Dynamic Multiple-genetic Algorithm Based On Pareto Solution Set Prediction

In real life,many practical problems can be abstracted into a Mult-iobjective Optimization Problem.However,most optimization problems change over time.Not only have multiple optimization goals,multiple constraints,high-dimensional decision variables,but also these optimization goals,constraints,and decision variables tend to change with time.This kind of problem comes down to dynamic multi-objective optimization problem(Dynamic Multi-objective).Optimization Problem).In order to solve this kind of optimization problem,the evolutionary algorithm enters the line of view of the research scholars because of its unique property--iteration,a set of solutions that satisfy all the objective functions.However,due to the complexity of the dynamic multi-objective optimization problem,the dynamic multi-objective optimization evolutionary algorithm is still not mature,and most of them are auxiliary strategies to increase the response environment change based on the static multi-objective optimization algorithm.However,these improved strategies increase the randomness of algorithmic searches after changes in the environment.As a result,the convergence rate of the population will decrease,making it impossible for the population to respond effectively to changes in the environment.Aiming at the above problems,this paper proposes an improved dynamic multi-objective evolutionary algorithm,and carries out Matlab simulation experiments through standard test functions to verify the effectiveness of the improved algorithm.The main work of this paper is as follows:Firstly,the definition and mathematical model of static and dynamic multi-objective optimization problems are briefly described.The main environmental response strategies in current dynamic multi-objective evolutionary algorithms are summarized,and the advantages and disadvantages of these five strategies and their scope of application are analyzed.The classification of dynamic optimization problems and the performance indicators of evaluating the performance of the algorithm are summarized.The basic principles and operation flow of each module of the classic non-inferior sorting genetic algorithm(NSGA-II)are analyzed in detail.Secondly,in order to improve the convergence speed of the evolutionary population and make the optimal solution set uniform,a dynamic multi-objective optimization algorithm based on pareto solution segmentation prediction is proposed.In the improved algorithm,the prediction strategy is applied in the whole process ofpopulation evolution to improve the convergence speed of the population,and the prediction accuracy is improved by the segmentation prediction strategy.According to the difficulty level of convergence of dynamic optimization problems,the number of pareto solution sets obtained by iteration is also different.The algorithm also adopts an adaptive mechanism--according to the difficulty of the optimization problem,adaptively generating random individuals around the predicted population to increase the diversity of the population.Third,the proposed new algorithm is evaluated for convergence and population diversity performance.The new algorithm and the classical dynamic non-inferior sorting genetic algorithm(DNSGA-II)algorithm are tested on three kinds of test functions,and the optimal frontier distribution map and performance curve of the proposed new algorithm and comparison algorithm are drawn.The experimental results are analyzed in detail,which shows that the new algorithm is competitive in maintaining population diversity and convergence.