A Parallel Genetic Algorithm Based Methodology for Network Reconfiguration in the Presence of Dispersed Generation

Distribution network planning and operation require the identification of the best topological configuration that is able to attend the power demand with minimum losses, maximum reliability level and adequate voltage profile. Many possible configurations may be obtained by opening and closing switching devices on the network in a process called reconfiguration. In the presence of dispersed generation, this problem gets more complex since questions related to flow direction, short circuit level, voltage regulation, islanding supply, etc are introduced. The network reconfiguration problem, under normal or contingency operation, is a non-linear combinatorial optimization problem of large dimensions. Several approaches for solving this problem have been proposed, but they usually require unacceptable practical simplifications or high computational effort. Modern heuristics methods like Genetic Algorithms (GA) have been successfully applied to this problem. Basically, these methods consist of randomly generating an initial set of candidate solutions and evolving this set by modifications on the candidates via genetic operators. This paper presents a methodology for network reconfiguration in the presence of dispersed generation, based on genetic algorithms. The problem is formulated as a multi-objective optimization problem, where both the electrical losses and the feeders’ voltage profile can be optimised. Since distribution systems usually operate with a radial configuration, it has been implemented an intelligent strategy that avoids generating unfeasible candidates by the GA in order to improve the computational performance. However, large systems optimal reconfiguration may be very time consuming especially if high precision models are used. In order to reduce the required processing time, a parallel multi-population genetic algorithm (PGA) has been implemented, in which a different population is allocated to each processor of the parallel platform and they interchange good candidate solutions to enhance the optimal solution search.