Dynamic Neighborhood Hybrid Particle Swarm Optimization for Constrained Optimization

Abstract

Particle swarm optimization (PSO) is simple and efficient, but there is serious premature convergence for solving constrained optimization problem. In order to control premature convergence, this paper proposed dynamic neighborhood hybrid particle swarm optimization (DNH_PSO), which firstly uses the dynamic neighborhood strategy that based on the random topology and the von Neumann topology to improve the global search capacity, secondly incorporates adaptive penalty function constraint handling mechanism, finally introduces Quasi-Newton method effectively enhanced the efficiency of local search ability and convergence speed. Through the experimental comparison with benchmark functions and results show that the algorithm had better global convergence in solving constrained optimization problems.