Abstract: An improved differential evolution algorithm is proposed to solve constrained optimization problems, which does not introduce penalty parameters to deal with constraints. In the process of evolution, the individuals generation based on good-point-set method is introduced into the evolutionary algorithm initial step. In order to improve global convergence and convergence speed of the proposed algorithm, DE/best/1 mutation scheme and DE/rand/1 mutation scheme are used to the feasible solution and the infeasible solution respectively. Several class Benchmark problems are tested, the results show that the proposed algorithm is an effective way for constrained optimization problems.

Key words: constrained optimization problems, Differential Evolution（DE） algorithm, good point set, mutation strategy

摘要： 提出一种改进的差分进化算法用于求解约束优化问题。该算法在处理约束时不引入惩罚因子，使约束处理问题简单化。利用佳点集方法初始化个体以维持种群的多样性。结合差分进化算法两种不同变异策略的特点，对可行个体与不可行个体分别采用DE/best/1变异策略和DE/rand/1策略，以提高算法的全局收敛性能和收敛速率。用几个标准的Benchmark问题进行了测试，实验结果表明该算法是一种求解约束优化问题的有效方法。

关键词: 约束优化问题, 差分进化算法, 佳点集, 变异策略