Electric Power Systems Research 77 (2007) 685–694

An improved TS algorithm for loss-minimum reconfiguration

in large-scale distribution systems
Dong Zhang ∗ , Zhengcai Fu, Liuchun Zhang
Room 617, No.7 Building, Shanghai Jiao Tong University, No.1954, Huashan Road, Shanghai 200030, China
Received 23 October 2005; received in revised form 30 May 2006; accepted 9 June 2006
Available online 18 July 2006

Abstract
This paper presents an improved tabu search (ITS) algorithm for loss-minimization reconfiguration in large-scale distribution systems. TS
algorithm is an efficient meta-heuristic searching algorithm. It has advantages of both high local search efficiency of hill-climbing method and
global search ability of intelligent algorithm. But tabu lengths and candidate neighborhood are two key parameters affecting searching performance
of TS algorithm, and these two parameters are hard to be effectively determined in advance. In ITS algorithm, mutation operation, a main operator
used in genetic algorithm, is introduced to weaken the dependence of global search ability on tabu length. In addition, the candidate neighborhood,
which only contains several optimal switch exchanges in each tie switch associated loop network, is designed to improve local search efficiency
and to save a large amount of computing time. The proposed ITS algorithm is applied to the sample system and numerical results well demonstrate
the validity and effectiveness of the proposed ITS algorithm.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Local search; Mutation operation; Network reconfiguration; TS algorithm

1. Introduction exchange resulting in maximum power loss reduction in a tie

switch associated loop network. Fan et al. [3] formulated the
Network reconfiguration is an important optimization tool in network reconfiguration problem as a LP (linear programming)
distribution network automation systems. With loads prediction problem and used a single-loop algorithm to search for the opti-
system and remote monitoring system, network structure can mal network structure. Kashem et al. [4] presented a geometrical
be real-time reconfigured, by changing open/closed states of approach for network power loss minimization, which searched
some tie and sectionalizing switches, to reduce power losses, every branch exchange in the largest zero loss-change circles to
balance overloads, improve power supply reliability and, in a find the maximum loss-reduction branch exchange. Chiang and
certain extent, increase the power supply capacity of distribution Jean-Jumeau [5] proposed a solution method for the network
systems. reconfiguration problem using the simulated annealing algo-
Since network reconfiguration problem is a complicated com- rithm which was extended later by Jiang and Baldick [6] and Jean
binatorial optimization problem, it is hard to get an optimal and Kim [7]. Ying and Saw [8] and Venkatesh et al. [9] proposed
configuration in large-scale distribution systems at a feasible a fuzzy adaptation of the evolutionary programming algorithm
computer time. By far, a lot of studies have been developed. to solve the network reconfiguration problem. Mori and Ogita
Civanlar et al. [1] presented a switch exchange method, in which [10] presented a parallel tabu search (PTS) method for network
a simple formula for the estimation of the loss reduction by reconfiguration. The PTS adopted four processors to do tabu
a particular switching option is developed. Baran and Wu [2] search simultaneously to form a parallel search. Mishima [11]
proposed the simplified DistFlow equations to simplify the cal- proposed a tabu search based solution algorithm for distribu-
culation of power flow and heuristic methods to find a switch tion system reconfiguration considering distributed generators.
Huang [12] proposed an enhanced genetic algorithm for distri-
bution network reconfiguration. Jeon and Kim [13] proposed a
∗ Corresponding author. combined algorithm of simulated annealing and tabu search for
E-mail address: xt [email protected] (D. Zhang). loss minimization reconfiguration in distribution systems.

0378-7796/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2006.06.005
686 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694

Algorithms proposed for network reconfiguration problem, objective function for ITS algorithm is as follows:
generally, can be classified into the following two main classes: nb

