Data-Driven Predictive Control Of Networked Nonlinear Systems

With the rapid development of industrial control and information technology,the communication network has deep integrated with the industrial control system,and then produce many networked control systems.Breaking through the closure of traditional industrial control systems,networked control systems are more flexible and efficient,and the wiring cost is low,but at the same time,the network also causes new challenges,such as data transmission delay,packet dropout,out of order and other problems.In addition,more and more complex production processes make it difficult to establish accurate models through physical and chemical mechanisms.In view of this,this paper takes complex nonlinear discrete networked systems as the research object,and datadriven networked predictive control methods are designed,which not only break the limitations of models which is difficult to obtain,but also actively compensate the communication constraints in networked control systems.The main research content of this paper is summarized as follows.1)Considering the character that the feedback channel delay can be obtained by the controller in real time,a one-way time delay model is built,and a data-driven networked predictive control method is proposed by equating complex nonlinear networked systems with the compact format data model.In this method,the control predictor obtains and compensates the time delay of the feedback channel in real time by using the timestamp technology,and then predicts the future control input by using the actual control input and the equivalent compact format dynamic linearization data model.Similarly,the forward channel delay is compensated in real time at the delay compensator.Finally,the theoretical analysis and numerical simulation validate the better compensation effect of the proposed method.2)The random time delay,packet dropout and disorder in feedback channel and forward channel are considered as round trip time delay.By analyzing the compensation mechanism of the delay compensator,two different packet selection methods are obtained,and then two data-driven networked predictive control methods are proposed.By analyzing the stability of these two methods and comparing them with the methods mentioned in the previous part,the method that has better performance is found whether the delay exists in the feedback channel or the delay exists in the forward channel.3)In order to overcome the problem that the compact format dynamic linearized data model is too simple,the optimal compensation idea of the above three data-driven networked predictive control methods is extended to the more accurate and moderate complexity of the partial format dynamic linearized data model,and then a partial format dynamic linearization data model networked predictive control method based on round trip time delay is proposed.The corresponding closed loop networked control system is analyzed.Finally,simulation results show that the proposed method can track the input reference signal more quickly and stably.