Abstract: Differential evolution (DE) is well known as a simple and efficient scheme for global optimization over continuous spaces. In this paper different variants of DE are investigated with five test functions, and the best one is given. Next, the appropriate population size is researched with the same test functions above, and the general and suitable value range of population size for high dimensional functions with more than ten variables is presented.

Key words: evolutionary algorithm, differential evolution, evolutionary variants, global optimization

摘要： 差异演化算法作为一种较新的演化算法，具有全局搜索能力强、控制参数少、算法简单、容易实现等特点，但在具体应用中仍面临着演化模式如何选择及最佳种群大小如何取值等问题。本文利用典型的五类多变量函数对差异演化算法常用演化模式进行了比较测试研究，给出了演化模式选择建议；在同时考虑种群大小对差异演化算法优化效率及优化结果两方面影响的情况下，利用测试函数对最佳种群大小的取值进行了研究，给出了差异演化算法用于多变量函数优化时种群大小的适用取值范围。

关键词: 演化算法, 差异演化, 演化模式, 全局优化