Predictive Control Research Of Nonlinear System With Time-delay Based On Neural Network Model

Time-delay,which refers to the phenomenon that the disturbance of the system can not be reflected in a timely manner,and the current control action will be delayed for some time to reflect on the object output.The phenomenon of time-delay exists widely in metallurgical,chemical,paper and other industrial production,and its existence will affect the quality of control,leading to poor stability of the system,and even oscillation,divergence.Therefore,it is of great practical significance to study the control problem of time-delay systems.The key to solve the time-delay control problem is to predict the output of the system,the predictive control can predict the future output according to the historical information and the current input,therefore,scholars carried out extensive research on predictive control of time-delay systems and achieved certain results.For the predictive control of nonlinear system with time-delay,the first we need to establish the prediction model which can describe the system dynamic characteristics,and adapt to the external environment changes.For a class of tools with complex nonlinear approximation capacity such as neural network,which cannot be directly applied in generalized prediction control of time-delay system.To solve this problem,this paper studied how to apply the neural network identification tool to identify prediction model effectively and improve the predictive control structure.The main work of this paper is as follows:1.The advantage of predictive control in time-delay system and the method of constructing neural network model are analyzed.Firstly,the structure and optimization algorithm of RBF neural network for identification of nonlinear systems with time-delay are analyzed.And through the method improvement of random Newton to estimate weights.Through the simulation experiments verify the improved neural network optimization algorithm can effectively describe the time-delay,training model convergence.2.In order to verify the effectiveness of the above proposed neural network model with delay,the predictive self-tuning PID control for time-delay systems is studied.The neural network is used to identify the forward prediction model,and then the controller is designed.The experimental results show that the proposed method has a good compensation for the NARMAX class of time-delay systems,but also has the potential instability of oscillation with the increase of system time-delay.3.Considering the inherent problem of predictive self-tuning controller,this paper improved the prediction model based on generalized implicit predictive control structure.Using neural network to approximate the nonlinear characteristics of the system,then through local linearization method to modify the predictive model,which makes control structure convert to the general form of implicit predictive control,and provides a solution of predictive control for a class of time-delay systems.Finally,the simulation experiments are provided,through compared with the experimental results of the self-tuning PID control scheme,we can see that the improved generalized predictive control strategy is more effective.