学习历史经验的粒子群优化算法  

Particle Swarm Optimization Based on Learning the Historical Experience

作　　者：靳雁霞[1] 薛敬[1] 张建华[1,2] 韩燮[1] 

机构地区：[1]中北大学计算机与控制工程学院,太原030051 [2]太原科技大学计算机科学与技术学院,太原030024

出　　处：《小型微型计算机系统》2016年第3期572-575,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61379080)资助;山西省自然基金项目(20130110177)资助

摘　　要：为解决基本粒子群算法收敛速度过慢、易于陷入局部最优等不足,提出一种对全局最优位置预测的粒子群算法.算法在基本粒子群的每一代搜索更新基础上,全局最优位置先通过反学习预测策略得到下一步进化的预测值和本次进化的经验,然后通过进化经验预测策略进一步预测全局最优位置,更精准更快速地寻到全局最优位置.最后,通过对几个基本测试函数寻优结果的对比说明,ALPSO算法可以在较短时间收敛,提高了收敛精度,且不易陷入局部极值,在工程应用上具有一定的实用价值.In order to deal with the problems of the possibility to fall into the local extreme and the slowglobal convergence rate for basic PSO algorithm,a Particle Swarm Optimization based on anti-learning predict policy（ ALPSO） is proposed. After searching in each generation based on particle swarm optimization,the anti-learning prediction strategy is used to predict the next step of the evolution for the global optimum position and meanwhile get this experience of the evolution,then through the evolutionary experience prediction strategy to make further prediction. Thus,particle swarm finds a faster and more accurate global optimal position in this algorithm. Finally,by a fewbasic test functions in comparing the results of the optimization description,ALPSO algorithm can converge in a shorter time,improve the convergence precision. It is difficult to fall into local minima. It has some practical value in engineering applications.

关 键 词：粒子群优化算法 全局最优位置 反学习 进化经验 预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]