Research On Multi-objective Evolutionary Optimization Strategy In Dynamic Environment

The dynamic multi-objective optimization problem is widely used in production practice and scientific research.Because the objective function,the number of objective functions,constraints and decision variables of the dynamic multi-objective optimization problem will change with the change of the environment,it is particularly important and difficult to solve the dynamic multi-objective optimization problem.When solving this kind of problems,it is required that the algorithm can not only meet the convergence and distribution,but also quickly and accurately respond to the changes of the environment.When dealing with complex practical problems,the performance requirements of the algorithm will also be improved.Therefore,this paper classifies and studies the dynamic multi-objective optimization problems,and solves three typical problems in the dynamic multi-objective optimization algorithm.The main contents are as follows:In order to quickly and accurately track the changing Pareto front and Pareto solution set in dynamic multi-objective optimization problems.In this paper,an algorithm based on the prediction strategy of reference line is proposed to solve the dynamic multi-objective optimization problem without relying on historical information.Firstly,the algorithm records the change of individual position of each reference line associated population at the initial stage of environmental change and a short period of time after individual self evolution.Furthermore,the direction of the optimal individual is predicted,and several extended individuals are evenly distributed in this direction.Then the non dominated individuals associated with each reference line are selected as the guiding individuals in the current environment.Finally,in the selected neighborhood of the guide individuals,a number of accompanying individuals are randomly generated to increase population diversity.Compared with the existing algorithm,the experimental results show that the new algorithm has a faster ability to respond to changes in the environment.In this paper,a high-dimensional dynamic multi-objective optimization algorithm based on reference points is proposed for the time-varying number of objective functions.When the number of objective functions changes,the set of reference points can provide a certain direction for population evolution.When the number of objective functions does not change,a series of reference points wit h good convergence performance and uniform distribution are generated adaptively to guide the population evolution,so as to improve the convergence speed and convergence of the algorithm.Compared with the existing algorithm,the superiority of the algori thm is proved.In this paper,a dynamic multi-objective optimization algorithm based on reference points is proposed to solve the problem of undetectable environment changes.In the framework of NSGA-II,the generation of reference points,the association of reference points and the updating of population files are added to guide the evolution of the population.At the same time,combined with the guide evolution strategy,first calculate the population of the current algebra and the popula tion center point after several generations of evolution,then judge the general evolution direction of the population,generate the guide individuals,predict the evolution of the future generations and accelerate the convergence speed of the algorithm,a nd finally obtain the optimal solution set.Experiments show that the proposed algorithm has obvious advantages in convergence and distribution.