PSO算法扰动优化策略及其收敛性研究  

Research on Perturbation Optimization Approach of PSO Algorithm and Convergence

作　　者：陶乾[1,2] 阮锦新[1] 常会友[3] 顾春琴[4] 陈强[1] 

机构地区：[1]广东第二师范学院计算机科学系,广州510310 [2]中国科学院深圳先进技术研究院,深圳518055 [3]中山大学软件学院,广州510006 [4]仲恺农业工程学院计算机系,广州510225

出　　处：《华南师范大学学报（自然科学版）》2014年第4期26-30,共5页Journal of South China Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(61370186);中国博士后基金项目(2013M542219);广东省重大科技攻关项目(2012A080104022);广东省高等学校学科与专业建设项目(2013LYM_0062)

摘　　要：为进一步提升求解精度、有效抑制早熟收敛,各类扰动(变异或跳转)优化策略常用来对粒子群优化(Particle Swarm Optimization,PSO)算法的pBest、gBest进行极值扰动,由此增强粒子在多维空间的搜索能力、提升算法性能.为分析扰动优化策略下粒子(PSO算法的搜索引擎)在多维空间的轨迹行为特性,采用级数对多维空间中粒子进行了理论分析并证明了扰动后粒子轨迹的收敛性;最后,结合项目调度问题在多维空间中对随机粒子运动轨迹进行了实证分析,验证了扰动后粒子轨迹的收敛性.In order to improve the precision of solution, to avoid the premature convergence, various perturbation （ mutation or jump） optimization approaches have been developed to realize the perturbation of the pBest or gBest, so as to enhance the search capability of the high-dimensional space and improve the performance of the PSO algorithm. To analyze the trajectory behavior of particles （ search engine of PSO algorithm） under perturbation optimization approaches in a multi-dimensional space, a theoretic analysis of particles is presented by series and the convergence of particle trajectory under perturbation optimization approaches is proved. Lastly, an empirical analysis of stochastic particles is also presented based on the project scheduling problem in a multi-dimensional space, and the experimental results are provided to prove the convergence of particle trajectory under perturbation optimization approaches.

关 键 词：微粒群优化 扰动 级数 收敛 项目调度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]