International Journal of

Electrical Power and Energy Systems

Manuscript Draft

Manuscript Number: IJEPES-D-13-01721

Title: Service Restoration for Unbalanced Distribution Networks Using a
Combination Two Heuristic Methods

Article Type: Research Paper

Keywords: service restoration, unbalanced distribution network, switches
index, graph-based, three phase load flow

Abstract: In this paper, two heuristic methods are proposed to find
effective and fast solution in unbalanced three phase distribution
networks. Switch selection indices based on analytically approach and
practicable heuristic graph-based method are proposed for solving the
service restoration problem in unbalanced distribution networks. The
problem formulation proposed, consists of three different objective
functions: First, minimizing the de-energized customers' load, second,
minimizing the number of switching operation, and finally, customer's
priority. A suitable assignment of switch indices to all tie switches
(ts) in networks are used to find best solution and decrease number of
switching operation. New graph-based approach for finding best
sectionalizes switch (ss) and minimizing voltage drop's amount is
utilized. The validity of these approaches has been tested on the two
unbalanced three phase distribution networks. Results have been presented
for modified IEEE 13-node and IEEE 37-node test case. The fastness and
effectiveness convergence of these approaches helps finding best solution
for service restoration problem.




1

Service Restoration for Unbalanced Distribution Networks Using a Combination Two
Heuristic Methods
Meysam Gholami
a,*
, Jamal Moshtagh
b
a,*
,Department of electrical engineering, university of Kurdistan, sanandaj, Iran.
b
, Department of electrical engineering, university of Kurdistan, sanandaj, Iran.
(
a
,m.gholami.67.86@gmail.com, tel:+98 9181706749)
(
b
,moshtagh79@yahoo.com, tel:+98 8716660073)
ABSTRACT
In this paper, two heuristic methods are proposed to find effective and fast solution in
unbalanced three phase distribution networks. Switch selection indices based on analytically
approach and practicable heuristic graph-based method are proposed for solving the service
restoration problem in unbalanced distribution networks. The problem formulation proposed,
consists of three different objective functions: First, minimizing the de-energized customers
load, second, minimizing the number of switching operation, and finally, customers priority.
A suitable assignment of switch indices to all tie switches (ts) in networks are used to find
best solution and decrease number of switching operation. New graph-based approach for
finding best sectionalizes switch (ss) and minimizing voltage drops amount is utilized. The
validity of these approaches has been tested on the two unbalanced three phase distribution
networks. Results have been presented for modified IEEE 13-node and IEEE 37-node test
case. The fastness and effectiveness convergence of these approaches helps finding best
solution for service restoration problem.
Key wordsservice restoration, unbalanced distribution network, switches index, graph-
based, three phase load flow.
*Manuscript
Click here to view linked References
2

1. Introduction
With significant extension of the modern power distribution networks in the world, the
likelihood of occurrence of fault and then blackout for one or more area will increase.
Therefore, customers satisfaction and service reliability are the important topic where most
of the paper localizes in this issue. Revenue earned by the Power Distribution Companies and
customers satisfaction is closely depending on reliability in distribution networks. In order to
satisfy users demand and maintain profit of power Supply Company, it is necessary to
restoring power service as soon as possible [1]. Due to a high number of switches, feeders
and branches in typical distribution systems, it is not easy to restore an out-of-service area
solely depending on the past experiences of human operators [2]. Therefore, with the advent
of quick computers and changing technology, to reduce the out of service area as efficiently
as possible, a computer aided decision supports assist the operators. How to arriving a fast
and effective service restoration in power distribution networks (PDNs), considering
unbalanced distribution network is of major concern in this paper. Protection devices in
network detect the fault location, when a fault is occurred in the PDN. After isolating the
fault by operation line switches, the PDN is divided to three sections: First, the upstream
section that is supplied from same feeder, second, the downstream un-faulted section that are
transferred to neighboring feeders according to presented approaches in this paper, and
finally, damage buses and lines that are isolated from network. In service restoration problem,
several issues must also be considered that are described as follow:
- Service restoration plan must be restored maximum safe out-of-service loads.
- Service restoration plan is implemented by changing switches state in PDNs,
therefore, the time taken by the service restoration depends on the number of
switching operation. Therefore, the number of switching operation should be kept to
minimum as possible.
3

