Research On Multiobjective Optimization Algorithm Based On Decision Variable Relationship

Multi-objective optimization problems are widely used in real life,where multiple objectives conflict with each other and it is impossible to optimize all objectives at the same time.Evolutionary algorithms have parallelism,and multi-objective evolutionary algorithms can obtain multiple optimal solutions in one run,and are often used to solve multi-objective optimization problems.However,in many complex cases,multi-objective optimization problems are affected by environmental changes,so dynamic multi-objective evolutionary algorithms have important research significance.This paper addresses both static and dynamic multi-objective optimization methods.The main research contents are as follows.Firstly,a multi-objective evolutionary algorithm based on hierarchical decision making,heuristic learning and historical environment is proposed for the problem of unbalanced search ability of evolutionary algorithms in multi-objective optimization problems from two perspectives of evolutionary history and environmental information.The evolutionary operator is guided by a hierarchical decision method based on distance-related fitness.A dynamic evolutionary heuristic learning method is introduced to solve the static optimization problem.Historical information obtained from the solution set surfaces is used to perform a comprehensive search for feasible regions.The effectiveness of the algorithm is verified using UF and WFG test sets,according to the experimental results,the algorithm exhibits strong convergence and diversity.Secondly,for dynamic multi-objective optimization problems,most dynamic multi-objective optimization algorithms use the same method to evolve decision variables,without considering the different characteristics of decision variables,and cannot handle the effective information generated during the evolution process.Therefore,a dynamic multi-objective evolutionary algorithm based on the decision variables relationship is proposed.Considering that decision variables contribute differently to convergence and diversity,the algorithm divides decision variables into two categories and uses different optimization methods for different types of decision variables.The algorithm’s efficacy is assessed through the utilization of F and DF test sets,and the experimental simulation results show that the algorithm can produce populations with better convergence and diversity compared with the other four algorithms.Finally,the rolling load distribution problem is optimized using the proposed dynamic multi-objective evolutionary algorithm.A mathematical model with equal power margin,minimum energy consumption,comprehensive slip factor and rolling force of the final stand as the objective functions is established,and the rolling speed is used as the environmental variation factor,and the stand depression rate,inlet and outlet tension of each stand are selected as the decision variables.The optimization results of five dynamic multi-objective evolutionary algorithms for the load distribution problem are investigated for the case of rolling variable speed.The results of simulation experiments show that the proposed algorithms have a better impact on the productivity of rolling load distribution than other algorithms.