惯性权值对粒子群算法收敛性的影响及改进_黄翀鹏.pdfVIP

第34 卷 第12 期 计 算 机 工 程 2008 年6 月 Vol.34 No. 12 Computer Engineering June 2008 ·博士论文· 文章编号：1000—3428(2008)12—0031—03 文献标识码：A 中图分类号：TP301.6 惯性权值对粒子群算法收敛性的影响及改进 黄翀鹏，熊伟丽，徐保国 (江南大学通信与控制工程学院，无锡 214122) 摘 要：研究惯性权值对粒子群算法(PSO)收敛性的影响，在分析线性权值递减策略基础上，提出一种基于各粒子适应值的递减策略—— FDIW 。标准测试函数对比实验表明，该策略可以使粒子在搜索初期获得更好的多样性，从而使粒子具有更强的摆脱局部极值的能力，在 搜索末期可以加快粒子收敛速度以提高PSO 算法的性能。 关键词：粒子群优化算法；惯性权值；递减策略；适应值 Influnce of Inertia Weight on Astringency of Particle Swarm Algorithm and Its Improvement HUANG Chong-peng, XIONG Wei-li, XU Bao-guo (School of Communication and Control Engineering, Jiangnan University, Wuxi 214122) 【Abstract 】Inspired by studying the effect of inertia weight in convergence of Particle Swarm Optimization(PSO) and analyzing advantages and disadvantages of the existing Linear Decreasing Inertia Weight(LDIW), a new modified strategy for inertia weight is proposed based on fitness—— FDIW. Comparative experiments of benchmark functions indicate that for most continuous optimization problems, FDIW has more varieties of the swarm at the early stages so that it can escape from local minimum more easily. FDIW can speed up the convergence of particles at the later stages to improve the performance of PSO. 【Key words 】Particle Swarm Optimization(PSO) algorithm; inertia weight; decreasing strategy; fitness 文献[1]提出粒子群算法(Particle Swarm Optimization, 2 惯性权值对粒子群算法性能的影响分析 PSO) ，具有计算简洁、易于实现、需调整参数少等特点。惯 2.1 理论分析 性权值w 的选取影响该算法性能和效率，先后出现了线性递 为便于分析，笔者将PSO 简化为一维、单个粒子的运动。 减w(LDIW)[2]、模