分步交叉差分进化算法及仿真应用  

Two Section Crossover Differential Evolution Algorithm and Its Application

作　　者：刘荣辉[1,2] 王斌斌[1] 郑建国[2] 

机构地区：[1]河南城建学院计算机科学与工程系,平顶山467036 [2]东华大学管理学院,上海200051

出　　处：《系统仿真学报》2013年第7期1549-1553,共5页Journal of System Simulation

基　　金：国家自然科学基金(70971020)

摘　　要：为了使得不同的优化问题能进行最佳寻优,基本差分进化算法需要有针对性的对控制参数进行设置,为此提出了分步交叉差分进化算法。该算法利用柯西分布随机数设计用于生成尺度因子和交叉概率,并把进化的过程分为两个步骤,不同步骤分别利用生成的交叉概率改善算法性能。为了加快收敛速度,避免算法寻优的盲目性,采用选中拥有优势解的随机向量作为基向量。通过挑选有代表性的Benchmark函数及典型运输问题进行仿真试验,并把其他经典算法参照,证明其拥有良好的收敛性的同时并显著提高寻优质量。Traditional differential evolution algorithm in solving a specific problem crucially depends on the setting of control parameters, new parameters adaptation updating manner was proposed. Mutation strategies and crossover strategies were respectively adapted by two operators in this algorithm based on Cauchy random number, and the evolutionary process was divided into two steps with different crossover strategies to improve the performance of the algorithm. At the same time, to enhance the convergent rate, a local search strategy was introduced to guide searching direction. Benchmark problems and transportation problem were used for algorithm verification and simulation results indicate that this algorithm is better than several other typical algorithms both in convergence rate and quality of optimization.

关 键 词：差分进化 柯西随机数 分步交叉 参数控制 运输问题 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]