A rapid particle swarm optimization algorithm with convergence analysis

Abstract

Confronting with complex and troublesome optimization problems, Particle Swarm Optimization (PSO) algorithm maybe easy to be led to local optimum and suffer the unacceptable phenomenon as premature convergence. This paper proposes a new rapid PSO algorithm (RPSO), utilizing all particles' individual best positions found so far to update its velocity; providing a distinguishing weight according to the particles' different positions; and an adaptive learning factors turning strategy based on particle's fitness value. Then Jury Criterion is adopted to make convergence analysis of our proposed algorithm. Ultimately, the optimization performance of RPSO is simulated by searching the optimums of four common benchmark functions and training a RBF network for approximating a nonlinear system. The simulation results reveal the satisfactory efficacy of our proposed algorithm.