A comparison of crossover operators in genetic algorithms for switch allocation problem in power distribution systems

Combinatorial optimization problems are commonly found in different stages of electric distribution systems deploying. Such problems demand the use of metaheuristics to find good solutions with acceptable computational effort. Among these metaheuristics, genetic algorithms prove to be an interesting method for this kind of problem due to the good solutions found by them in several applications. From this context, the present paper proposes an analysis of the effect of different crossover operators on the quality of obtained solutions in a genetic algorithm applied to the switch allocation problem in power distribution systems. The comparisons were conducted based on a hypothetical system from the literature with 135 buses and 1 feeder. The experiments showed which the restriction degree imposed on the search space influences the differences between crossover operators. The results suggest which exists an ideal number of cut points for the multi-points crossover operator which found better results than the one-point, uniform, and other crossovers.