关键词: 粒子群优化, 相似度, 阈值控制, 高斯噪声扰动

Abstract: The biggest flaw of PSO（Particle Swarm Optimization） is easy to premature convergence, although some improved PSO algorithms can increase the ability of convergence, but they can not fundamentally solve the problem of premature convergence. In this paper, a new concept of similarity between particles is proposed to measure the degree of the diversity of particle swarm, and for the purpose of gaining a gradual progress of global optimal solution, adaptive thresholds are used to control the adjustment of convergence rate of particle swarm algorithm. In each iteration stage, Gaussian noise and other disturbances based on the similarity are also used to readjust the position of the particle in order to avoid the particle plunging local optimum. Experimental results and theoretical discussion show that new algorithm can effectively improve the globally searching ability of PSO, and effectively avoid the premature convergence.

Key words: Particle Swarm Optimization（PSO）, similarity, threshold control, disturbance of Gaussian noise