1

A minimal spanning tree algorithm for distribution

networks configuration
Diana P. Montoya, Juan M. Ramirez, Member, IEEE.

AbstractA minimal spanning tree algorithm for distribution reconfiguration purposes. The proposition is based on power
networks configurations is proposed. The paper approaches the losses minimization taken the switches status into account, a
formulation by the power losses minimization in order to select radial configuration, and satisfying the demand. The Kruskals
a new configuration, keeping a radial configuration. This type
of problem becomes a non-differentiable, mixed integer, and algorithm [12] is used to determine the minimum spanning
highly combinatorial in nature. The Kruskals algorithm is tree.
used to determine the minimum spanning tree, which has the
characteristic to run in polynomial time. II. F ORMULATION
Index TermsDistribution systems reconfiguration, Power loss The reconfiguration problem is formulated by a real
reduction, Minimal Spanning Tree.
power losses minimization while satisfying the grids
radiality constraint and satisfying the demand. The three-phase
I. I NTRODUCTION
distribution system is assumed to be balanced and all loads are
ISTRIBUTION networks are generally structured in
D mesh, but operated in a radial configuration for effective
co-ordination of the protective schemes, and to reduce the
assumed to be of constant power type. The reconfiguration
problem maybe formulated as follows:

fault level. Reconfiguration of a distribution system is a E

 Pn2 + Q2n
process that alters feeder topological structure by managing M inimize Rn (1)
the open/closed status of sectionalizing and tie-switches, n=1
|Vn2 |
under either contingencies or normal operating conditions.
where E is the total number of lines; Rn , Pn and Qn are the
There are many alternatives available for reducing losses at
resistance, active power flow and reactive power flow through
the distribution level, for instance: (i) reconfiguration, (ii)
line n, respectively; Vn is the n-th bus voltage.
capacitor installation, (iii) load balancing, (iv) the use of
Eq. (1) corresponds to the objective function and represent
higher voltage levels. This paper focuses on the reconfiguration
the total real power losses of the grid. The objective function
alternative.
(1) is subject to:
The problem of distribution network reconfiguration is
highly complex, combinatorial, and nondifferentiable due
to the large number of discrete switching elements. In Vmin V Vmax (2)
addition, the radiality constraint typically introduces additional =0 (3)
complexity in the reconfiguration problem of large-sized
distribution networks. Eq. (2) refers to voltage constraints for each bus. Eq. (3),
Although the branch and bound, branch exchange, and deals with the radial topology constraint.
the expert systems can solve the problem with rather
less computational burden, the calculated results are only III. BASICS
approximations and local optima [1]. In the open research,
A. Graph Theory
several heuristic techniques have demonstrated to solve the
reconfiguration problem by losses minimization. The most According to the graph theory [13], a graph (G) refers to a
used heuristic methods are Genetic Algorithm (GA) [1], [2], collection of vertices or nodes, and a collection of edges that
[3], Tabu Search (TS) [4], Ant Colony Optimization (ACO) connect pairs of vertices. A subgraph is a subset of branches
[5], [6], [7], [8], Particle swarm optimization (PSO) [8], (edges) and nodes (vertices) of a graph. Graphs are so named
[9], [10], [11]. Likewise, other heuristic methods work well because they can be represented graphically, and it is this
handling small systems. In general, it is difficult to find global graphical representation which helps understand many of their
optimum in actual systems that would have a large number of properties.
switches. The symbolic representation V (G) and E (G) denote the
An efficient minimal spanning tree algorithm for distribution set of vertices and the set of edges in a graph G, respectively.
network configuration is proposed in this paper with v (G) and (G) denote the number of vertices and the edges
in graph G.
Diana Montoya is a M. Sc. student in Department of Electrical Power A tree is a connected subgraph of a connected graph
System, CINVESTAV-Guadalajara. (e-mail:[email protected])
Juan M. Ramirez is with CINVESTAV-Guadalajara. MEXICO. (e- containing all the nodes of the graph, but containing no cycles,
mail:[email protected]) i.e., a tree is a connected acyclic graph [13].

