Research On Dynamic Multi Objective Optimization Algorithm Based On Kalman Filter

As a stochastic optimization problem with multiple objectives,dynamic multi-objective optimization has been widely used in thrust distribution of ships,engineering optimization,investment management and many other optimization fields in recent years.The difficulty of this study is how to start the environmental response strategy in time after the detection of environmental changes,design a more comprehensive and efficient environmental response strategy to respond to environmental changes in time and quickly,further improve the optimization effect and solve the optimization problem.In this thesis,through the summary and analysis of the relevant literature at home and abroad in recent years,and aiming at the shortcomings of the existing algorithm environment response strategy and other problems,we carry out the research of dynamic multi-objective optimization of the historical center point prediction,improve the convergence and diversity of the algorithm.Firstly,through the analysis of a large number of relevant literatures at home and abroad,the primary domain of research and research progress of dynamic multi-objective optimization algorithm in recent years are summarized,the application of dynamic multi-objective optimization in the field of engineering optimization is clarified,the existing problems are pointed out,and prospected future direction.The research work is summarized in the research progress of dynamic multi-objective optimization algorithm.Secondly,the key to the design of the algorithm is how to efficiently use the existing historical information to respond to the changes when the environment changes.To solve some problems like this,a dynamic multi-objective optimization algorithm(KFPS)based on modified Kalman filter prediction is designed.The Kalman filter model is established to predict the changed population center point,and then the center point of the approximate ideal Pareto optimal solution set is used to modify the prediction result,so as to avoid the prediction model guiding evolution falling into the preference field and affecting the optimization effect.Finally,reinitialize the population to make a rapid response to environmental changes,so as to improve the optimization ability of the algorithm.The results prove the effectiveness of the proposed algorithm.Finally,in view of the insufficient performance of Kalman filter prediction model in some nonlinear optimization problems,an ensemble Kalman filter(En KFPS)prediction method suitable for nonlinear and high real-time optimization problems is proposed.En KFPS uses the historical information to build a mathematical model,forecasts the population center through the ensemble Kalman filter model,and reinitializes the population with the predicted information when the environment changes.The algorithm overcomes the shortcoming that Kalman filter only deals with linear problems and saves more computing resources.The experimental results verify the effectiveness of the algorithm.This thesis focuses on the summary and refinement of various research contents at home and abroad,and on the basis of existing research,puts forward the Kalman filter prediction method suitable for solving linear problems and the ensemble Kalman filter prediction method which makes up for its shortcomings and performs better on nonlinear problems.Through the experimental design,compared with other existing classical algorithms in some standard test functions,the feasibility and effectiveness of the proposed method are proved.