A Multiobjective Evolutionary Algorithm for Distribution System Reconfiguration Problems

Abstract

Real-world multiobjective optimization problems (MOPs) are generally constrained and multiobjective evolutionary algorithms (MOEAs) have been proposed for solving constrained MOPs (CMOPs). When solving CMOPs observed in the real world, which are computationally demanding and have severe constraints, finding a well-converged population to constrained Pareto front is not easy. In such cases, infeasible nondominated solutions with slight constraint violations can provide useful information for decision makers. In this paper, we propose a MOEA that can provide decision makers with trade-off information between constraint violations and objective values of those untractable CMOPs. The proposed MOEA partitions the objective space of a CMOP, with objectives to minimize constraint violations, into multiple subspaces and maintains the diversity of the population by preserving feasible and infeasible solutions in the subspaces. Solutions with slight constraint violations and small objective values are preserved in the population using different selection pressures for the feasible and infeasible populations. Computational experiments for solving a distribution system reconfiguration problem, which is a computationally demanding and severely constrained CMOP in the real world, are performed to demonstrate the effectiveness of the proposed MOEA.