978-1-4673-2729-9/12/$31.00 2012 IEEE

 2

The Binet-Cauchy theorem [13] says that the total number Selection: Select the edge from the pool of candidates
of trees, Nt , is given by the determinant of A.AT , i.e., with less weight.
Objective Function: To determine the total length of
    selected edges (minimum).
det AAT = Products of corresponding majors of A and AT
Kruskals algorithm run in O (E log E) time or equivalently,
all
majors O (E log V ) time, where E is the number of edges in the

= (all nonzeros majors of A)
2
(4) graph and V is the number of vertices [18]. The big O
notation classifies algorithms by how they respond (e.g. in their
= Nt
processing time or working space requirements) to changes in
where matrix A is called the incidence matrix, and is input size. These routines are equivalent to the following:
obtained from the matrix A (complete incidence matrix or 2 2
E is at most V and log V = 2logV is O (log V ).
the complete node-branch incidence matrix) by deleting the If we ignore isolated vertices, V E + 1, so log V is
column corresponding to the reference node. O (log E).
The reconfiguration problem could be solved by finding The Kruskals algorithm pseudocode is presented below
all the trees of a graph and evaluate the losses. Notice that [18]:
according to (4), the problem of finding the trees is a no 1 function Kruskal (G = N, A : graph; length :
polynomial problem (NP) respect to the number of vertices. A R+ ): set of edges
2 Define an elementary cluster C (v) {v}
B. Minimal Spanning Tree (MST) 3 Initialize a priority queue Q to contain all edges in G,
using the weights as keys.
The minimal spanning tree problem is one of the oldest and
4 Define a forest T
most basic graph problems in theoretical computer science.
5 //T will ultimately contain the edges of the MST
Its history dates back to Boruvkas algorithm in 1926 [12],
[14]. Nowadays, there are two algorithms commonly used: (i) 6 //n is total number of vertices
Prim, and (ii) Kruskal. Both are greedy algorithms that run in 7 while T has fewer than n 1 edges do
polynomial time. 8 //edge (u, v) is the minimum weighted route from u
The Prims algorithm has been used to solve the restoration to v
problem in distribution networks [15], where an assignment of 9 (u, v) Q. removeM in ()
the weighting coefficient to all branches are used to decrease 10 // prevent cycles in T . add (u, v) only if T does not
the number of switching operations. already contain a path between u and v.
in the proposed algorithm was employed the greedy 11 // the vertices has been added to the tree.
Kruskals strategy the greedy Kruskals algorithm was 12 Let C (v) be the cluster containing v, and let C (u)
employed to solve the reconfiguration problem and because be the cluster containing u.
it finds a minimum spanning tree by a global search in the 13 if C (v) = C (u) then
network. If the edges weight are integers, then deterministic 14 Add edge (v, u) to T .
algorithms are known to solve the problem in O (m + n) 15 Merge C (v) and C (u) into one cluster,
integer operations [16]. that is, union C (v) and C (u).
Given a weighted edge graph, this problem calls for finding 16 return tree T
a subgraph (a tree) spanning all the vertices which total weight
is minimal. In either case the selected links will have to form IV. P ROPOSITION
a tree (assuming all weights are positive and undirected). On Distribution networks are operated in radial configurations,
the contrary, then the selected links contain a cycle [17]. despite they have a mesh structure. The radial configuration
decreases the fault level and it is appropriate from the
C. Kruskals algorithm protection point of view. In this paper, the network
configuration is based on this prerequisite. Additionally, the
Kruskals algorithm is based directly on the generic
algorithm looks for minimum losses. The proposed algorithm
minimum-spanning-tree algorithm. In the Kruskals algorithm
looks for the minimum losses.
the edges are selected and added to the spanning tree in
It is assumed that all data are available. These includes:
increasing order of their weights. An edge is added to the
(i) topology; (ii) switches status (open/closed); (iii) feeders
tree only if it does not create a cycle [12].
parameters (resistance, inductive reactance). With this
Thus, the Kruskals algorithm works by growing the tree,
information, each branchs weight is assessed. Then, the
one edge at a time, adding the smallest edge that does not
minimum spanning tree is found.
create a cycle.
Kruskals algorithm can be summarized in the following
way: A. Weighting
Candidates: Edges. After several trials, the weights used in the MST are the
Solution: build a tree. absolute value of the active power flows through branches,
Restriction: The set of edges contains any cycle. calculated by a load flow run using the configuration of the
 3

entire network (lines normally open and closed). The bus

voltage in the source (substation) is set to 1.0 p.u. After the
load flow convergence, all active power flows are calculated
through each branch, Fig. 1.

