Abstract: A new Particle Swarm Optimization（PSO） algorithm with global optimization of parameters for adaptive mutation is proposed around two key parameters[w]and[pgd]which all affect PSO algorithm performance to avoid the possible problems about local convergence and low precision. The concept of particle entropy set is defined which can accurately reflect the PSO data global aggregation behavior. The regression variance for inertia weight[w]of swarm dimensional data and the random variance for global variable[pgd]are determined by the particle entropy of every dimension data, and the method of using mutation frequency factor is used to avoid divergence in the algorithm. Simulation results show that compared with the conventional algorithm there are great advantages in optimization precision, getting rid of local traps, stability, etc, and good performance in solving complex multimodal problems with this algorithm.

Key words: particle entropy set, inertia weight, global optimal location, adaptive mutation, Particle Swarm Optimization（PSO）

摘要： 为了避免普通粒子群算法（PSO）可能出现的局部收敛及精度不高现象，围绕影响PSO算法性能的两个重要参数[w]和[pgd]，提出了一种面向全局优化的参数自适应变异PSO改进算法。算法定义了粒子熵集概念，可以精确反映粒子群数据的全局聚集特性，由粒子群各维数据的熵值大小决定是否对各维数据的惯性权重[w]进行回归变异，对全局变量[pgd]进行随机变异，并采取引入变异次数因子等方法来避免寻优发散。仿真研究表明该算法比常用算法在寻优精度、摆脱局部陷阱、稳定性等方面均有明显提高，在求解复杂多峰问题上有着良好的表现。

关键词: 粒子熵集, 惯性权重, 全局最优位置, 自适应变异, 粒子群优化算法