- Hospital, police station, firehouse, etc, are the highest priority in PDNs. This issue
must be considered in service restoration plan that the supply must be restored to
highest priority customers.
- In each PDN, most important constraint is radial structure due to various reasons,
such as ease of fault location detection, fault isolation and protective devices
coordination. When the structure of the network is changed during the service
restoration, this constraint must be kept on.
- Buses voltage, lines current and elements loading also changes during the service
restoration plan. Therefore, it is important that these constraints dont cross their
respective operational limits.
- Customers satisfaction and reliability of distribution networks are closely dependent
on interruption frequency and duration. Therefore, the restoration plan runtime must
be minimized for finding a quick solution.
In past years, many methods have been proposed to find solution for restoration problem
from different perspective. Considering complexity PDNs, analytic method for solving the
restoration problem can hardly be applied. Therefore, heuristic search method [3-8] or expert
system approach [9] have been adopted. In [10] G-net inference mechanism with operation
rules is applied. In [11] Petri Net combined with a rule-based expert systems have been
applied to implement the service restoration. In [12] fuzzy cause-effect networks are used to
model the heuristic knowledge inference involved in the restoration plan. In [13] a fuzzy
decision-making approach has been applied to determine the most desirable restoration plan
with consideration different practical factors, but fuzzy method doesnt guarantee the optimal
solution. In [14] non- dominated sorting genetic algorithm-II (NSGA-II) for solving the
service restoration problem is presented and to reduce the software runtime, a faster version
of NSGA-II has been implemented. In [15] mathematical programming is presented to
4

reconfigure the network to restore un-faulted section of the system. In [16, 17] combination
methods are applied. In [16] objective functions are modeled with fuzzy sets, and then
optimization problem is solved by the Genetic Algorithm (GA). [17] consist of two stages:
First, the fuzzy multi criteria evaluation, and afterward, the Grey relational analysis. In [18],
service restoration with Load curtailment of in-service customers via direct load control has
been implemented. In [19], reliability assessment of complex radial distribution systems is
presented, however this paper simplifies the restoration plan. In [20, 21], restoration problem
in distribution network with dispersed generation is implemented. In this paper, a fast and
effective methods based on two new heuristic algorithms for service restoration in
unbalanced PDNs is presented. Unbalanced distribution network, customers priority, buses
voltage, lines current, equipments loading and minimum software runtime consideration, are
the main features of the proposed method.
This paper is organized as follows: Section 2 describes the problem formulation of a
typical restoration problem. In section 3, indices for ranking the networks switch are
described. In section 4, graph-based method is described. Section 5 reviews two new heuristic
algorithms for service restoration. Section 6 briefly describes three-phase load flow program
for fast response to the network change inducted by system reconfiguration. Section 7 shows
a numerical example to demonstrate the fastness and effectiveness of the proposed methods
and the conclusion are drawn in section 8 finally.

2. Problem formulation
Service restoration in unbalanced distribution networks considering customers priority
are formulated as multi-constraint and multi-objective optimization problem. In this paper,
three different objective functions are presented. Maximizing total load to be restored,
minimizing the number of the switching operations and maximizing priority load restored are
5

these objective functions. Besides, important constraints consists of network radial structure,
buses voltage, lines current, equipments loading have also been considered in this paper.
Objective function briefly:
(1)
(2)
(3)
Where
Energized loads in network;
Total buses are restorable;
Buses with high priority those are restorable;
Number of switching operation;
Constraints:
1) Radial network structure should be maintained.
2) Bus voltage limits (for all buses):
(4)
Where
Minimum acceptable bus voltage;
Voltage at bus k, phase p;
Maximum acceptable bus voltage.
3) Line current limits (for all lines):
(5)
Where
Minimum acceptable line current;
Current in line j, phase p;
Maximum acceptable line current.