R + jX
Vk k Vm m
Ik Im

Yk Yk
2 2

Fig. 1. Tie-lines power flow

The voltages at the sending node Vk and the receiving one

Vm , are given by (5) and (6)

Vk = |Vk | cos (k ) + j |Vk | sin (k ) (5)

Vm = |Vm | cos (m ) + j |Vm | sin (m ) (6)
The injected current in the k-th bus becomes

 
Vk Vm Gk + jBk
Ik = + Vk (7)
R + jX 2
 
Vk Vm Yk
= + Vk
Zkm 2
and the current thorough the line m:
 
Vm Vk Gm + jBm
Im = + Vm (8)
R + jX 2
 
Vm Vk Yk
= + Vm
Zkm 2
Therefore, the power flow is given by:

Se = Vk Ik (9)
  
Vk Vm Yk
= Vk + Vk
Zkm 2


2 1 Vk Vm Y
= |Vk | + k
R jX 2
Pe = (Se ) (10)
Fig. 2. Flow chart of the proposed algorithm
Qe = (Se ) (11)
V. E XAMPLES
In this paper, the branchs weight is proposed as follows,
A. 33-Bus System
1 1 The first test system is a hypothetical 12.66 kV system with
W = =
 
(12)
|Pe |  |Vk |

2 1Vk Vm Yk
33 nodes, 37 edges, and 5 looping branches (tie lines). The
RjX + 2
total real and reactive power loads on the system are 3715 kW
The flow chart of the proposed algorithm for losses and 2300 kVAr, respectively. The total loss is about 8% of
minimization is shown in Fig. 2. The proposed algorithm was the total load. A lossy system is selected because the loss
developed using Matlab, and its performance is assessed by reduction is expected to be appreciable. The system data is
two distribution systems (33-bus system and 69-bus system) given in [19]. The single-line diagram of the system is depicted
[19], [20]. in Fig. 3.
 4

1.02
Before reconfiguration
After reconfiguration
1

Bus Voltage [p.u.]

0.98

0.96

0.94

0.92

0.9
5 10 15 20 25 30
Bus No.

Fig. 4. Voltage profile for the 33-bus test system

Wang, et al [24], developed and immune genetic algorithm

Fig. 3. Single line diagram of 33-bus system fo distribution networks reconfiguration (IIGA) by analogy
to biological immune mechanisms. Choi [25] solves the
To carry out the power flow study, the selected base power problem with a heuristic based switching scheme to solve the
and base voltage are 100 MVA and 12.66 kV, respectively. The distribution feeder loss reduction problem.
slack bus is the main substation (bus 1), and its voltage angle Simulation results for the test system are summarized in
and voltage magnitude are set to 1.00 p.u. The remaining Table I and Fig. 5.
buses are load buses (PQ type), where for the flat starting, all
TABLE I
voltage angles are setting to zero and all voltage magnitudes C OMPARISON RESULTS FOR TEST 33- BUS SYSTEM
are 1.0 p.u.
The real-power losses before reconfiguration (initial Methods Configuration Ploss [kW] Vmin [p.u]
conditions of test systems, Fig. 3) become 197.379 kW, while Initial conf. 33-37 197.379 0.914
Varying Degree [21] 7, 10, 14, 27, 30 158.063 0.904
the minimum voltage is Vmin = V18 = 0.914 p.u. Branches FMGA [19] 7, 9, 14, 32, 37 136.25 0.939
with greater losses are the closest to the most loaded buses. L-E [22] 8, 14, 28, 32, 33 142.45 0.939
Higher loads (greater than 200 kW) are connected to nodes HSA [23] 7, 10, 14, 36, 37 138.453 0.935
IIGA [24] 7, 9, 14, 32, 37 136.25 0.939
8, 9, 24, 25, 30 and 21; in the neighboring branches respect Sw Scheme [25] 7, 9, 14, 32, 37 136.25 0.939
to these nodes are the highest proportion of losses. When Proposed Method 7, 10, 14, 28, 36 138.825 0.939
applying the MST algorithm, the configuration that presents
the fewer losses is that where branches 7, 10, 14, 28 and 36
are open.
Fig. 4 illustrates the voltage profile improvement achieved
by the proposed strategy. Notice that magnitudes of most of
the bus voltages have been improved after reconfiguration.
The minimum bus voltage is equal to 0.9178 p.u, and after
reconfiguration it is raised to 0.9593 p.u.
The real-power losses with the proposed configuration
become 138.825 kW, and the minimum voltage magnitude is
Vmin = V60 = 0.939 p.u. The results were compared with the
configuration proposed in [19], [21], [22], [23], [24] and [25].
Venkatesh, et al [19] solve the configuration problem with
two algorithms: Fuzzy Mutated Genetic Algorithm (FMGA)
and Evolutionary Programming (EP). Babu, et al [22], present
an algorithm based upon Loop Eliminating method (L-E); the
L-E method is a computationally heuristic method. Srinovasa
Rao, et al [23], solve the problem by a Harmony Search
Algorithm (HSA), this is an algorithm which is conceptualized
Fig. 5. Comparison results for the 33-bus test system
using the musical process of searching for a perfect state
of harmony; it uses a stochastic random search instead of
a gradient search which eliminates the need for derivative The weight in branches before and after reconfiguration are
information. shown in Table II.
 5

