Contents lists available at ScienceDirect Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr An initial-point strategy for optimizing distribution system reconfguration Hossein Karimianfard, Hossein Haghighat ⁎ Islamic Azad University, Jahrom Branch, Jahrom, Iran ARTICLEINFO Keywords: Distribution system Reconfguration Optimization Initialization strategy Evolutionary algorithms ABSTRACT This study presents a simple and fast strategy for selecting the candidate solutions of the distribution system reconfguration problem without solving optimization based or load fow based programs. It consists of con- structing a node-node adjacency matrix of the original network topology in which specifc diagonal entries corresponding to the of-diagonal ones represent the favourable candidate solutions. This strategy needs only primitive network information and is applicable to any optimization method and network dimension without afectingthecomputationaltractabilityoftheproblem.Throughnumericalexperimentsonvarioustestsystems, the efcacy of the proposed scheme is illustrated. 1. Introduction Rconfguration is a widely used for reducing grid losses and miti- gating operational limits violations of the distribution network. Early studies [1,2] have shown that by changing the status of tie and sec- tionalizingswitches,aradialconfgurationwithminimumlossescanbe achieved. Originally, it was developed for reducing losses and hence- forth was extensively employed by power engineers and researchers to improve the operational performance of the distribution system. The primary objective of network reconfguration is to determine the best network topology, with respect to a given objective, while satisfying the operational standards and radiality of the distribution system. This problem is formulated as a constrained nonlinear optimi- zation problem and solved using mathematical programming methods orevolutionaryalgorithms.Thebinarynatureofthedecisionvariables, associated with the on/of status of the switches, gives rise to a mixed integer nonlinear program, which is known to be difcult to solve especiallyinlarge-scaledistributionsystems.Muchhasbeendonesofar in relation to the applications of distribution system reconfguration and improving the problem formulation, modeling accuracy, and computational performance of the solution algorithms. Mathematical programming methods including linear and nonlinear formulations [3–11], and meta heuristic algorithms including simulated annealing [12,13],geneticalgorithms [14],PSO [15],artifcialimmune [16],and harmonysearchalgorithm [17] havebeenemployed.Avarietyofsingle or multiple objectives involving loss reduction [18–20], maximization of load balance among feeders [21], reducing the number of switching actions [8], etc., are adopted. Giventhevaryingloadconditionsaswellastheuncertaintyofinput parameters and data, fuzzy methods [22], stochastic programming [23], and robust optimization [24] have been applied to solve the problem. In addition to uncertainty modeling, multi-objective pro- gramming and fuzzy approach were extensively used to optimize sev- eral conficting objectives at the same time [25]. Although various modeling and computing aspects of the distribu- tion system reconfguration problem have been addressed in the lit- erature, so far, little has been done on the initialization strategy and candidate solution selection techniques. It is generally known that by augmenting the solution algorithms (mathematical or evolutionary) with a simple initialization technique, often speed and accuracy im- prove quite dramatically. Surprisingly, however, no systematic work seemstohavebeendoneonviablewaysofchoosingthestartingpoints of the reconfguration problem with the possible exception of [5]. This recent reference introduces such a scheme and develops a minimum spanningtreealgorithmtofndaninitialsolutionofthereconfguration problem with the objective of minimizing losses. This approach, frst, formsanedge-weightedgraphofthenetworkandsubsequentlyseeksa spanning tree whose sum of edge weights is minimal. However, the complexity and run-time of this algorithm depend on the network di- mension making it less appealing in large-scale networks. In this paper, a fast and simple strategy is proposed which is con- structed based on the node-node adjacency matrix and needs only to- pological network information without relying on any optimization or loadfowprogram.Choosinginitialsolution(i.e.,startingpoints)inthis wayisbothfastandsimplewhileitscomputationisindependentofthe network dimension which makes it appealing in practice. The required information includes original network topology consisting of branch connections and tie switch locations, which are naturally available for https://doi.org/10.1016/j.epsr.2019.105943 Received 18 December 2018; Received in revised form 6 July 2019; Accepted 11 July 2019 ⁎ Corresponding author. E-mail address: haghighat@jia.ac.ir (H. Haghighat). Electric Power Systems Research 176 (2019) 105943 Available online 07 August 2019 0378-7796/ © 2019 Elsevier B.V. All rights reserved. T