一种引入密度因子的改进粒子群优化算法  

Improved particle swarm optimization algorithm with density factor

作　　者：孙锋利[1,2] 何明一[1] 高全华[2] 

机构地区：[1]西北工业大学电子信息学院,西安710077 [2]长安大学理学院,西安710064

出　　处：《计算机应用研究》2011年第8期2828-2830,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60736007);长安大学中央高校专项科研基金资助项目(CHD2010JC133)

摘　　要：在传统的线性递减惯性权重(LDW)粒子群算法的基础上,提出一种新的引入粒子密度因子的粒子群算法。该算法根据粒子平均适应度值和社会最优适应度值,采用径向基函数形式来度量粒子群在最优值附近的聚集程度。在进化过程中,当密度因子大于一定值时,在LDW惯性权重因子中加入扰动项,使粒子群重新散开,从而跳出局部极值,避免算法出现早熟现象。基于Benchmark函数库的仿真实验表明,该算法一定程度上避免了算法过早收敛,尤其是在高维和多极值情况下性能明显优于传统PSO算法。Based on the conventional linear decreased weight particle swarm optimization algorithm,proposed a novel improved PSO with a so-called density factors involved.Defined the density of one generation in the form of radial basis function with the average fitness value and the best one of the whole swarm,which was used as a metric of the assembling degree around the best fitness value.In the process of evolution,introduced a disturbing term in the LDW factor formula when the density factor was larger than a particular constant in order to scatter the particle swarm and leap the local minimum.The simulation tests show that the new PSO algorithm avoids the premature phenomenon in a sense especially for high dimensional and multiple extremum scenarios.

关 键 词：粒子群优化 密度因子 线性递减惯性权重 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]