TABLE II
W EIGHTS FOR THE 33- BUS TEST SYSTEM

Branch Weight Before Weight Afer

1 26.07 25.95
2 36.99 40.42
3 80.13 160.45
4 89.02 199.15
5 94.39 226.56
6 226.96 499.85
7 416.05 -
8 276.23 254.54
9 1427.42 1665.88
10 9978.52 -
11 2858.80 2222.22
12 452.52 553.93
13 624.15 832.81
14 2496.41 -
15 476.75 474.97
16 669.81 665.42
17 1126.58 1110.51
18 97.97 78.86
19 107.63 85.03
20 120.98 93.46
21 244.22 264.58
22 74.87 57.70
23 80.75 61.37
24 124.07 84.41
25 181.38 555.23
26 203.78 832.93
27 232.56 1665.90
28 271.84 -
29 159.79 159.89
30 236.46 237.12 Fig. 6. Single line diagram of 69-bus system
31 368.08 370.07
32 1625.44 1666.32
33 310.63 168.67 Fig. 7 illustrates the voltage profile improvement achieved
34 433.23 368.12
35 316.48 350.22
by the proposed strategy. Notice that magnitudes of most of
36 69983.01 - the bus voltages have been improved after reconfiguration.
37 263.02 133.38 The minimum bus voltage is equal to 0.9178 p.u., and after
reconfiguration it is raised to 0.9593 p.u.
The real-power losses with the proposed configuration
B. 69-Bus System become 80.868 kW, and the minimum voltage magnitude is
Vmin = V60 = 0.9593 p.u. The results are compared respect
The second test system is a hypothetical 12.66 kV system to the configurations proposed in [1], [20], [22] and [24].
with 69 nodes, 73 edges, and 5 looping branches (tie lines).
The total real and reactive power loads on the system are
1.02
3801.5 kW and 2694.6 kVAr, respectively. The system data is Before reconfiguration
After reconfiguration
given in [20]. The schematic diagram of the system is depicted
1
in Fig. 6.
To carry out the power flow study, the selected base power
Bus Voltage [p.u.]

0.98
and base voltage are 10 kVA and 12.66 kV, respectively. The
slack bus is the main substation (bus 1), and its voltage angle 0.96
and voltage magnitude are set to 1.00 p.u. The remaining
buses are load buses (PQ type), where for the flat starting, all 0.94
voltage angles are setting to zero and all voltage magnitudes
are 1.0 p.u. 0.92

The real-power losses before reconfiguration (initial

conditions of test systems, Fig. 6) become 204.45 kW, while 0.9
10 20 30 40 50 60
the minimum voltage is Vmin = V65 = 0.9178 p.u. Branches Bus No.
with greater losses are the closest to the most loaded buses.
Higher loads (greater than 100 kW) are connected to buses 10, Fig. 7. Voltage profile for test 69-bus system
11, 48 and 49; in the neighboring branches respect to these
buses are allocated the highest proportion of losses. When Savier and Das [20] solve the configuration problem with
applying the MST algorithm, the configuration that presents an algorithm based on a fuzzy multiobjective approach,
less losses is that where branches 13, 20, 58, 61, and 69 are and the max-min principle is adopted for the multiobjective
open. optimization in a fuzzy framework. Swarnkar, et al [1] solve
 6

