Dynamic Multi‐Objective Optimisation of Environmental Change Degree and Similarity Based on Nonparametric Tests

Dynamic multi‐objective evolutionary algorithms (DMOEAs) are widely used to solve dynamic multi‐objective optimisation problems. However, most existing DMOEAs rarely consider the severity of environmental changes and the varying impact of these changes on different decision variables, which restricts their applicability in complex dynamic scenarios. Therefore, we propose a dynamic multi‐objective optimisation of environmental change degree and similarity based on nonparametric tests, called ECDSM, which consists of two main components. First, Cliff's delta is employed to quantify the degree to which environmental changes affect each decision variable, enabling the algorithm to employ variable‐specific strategies and thus achieve more fine‐grained adaptation. Second, a prediction strategy based on a similarity mechanism, implemented through the Wilcoxon rank‐sum test, reintegrates historically optimal solutions from the most similar past environment, thereby accelerating convergence. To evaluate the effectiveness of ECDSM, we conducted comparative experiments on 14 dynamic test problems and one application problem. The results demonstrate that ECDSM exhibits high overall performance.