基于内部罚函数的进化算法求解约束优化问题

为解决现有约束处理方法可行解的适应度函数不包含约束条件的问题,提出了一种内部罚函数候选解筛选规则.该候选解筛选规则分别对可行解和不可行解采用内部罚函数和约束违反度进行筛选,从而达到平衡最小化目标函数和满足约束条件的目的.以进化策略算法为基础,给出了基于内部罚函数候选解筛选规则的进化算法的一个实现.进一步地,从理论和实验角度分别验证了内部罚函数候选解筛选规则的有效性:以(1+1)进化算法为例,从进化成功率方面验证了内部罚函数候选解筛选规则的理论有效性;通过13个测试问题的数值实验,从进化成功率、候选解后代是可行解的比例、进化步长和收敛速度方面验证了内部罚函数候选解筛选规则的实验有效性.

Interior Penalty Rule Based Evolutionary Algorithm for Constrained Optimization

In order to combine constraints into the evaluation of feasible solutions, a set of interior penalty rules is proposed to improve the efficiency of evolutionary algorithms in solving the constrained optimization problems. In these rules, interior penalty functions are used to evaluate feasible solutions and constraint violations are used to evaluate infeasible solutions. In addition, an interior penalty rule based evolution strategy algorithm is derived to solve constrained optimization problems. The theory validity of these rules is analyzed based on the successful rate of an (1+1) evolutionary algorithm. The proposed approach is tested with 13 benchmark problems. The results indicate that the presented approach is competitive with two existing state-of-the-art techniques.