基于惯性权重非线性动态变化的微粒群算法  

Improved Particle Swarm Optimizer Based on Nonlinear Inertia Weight Dynamic Changing

作　　者：王辉[1] 钱锋[1] 

机构地区：[1]华东理工大学化学工程联合国家重点实验室,上海200237

出　　处：《计算机科学》2008年第4期146-148,共3页Computer Science

基　　金：国家杰出青年科学基金(60625302);国家973计划(2002CB3122000);国家863计划项目(20060104Z1081);上海市科委重大基础研究(05DJ14002);上海市自然科学基金(05ZR14038)基金

摘　　要：本文提出一种惯性权重非线性动态变化的微粒群算法(NDPSO),根据微粒在群体中距群体最优位置的远近,选择不同的惯性权重非线性下降指数,每个微粒根据个体状况选择不同的惯性权重。在NDPSO算法中,距最优位置较远区域惯性权重选择较小的下降指数,这样利于微粒较快地飞向群体最优位置,提高算法的全局搜索性能;当微粒飞到距最优位置较近区域时,惯性权重采用较大的下降指数,这样微粒在此区域进行细致的搜索,提高算法的收敛速度。为了研究NDPSO算法的性能,对几种典型高维非线性函数进行了测试。测试结果表明,与其它几种典型的微粒群算法相比,NDPSO明显地提高了算法的性能。This paper proposes a particle swarm optimizer （PSO） with nonlinear inertia weight dynamic changing （NDPSO） to improve the speed of convergence and fine tune the search in the multidimensional space. The decreasing exponent of inertia weight is selected by the region which particle stays in the swarm and each particle will be having different inertia weight according to the place which it stays. If the particle is far from the optimal point of the swarm the exponent of inertia weight will be smaller in order to fly to the optimal position quickly, on the other hand if particle flies near to the optimal position the exponent of inertia weight will be larger for particle searching for the optimal value carefully. The propose NDPSO is tested with a set of 4 benchmark problems and compares with linearly decreasing weight PSO （SPSO）, nonlinearly decreasing weight PSO （C＆S-PSO）. Experiment results indicate that the proposed PSO improves the search performance on the benchmark functions significantly.

关 键 词：微粒群算法 惯性权重 非线性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] O174.41[自动化与计算机技术—计算机科学与技术]