基于动态距离阈值和重生机制的动态微粒群优化  

Dynamic particle swarm optimization based on dynamic distance threshold and regeneration mechanism

作　　者：张捷[1,2] 封俊红[1] 

机构地区：[1]广州大学松田学院信息科学与工程系,广州511370 [2]西安电子科技大学计算机学院,西安710071

出　　处：《计算机应用研究》2009年第12期4526-4529,共4页Application Research of Computers

基　　金：广州大学松田学院科研基金资助项目(GZDXSTXY06-03)

摘　　要：通过给标准微粒群引入动态距离阈值,将微粒分为最佳位置附近和最佳位置之外两类,让最佳位置附近的微粒进行集中搜索,让之外的微粒进行分散搜索,合理地平衡了两者的矛盾,使得在微粒多样性保持基本稳定的情况下,实现了收敛速度的提高。对没有任何贡献的死亡微粒进行重生,既可以有效地抑制微粒多样性的减少,又能使搜索跳出局部最优。通过仿真实验证实了这种算法是既能增加收敛性又能提高微粒的多样性。Through introducing dynamic distance threshold to standard PSO, two kinds particle between inside and outside the best location were divided into, the former executed concentrated search, the latter did scatter search, which reasonably balanced the both contradiction. This result in that under the circumstance of keeping fundamental stable condition in particle diversity, convergence speed was improved. By regenerating the dead particle without any contribution, this both might suppress effectively reductions of the particle diversity, and could cause the search to jump out local optimization. The simulating experiment have certified that the algorithm can improve not only the convergence but also the particle diversity.

关 键 词：微粒群 动态距离阈值 多样性 收敛性 死亡微粒 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]