Performance of multiple objective evolutionary algorithms on a distribution system design problem - computational experiment

Abstract

The paper presents a comparative experiment with four multiple objective evolutionary algorithms on a real life combinatorial optimization problem. The test problem corresponds to the design of a distribution system. The experiment compares performance of a Pareto ranking based multiple objective genetic algorithm (Pareto GA), multiple objective multiple start local search (MOMSLS), multiple objective genetic local search (MOGLS) and an extension of Pareto GA involving local search (Pareto GLS). The results of the experiment clearly indicate that the methods hybridizing recombination and local search operators by far outperform methods that use one of the operators alone. Furthermore, MOGLS outperforms Pareto GLS.