一种基于差异演化的协同粒子群优化算法  

A Cooperative Particle Swarm Optimization Algorithm Based on Differential Evolution

作　　者：毛恒[1] 王永初[1] 

机构地区：[1]华侨大学机电及自动化学院,福建泉州362021

出　　处：《信息与控制》2008年第2期176-180,185,共6页Information and Control

基　　金：国务院侨务办公室科研基金资助项目(03QZR13)

摘　　要：提出一种协同进化PSO算法,用于保持粒子种群的多样性并避免发生"早熟"的问题.该方法采用两个不同的分群;其中分群一的粒子采用标准PSO算法进行搜索寻优,分群二的粒子采用差异演化算法进行搜索和寻找最优解.在搜索过程中,如果标准PSO算法的适应度变化率低于一个阈值,则按照黄金分割率用分群二中的若干优势粒子取代分群一中的劣势粒子.用所提出的PSO算法和标准PSO算法对4种常用函数进行优化.结果表明,该粒子群优化算法比标准粒子群优化算法更容易找到最优解,而且优化效率和优化性能明显提高.This paper proposes a cooperative evolution particle swarm optimization（PSO） algorithm to preserve the variety of particle swarms and to avoid ＂premature＂ problem.The new algorithm uses two different particle swarms to search and find optimal value： sub-swarm one uses the standard PSO,and sub-swarm two uses the differential evolution algorithm.During the search,if the variety rate of fitness of the standard PSO is lower than a threshold,the bad particles in sub-swarm one are replaced with some good particles in sub-swarm two based on the principle of golden section.Then,the proposed PSO and the standard PSO are used to optimize four commonly-used functions.And the results show that the new PSO has higher efficiency,better performance and can find optimal value more easily than the standard PSO.

关 键 词：粒子群优化(PSO)算法 差异演化 黄金分割率 协同进化 适应度变化率 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]