拥有领导机制的改进粒子群算法  

Improved particle swarm optimization with leadership

作　　者：周龙甫[1,2] 师奕兵[2] 张伟[2] 

机构地区：[1]解放军第452医院信息科,成都610061 [2]电子科技大学自动化工程学院,成都611731

出　　处：《控制与决策》2010年第10期1463-1468,1473,共7页Control and Decision

基　　金：教育部新世纪人才支持计划项目(NCET-05-0804);国家863计划项目(2006AA06Z222)

摘　　要：为了提高粒子群算法的全局收敛能力和收敛速度,在以往文献的基础上提出一种改进粒子群算法.受生物学研究成果的启发,引入领导机制,将粒子群搜索过程分为领导粒子带领下的探索性搜索和所有粒子共同参与的开发性搜索两部分.通过"变异"机制来增强群体多样性,采用5个标准函数对算法性能进行分析,应用Bonferroni多重比较法,将改进算法与两种经典算法进行基本性能对比.实验结果表明,所提出的改进算法探索速度快,全局搜索能力优.In order to enhance the global convergence ability and the convergence speed of particle swarm optimization（PSO）,based on the results of reference,an improved strategy is proposed.Inspired by the effect of individual experience in animal groups on the move,the whole searching process is departed into two steps,the exploration search with leadership and the exploitation search.The mutation operator is used to add the diversity of swarm.By analyzing five benchmark functions,the improved method is compared with two famous PSO methods.And the variance analysis of statistic theory is applied to compare the performance of the three methods.Experimental simulation results show that the proposed algorithm can not only significantly speed up the convergence,but also effectively solve the premature convergence problem.

关 键 词：粒子群 领导机制 变异 方差分析 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]