Research On The Control Method Of Nonlinear System Based On Unmodeled Dynamic Compensation

In the industrial control,all the controlled objects are nonlinear,such as ball mill system,water irrigation level system and wind tunnel system and so on.It is difficult to obtain better control performance by applying traditional linear controller to these nonlinear systems.Therefore,it is very important to study the control problem of nonlinear systems.The main difficulty encountered in the study of nonlinear systems is that the lack of appropriate methods to eliminate the effects of nonlinear terms and uncertain terms on the system.Fortunately,with the development of neural network theory,the nonlinear adaptive control method based on neural network has made rapid progress.The core idea of this method is to estimate and eliminate the influence of nonlinear and uncertain terms on the system by using neural network.However,this method requires a large number of system state data,and the traditional back propagation neural network has the shortcomings of slow convergence rate and easily falling into local minimum.In order to overcome the above shortcomings,the specific research work in this paper is described as follows:(1)This thesis studies the problem of adaptive control for a class of SISO nonlinear systems.A new control framework is designed for the controlled object.The control framework includes a linear controller and a proportional-integral-derivative neural network compensation controller.In this control framework,the linear controller makes the system output gradually converge to the vicinity of the given signal.The proportionalintegral-derivative neural network compensation controller estimates and eliminates the influence of unmodeled dynamics on the system,so that the system outputs can better track the given signal.The control framework has the advantages of needing less online measurement data and good robustness.A numerical simulation and a simulation experiment of a tank level control system verify that the method has better control effect.(2)Based on the above control framework of SISO nonlinear systems,the adaptive control problem of a class of MIMO strongly coupled nonlinear systems is studied.Different from SISO nonlinear systems,MIMO nonlinear systems need to consider the effect of coupling on the system.Therefore,in this paper,both the linear controller and unmodeled dynamic compensation controller consider the decoupling problem.A numerical simulation verifies the effectiveness of the proposed method.(3)The unmodeled dynamic compensation controller designed in this paper is a proportional-integral-derivative neural network.Compared with the traditional BP neural network,the hidden layer of the proportional-integral-derivative neural network incorporates proportional neurons,integral neurons and differential neurons.In this case,the proportional-integral-derivative neural network inherits the characteristics of "memory" and "prediction" of PID,so that the proportional-integral-derivative neural network is more intelligent than the traditional BP neural network.And because of the effect of the integral neurons,the proportional-integral-differential neural network is not easy to fall into the local minimum.Furthermore,the connection method of the proportional-integral-differential neural network is designed according to the basic principle of the PID control law,the rapid convergence of the system thus is guaranteed.