Dynamic multiobjective optimization problems exist in daily life and industrial practice. The objectives of dynamic multiobjective optimization problems conflict with each other. In most dynamic multiobjective optimization algorithms, the decision variables are optimized in the same way, without considering the different characteristics of the decision variables. To better track Pareto-optimal front and Pareto-optimal set at different times, a dynamic multiobjective optimization algorithm based on decision variable relationship (DVR) is proposed. Firstly, the decision variables are divided into two categories based on the detection mechanism of the contribution of decision variables to diversity and convergence. Secondly, different optimization methods are used for different types of decision variables. And a diversity maintenance mechanism is proposed. Finally, the individuals generated by these two parts and the perturbed individuals are combined. The combination individuals are nondominated sorted to form a population in the new environment. To verify the performance of the proposed algorithm, DVR is compared with five state-of-the-art dynamic multiobjective optimization evolutionary algorithms on 15 benchmark instances. The experimental results show that the DVR algorithm obtains 24 inverse generation distance optimal values in 45 groups of test data.
原文链接
分享
引用