(1) heuristic algorithms [1–4], such as switch exchange algo- min f = Ploss i + k1 fv + k2 fc (2)
rithm, optimal flow pattern algorithm, etc. (2) Intelligent algo- i=1
rithms [5–13], such as genetic algorithm (GA), simulated
annealing (SA) algorithm, tabu search (TS) algorithm, etc. where f is the objective function, fv the penalty function for
Among these algorithms, heuristic algorithms are all greedy violation on the nodal voltage magnitude bound, fc the penalty
search algorithm. They are easy to be implemented and with function for violation on branch capacity bound, and k1 and k2
high searching efficiency, but generally cannot converge to the are the penalty coefficients.
global optimum solution in the large-scale distribution systems.
GA and SA can direct searching process to the global optimum 3. TS algorithm
at the probability 1 in theory. But they all inevitably involve a
large amount of computation requirements and really are time TS algorithm is an extended local search algorithm. Through
consuming methods. Furthermore, in the definite iterative gen- introducing a memory system called tabu list to record the latest
erations, they may lose global optimum solution too. moves, TS algorithm can escape from the current local optimum
TS algorithm is an efficient meta-heuristic searching algo- by using tabu information. TS algorithm has both advantages
rithm compared with GA and SA. However, tabu length and of high search efficiency of local search algorithm and global
candidate neighborhood are two key parameters to affect search- search ability of intelligent algorithm.
ing property of TS algorithm, and they are hard to be effectively The searching procedure of TS algorithm, generally, can be
determined in advance, especially tabu length. described as follows:
In this paper, improvements about tabu length and design of
candidate neighborhood are developed to make TS algorithm • Step 1. Form an initial solution and initialize control param-
more effective and more efficient for loss-minimum reconfigu- eters.
ration in large-scale distribution systems. • Step 2. If stop criterion is satisfied, then stop searching process
and output the final results. Otherwise, select feasible solu-
tions that meet tabu restrictions to form the candidate neigh-
2. Problem formulation borhood N(H, xnow ) around current solution xnow . Find the
best one xnext among the candidate neighborhood to update the
The loss-minimum reconfiguration problem, generally, can current solution(xnow = xnext ), and update tabu list by adding
be mathematically expressed as this new tabu move, from xnow to xnext , into tabu list and
releasing the oldest move. Then go to step 2.
nb

min F= Ploss i where H denotes the whole neighborhood around current solu-
i=1 tion. N(H, xnow ) denotes the selected candidate neighborhood
subject to g(x) = 0, Vmin ≤ Vi ≤ Vmax , Si ≤ Si max around current solution.

(1) 4. Improved TS algorithm

where nb is the total number of branches, F the total power To take advantage of the features of network structure itself
loss of the system, Ploss i the power loss of branch i, x the and make TS algorithm be more suitable for network reconfig-
nodal voltage vector, g(x) the power flow equation, Vi the nodal uration problem, in this section, ITS algorithm, with mutation
voltage magnitude of node i, Vmin (Vmax ) the lower(upper) bound operation and high quality candidate neighborhood, is devel-
of nodal voltage and Smax is the upper bound of the branch oped.
capacity.
Equality constraint is the operating constraint, i.e. power flow 4.1. Mutation operation
equation.
Inequality constraints are nodal voltage and branch capacity A multi-peak search problem, as shown in Fig. 1, has four
constraints. local optima distributed in the A, B, C and D four regions, respec-
Besides equality and inequality constraints mentioned above, tively. After the local optimum, e.g. in region A, is reached, for
there are still constraints in topological structure. The operating TS algorithm, tabu length is vital to problem that whether tabu
structure of distribution network should be a radial configuration information is enough to get search process out of the local min-
and there should be no nodes without power supply path. imum region. Obviously, different local optima need different
The topological structure constraint can be easily satisfied tabu length to efficiently get out of it.
by using the newly designed candidate neighborhood in the To this problem, mutation operation is introduced to make
proposed method. The constraints of nodal voltage and branch search process can directly jump out of current region at a cer-
capacity can be included into the objective function by using tain probability after the local optimum is reached. The rule of
penalty function method to relax inequality constraints. The mutation is given as follows:
 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694 687

Fig. 1. Distribution of local optima.

(1) When current-best solution, which is the initial solution at where Dj (i) is the set of ns switch exchanges in the jth tie switch
the beginning of the program or the solution produced by associated loop network at the ith iteration round and ns is the
mutation operation, is evolving, mutation operation is dis- number of selected switch exchanges according to power loss
abled. The search process is just like that of the conventional reduction order in each tie switch associated loop network.
TS algorithm.
(2) When current-best solution stops evolving, which generally
means that the local optimum is reached and search process
begins to escape from current region, mutation operation is
enabled at the threshold mutation probability fset .

