Abstract: In order to deal with the problems that the diversity of particle swarm is low and it is easy for particle swarm to fall in local optimum solution, this paper proposes a novel Particle Swarm Optimization （PSO） algorithm based on self-adaptive mutation, which combines with the optimal and other particles’ different role in the population. In the proposed algorithm, according to the evolution degree, the optimal particle can adaptively adjust its adjacent search domain size so as to strengthen the local search capacity and for the non-optimal particles, their locations can initialize randomly in low probability in order to increase the diversity of particle swarm and enhance the global search capacity when its speed is zero. In simulation, the algorithm is applied to the optimization problems of six typical complex functions, and comparing its performance with the other mutation PSO algorithms. The simulation results show that the proposed algorithm not only enhances population diversity, but also strengthens the local search capacity.

Key words: Particle Swarm Optimization（PSO）, local convergence, self-adaptive, mutation, swarm intelligence

摘要： 为了解决粒子群种群多样性低、容易陷入局部最优的缺点，结合最优粒子和其他粒子在种群中的不同作用，给出了一种自适应变异粒子群算法。算法中最优粒子根据种群进化程度，自适应调整自身搜索邻域大小，增强种群的局部搜索能力；对非最优粒子的位置进行小概率的随机初始化，当其速度为零时，速度自适应变化，以便增强种群多样性和全局搜索能力。仿真实验中，将算法应用于6个典型复杂函数优化问题，并与其他变异粒子群算法比较，结果表明，增强种群多样性的同时提高了局部搜索能力。

关键词: 粒子群算法, 局部收敛, 自适应, 变异操作, 群体智能