基于聚类分析的差分算法协作研究

Cooperation Research on Differential Evolution Algorithm Based on Clustering Analysis

针对差分进化算法存在的收敛速度慢、易陷入局部最优等不足,本文提出一种融入聚类分析的差分进化算法。首先,利用聚类分析方法将差分算法的种群进行聚类分类,抽取代表元个体,利用新的个体来替换原种群中的较差个体,去除种群中的冗余信息将种群进行优化更新,从而使得整个种群可以快速准确地收敛于全局最优解。最后本文利用MATLAB编程模拟仿真,基于CEC2005测试函数库进行了模拟实验,结果表明加入了聚类分析替换策略的差分进化算法不仅有效地抑制了早熟收敛、提高了收敛速度,还有着简洁高效、鲁棒性强等特性。

Aiming at the shortcomings of differential evolution （DE） algorithm, such as slow convergence speed and easy to fall into local optimal solution, a differential evolution algorithm that incorporates cluster analysis is proposed in this paper. Cluster analysis is firstly used to classify populations, and typical new individuals are ex-tracted. New individuals are used to replace poor individuals in the original population and redundant information in the population is removed. The population is optimized and updated so that the entire population can quickly and accurately converge to the global optimum. Finally, this paper does simulation experiments with CEC2005 test function using MATLAB simulation. The results show that differential evolution algorithm with cluster analysis re-placing strategy not only effectively inhibits premature convergence, improves the convergence speed, but also has simple, efficient, and robust characteristics.