基于均匀设计的差分进化算法的参数设定

差分进化算法参数的设定多采用经验选取方式,其缺点是试验运行量大以及难以得到最优参数组合,从而在很大程度上影响了算法的寻优能力。将均匀设计的试验方法引入差分进化算法的参数设定中,通过对单峰函数、多峰函数和病态函数等3种不同类型的标准测试函数进行均匀设计试验,找出适合不同类型标准测试函数的最优参数组合,从而达到对差分进化算法的参数进行设定的目的。结果显示,将经过均匀设计试验得到的两组最优的参数组合用于差分进化算法时,所获得的平均全局最优解为4.3215,平均标准差为3.650。可见,利用均匀试验设计方法对基本差分进化算法的参数进行设定是可行且有效的,同时具有较好的稳定性。

Parameter Establishment of Differential Evolution Algorithm Based on Uniform Design

The parameter establishment of differential evolution algorithm is generally determined by the experience selection method,whose shortcomings include the massive operational parameters,the difficulty in obtaining the best parameter combination,and obstacle in improving the optimization ability of the algorithm to a great extent.The article introduced the uniform design method to differential evolution algorithm in parameter establishment.The optimal parameters combination which can be applied to different types of standard test functions is discovered by means of the uniform design test for three different types of standard test function:the unimodal function,multi-peak function,and morbid function.Finally the differential evolutionary algorithm for parameter establishment can be specified.The result is as follows.When the two groups of optimal parameter combination obtained by the uniform design experiment are applied to the differential evolution,the average global optimal solution is 4.3215 and the average standard deviation is 3.650.It follows that the method of uniform experimental design is feasible and effective to set the parameters of the differen-tial evolution algorithm and the method offers good stability.