Syst 2010:1.
generation of frequent item sets. J Parallel Distrib Comput 2001;61:350–71.
[11] Sathish Kumar K, Jayabarathi T. Power system reconfiguration and loss
[27] Karaboga D, Basturk B. A powerful and efficient algorithm for numerical
minimization for an distribution systems using bacterial foraging optimization
function optimization: artificial bee colony (ABC) algorithm. J Global Optim
algorithm. Int J Electr Power Energy Syst 2012;36:13–7.
2007;39:459–71.
[12] Yuan-Kang W, Ching-Yin L, Le-Chang L, Shao-Hong T. Study of reconfiguration
[28] Venkatesh B, Ranjan R, Gooi H. Optimal reconfiguration of radial distribution
for the distribution system with distributed generators. IEEE Trans Power
systems to maximize loadability. IEEE Trans Power Syst 2004;19:260–6.
Deliv 2010;25:1678–85.
[29] Thukaram D, Wijekoon Banda H, Jerome J. A robust three phase power flow
[13] de Oliveira LW, Carneiro Jr S, de Oliveira EJ, Pereira JLR, Silva Jr IC, Costa JS.
algorithm for radial distribution systems. Electric Power Syst Res
Optimal reconfiguration and capacitor allocation in radial distribution systems
1999;50:227–36.
for energy losses minimization. Int J Electr Power Energy Syst 2010;32:840–8.
[30] Baran M, Wu FF. Optimal sizing of capacitors placed on a radial distribution
[14] Zin AAM, Ferdavani AK, Khairuddin AB, Naeini MM. Reconfiguration of radial
system. IEEE Trans Power Deliv 1989;4:735–43.
electrical distribution network through minimum-current circular-updating-
mechanism method. IEEE Trans Power Syst 2012;27:968–74.