解非线性规划问题的非参数罚函数多目标正交遗传算法

Non-parameter Penalty Function Multi-Objective Orthogonal Genetic Algorithm for Nonlinear Programming Problem

对非线性规划问题的处理通常采用罚函数法,使用罚函数罚的困难在于参数的选取。本文提出了一种解非线性规划问题非参数罚函数多目标正交遗传算法,对违反约束的个体进行动态的惩罚以保持群体中不可行解的一定比例,从而不但有效增加种群的多样性,而且避免了传统的过度惩罚缺陷,使群体更好地向最优解逼近。数据实验表明该算法对带约束的非线性规划问题求解是非常有效的。

Penalty functions are often used in constrained optimization, but it is difficult to choose parameter property. In this paper, a new non-parameter penalty function multi-objective orthogonal genetic algorithm is presented to solve the nonlinear programming problem. It puts penalty to constraint violations in order to keep a ratio of infeasible solutions in population. As a result, it can not only increase the diversity of population but also avoid the defects of over-penalization. This makes the group approach optimal solution easy. The numerical experiment shows that this algorithm is effective in dealing with the nonlinear programming problem.