Novel Particle Swarm Optimization with Differential Evolution Operators

This paper proposes a hybrid PSO variant equipped with differential evolution operators (PSODEO). In PSODEO, three operations (mutation, crossover, and selection) of differential evolution (DE) are organically embedded into PSO to generate novel learning exemplars. Specifically, mutation and crossover operations introduce diverse but not blind information into personal learning exemplars, which is beneficial to enhance exploration ability and preserve population diversity. Selection operation collects the information of particles with better performance and uses them to generate new social learning exemplar, which is conducive to avoid premature convergence. Based on CEC2017, experimental results are obtained by comparing the proposed PSODEO with 17 PSO and 11 non-PSO methods. The comparison results indicate that, by embedding the DE operators, the proposed PSODEO can offer stronger search ability, stabler convergence speed, higher solution accuracy, better population diversity, and shorter time consumption.