带惯性变异与正交设计的差分进化改进算法  

Improved differential evolution algorithm using inertial mutation and orthogonal design

作　　者：刘进[1] 覃洁萍[2] 

机构地区：[1]广西师范学院,南宁530023 [2]广西中医学院,南宁530001

出　　处：《计算机工程与应用》2011年第34期34-38,60,共6页Computer Engineering and Applications

基　　金：广西自然科学资金计划资助项目(No.桂科自1013054)

摘　　要：针对目前差分进化算法收敛速度慢,易出现早熟的问题,提出一种新的带惯性变异与正交设计的差分进化改进算法。在经典差分进化算法的基础上,对每一代群体中优于前一代的个体执行惯性变异,寻求发现更优个体,在每一代群体更新结束后,对群体中最优个体在其局部邻域内使用正交设计方法进行搜索,加快发现最优解的速度。通过对7个常用的基准测试函数进行测试,结果表明提出的算法在求解精度、收敛速度、稳定性和扩展性等方面均有很好的表现,明显优于经典差分进化算法和目前常见的差分进化改进算法。In order to solve the problem of slow and premature convergence for the classical differential evolution algorithm, a novel improved differential evolution algorithm is proposed using inertial mutation and orthogonal design.On the basis of classical differential evolution algorithm,an inertial mutation is applied to the local neighborhood of better individuals of each generation to get better ones.At the end of each generation updating,orthogonal design is applied to the neighborhood of the best individual to accelerate convergence.A set of 7 standard benchmark functions is adopted to test the performance of the proposed algorithm.The results show that the algorithm has a good solution quality,convergence rate,stability and scalability compared with classical differential evolution algorithm and current common using the improved differential evolution algorithm.

关 键 词：差分进化 惯性变异 正交设计 全局优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]