克服早熟收敛现象的粒子群优化算法  

Particle swarm optimization for breaching local convergence

作　　者：范培蕾[1] 张晓今[1] 杨涛[1] 

机构地区：[1]国防科学技术大学航天与材料工程学院,长沙410073

出　　处：《计算机应用》2009年第B06期122-124,148,共4页journal of Computer Applications

基　　金：武器装备预研基金资助项目(51400010304KG0178)

摘　　要：为了在不了解优化函数是否存在多个局部点对早熟收敛现象进行准确判定,提出了早熟收敛判定方法;为了突破局部极值的限制以再次寻优计算,引入了基于混沌变异的粒子群算法,对出现早熟收敛现象的粒子群进行混沌变异,使得最优点不会在一点重复出现,并采用平均截止代数和截止代数分布熵组成的平面测度对算法的优化效率进行度量。仿真结果表明此算法能有效地克服早熟收敛现象,全局寻优能力较强,寻优速度较快,有效地解决了收敛性能和全局寻优能力之间的矛盾。In order to judge nicely local convergence phenomena when the objective-function has many local optimization points, a new method of criterion was proposed, and new Particle Swarm Optimization （PSO） algorithm based on chaos mutation was adopted to breach the restrictions of local optimization points. The particle swarm must be mutated for avoiding optimization point clustering together in the same region/point. Then mean truncated generation and distribution entropy were used to evaluate its performance and efficiency. Simulation results indicate that it can effectively breach local convergence, and it also have strong overall search ability while improving optimization speed and solving some conflicts of convergence and overall optimization ability.

关 键 词：早熟收敛分析 混沌变异 平均截止代数 分布熵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]