Abstract: Differential evolution (DE) is one of the stochastic optimization algorithms, which has advantages of fast convergence speed and good robustness. Currently there are several crossover and mutation strategies for DE, and they exhibit different performance in optimizing different problems. This paper introduces 10 different DE update strategies and tests them on a suit of widely used benchmark problems, in order to thoroughly evaluate and compare their performance. By analyzing and comparing the convergence curves of those DE algorithms in various problem landscapes and optimization states, this paper summarizes the search behavior and performance of these update strategies in different environments. This paper not only provides guidance for engineering application that helps researchers choose a suitable strategy for a given application of DE, but also provides theoretical basis for the design of new update rules and adaptive DE.

Key words: differential evolution (DE), evolutionary variants, update strategies, evolutionary computation, global optimization

摘要： 差分演化算法（differential evolution，DE）是一种模拟生物演化过程的随机搜索方法，具有收敛速度快，鲁棒性好等优点。目前DE有多种交叉和变异策略，它们在求解各类优化问题时表现出各自不同的性能。介绍了10种差分演化算法的更新策略，并利用标准测试函数集对它们进行了全面与系统的实验比较。通过分析采用这些策略的DE算法在不同解空间及进化各阶段的收敛曲线特点，对比总结了不同版本的DE算法在各类环境下的搜索性能。该研究一方面能够为DE算法的实际应用提供技术指导，帮助学者选择合适的DE更新策略以更好地解决工程问题；另一方面能够为新型DE更新策略的开发和自适应DE算法的设计提供理论基础。

关键词: 差分演化算法（DE）, 演化模式, 更新策略, 演化计算, 全局优化