e
t
N k
k
L max

e
HP
N k
k
L max
op
N min
k
L
t
N
HP
N
op
N
max min
k
p
k k
V V V < <
min
k
V
p
k
V
max
k
V
max min
j
p
j j
I I I < <
p
j
I
max
j
I
min
j
I
6

4) Equipment loading limits (for transformers):
(6)
Where
Loading for i transformer;
Rated loading for i transformer.
Operational constraint can be obtained from three phase load flow calculation. In actual
practice, the minimum limit of the line current should not be taken and, in fact, this limit has
been taken as zero [14]. In this study, we are used two heuristic approaches based on switch
indices and graph-based for finding best solution to restore maximum total customers in de-
energized area. In most restoration plan, there are several plans available for a restoration
problem, however, how to select best plan, two methods must describe. These methods are
presented in next sections.

3. Switch ranking
In this paper, two switch indices for best selection have been used. the base of the
proposed algorithm for these indices is voltage drop. A first and most important index is VD
that is proportionate with voltage drop between substation bus and primary side of each tie
switch (ts). For each ts, VD is defined as follow:
(7)
Where
Sum of active loads between substation bus and primary side of tie switch i, for each
three phase;
Sum of reactive loads between substation bus and primary side of tie switch i, for
each three phase;
Sum of real segment of positive impedance sequence of lines between substation bus
max
i
tr
i
tr
max
i i
tr tr <
c b a p
V
X Q R P
VD
i
p
i i
p
i
, , =
+
=
p
i
P
p
i
Q
i
R
7

and primary side of i tie switch;
Sum of imaginary segment of positive impedance sequence of lines between
substation bus and primary side of i tie switch;
and V is substation voltage.
This index is shown in Fig.1. Suppose that one fault took place at point A. Therefore,
area1 is downstream un-faulted area and ts1 is one candidate switch for service restoration
implementation. For ts1, VD is obtained from node number 5, 6, that is proportionate with
direction1 (dir1). ts3 is another candidate switch for service restoration implementation. For
ts3, VD is obtained from node number 10, 12 and 13 that is proportionate with direction2
(dir2). A second index (Z
path
) is direction impedance (per-unit) for lines lying in the path
between the secondary side of each ts and end buses in network. For each ts, this index is
defined as follow:
(8)
Where Z
b
is positive impedance sequence of branch b and N
br
is lines lying in the path
between the secondary side of each ts and end buses in network. This index is shown in Fig.1.
Suppose that one fault took place at point B. Therefore, area2 is downstream un-faulted area
and ts2 is one candidate switch for service restoration implementation. For ts2, Z
path
is
impedance of direction3 and direction4 (dir3 and dir4). These utilized indices, help both
reducing the solution search space and ranking switches to find best solution for service
restoration plan. Indeed, when search space is reduced then runtime software will decrease
and this issue makes guarantee to fast service restoration implementation. In next section
graph-based method is described and how to utilize these methods is described in section 5.

4. Graph-based method

e
=
br
N b
b path
Z Z
i
X
8

Before description this method for finding the minimum voltage drop in network (tree or
graph) we must define three groups. For any bus (node) belongs to the graph, the following
definition is valid:
- Sub-graph D, which consists of nodes that have already been added to the graph.
- Sub-graph UD, which consists of nodes that have not been added to the graph.
- Sub-graph B, which consists of branches that can connect nodes from sub-graph D.
When one branch from tree with minimal weights is selected in each iteration, one node
(node S) of each branch is in sub-graph D and the other node (node R) is in sub-graph UD.
Therefore, node R is added to the sub-graph D and deleted from sub-graph UD. This method
is shown in Fig. 2. This approach is utilized for finding best sectionalizes switch (ss) after
finding best tie switches (ts). The evaluation of weighting coefficient has been described in
follow:
(9)
Where
X is branch state (0 for close branches and inf for open branches), Z
b
is positive sequence
impedance of branch b, N
br
is branches lying in the path between the secondary node (node
R) of branch is in sub-graph UD and substation node. The starting state and the major steps in
the first iteration has been described as follows:
Starting state:


1, 2 are branches between sub-graph D and sub-graph UD.
First iteration (selection of branch with minimum weight, (a) in Fig. 2):
Weighting coefficient for branch 1 and 2:
W
1
= X
1
+Z
1
and W
2
= X
2
+Z
2

e
+ =
Nbr b
b
Z X W
{ }
{ }
{ } B
UD
D
e
e
e
2 , 1
12 , 11 , 10 , 9 , 8 , 7 , 6 , 5 , 4 , 3 , 2
1
9

Therefore, branch with minimum of W is selected (suppose this is W
1
).
Updating sub-graphs:


Second iteration (selection of branch with minimum weight, (b) in Fig. 2):
Weighting coefficient for branch 2, 3 and 4:
W
2
= X
2
+Z
2
, W
3
=X
3
+Z
1
+ Z
3
and W
4
= X
4
+Z
1
+ Z
4

Therefore, branch with minimum of W is selected (suppose this is W
2
).
Updating sub-graphs:


Updating D, UD and B will be done for next iteration. When sub-graph UD was emptied,
the iteration procedure is finished and after last iteration, all branches belong sub-graph B
must be opened. This state that is final structure is shown in (f), Fig. 2.

5. Service restoration algorithm
To describe the problem of multi-objective service restoration in distribution networks,
the faults scenario must be described. When a short-circuits fault is occurred on the feeder,
circuit breaker at the outset of feeder is operated to clear the fault. All boundary line switches
are operating to isolate the faulted area. The feeders circuit breaker is then closing to restore
the upstream customers. For the downstream area, best switch indices for best switches
selection based on first heuristic approach is implemented. This approach is described in this
section. The proposed approach is calculated fast and implemented using remotely controlled
switches in unbalanced distribution networks. In this paper, three objective functions are
prioritizing as: 1) maximizing the amount of total load to be restored, 2) minimizing the
{ }
{ }
{ } B
UD
D
e
e
e
4 , 3 , 2
12 , 11 , 10 , 9 , 8 , 7 , 6 , 5 , 4 , 3
2 , 1
{ }
{ }
{ } B
UD
D
e
e
e
5 , 4 , 3
12 , 11 , 10 , 9 , 8 , 7 , 6 , 5 , 4
3 , 2 , 1
10

number of the switching operation and 3) customers priority consideration. Proposed
algorithm considering five steps are described in follow.
Step 1) isolation the fault;
Step 2) creation the candidate tie switches list;
Step 3) selection one ts due to first proposed algorithm, three phase unbalanced load flow
implementation, and network constraints survey. If no constraints violation
exists, go to step 5;
Step 4) selection next ts due to first proposed algorithm, selection respective sectionalize
switch (ss) due to graph-based method, three phase unbalanced load flow
implementation and network constraints survey. If no constraints violation exists
go to step 5, else repeat this step;
Step 5) return best service restoration plan.
Step 1) isolating the fault
When a fault is occurring in each PDN, faults line, sending and receiving bus sides for
this line are detected. The adjacent buses and lines are wended in each direction sequentially
and to clear the fault, first circuit breaker in these directions is founded and operated.
Therefore, feeder that fault has been occurred on it and feeders circuit breaker is detected.
For fault isolation, the adjacent buses and lines are wended in each direction sequentially.
The first switches in each direction is found and operated. Therefore three areas are formed in
network: First, the upstream out-of-service area that is first restored by closing the feeders
circuit breaker, second, the damage area that must been repaired, and finally, the downstream
un-faulted area then is transferred to the neighboring feeders according to the proposed
algorithms.
Step 2) creation candidate switch lists
11

