改进的差分演化算法及其在函数优化中的应用

Improved Differential Evolution Algorithm and Its Application in Function Optimization

差分演化算法是一种简单、高效的函数优化方法,但也存在算法结果不稳定,容易陷入局部最优解等情况,针对上面问题,提出了混合混沌和逆向学习算子来初始化种群,保持种群的多样性,加快算法的收敛速度;提出了随机排序的选择策略,提高种群演化后期的多样性,避免算法提前陷于局部最优,通过对国际上通用22个标准测试函数进行验证,本文提出的算法在最优解的质量与稳定性优于其它被比较的算法.

Differential evolution algorithm is a simple and efficient function of optimization method,whose result is not stable but easy to fall into local optimal solutions. Aiming at the above problems： 1） It is proposed in this paper the chaos and reverse learning operator to initialize the population,to keep the diversity of population. 2） The selection strategy of random order is proposed to improve the diversity of the population in the later period,so as to avoid the local optimization in advance. By testing the standard test function,the algorithm proposed in this paper is superior to other algorithms compared with other ones.