the configuration problem with a GA; Babu, et al [22] with manner. Nowadays, sub-optimal results have been obtained,
an L-E method; and Wang, et al [24] with an immune genetic and will be presented in a future paper.
algorithm IIGA.
Simulation results for this test system are summarized in VI. C ONCLUSION
Table III and Fig. 8. It is worth noting that the results obtained This paper presented the application of the minimal
by the proposed algorithm are comparable or better that results spanning tree algorithm to solve the distribution network
obtained with the above mentioned strategies. reconfiguration problem. Some rules are framed by the
TABLE III application of the graph theory to create only feasible radial
C OMPARISON RESULTS FOR TEST 69- BUS SYSTEM topologies. The algorithm is easy to apply and to reply. It
is deterministic, which means that, under the same conditions,
Methods Configuration Ploss [kW] Vmin [p.u.] the same result is guaranteed for different runs. This represents
Initial conf. 69-73 204.45 0.9178
Heuristic [20] 14, 56, 61, 69, 70 85.931 0.9567 a difference respect to the evolutionary algorithms that have
GA [1] 14, 56, 61, 69, 70 85.931 0.9567 been applied. Weighting branches through the reciprocal of
L-E Method [22] 13, 20, 64, 58, 69 91.671 0.949 its active power flow it is an appropriate strategy to attain a
IIGA [24] 14, 58, 61, 69, 70 79.959 0.959
Proposed 13, 20, 58, 61, 69 80.868 0.9593 radial topology with minimum losses. The proposed algorithm
is tested on two distribution systems test (33-bus and 69-
bus) for loss minimization showing promising results. As a
secondary advantage, in general, voltage magnitudes become
better than those obtained with other formulations. The use
of evolutionary techniques requires high computational time
and burden, while the proposed algorithm runs in polynomial
time.

R EFERENCES
[1] N. Swarnkar, N. Gupta and K. Niazi, A novel codification for
meta-heuristic techniques used in distribution network reconfiguration,
Electric Power Systems Research - ELSEVIER, vol. 81, pp. 16191626,
2011.
[2] Y. Hu, N. Hua, C. Wang, J. Gong, and X. Li, Research on distribution
network reconfiguration, in Proc. Int Computer, Mechatronics, Control
and Electronic Engineering (CMCE) Conf, vol. 1, 2010, pp. 176180.
[3] Y.-Y. Hong and S.-Y. Ho, Determination of network configuration
considering multiobjective in distribution systems using genetic
algorithms, IEEE Transactions on Power Systems, vol. 20, no. 2, pp.
Fig. 8. Comparison results for test 69-bus system 10621069, 2005.
[4] A. Abdelaziz, F. Mohamed, S. Mekhamer, and M. Badr, Distribution
system reconfiguration using a modified tabu search algorithm, Electric
Likewise, although our results are not the best, are quite Power Systems Research, vol. 80, no. 8, pp. 943 953, 2010.
[5] E. Carpaneto and G. Chicco, Distribution system minimum loss
competitive. The advantage of this algorithm is its reliability reconfiguration in the hyper-cube ant colony optimization framework,
and velocity in comparison to evolutionary algorithms that Electric Power Systems Research, vol. 78, no. 12, pp. 2037 2045,
exhibit a slow process of convergence. 2008, special Issue Papers Presented at the 6th World Energy System
Conference, 6th World Energy System Conference.
The proposed algorithm has been implemented into a [6] T. Niknam, A. M. Ranjbar, A. R. Shirani, and A. Ostadi, A new
software package in MATLAB 7.10.0.499, executed on a approach based on ant colony algorithm to distribution management
desktop with 3 GHz Intel Core 2 Duo E6850 Processor and system with regard to dispersed generation, in Proc. 18th Int. Conf.
and Exhibition Electricity Distribution CIRED 2005, 2005, pp. 15.
2 GB System Memory, and applied to several distribution [7] T. Niknam, A new approach based on ant colony algorithm to
systems. In this section some numerical results are presented to distribution management system with regard to dispersed generation,
illustrate the performance of the proposed solution algorithm. in Energy Conversion and Management 49(12): 3417-3434, 2008, pp.
15.
The computational time are shown in Table IV. [8] R. K. T. Niknam and B. B. Firouzi, A hybrid evolutionary algorithm for
distribution feeder reconfiguration, SADHANA, vol. 35, pp. 135162,
TABLE IV 2010.
C OMPUTATIONAL TIME [9] R. Kennedy, J.; Eberhart, Particle swarm optimization, IEEE
International Conference on Neural Networks, vol. 4, pp. 1942 1948,
System Time [seconds] 1995.
33-bus system 0.250665 [10] Z.-L. Gaing, A particle swarm optmization approach for optimum
design of pid controller in avr system, IEEE Transactions on Energy
69-bus system 0.436594 Conversion, vol. 19, no. 3, pp. 384391, 2003.
[11] W.-C. Wu, M.-S. Tsai, and F.-Y. Hsu, A new binary coding particle
swarm optimization for feeder reconfiguration, in Proc. Int. Conf.
The next step in this application is to take distributed Intelligent Systems Applications to Power Systems ISAP, 2007, pp. 16.
generation into account, since the smart grid philosophy will [12] J. J.B. Kruskal, On the shortest spanning subtree of a graph and the
be a reality in the near future. With this purpose, the spanning traveling salesman problem, Proceedings of the American Mathematical
Society, vol. 7, p. 4850, 1956.
tree algorithm is being prepared to function based on self [13] N. Balabanian and T. Bickart, Electrical Network Theory, Hardcover,
organization theory, so that it will work in a distributed Ed. John Wiley & Sons Inc, 1969.
 7

