Strong Pareto-based multiobjective differential evolution algorithm

Abstract

SPEA2 is a multiobjective optimization algorithm based on the Pareto advantage. Since SPEA2 uses traditional genetic operators, the algorithm cannot solve problems in various complex scenarios. For this reason, this article proposes a multiobjective differential evolution algorithm based on strong Pareto (SPEA2-DE) to solve the above problem with several main strategies. First, the new proposed mechanism divides individuals into nondominated and dominated individuals, and it uses a difference operator based on the optimal vector for nondominated individuals to improve the local exploitation capability. To improve the global exploration capability for dominated individuals, this mechanism also uses a difference operator that is based on the current vector and the optimal individual. Second, the learning factor uses an adaptive strategy based on individual adaptation values to improve both the global exploration capability and the local exploitation capability. Furthermore, an adaptive parameter strategy for crossover probabilities for different evolutionary stage characteristics is employed to improve convergence. Third, the improved environmental selection strategy in SPEA2 makes full use of individual density information to increase population diversity, and the convergence of the proposed algorithm is analyzed. To evaluate the performance of SPEA2-DE, comparison experiments are conducted with six state-of-the-art multiobjective algorithms to solve 16 functions in the DTLZ and WFG benchmark suite and address five real-world engineering application scenarios. The comparison results indicate that the proposed SPEA2-DE is significantly better than, or at least comparable to, its competitors.