摘要： 对约束优化问题的处理通常使用惩罚函数法,使用普通惩罚函数法的困难在于参数的选取。该文提出一种基于演化算法的非参数罚函数算法,对违反约束条件动态地进行惩罚,由适应值的设定来平衡群体中可行解和不可行解的比例,使群体较好地向最优解逼近。使用实数编码的多父体单形杂交演化策略来实现新算法,通过对测试函数的检验,该算法具有稳健、高效、简洁易于实现等特点。

关键词: 约束优化, 演化算法, 惩罚函数, 单形杂交

Abstract: Penalty functions are often used in constrained optimization, however it is difficult to chose parameter properly. This paper introduces a new nonparameter penalty function for evolutionary optimization. It puts dynamic penalty to constraint violations, and balances feasible solutions and infeasible solutions through fitness. This makes group approach optimal solution easily. The new algorithm is tested using multi-parent crossover evolution strategy on 5 benchmark problems.Results show that the new method is efficient ,robust and easy to realize.

中图分类号: 

• TP301.6