Candidate tie switches are identified from energized feeder that can connect directly into
the out-of-service area. Candidate sectionalizes switches (ss) that are located in the out-of-
service area and identified from graph-based method. For each candidate ts, VD and Z
path
are
obtained. In this section, new weighting factor is utilized to converts of these two indices into
an equivalent single index. Final index has been described as follow:
(10)
Where and are two weight factors, that have two continues amounts between 0 and 1
(0< and <1). First and most important index has been greater weighting than second index
in this paper. Weighting factors amount can be initialized with PDN operators. FI list is
sorted in increasing order amount of FI and ts lists are formed.
Step 3) operates one ts.
One ts due to least member of FI (first part of ts list) operation and three phase unbalance
load flow calculation are attempted and network constraints are checked.
3.1) if all network constraints are satisfied, restoration plan is identified, therefore, set
N
op
=1, and go to step 5.
3.2) if an overload or voltage violations exists, go to step 4.
Step 4) selection next ts and respective ss
4.1) next ts from ts list is selected and operated. Therefore, one direction in distribution
network is being existed that making one loop in network. For remove this loop, one ss in this
direction must be opened to maintain radial structure. ss choosing is based on graph-based
method for minimizing voltage drop in distribution network.
4.2) ss is opened, ts is closed, three phase unbalance load flow calculation is attempted and
network constraints are checked.
4.3) if all network constraints are satisfied, set N
op
=N
op
+2 and go to step 5.
) max(
path
Z VD FI | o + =
12

4.4) if an overload or voltage violation exists, open previously ts, close previously ss and
repeat step 4.1, 4.2, 4.3, until iteration procedure is finished. If no restoration plan exists, go
to step 4.5.
4.5) first switch pair (ts, ss) in the step 4 is operated and for restoration procedure
continuance, this step (step 4) is repeated.
Step 5) restoration plan (final step).
In this step sequence operations for ts and ss that have been selected is described. First, all
ss that have been selected must be opened and afterward, all of ts that have been selected
must be closed. To illustrate the fastness and effectiveness of the proposed algorithm, two
unbalanced distribution networks consists of: modified IEEE 13-node and IEEE 37-node
have been tested and presented in section 7.

6. Fast load flow technique
After network restoration, the three phase unbalanced distribution load flow has to be
calculated to examine the voltage, current and capacity constraints for feeders, lines and
elements with additional of new load points. In this paper, we are used fast load flow
technique for fast service restoration. For receipt more information about this technique,
please refers to [22]. In this section, summary of this technique [22] is described. The
fundamental idea discussed here is how to obtain the power flow solution by using the
elements of a unique quasi-symmetric matrix called TRX in the iterative process. The
proposed TRX matrix constitutes a complete database by including information of network
topology structure as well as branch impedances of the distribution feeder. The method is
described in six steps: data preparation, initialization, current and voltage calculations, quasi-
symmetric matrix calculation, and convergence process. This formulation is given including
three phase line shunt-admittances and loads are modeled as constant power. The input data
13

is given by three-phase per-unit node-branch oriented information. The basic data required is:
three-phase injected powers and sending and receiving nodes of a given line impedance. The
branch impedances are given as a rectangular 3nx3 phase impedance matrix Z
abc
.
(11)
Where is the 3-phase matrix impedance corresponding to ij line section:

(12)

Shunt admittances modeled by a rectangular 3nx3 matrix Y
abc
:
(13)
Fig. 3 shows a radial distribution network with n+1 nodes, and n branches and a single
voltage source at the root node 0. Under the unbalanced approach, nodal power injection
vector S is given per node and per phase.

