A Gemetic Algorithm for Distribution Feeders Reconfiguration for Loss Minimization

In this paper the application of a genetic optimization algorithm for distribution network loss minimization is proposed. In general, distribution networks are operated as radial networks; however, their design is such that their configuration could be changed time to time. Computing the optimal radial configuration that minimizes the total network losses for a given load distribution involves solving a large discrete nonlinear programming. Since these type of problems are very hard to solve in an on-line environment, as in practical distribution systems, approximate near optimal solutions are sought and proposed by different authors. Genetic algorithms are general purpose stochastic optimization methods which are suitable for solving discrete nonlinear programming problems. It is shown how to transform the distribution network reconfiguration for loss minimization problem to a form suitable for genetic algorithm application. In order to accomplish this an efficient Kruskal-like algorithm is developed that constructs spanning trees and computes line losses for feeders with different radial configurations. An important part of the Kruskal-like algorithm is a very efficient and fast technique for recognizing loops and constructing spanning trees which is known as "fast disjoint-set union" algorithm.