In the mutation operation, randomly select mt (1 ≤ mt < nt )

tie switches, where nt is the total number of tie switches.
For each selected tie switch, randomly open a sectionalizing
switch from the corresponding loop network and close this tie
switch.
Flowchart of mutation operation is given in Fig. 2 and fi is
mutation probability generated in the ith iteration round.

4.2. Design of candidate neighborhood

Distribution network is normally operated radially and each

tie switch associates to one loop network. To maintain radially
structure, a tie switch is closed and a sectionalizing switch, in
the corresponding loop network, must be opened. This switch
exchange will result in the changes of power loss of the system.
If the power loss reduction order of these switch exchanges for
each tie switch can be easily determined and several best ones
are selected as candidate neighborhood of tie switch, the local
search efficiency and search performance of TS algorithm will
be improved greatly and a large amount of computing time will
be saved.
Combined with the structure features of distribution network,
the candidate neighborhood, in this paper, is defined as:
Ω(x) = {Dj (i)|j = 1, 2, . . . , nt } Fig. 2. Flowchart of mutation operation.
688 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694

Once optimal transfer load Pk + jQk is obtained, the power

loss reduction order in the loop network can easily be deter-
mined. The steps to determine this order are given as follows.
First, find the switch exchange with the maximum power loss
reduction, i.e. the no. 1 switch exchange.

(1) If Pk + jQk > 0, the optimal transfer power flow Pk + jQk is

really transferred from set L to set R. The branch in set L with
power flow being closest to Pk + jQk is the matched branch.
The switch associated to this branch and the tie switch is the
switch exchange with no. 1 power loss reduction order.
Fig. 3. A tie switch associated loop network. (2) If Pk + jQk < 0, the optimal transfer power flow Pk + jQk is
really transferred from set R to set L. The branch in set L with
Here, an optimal transfer load algorithm is developed to sim- power flow being closest to Pk + jQk is the matched branch.
plify the determination of these switch exchanges in a tie switch The switch associated to this branch and the tie switch is the
associated loop network and is described as follows. switch exchange with no. 1 power loss reduction order.
Fig. 3 shows a tie switch associated loop network, the left
and right branch set, L and R, of the loop network are defined as Second, find the nos. 2, 3, 4,. . . switch exchanges.
follows: Obviously, the sectionalizing switch in the no. 2 switch
exchange must be the switch in the adjoining branch which either
L = {L1 , L3 , . . . , Ln }, R = {L2 , L4 , . . . , Lm } in the left or in the right of no. 1 switch.
Using the DistFlow [2], branch power flow can be approxi- The sectionalizing switch in the no. 3 switch exchange must
mately calculated by Eq. (3): be the switch in the adjoining common branch which either in
  the left or in the right of no. 1 switch and no. 2 switch.
Pi = PLj , Qi = QLj (3) The switch exchanges with other power loss reduction order
j∈i j∈i can be determined in the same way.
where Pi and Qi are the power flow at branch i, and PLj and QLj For TS algorithm, if all feasible switch-exchange combina-
are the downstream node power loads. tions are selected as candidate neighborhood, the total amount
Resistance of tie switch k associated branch is rk . When use of computation indetermining an optimal system configura-
the fact that Ui ≈ 1 p.u., power loss in the loop network can be tion is at most nd ni=1 t
Sij unit time, where nd is the maximum
approximately calculated as follows: generations of iteration, and Sij is the number of sectionalizing
  switches of the loop network corresponding to jth tie switch at
Ploss k = (Pi2 + Q2i )ri + (Pi2 + Q2i )ri (4) ith iteration. Whereas, for ITS algorithm, the total amount of
i∈L i∈R computation required is nd × ns × nt unit time. Since ns  Sij ,
ns × nt  ni=1 t
Sij , the amount of computation required in ITS
Assume that tie switch k and a sectionalizing switch in branch
algorithm is far less than that of the conventional TS algorithm.
set L change open/closed status and the transfer load Pk + jQk
Therefore, searching efficiency of ITS algorithm is enhanced
is a continuous variable. Power loss of the new structure, after
greatly.
switch exchange, is calculated as follows:

Pch = ((Pi − Pk )2 + (Qi − Qk )2 )ri 4.3. Procedure of the proposed method
i∈L
 The procedure of the proposed ITS algorithm is summarized
+ ((Pi + Pk )2 + (Qi + Qk )2 )ri + (Pk2 + Q2k )rk (5)
as follows:
i∈R

