求解复杂约束优化问题的集成多策略差分进化算法

Differential Evolution with Ensemble Multi-strategies for Solving Complex Constrained Optimization Problems

工程和科学领域中的优化问题常常具有大量的约束限制,称为约束优化问题.这类问题要求算法有能力在可行域中寻找问题的最优解.本文针对约束优化问题提出一种集成多策略的差分进化算法(Differential Evolution with Ensemble Multi-Strategies,EMSDE).首先,提出一种用于约束优化的参数自适应策略,利用归一化罚函数作为权重引导参数自适应地生成.其次,结合约束和动态罚函数法设计一种新的约束处理技术.最后,采用CEC2017约束优化基准函数来测试EMSDE和7种经典的约束优化算法.实验结果表明,相比7种经典的算法,EMSDE算法具有很强的竞争力.

Optimization problems in engineering and science often have a large number of constraints,which are called constrained optimization problems.This type of problem requires the algorithm to be able to find the optimal solution in the feasible domain.In this paper,we propose the differential evolution with ensemble multi-strategies(EMSDE)for constrained optimization problems.First,a parameter adaptation strategy for constrained optimization is proposed,using a normalized penalty function as a weight to guide the adaptive generation of parameters.Second,a newconstraint handling technique is designed by combining the-constraint and the dynamic penalty function method.Finally,the CEC2017 constrained optimization benchmark functions are used to test EMSDE and seven typical constrained optimization algorithms.The experimental results demonstrate that the EMSDE algorithm is competitive compared with the seven typical algorithms.