混合差分进化算法  

Hybrid differential evolution algorithm

作　　者：李丽蓉[1] 高卫峰[2] 

机构地区：[1]山西警官高等专科学校计算机科学与技术系,山西太原030021 [2]西安电子科技大学应用数学系,陕西西安710071

出　　处：《计算机工程与设计》2012年第6期2446-2450,共5页Computer Engineering and Design

摘　　要：为了克服差分进化算法容易出现早熟和收敛速度慢的问题,提出了一种混合差分进化算法。该算法在趋药性差分进化算法(CDE)的基础上,通过对较优个体进行变异操作,维护了种群多样性、避免早熟;通过将较差的个体与较优个体进行杂交,提高了开采能力、加快了收敛速度。基于这两种策略,算法的开采能力与探索能力达到了平衡。用该算法解决标准函数优化问题,并将仿真结果与其他算法进行比较,数值结果表明该文算法具有较快的收敛速度和很强的跳出局部最优的能力。To overcome the problems of premature convergence frequently appeared in differential evolution（DE） and its poor convergence,a hybrid differential evolution is proposed.Based on the chemotactic differential evolution algorithm,a mutation operation is added to the better individuals to keep the diversity and avoid the premature convergence,and a crossover operation is added to the worse individuals to increase the exploitation and enhance convergence rate.Due to the two strategies,the exploration and exploitation of the algorithm can be well balanced.Finally,a suite of 12 benchmark functions is used to verify the proposed algorithm and the result of simulation,which is compared to other well-known algorithms,indicates the proposed approach is shown to have better convergence rate and great capability of preventing premature convergence.

关 键 词：差分进化算法 趋药性差分进化算法 杂交操作 变异操作 早熟 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]