To seek an optimal transfer load Pk + jQk that makes Pch a • Step 1. Input system data and initialize control parameters,
minimum, the partial differentials of Pch with respective to Pk such as tabu length, mutation probability fset , select number
and Qk , respectively, are set to be zero, as follows: ns , etc.
• Step 2. Find ns switch exchanges in each tie switch associated
∂Pch ∂Pch
= 0, =0 (6) loop network and form the candidate neighborhood of current
∂Pk ∂Qk solution.
Solving Eq. (6) yields a solution of the optimal transfer load • Step 3. Move the current solution to the best one in the candi-
as Eq. (7): date neighborhood while meet the tabu restriction and update
    tabu list.
L Pi r − i ∈ R P i ri L Qi r − i ∈ R Qi r • Step 4. Perform mutation operation, as shown in Fig. 2.
Pk = i ∈ , Qk = i ∈
i ∈ L∪R ri + rk i ∈ L∪R ri + rk • Step 5. Go to step 6 if stop criterion is satisfied. Otherwise,
(7) go to step 2.
 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694 689

Fig. 4. Initial configuration of the 119-bus test system.

Table 1
Results comparison between different methods
Initial structure Branch exchange [2] TS algorithm [11] ITS algorithm
Tie line set 48 27 45 44 45 44 45 44
17 27 17 27 27 26 27 26
8 24 23 24 22 23 23 24
56 45 53 52 54 53 54 53
65 51 51 50 51 50 65 51
38 65 64 65 64 65 61 62
9 42 41 42 41 42 41 42
61 100 61 100 61 100 95 100
76 95 76 77 76 77 77 78
91 78 74 75 74 75 74 75
103 80 79 80 80 79 101 102
113 86 85 86 85 86 86 113
110 89 89 110 89 110 89 110
115 123 114 115 114 115 114 115
25 36 35 36 33 34 35 36
Tabu length 9 9
Mutation probability 0 0.01
Power losses (kW) 1294.3 885.56 884.163 865.865
Lowest voltage (p.u.) 0.9321 0.9321 0.9321 0.9323
Lowest voltage node 116 116 116 116
690 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694

Fig. 5. Searching process of TS algorithm and ITS algorithm.

• Step 6. Output the optimization network structure and calcu-

late system power loss.
Table 2
Relationship between searching performance and tabu length
5. Numerical results
Tabu length TS algorithm ITS algorithm

Power loss (kW) FRGS Power loss (kW) FRGS Test system is a 11 kV distribution system with 118 section-
alizing switches and 15 tie switches as shown in Fig. 4. The total
1 884.163 21 880.353 464
power loads are 22709.7 kW and 17041.1 kvar. The detailed data
2 884.163 21 880.353 471
3 884.163 21 880.353 475 is given in Appendix A. In the test, ns = 3, mt = 1 and fset = 0.01.
4 884.163 21 880.353 483 Reconfiguration results of switch exchange method [2], TS
5 884.163 21 880.353 505 algorithm [11] and ITS algorithm are given in Table 1. It is
6 884.163 22 865.895 567 clear that the optimization network structures of ITS algorithm
7 884.163 22 880.353 518
are different with that of switch exchange method and TS algo-
8 884.163 22 874.459 565
9 884.163 22 865.865 598 rithm. And ITS algorithm gets better optimization results, both
10 884.163 22 866.483 593 in power loss reduction and nodal voltage profile, than that
11 884.163 22 865.895 589 of switch exchange method and TS algorithm. At a Celeron
12 884.163 22 871.567 595 1.3 GHz CPU computer system, 600 iterations of tabu search
13 884.163 22 865.865 379
with mutation operation take 9.038 s.
 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694 691

first time reaching the final optimum. From Table 2, it is noted

that global search ability of TS algorithm (fset = 0) is rely more
on tabu length than that of the ITS algorithm (fset = 0.01). In
ITS algorithm, the relationship between global search ability
and tabu length is relaxed and the capacity of global search is
improved (Fig. 6).

