Research On Dynamic Multi-Objective Optimization Algorithm Based On Prediction

In real life,there were many optimization problems,which usually include multiple conflicting goals that must be optimized at the same time.Due to the conflicts between multiple goals,there was no single method to optimize all goals.Multi-objective optimization algorithm can obtain multiple optimal solutions in a single run,which made multi-objective optimization algorithm the best solution.However,in practice,many optimization problems were dynamic changes.In the face of complex dynamic changes,the static multi-objective optimization method has obvious limitations.Therefore,it was of great significance to study the dynamic multi-objective optimization method.In the dynamic optimization technology,the prediction method was greatly affected by the accuracy of the prediction model.In order to further improve the prediction accuracy and the response speed to environmental changes,the dynamic multi-objective optimization algorithm was studied.The main research works were as follows:Aiming at the problem that the number of representative elite individuals generated based on the central Point prediction method was small and it was difficult to obtain the optimal solution set,a dynamic multi-region Keen Point Strategy(MRKPS)was proposed based on the multi-region knee Point Strategy.Special points and multi-region knee point individuals were used as guide population to improve the response speed of predicted population to environmental change.The combination of fast nondominant sorting and forward center feedback method can update the progeny population and improve the prediction accuracy of progeny individuals.In addition,the diversity maintenance strategy was adopted to maintain the diversity of the initial population.The algorithm was simulated using F,FDA and d MOP standard functions,and the experimental results showed that the MRKPS algorithm as a whole showed good convergence and distribution.In order to improve the accuracy of Prediction of progeny population,a dynamic multi-objective optimization algorithm(Representative Individual Prediction Strategy,RIPS)was proposed.On the basis of the special point and multi-region knee point strategy,the inverse model method was adopted to select ideal individuals,and the forward feedback method was used to predict the progeny population.In addition,individuals with uncorrelated weight vectors obtained progeny by mutation recombination,and the progeny population obtained by combination of inverse model and recombination variation.Experimental results showed that the solution set predicted by RIPS algorithm has good convergence and distribution.Finally,the algorithm was applied to the optimization of the cold rolling schedule.In the process of the optimization,the speed was taken as the environmental change factor,the objective function of preventing slipping and the objective function of equal relative load were adopted.The experiment showed that the dynamic multi-objective optimization algorithm was superior to the original procedure in the process of the optimization of the cold rolling schedule.