一种新的基于正交实验设计的约束优化进化算法  

A Novel Constrained Optimization Evolutionary Algorithm Based on Orthogonal Experimental Design

作　　者：蔡自兴[1] 江中央[1] 王勇[1] 罗一丹[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机学报》2010年第5期855-864,共10页Chinese Journal of Computers

基　　金：国家基础研究项目(A14200060159);国家自然科学基金(60805027;90820302);教育部博士点基金(200805330005);湖南省研究生创新基金(CX2009B039)资助~~

摘　　要：提出了一种新的基于正交实验设计的约束优化进化算法.新算法的主要特点是:在搜索机制方面,利用正交实验设计方法安排多个父代个体的交叉操作,提出了一种新的多父体正交交叉算子,新的交叉算子能够有效利用多个父代个体所携带的信息产生新的具有代表性的子代个体.此外,利用单形交叉算子对父代种群进行并行搜索,以协调算法的勘探和开采能力.在约束处理技术上,新算法引入了一个衡量个体优、劣的新比较准则.通过13个标准的测试函数验证了算法的通用性和有效性.A novel constrained optimization evolutionary algorithm based on orthogonal experimental design, referred as COEA/OED, is proposed in this paper for constrained optimization problems. The primary features of the algorithm proposed are as follows. As for search mechanism, COEA/OED utilizes orthogonal experimental design method to arrange the crossover operation of several parents and, as a result, a new multi-patent orthogonal crossover operator is proposed, which can effectively make use of the information carried by the parents and generate representative offspring. In addition, the simplex crossover is used to enrich the exploratory and exploitative abilities of the algorithm proposed. As for constraint-handing technique, a novel individual comparison criterion is introduced. COEA/OED is tested on 13 well-known benchmark functions, and the empirical evidence demonstrates that COEA/OED is generic and effective.

关 键 词：约束优化 进化算法 正交实验设计 约束处理技术 单形交叉算子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]