一类推广的差异演化算法及其应用

A Novel Extended Differential Evolutionary Algorithm and Its Applications

针对差异演化算法的局部收敛性问题,从Minimax优化的角度,提出求解非线性多峰函数优化问题的一类推广的差异演化算法(EDEA).该算法利用均匀设计方法在可行域内产生初始群体,增加种群的差异性,具有大范围收敛的性质;并且动态收缩可行域,有效地抑制了粒子群优化算法易收敛到局部最优的缺陷;给出应用该方法到典型非线性优化和不稳定周期点的求解的具体步骤,通过仿真实验证明该算法是鲁棒的.

To investigate the problem of local convergence of differential evolution （DE） algorithm, a novel extended differential evolutionary algorithm （EDEA） is proposed for nonlinear multi-moda tions＇ optimization in the way of Minimax optimization. It generates the initial population in feasib funce field by uniform design method for larger diversity of the population, so it has the property of convergence in large-scale. And it restrains DE＇s local convergence limitation virtually through contracting feasible space dynamically. Finally details of applying the proposed method into typical nonlinear optimization and unstable periodic points are given, and experiments done show the improved teehnique＇s robustness.