6. Conclusions

This paper presents an improved TS algorithm based

method for loss-minimum reconfiguration in distribution sys-
tems. Improvements are aimed at improving global search ability
and local search efficiency of TS algorithm. Improvement of
global search ability results from the introduction of mutation
operation. Mutation operation can effectively relax the relation-
ship between global search ability and tabu length, and make
Fig. 6. Relationship between searching performance and tabu length. searching process easily get out of current region. Improvement
of local search efficiency of TS algorithm results from a newly
Fig. 5 gives the searching processes of TS algorithm (Fig. 5a designed candidate neighborhood, which has features of forming
and b) and ITS algorithm (Fig. 5c and d). From (b), it is noted easily and containing only the several best power loss reduction
that the searching process of TS algorithm is cycling around switch exchanges in each tie switch associated loop networks.
the local minimum and cannot escape from it. Whereas in ITS The validity and effectiveness of the proposed ITS algorithm is
algorithm, (c) and (d), when current-best solution stop evolving well proved by the sample test systems.
for a certain generations, the mutation operation is activated to
effectively break cycling. Appendix A
In Table 2, the relationship between searching performance
and tabu length is given. FRGS denotes the generations of the Network data of the test system are shown in Table A.1 .

Table A.1
Branch number Branch Branch impedance Loads

Rc. Nd. Sn. Nd. r () x () PL (kW) QL (kvar)

Sectionalizing lines
1 0 1 0 0 0 0
2 1 2 0.036 0.01296 133.84 101.14
3 2 3 0.033 0.01188 16.214 11.292
4 2 4 0.045 0.0162 34.315 21.845
5 4 5 0.015 0.054 73.016 63.602
6 5 6 0.015 0.054 144.2 68.604
7 6 7 0.015 0.0125 104.47 61.725
8 7 8 0.018 0.014 28.547 11.503
9 8 9 0.021 0.063 87.56 51.073
10 2 10 0.166 0.1344 198.2 106.77
11 10 11 0.112 0.0789 146.8 75.995
12 11 12 0.187 0.313 26.04 18.687
13 12 13 0.142 0.1512 52.1 23.22
14 13 14 0.18 0.118 141.9 117.5
15 14 15 0.15 0.045 21.87 28.79
16 15 16 0.16 0.18 33.37 26.45
17 16 17 0.157 0.171 32.43 25.23
18 11 18 0.218 0.285 20.234 11.906
19 18 19 0.118 0.185 156.94 78.523
20 19 20 0.16 0.196 546.29 351.4
21 20 21 0.12 0.189 180.31 164.2
22 21 22 0.12 0.0789 93.167 54.594
23 22 23 1.41 0.723 85.18 39.65
24 23 24 0.293 0.1348 168.1 95.178
25 24 25 0.133 0.104 125.11 150.22
26 25 26 0.178 0.134 16.03 24.62
27 26 27 0.178 0.134 26.03 24.62
692 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694

Table A.1 (Continued )

Branch number Branch Branch impedance Loads
Rc. Nd. Sn. Nd. r () x () PL (kW) QL (kvar)

