Research On Predictive Control Of Multi-input And Multi-output Nonlinear Systems Based On Data-driving

With the development of network technology and computer technique,computer communication extends into the industrial field,resulting in the emergence of network control system comes into being.Because the network participates in the transmission of information,the network control system will not only be affected by the constraints of traditional control systems,but also by the network induction phenomenon.At the same time,the rapid development of network has also led to frequent network attacks,making network security an important issue in industrial production.Therefore,the networked control systems submit to constraints become research highlights.Network predictive control,as a common research method in network control systems,actively compensates for the influence of network constraints by predicting control data.Generally,network prediction control methods are designed based on accurate system models.However,the vigorous development of industry has aroused the increasing complexity of the controlled system,which makes it difficult to derive precise mathematical modeling of the system.Therefore,the implementation of data-driven control of network control system has important exploration value and profound research significance.In this paper,a corresponding data-driven network predictive control scheme is designed for multi-input and multi-output nonlinear system with multiple constraints.The specific working arrangements are as follows:1.A class of nonlinear systems with packet dropout,random delay and resulting packet disorder in dual-channel networks is studied.The model of the system is unknown and hence cannot be exactly modeled.Therefore,the influence of various constraints on the system data is analyzed,and the commonalities are summarized.Then,a data-based predictive control scheme is designed to compensate for the induced phenomena in the network.Firstly,dynamic linearization of unknown systems is implemented based on model-free adaptive predictive control.Then,the three induced phenomena are summarized with a redefined random round-trip delay.Subsequently,data-driven control is realized by using the input and output of system.At the same time,a predictive control method is proposed to actively compensate the round-trip delay.Eventually,a numerical simulation example is utilized to compare the tracking effect of the system and verify the effectiveness of the proposed scheme.2.For nonlinear systems with unknown models,false data injection attacks on dual channels are studied.The measurement noise in the output data due to environmental impact is also considered.A data-driven attack detection and prediction control scheme is designed for this purpose.Firstly,the nonlinear system is dynamically linearized based on the model-free adaptive predictive control algorithm.A prediction mechanism is devised to acquire the predicted value of the system control signal.Secondly,the system control residuals are calculated according to the obtained control and prediction signals.An attack detection mechanism that combines residual test and Lilliefors test is used to detect whether the data is attacked or not.The data which unattacked continues to be transmitted and attacked is discarded directly.Then,the discarded attacked data is compensated by the compensation strategy.Finally,numerical simulation example is used to validity the effectiveness of the proposed scheme when single channel or dual channels are attacked.3.For multi-sensor networking systems,the saturation constraints of sensor devices affected by hardware conditions are considered.At the same time,the transmission of system data in each network channel is affected by time delay.First,after the system is linearized equally,a control scheme which only using the input and output data is presented for each sensor device with output saturation.Then,in order to compensate for various time delays in multi-sensor channel and forward channel,a predictive control tactic is constructed to guarantee the tracking performance of the system.Ultimately,an example of numerical simulation is used to demonstrate the output tracking performance of this proposed scheme.