Three Sub-Swarms Exchange Particle Swarm Optimization Based on DifferentEvolvement Model

摘要

Particle swarm optimization and its modifications appear premature convergence for complex optimization problem, because particles' performance becomes same in seeking later period. In this paper, a new model is proposed to avoiding particles' performance same and possessing strong exploration capacity. Considering exploration and exploitation capacity diverse in different stage, the particle swarm is divided into three identical sub-swarms, with the first adopting the standard PSO model, the second adopting the cognition-only model, and the third adopting the proposed model. When the three sub-swarms evolve steady states independent, we exchange some particles between the three different sub-swarms, which can increase the information exchange between the sub-swarms, improve the population diversity and reduce the possibility of getting local extreme value. Four complex testing functions' results indicate that the proposed algorithm has greater globally optimal solution, better optimal efficiency and better performance than PSO and TS-PSO[1] in many aspects