28 4 29 0.015 0.0296 594.56 522.62

29 29 30 0.012 0.0276 120.62 59.117
30 30 31 0.12 0.2766 102.38 99.554
31 31 32 0.21 0.243 513.4 318.5
32 32 33 0.12 0.054 475.25 456.14
33 33 34 0.178 0.234 151.43 136.79
34 34 35 0.178 0.234 205.38 83.302
35 35 36 0.154 0.162 131.6 93.082
36 31 37 0.187 0.261 448.4 369.79
37 37 38 0.133 0.099 440.52 321.64
38 30 40 0.33 0.194 112.54 55.134
39 40 41 0.31 0.194 53.963 38.998
40 41 42 0.13 0.194 393.05 342.6
41 42 43 0.28 0.15 326.74 278.56
42 43 44 1.18 0.85 536.26 240.24
43 44 45 0.42 0.2436 76.247 66.562
44 45 46 0.27 0.0972 53.52 39.76
45 46 47 0.339 0.1221 40.328 31.964
46 47 48 0.27 0.1779 39.653 20.758
47 36 49 0.21 0.1383 66.195 42.361
48 49 50 0.12 0.0789 73.904 51.653
49 50 51 0.15 0.0987 114.77 57.965
50 51 52 0.15 0.0987 918.37 1205.1
51 52 53 0.24 0.1581 210.3 146.66
52 53 54 0.12 0.0789 66.68 56.608
53 54 55 0.405 0.1458 42.207 40.184
54 54 56 0.405 0.1458 433.74 283.41
55 30 58 0.391 0.141 62.1 26.86
56 58 59 0.406 0.1461 92.46 88.38
57 59 60 0.406 0.1461 85.188 55.436
58 60 61 0.706 0.5461 345.3 332.4
59 61 62 0.338 0.1218 22.5 16.83
60 62 63 0.338 0.1218 80.551 49.156
61 63 64 0.207 0.0747 95.86 90.758
62 64 65 0.247 0.8922 62.92 47.7
63 1 66 0.028 0.0418 478.8 463.74
64 66 67 0.117 0.2016 120.94 52.006
65 67 68 0.255 0.0918 139.11 100.34
66 68 69 0.21 0.0759 391.78 193.5
67 69 70 0.383 0.138 27.741 26.713
68 70 71 0.504 0.3303 52.814 25.257
69 71 72 0.406 0.1461 66.89 38.713
70 72 73 0.962 0.761 467.5 395.14
71 73 74 0.165 0.06 594.85 239.74
72 74 75 0.303 0.1092 132.5 84.363
73 75 76 0.303 0.1092 52.699 22.482
74 76 77 0.206 0.144 869.79 614.775
75 77 78 0.233 0.084 31.349 29.817
76 78 79 0.591 0.1773 192.39 122.43
77 79 80 0.126 0.0453 65.75 45.37
78 67 81 0.559 0.3687 238.15 223.22
79 81 82 0.186 0.1227 294.55 162.47
80 82 83 0.186 0.1227 485.57 437.92
81 83 84 0.26 0.139 243.53 183.03
82 84 85 0.154 0.148 243.53 183.03
83 85 86 0.23 0.128 134.25 119.29
84 86 87 0.252 0.106 22.71 27.96
85 87 88 0.18 0.148 49.513 26.515
86 82 89 0.16 0.182 383.78 257.16
87 89 90 0.2 0.23 49.64 20.6
88 90 91 0.16 0.393 22.473 11.806
89 68 93 0.669 0.2412 62.93 42.96
 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694 693

Table A.1 (Continued )

Branch number Branch Branch impedance Loads

Rc. Nd. Sn. Nd. r () x () PL (kW) QL (kvar)

90 93 94 0.266 0.1227 30.67 34.93

91 94 95 0.266 0.1227 62.53 66.79
92 95 96 0.266 0.1227 114.57 81.748
93 96 97 0.266 0.1227 81.292 66.526
94 97 98 0.233 0.115 31.733 15.96
95 98 99 0.496 0.138 33.32 60.48
96 95 100 0.196 0.18 531.28 224.85
97 100 101 0.196 0.18 507.03 367.42
98 101 102 0.1866 0.122 26.39 11.7
99 102 103 0.0746 0.318 45.99 30.392
100 1 105 0.0625 0.0265 100.66 47.572
101 105 106 0.1501 0.234 456.48 350.3
102 106 107 0.1347 0.0888 522.56 449.29
103 107 108 0.2307 0.1203 408.43 168.46
104 108 109 0.447 0.1608 141.48 134.25
105 109 110 0.1632 0.0588 104.43 66.024
106 110 111 0.33 0.099 96.793 83.647
107 111 112 0.156 0.0561 493.92 419.34
108 112 113 0.3819 0.1374 225.38 135.88
109 113 114 0.1626 0.0585 509.21 387.21
110 114 115 0.3819 0.1374 188.5 173.46
111 115 116 0.2445 0.0879 918.03 898.55
112 115 117 0.2088 0.0753 305.08 215.37
113 117 118 0.2301 0.0828 54.38 40.97
114 105 119 0.6102 0.2196 211.14 192.9
115 119 120 0.1866 0.127 67.009 53.336
116 120 121 0.3732 0.246 162.07 90.321
117 121 122 0.405 0.367 48.785 29.156
118 122 123 0.489 0.438 33.9 18.98

Branch number Branch Branch impedance Loads

Rc. Nd. Sn. Nd. r () x () Rc. Nd. Sn. Nd.