(14)


At given iteration k, the relationship between injected currents I
k
and branch currents J
k
is
set through an upper triangular matrix T accomplishing the Kirchhoff Current Laws (KCL) as
follows:
(15)
(16)

Update voltage:
(17)
] [
) 1 ( 01 n n
abc
ij
abc abc abc
Z Z Z Z

=
ij
abc
Z
(
(
(
(

=
ij
cc
ij
cb
ij
ca
ij
bc
ij
bb
ij
ba
ij
ac
ij
ab
ij
aa
ij
abc
Z Z Z
Z Z Z
Z Z Z
Z
] [
) 1 ( 01 n n
abc
ij
abc abc abc
Y Y Y Y

=
c b a p
jQ P
jQ P
jQ P
S
S
S
S
T
n p n p
i p i p
p p
T
n p
i P
p
, ,
1 1
1
=
(
(
(
(
(
(
(

+
+
+
=
(
(
(
(
(
(
(

k
abc
k
abc
I T J . =

|
|
.
|

\
|
=
c b a p
k
i p
ij
pp
k
i p
i p
k
i p V Y
V
S
I
, ,
*
k
abc abc abc
T
abc abc
k
abc
I T Z T V V . . .
0
1
+ =

+
14

Where, V
abc-0
is initial voltage vector.
Convergence check-in and final calculations:
(18)

7. Numerical example
To illustrate effectiveness of proposed algorithms, two different unbalanced distribution
networks have been considered. These methods have been coded in MATLAB software. this
section, is demonstrated performance of method. When a fault is occurred on network,
protection devices are operated immediately for detecting and isolating the fault. In this
paper, total numbers of switching operations for isolating the fault and service restoration are
obtained. The impedance of lines, phase impedance matrix and phase admittance matrix are
calculated from data of these networks in [23]. In some distribution networks, some loads are
modeled as distributed load; therefore, if every load point is modeled as a node, then systems
will have a large number of nodes. Thus, these loads are modeled as spot load in this paper.
In these test cases, one, two, or three phase loads with wye or delta connections can exist. In
this paper some assumptions for unbalanced distribution networks have been considered that
are described as follow:
1) All load are modeled based on constant power model;
2) The Regulator and Capacitors components is removed from networks;
3) For all branches in networks, one switch in send side of branch is considered;
4) Some tie switches are introduced in the networks for illustrating restoration plan.
5) Amount of loads are modified in order to performance restoration plane and regard
networks constraints.
6) For conversing distributed loads to spot loads, virtual nodes 2, 3 are introduced in
IEEE 13-node network.
c b a p n i V V
k
i p
k
i p , , ,..., 1
1
= = s
+
c
15

Fig. 4 and 5 are shows the one-line diagram of modified IEEE 13-node and IEEE 37-node
networks respectively. Introduced tie lines (or tie switches) for service restoration, actual and
with loss are considered. In this paper, two weighting (, ) are considered 0.8 and 0.2
respectively. Voltage magnitude range must be within limits of 0.95 and 1.05 per-unit. For
demonstrates validity of service restoration algorithm, fault simulation in several locations
are considered.
7.1.Service restoration results
Table 1 and table 2 displays the restoration algorithms results for IEEE 13-node and IEEE
37-node unbalanced distribution networks respectively. In these tables, following information
is provided:
- Total number of switching operations for isolating the fault and service restoration;
- Runtime software for service restoration plan.
When fault is occurred on the branch 2-3 in IEEE 13-node network, two switch
operations have been required to isolate the fault and three switch operations have been
required to full service restoration implementation. Bus number 3 is damaged bus that cant
be restored. Bus numbers 4, 5, 10, 11, 12, 13 and 14 are de-energized bus that must be
restored from neighboring tie lines (ts 6-11 and 9-12). For removing loop in network, one
switch must be opened; therefore switch number 4-10 is obtained from graph-based method.
Voltage magnitude per-unit for fault on this branch is shown in Fig. 6. This Fig shows that
voltage for all buses are in definition limits. When fault is occurred on the branch 4-12 in
IEEE 13-node network, three switch operations have been required to isolate fault. In this
case, bus number 12, 13 cant be restored.
When fault is occurred on the branch 3-4 in IEEE 37-node network, two switch
operations have been required to isolate the fault and three switch operations have been
16

required to full service restoration implementation. Voltage magnitude per-unit for all bus for
this case is shown in Fig. 7.

