协同粒子群优化算法  

Cooperative particle swarm optimization

作　　者：刘怀亮[1] 苏瑞娟[2] 许若宁[3] 高鹰[1] 

机构地区：[1]广州大学计算机科学与教育软件学院,广州510006 [2]广东科贸职业学院信息工程系,广州510430 [3]广州大学数学与信息科学学院,广州510006

出　　处：《计算机应用》2009年第11期3068-3073,共6页journal of Computer Applications

基　　金：广东省自然科学基金资助项目(8451009101001040)

摘　　要：为解决粒子群优化算法易陷入局部最优的问题,提出了两种新方法协同处理粒子群优化算法:对比平均适应度值差的粒子,用动态Zaslavskii混沌映射公式改进粒子惯性权重与速度矢量,在复杂多变的环境中逐步摆脱局部最优值,动态寻找全局最优值;对好于或等于适应度平均值的粒子,用动态非线性函数调整粒子惯性权重与速度矢量,在保存相对有利环境的基础上逐步向全局最优处收敛。两种方法相辅相成、动态协调,使两个动态种群相互协作、协同进化。实验表明该算法在多个标准测试函数下都超越了同类著名改进算法。To solve the premature convergence problem of Particle Swarm Optimization （PSO）, two new methods were introduced to improve the performance cooperatively： When particles＇ fitness values were worse than the average, the dynamic Zaslavskii chaotic map formula was devised to modify the inertia weight and velocity, which can make particles break away from the local best and search the global best dynamically; On the contrary, when fitness values were better than or equal to the average, the introduced dynamic nonlinear functions were used to modify the inertia weight and velocity, which can make particles retain favorable conditions and converge to the global best continually. Two methods coordinate dynamically, and make two dynamic swarms cooperate to evolve. Experimental results demonstrate that the new introduced algorithm outperforms several other famous improved PSO algorithms on many well-known benchmark problems.

关 键 词：粒子群优化 速度矢量 动态Zaslavskii混沌映射公式 动态非线性函数 协同进化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]