基于改进种群多样度的差分进化算法  

Differential Evolution Algorithms Based on Improved Population Diversity

作　　者：陈爱华[1] 董新民[1] 董志[2] 刘棕成[1] 

机构地区：[1]空军工程大学航空航天工程学院,西安710038 [2]中国人民解放军驻沈阳飞机工业集团有限公司军事代表室,沈阳110850

出　　处：《电光与控制》2012年第7期80-84,共5页Electronics Optics & Control

基　　金：航空科学基金(2008ZC01006)

摘　　要：针对差分进化算法进化后期易出现早熟收敛而陷入局部最优的缺陷,提出了一种基于改进种群多样度的差分进化算法。对进化算法种群多样度进行了研究,经过数学推导,证明了种群多样度与算法全局寻优性能的关系,提出了一种随机变异策略,更好地保持了寻优过程中种群的多样性,增强算法的全局搜索能力。典型测试函数实验表明,改进后的差分进化算法相对于标准差分进化算法具有更好的种群多样性和抑制早熟收敛的能力。Aiming at the premature convergence problem at evolutionary anaphase of Differential Evolution （DE） algorithms, a modified DE algorithm （called DIDE ） was proposed, which used diversity theory approach to improve the population diversity. The relationship between changes of population diversity and performance of DE was proved mathematically. A random mutation method was proposed according to the relationship, which could make the algorithms keep the diversity much better and could enhance its global searching ability. The performance of DIDE was evaluated on a test bed of two functions. The numerical results were compared with that of the original differential evolution method, which indicated that this modification enables the algorithm to get a better transaction between the convergence rate and robustness.

关 键 词：差分进化算法 种群多样度 早熟收敛 优化 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP301.6[自动化与计算机技术—计算机系统结构]