8. Conclusions
In this paper, fast service restoration in unbalanced PDNs, with consideration to priority
customers as multiple objective functions consists of: 1) maximizing the amount of total load
to be restored, 2) minimizing the number of the switching operation, 3) customers priority
considering are implemented. For this work, we are used two new heuristic algorithm based
on two important indices and graph-based method. Core of the proposed first algorithm is
voltage drop between candidate ts and substation bus and second algorithm is based on
graph-based method to minimize voltage dropping in networks. Fast load flow technique
based on a real quasi-matrix [22] has been utilized. Finally, the proposed algorithm has been
implemented and tested on two unbalanced distribution networks and results that have been
obtained, summarized as follow:
1) Total number of switching operation for isolating the fault and service restoration;
2) Buses that not restored.
3) Sequence operation for selected ts and ss.

References
[1] L. Zhang and L. Sun, "Multi-Objective Service Restoration for Blackout of Distribution
System with Distributed Generators based on Multi-agent GA," ELSEVIER, Energy
Procedia, 12 (2011) 253-262.
[2] C. Ming Huang, C. Tao Hsieh and Y. Shan Wang, "Evolution of radial basic function
neural network for fast restoration of distribution systems with load variations,"
ELSEVIER, Electrical Power and Energy Systems, 33 (2011) 961-968.
17

[3] S.Dimitrijevic and N. Rajakovicb, "An innovative approach for solving the restoration
problem in distribution networks," ELSEVIER, Electric Power Systems Research, 81
(2011) 1961-1972.
[4] K. N. Miu, H.-D. Chiang, B. B. Yuan, and G. Darling, Fast service restoration for large-
scale distribution systems with priority customers and constraints, IEEE Trans. Power
Syst., 13 (3) (Aug. 1998) 789795.
[5] S. Toune, H. Fudo, T. Genji, Y. Fukuyama, and Y. Nakanishi, Comparative study of
modern heuristic algorithms to service restoration in distribution systems, IEEE Trans.
Power Del., 17 (1) (Jan. 2002) 173181.
[6] K. Miu, H.-D. Chiang, and R. J. McNulty, Multi-tier service restoration through network
reconfiguration and capacitor control for large-scale radial distribution systems, IEEE
Trans. Power Syst., 15 (3) (Aug. 2000) 10011007.
[7] V, Susheela Devi, D. P. Sen Gupta, and G. Anandalingam, Optimal restoration of power
supply in large distribution systems in developing countries, IEEE Trans. Power Del., 10
(1) (Jan. 1995) 430438.
[8] Y. Y. Hsu and H. C. Kuo, A heuristic based fuzzy reasoning approach for distribution
system service restoration, IEEE Trans. Power Del., 9 (2) (Apr. 1994) 948953.
[9] S. J. Lee, K. H. Kim, K. Y. Nam, and J. K. Lee, Service restoration expert system
adopting branch pattern based grouping strategy in distribution systems, Expert Syst.
Appl.Power Syst. IV, (Jan. 1994) 423427.
[10] Y. L. Ke, Distribution feeder reconfiguration for load balancing and service restoration
by using G-nets inference mechanism, IEEE Trans. Power Del., 19 (3) ( Jul. 2004)
14261433.
18