Tie lines
119 48 27 0.5258 0.2925 0 0
120 17 27 0.5258 0.2916 0 0
121 8 24 0.4272 0.1539 0 0
122 56 45 0.48 0.1728 0 0
123 65 56 0.36 0.1296 0 0
124 38 65 0.57 0.572 0 0
125 9 42 0.53 0.3348 0 0
126 61 100 0.3957 0.1425 0 0
127 76 95 0.68 0.648 0 0
128 91 78 0.4062 0.1464 0 0
129 103 80 0.4626 0.1674 0 0
130 113 86 0.651 0.234 0 0
131 110 89 0.8125 0.2925 0 0
132 115 123 0.7089 0.2553 0 0
133 25 36 0.5 0.5 0 0

References [4] M.A. Kashem, V. Gannpathy, G.B. Jasmon, A geometrical approach

for network reconfiguration based loss minimization in distri-
[1] S. Civanlar, J.J. Grainger, S.S.H. Le, Distribution feeder reconfiguration bution systems, Electrical Power Energy Syst. 23 (2001) 295–
for loss reduction, IEEE Trans. Power Delivery 3 (1988) 1217–1223. 304.
[2] M.E. Baran, F.F. Wu, Network reconfiguration in distribution systems for [5] H.D. Chiang, R.M. Jean-Jumeau, Optimal network reconfiguration distri-
loss reduction and load balancing, IEEE Trans. Power Delivery 4 (2) (1989) bution system. Part 1. A new formulation and a solution methodology, IEEE
1401–1407. Trans. Power Delivery 5 (1990) 1902–1909.
[3] J.Y. Fan, L. Zhang, J.D. McDonald, Distribution network reconfigura- [6] D. Jiang, R. Baldick, Optimal electric distribution system switch recon-
tion: single loop optimization, IEEE Trans. Power Syst. 11 (3) (1996) figuration and capacitor control, IEEE Trans. Power Syst. 11 (1996)
1643–1647. 890–897.
694 D. Zhang et al. / Electric Power Systems Research 77 (2007) 685–694

[7] Y.J. Jean, J.C. Kim, An efficient simulated annealing algorithm for network Dong Zhang received his M.Sc. degree from Xi’an University of Technol-
reconfiguration in large-scale distribution systems, IEEE Trans. Power ogy, Shaanxi, P.R. China in Electrical Engineering in 2001. Presently, he is
Delivery 17 (2002) 1070–1078. a Ph.D. student in the Department of Electrical Engineering at Shanghai Jiao
[8] Y.H. Ying, Y.H. Saw, Genetic algorithm based network reconfiguration for Tong University.. His most recent research interests include distribution system
loss minimization in distribution systems, Power Eng. Soc. Gen. Meet. automation, optimization theory and overvoltage protection of the low voltage
IEEE 1 (2003) 13–17. distribution systems.
[9] B. Venkatesh, R. Ranjan, H.B. Gooi, Optimal reconfiguration of radial
Zheng-Cai Fu received the Ph.D. degree in Electrical Engineering from Shanghai
distribution systems to maximize loadability, IEEE Trans. Power Syst. 19
Jiao Tong University, Shanghai, P.R. China in 2001. He is currently a professor
(1) (2004) 260–266.
of electrical engineering at Shanghai Jiaotong University. His research interests
[10] H. Mori, Y. Ogita, A parallel tabu search based method for reconfigurations
include power systems optimization, electromagnetic compatibility theory and
of distribution systems, Power Eng. Soc. Summer Meet. IEEE 1 (2000)
lightning protection.
73–78.
[11] Y. Mishima, K. Nara, T. Satoh, T. Ito, Method for minimum-loss reconfig- Liu-Chun Zhang received his M.Sc. degree from Taiyuan University of Sci-
uration of distribution system by tabu search, Electrical Eng. Jpn. 152 (2) ence & Technology, Shanxi, P.R. China in electrical engineering in 2001. She
(2005). is currently a Ph.D. student in the Department of Electrical Engineering at
[12] Y.C. Huang, Enhanced genetic algorithm-based fuzzy multi-objective Shanghai Jiao Tong University. Her recent research interests include distribution
approach to distribution network reconfiguration, IEE Proc.-Gen. Transm. system automation and overvoltage protection of the low voltage distribution
Distrib. 149 (5) (2002). systems.
[13] Y.-J. Jeon, J.-C. Kim, Application of simulated annealing and tabu search
for loss minimization in distribution systems, Electrical Power Energy Syst.
26 (2004) 9–18.