An amelioration Particle Swarm Optimization algorithm

Abstract

a new amelioration Particle Swarm Optimization (SARPSO) based on simulated annealing (SA), asynchronously changed learning genes (ACLG) and roulette strategy was proposed because the classical Particle Swarm Optimization (PSO) algorithm was easily plunged into local minimums. SA had the ability of probability mutation in the search process, by which the search processes of PSO plunging into local minimums could be effectively avoided; ACLG could improve the ability of global search at the beginning, and it was propitious to be convergent to global optimization in the end; the roulette strategy could avoid prematurity of the algorithm. The emulation experiment results of three multi-peaking testing functions had shown the validity and practicability of the SARPSO algorithm.