[14] S. Chung and A. Condon, Parallel implementation of Bouvkas Int Information Management, Innovation Management and Industrial
minimum spanning tree algorithm, in Proc. 10th Int. Parallel Engineering Conf, vol. 3, 2009, pp. 218223.
Processing Symp., IPPS, 1996, pp. 302308. [25] S. Y. Choi, Heuristic approach for loss reduction in distribution
[15] S. Dimitrijevic and N. Rajakovic, An innovative approach for solving networks, in Proc. IEEE Power Engineering Society General Meeting,
the restoration problem in distribution networks, Electric Power Systems 2006.
Research, vol. 81, no. 10, pp. 1961 1972, 2011.
[16] M. L. Fredman and D. E. Willard, Trans-dichotomous algorithms for
minimum spanning trees and shortest paths, Journal of Computer and
System Sciences, vol. 48, no. 3, pp. 533 551, 1994.
[17] J. L. Bentley, A parallel algorithm for constructing minimum spanning Diana P. Montoya obtained her degree in
trees, Journal of Algorithms, vol. 1, no. 1, pp. 51 59, 1980. Electrical Engineering from Universidad Nacional
[18] C. R. R. Cormen, T.H. Leiserson and C. Stein, Introduction to de Colombia, Medelln-Colombia in 2008. She is
Algorithms. McGraw-Hill, 2009. currently pursuing her Master of Science degree in
[19] B. Venkatesh, S. Chandramohan, N. Kayalvizhi, and R. P. Electrical Engineering in the Centro de Investigacion
Kumudini Devi, Optimal reconfiguration of radial distribuion system y Estudios Avanzados del Instituto Politecnico
using artificial intelligence methods, in Proc. IEEE Toronto Int Science Nacional (CINVESTAV), Guadalajara, Mexico. Her
and Technology for Humanity (TIC-STH) Conf, 2009, pp. 660665. research interest is Smart-Grid and Electric Power
[20] J. S. Savier and D. Das, Impact of network reconfiguration on loss Systems.
allocation of radial distribution systems, IEEE Transactions on Power
Delivery, vol. 22, no. 4, pp. 24732480, 2007.
[21] M. E. Baran and F. F. Wu, Network reconfiguration in distribution
systems for loss reduction and load balancing, IEEE Transactions on
Power Delivery, vol. 9, no. 4, pp. 101102, 1989.
[22] P. R. Babu, M. P. R. Vanamali, M. P. V. V. R. Kumar, and V. S.
Juan M. Ramirez (M1986) obtained his BS
Hemachandra, Network reconfiguration in distribution systems using
in Electrical Engineering from Universidad de
l-e method, in Proc. Annual IEEE India Conf. (INDICON), 2010, pp.
Guanajuato, Mexico in 1984; M. Sc. in Electrical
14.
Engineering from UNAM-Mexico in 1987; Ph. D. in
[23] R. Srinivasa Rao, S. V. L. Narasimham, M. Ramalinga Raju, and
Electrical Engineering from UANL-Mexico in 1992.
A. Srinivasa Rao, Optimal network reconfiguration of large-scale
He joined the department of Electrical Engineering
distribution system using harmony search algorithm, IEEE Transactions
of CINVESTAV in 1999, where he is currently a full
on Power Systems, vol. 26, no. 3, pp. 10801088, 2011.
time professor. His areas of interest are in operation
[24] C. Wang, A. Zhao, H. Dong, and Z. Li, An improved immune
and control of electric power systems.
genetic algorithm for distribution network reconfiguration, in Proc.