改进的粒子群算法及收敛性分析  

Modified particle swarm optimization algorithm and its convergence analysis

作　　者：谢铮桂[1,2] 钟少丹[1,2] 韦玉科[2] 

机构地区：[1]韩山师范学院数学与信息技术系,广东潮州521041 [2]广东工业大学计算机学院,广州510090

出　　处：《计算机工程与应用》2011年第1期46-49,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.30472122~~

摘　　要：针对PSO算法对多峰值函数搜索易陷入局部极值点的缺点,提出一种改进的粒子群(MPSO)算法。MPSO算法采用逃逸策略和免疫学习策略来保证种群多样性,使算法能有效进行全局搜索。并讨论MPSO算法的收敛性,证明其能以概率1全局收敛。最后用3个常用的测试函数进行仿真,实验结果表明MPSO算法比PSO算法有更好的收敛性和更快的收敛速度。This paper proposes a modified particle swarm algorithm in allusion to the defect that the PSO algorithm easily plunges into the local optimization for multi-peak function optimization problem.MPSO adopts escape strategy and immune learning strategy to guarantee the particles diversity and to make particles explore the global optimization more efficiently.The convergence of MPSO algorithm is discussed and is proved to converge to the global optimization with probability one.Finally,three familiar test functions are simulated to show that MPSO achieves better and faster convergence.

关 键 词：粒子群算法 逃逸 免疫学习 全局优化 收敛性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]