基于自动重组机制的多目标粒子群优化算法研究  

Multi-object Particle Swarm Optimization Based on Automatic Reorganization Mechanism

作　　者：袁靖[1] 袁丹 彭道刚[1] 张浩[1] 

机构地区：[1]上海电力学院自动化工程学院,上海200090 [2]国网浙江宁波市鄞州区供电公司,浙江宁波315100

出　　处：《广东电力》2014年第10期32-37,共6页Guangdong Electric Power

基　　金：上海市"科技创新行动计划"高新技术领域重点科研项目(14511101200)

摘　　要：粒子群优化一直存在过早收敛的问题。为此,提出一种自动重组机制,当检测到粒子群过早收敛时,会自动触发各粒子的位置重组,从而解放陷入局部最优的粒子群,使其能回到搜索全局最优的正轨。另外,粒子群重组后的新搜索空间的建立,会根据粒子位置与全局最优位置之间偏差的不确定性程度,按照一定比例在各维度上取得新的搜索范围。实验结果表明,在运用于二维Rastrigin函数的基准测试中,该优化算法取得了近似全局最优解。Aiming at premature convergence of particle swarm optimization,a kind of automatic reorganization mechanism was proposed.When detecting premature convergence of particle swarm,it was able to automatically trigger position reor-ganization and consequently liberate particle swarm in partial optimization and make it back to the right path of searching for global optimum.In addition,establishment of the new searching space after particle swarm reorganization would be on the basis of uncertainty degree of deviation between position of the particle and global optimum position.Meanwhile,ac-cording to a certain proportion,a new searching range was able to be acquired.Experimental results indicated that in bench-mark testing applying two-dimension Rastrigin function,this optimization algorithm was able to acquire approximate global optimum solution.

关 键 词：进化算法 粒子群优化 多目标优化 过早收敛 自动重组机制 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]