摘要
惯性权重对粒子群优化算法的性能有重要影响,但目前的改进方法大多将惯性权重按照设定规律变化,未能与新产生个体的质量建立有效联系。为此,提出了一种基于学习机制和局部搜索机制的混合策略来调整惯性权重,以进一步提高算法的性能。基于学习机制的策略通过统计一定代数内新个体的优劣情况,建立了用以生成惯性权重的分布模型;基于局部搜索机制的策略以惯性权重为变量,利用低维下的局部搜索来调整其取值。在5个标准函数和1个输电网扩展规划问题函数上的测试结果表明,该算法在其中5个函数上取得了优于对比算法的测试结果。
Inertia weight was a very important parameter for the performance of particle swarm optimizer.Many algorithms ad-justed inertia weight in a given way.However,these methods had not well established the relationship to the quality of new solu-tions.A hybrid strategy based on learning process and local search used to adjust the inertia weight was propsed to improve the algorithm’s performance.The learning process counted the quality of new solutions during a predefined generation interval to ob-tain the distribution model of inertia weight,and then the model was used to generate new values of inertia weight.The local search process set the inertia weight as the variable and tuned it by executing the searching in the low dimensional space.The ex-perimental results on five standard functions and one transmission network expansion planning problem function showed that the proposed algorithm on five of six functions obtained better results than others algorithms.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第7期2552-2556,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61142010)
河南省科技计划基金项目(122300410174)
关键词
粒子群优化
惯性权重
自适应
分布模型
局部搜索
particle swarm optimizer
inertia weight
self-adaptive
distribution model
local search
作者简介
于海鹏（1979-），男，河南平顶山人，硕士，讲师，研究方向为计算机应用与图像处理；
翟红生（1975-），男，河南巩义人，硕士，讲师，研究方向为计算机应用技术。E-mail：yuhp1979@126．com