[11] C. Chao-Shun, C.-H. Lin, and T. Hung-Ying, A rule-based expert system with colored
petri net models for distribution system service restoration, IEEE Trans. Power Systems,
17 (Nov. 2002) 10731080.
[12] C. M. Huang, Multi-objective service restoration of distribution systems using fuzzy
cause-effect networks, IEEE Trans. Power Syst., 18 (2) (May 2003) 867874.
[13] S. J. Lee, S. I. Lim, and B. S. Ahn, Service restoration of primary distribution systems
based on fuzzy evaluation of multi-criteria, IEEE Trans. Power Syst., 13 (3) (Aug. 1998)
11561163.
[14] Y. Kumar, B. Das, and J. Sharma, Multiobjective, multiconstraint service restoration of
electric power distribution system with priority customers, IEEE Trans. Power Del., 23
(1) (Jan. 2008) 261270.
[15] K. L. Butler, N. D. R. Sarma, and R. Prasad, Network reconfiguration for service
restoration in shipboard power distribution systems, IEEE Trans. Power Systems, 16
(Nov. 2001) 653661.
[16] Y. Hsiao and C. Chien, Enhancement of restoration service in distribution systems using
a combination fuzzy-GA method, IEEE Trans. Power Syst., 15 (Nov. 2000) 13941400.
[17] W. H. Chen, M. S. Tsai, and H. L.Kuo, Distribution system restoration using the hybrid
fuzzy-grey method, IEEE Trans. Power Syst. 20 (1) (Feb. 2005) 199255.
[18] M. R. Kleinberg, K. Miu, H. D. Chiang, Improving service restoration of power
distribution systems through load curtailment of in-service customers IEEE Trans.
Power Syst., 26 (3) (Aug. 2011) 11101117.
[19] W. Li, P. Wang, Z. Li, Y. Liu, Reliability evaluation of complex radial distribution
systems considering restoration sequence and network constraints IEEE Trans. Power
Del., 19 (2) (April. 2004) 753758.
19

[20] T. T. H. Pham, Y. Bsanger, and N. Hadjsaid, New challenges in power system
restoration with large scale of dispersed generation insertion, IEEE Trans. Power Syst.,
24 (1) (Feb. 2009) 398406.
[21] V. Kumar, R. H. C. Kumar, I. Gupta, and H. O. Gupta, DG integrated approach for
service restoration under cold load pickup, IEEE Trans. Power Del., 25 (1) (Jan. 2010)
398406.
[22] P.M. De Oliveira-De Jesus, M.A. Alvarez, J.M. Yusta, " Distribution power flow method
based on a real quasi-symmetric matrix" ELSEVIER, Electric Power Systems Research,
95 (2013) 148-159.
[23] Radial distribution test feeders.

Table 1: Restoration Results for Modified IEEE 13-Node Network.
case
fault
location
switch operation to fault
isolation
switches operation to
SR
runtime
(S)
1 4-10 4-10, 10-11 6-11 0.35
2 2-3 2-3, 3-4 6-11, 9-12,4-10 0.52
3 4-12 4-12, 12-14, 12-13 5-14 0.3
4 1-8 1-8, 8-9 9-12 0.39

Table 2: Restoration Results for Modified IEEE 37-Node Network.
case
fault
location
switch operation to fault
isolation
switches operation to
SR
runtime
(S)
1 2-26 2-26, 26-27 36-27 0.45
2 27-30 27-30, 30-31, 30-33 26-35 0.43
3 2-23 2-23, 23-24, 23-25 20-25 0.35
4 3-4 3-4, 4-5 27-36, 22-16, 7-8 1.12
5 5-6 5-6, 6-14, 6-7 22-16, 36-12, 9-10 0.96
6 7-8 7-8, 8-9, 8-15 36-12, 22-16 0.86
7 3-19 3-19, 19-20 25-20 0.52

20


Fig.1. A 16-bus distribution network
.

Fig. 2. Example of graph-based method.

21


Fig.3. The branch and node numbering of a radial distribution network.


Fig.4. modified IEEE 13-node network.


Fig.5. modified IEEE 37-node network.
22



Fig.6. voltage after restoration for modified IEEE 13-node network.


Fig.7. voltage after restoration for modified IEEE 37-node network.