Adaptive Optimal Decoupling Control For A Class Of Nonlinear Systems And Its Application To A Ball Mill Coal-pulverizing System

The problem of decoupling control of multivariable systems has always been a research hotspot in the control field.On the one hand,the existing decoupling control method does not consider the performance requirements of the system,and cannot meet the increasingly improved performance index.On the other hand,the existing multivariable optimal tracking control method considers the requirements of the system for performance index,but cannot achieve the decoupling control of multivariable strong coupling systems.This paper proposed an approximate optimal decoupling control method for a kind of nonlinear multivariable strong coupled system supported by the National Natural Science Foundation of China project"Nonlinear multivariable adaptive optimal decoupling control and application in steel ball milling mechanism powder system(61573090)",which can achieve decoupling control of a strongly coupled system and minimize specified performance index.The main research contents of this paper are summarized as follows:(1)An approximate optimal decoupling control method based on neural network disturbance observer is proposed for a class of continuous time linear systems with unknown external disturbances.Firstly,the recurrent neural network is used to estimate the external disturbance,then the virtual intermediate variable is introduced,and the system is decoupled by the method of feedforward and output feedback.Then,for the decoupled system,the infinite time optimal tracking control method is adopted to realize the tracking of any reference input by the system.Finally,the effectiveness of the proposed method and the superiority of the approximate optimal tracking control method based on external disturbance compensation are verified by simulation.(2)For a class of continuous-time nonlinear systems with unknown dynamics,an approximate optimal tracking control method based on adaptive dynamic programming is proposed.Firstly,the recursive neural network is used to establish the controller design model,and then the evaluation neural network is used to estimate the optimal performance index,so as to obtain the estimated value of the partial derivative of the optimal performance index,and then obtain the approximate optimal tracking controller,and finally robust terms are designed to compensate for neural network modeling errors by using tracking errors between the system output and reference inputs.This method can overcome the limitations of the existing approximate optimal tracking control method that can only track continuous differentiable input.The simulation results verify the effectiveness and superiority of the proposed method.(3)For a class of continuous-time nonlinear multivariable systems with unknown dynamics,a robust approximate optimal decoupling control method is proposed based on the optimal decoupling performance index under the framework of the approximate optimal tracking control method.The simulation experiment was carried out with the mechanism model of a ball mill coal-pulverizing system.Firstly,using the mechanism model of the steel ball milling mechanism,the input and output data are collected near the operating point,and then the controller design model is established by recurrent neural network.Finally,the robust approximate optimal decoupling controller is designed for the model.Simulation results verify the feasibility and effectiveness of the proposed method.