Research And Application Of Objective Optimization Algorithm Based On Neurodynamics

This thesis mainly studies single objective,multi-objective,and multimodal multiobjective optimization problems.Neurodynamic methods are used in many optimization problems due to their outstanding precise optimization ability.In order to better solve the target optimization problem,this paper designs an optimization mechanism based on neural dynamics methods and integrates it with the optimization algorithm.The experimental results show that the three algorithms designed in this paper are significantly superior to other similar algorithms.The innovative work of this thesis is summarized as follows:(1)A single objective optimization algorithm based on neural dynamics(NDHPSO)is proposed for single objective optimization problems.Different from ordinary neurodynamic optimization methods,it effectively integrates heterogeneous integrated particle swarm optimization algorithm,and designs a strategy of dynamic population change.The effective combination of neurodynamic method and single objective algorithm can greatly improve the convergence performance of the algorithm.The experimental results show that the algorithm performs well on the test set.(2)A multi-objective optimization algorithm based on neural dynamics(NDMPEA)is proposed for multi-objective optimization problems.It introduces a power type time-varying factor strategy to enhance the optimization performance of neural dynamics,and designs a multi group co evolution strategy that effectively combines neural dynamics methods with multi-objective optimization algorithms.At the same time,in order to ensure the information exchange among multiple populations,a variety of information exchange strategies are proposed to promote the flow of information.The experimental results show that multi-objective optimization algorithms that integrate neural dynamics methods have better optimization performance than general multi-objective optimization algorithms.(3)A multi-modal and multi-objective fireworks algorithm based on neural dynamics(FWMMO＿ND)is proposed for multimodal and multi-objective problems.It is an improvement on the fireworks algorithm,which improves the explosion form of fireworks and the calculation method of spark generation.It uses neural dynamics methods to generate dynamic sparks,and uses a dual grid partitioning strategy to maintain population diversity in space.Finally,individuals with better performance are selected through a grid based crowding mechanism.The experimental results show that this algorithm can find more equivalent Pareto optimal solutions in the decision space.Finally,this thesis also applies the designed algorithm to the parameter estimation problem of photovoltaic models and the solution problem of nonlinear equations,and verifies through experiments that the algorithm designed in this thesis also maintains good performance in solving practical application problems.