Over recent decades, several efficient constraint-handling methods have been proposed in the area of evolutionary computation, and the \(\varepsilon \) constraint method is considered as a state-of-the-art method for both single and multiobjective optimization. Still, very few attempts have been made to improve this method when applied to the differential evolution algorithm. This study proposes several novel constraint-handling methods following similar ideas, where the \(\varepsilon \) level is defined based on the current violation in the population, individual \(\varepsilon \) levels are maintained for every constraint, and a combination of fitness and constraint violation is used for determining infeasible solutions. The proposed approaches demonstrate superior performance compared to other approaches in terms of the feasibility rate in high-dimensional search spaces, as well as convergence to global optima. The experiments are performed using the CEC’2017 constrained suite benchmark functions and a set of Economic Load Dispatch problems.

在最近的几十年中，在进化计算领域已经提出了几种有效的约束处理方法，并且\（\ varepsilon \）约束方法被认为是用于单目标和多目标优化的最新方法。当应用于差分演化算法时，仍很少尝试改进此方法。这项研究遵循相似的思想提出了几种新颖的约束处理方法，其中\（\ varepsilon \）的级别是基于总体中当前的违规行为来定义的，个体\（\ varepsilon \）为每个约束条件保持级别，并结合适用性和约束条件违反条件来确定不可行的解决方案。就高维搜索空间中的可行性率以及向全局最优收敛而言，所提出的方法与其他方法相比表现出优越的性能。使用CEC'2017约束套件基准功能和一组经济负荷分配问题进行了实验。