Font Size: a A A

Research And Application Of Robust Model Predictive Control

Posted on:2020-08-26Degree:DoctorType:Dissertation
Country:ChinaCandidate:H B CaiFull Text:PDF
GTID:1488306740471934Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
The process industry is one of the main industries in the national economy.With the development of the global industrialization,the process industry has greatly become more and more important.However,it has also encountered some challenges,such as poor production quality,high production cost and weak production capacity.In order to improve production quality,enhance production capacity and adapt to market demand,some researchers should actively deal with these challenges.It is well known that the development of advanced control method is one of the effective ways to cope with these challenges in the process industry.Model predictive control(MPC)is a new class of computer control method.Its characteristic is that it can predict the future dynamic behavior of the system based on the prediction model,and it uses online rolling optimization to calculate the control law,and it takes advantage of the feedback correction strategy to correct and compensate the deviation of prediction model.Moreover,it can also deal with the system constraints.Therefore,MPC is suitable for the complex industrial process control.About several decades,MPC has been obtained many applications.On the other hand,a robust model predictive control(RMPC),which is an important branch of MPC,has been already received great attention in recent years because it can cope with the uncertainties and external disturbances.Therefore,the study of industrial process control based on RMPC is a new idea.In this paper,for having the external disturbances,system faults,actuator saturation,networked packet loss and system modes random switching in industrial process,the new RMPC methods will be proposed so as to guarantee the stability of system and improve the control performance of system,and a continuous stirred tank reactor,which is the most representative in industrial process,can be used as the simulation research.The main research works are as follows:1.For the system subject to both input disturbances and output disturbances,a double-layered RMPC is proposed.Firstly,the Kalman filter is used to estimate the disturbances in state and disturbance estimation layer.Furthermore,based on the estimated disturbances,a new steady-state target is calculated in the steady-state target calculation layer,which is the set point of the dynamic control layer.In dynamic control layer,the controller can be designed in order to reject disturbances.Moreover,for the proposed method,the characteristic of the offset free control is proved.2.For a class of discrete time-varying and time delay linear systems subject to system faults and bounded disturbances,an integrated scheme both fault detection and fault-tolerant control is proposed.On the one hand,a H? filter is treated as the residual generator.The new fault detection system is established by augmenting both the system states and H? filter states.Then,the fault detection problem is transformed into the filter design.The filter parameters are obtained by solving a set of linear matrix inequalities(LMIs).On the other hand,the mixed H2/H? RMPC is treated as a fault-tolerant controller.In order to compensate the influence of time delay,the state feedback control law with memory function is considered.Moreover,the recursive feasibility and robust stability of optimization problem is proved when disturbance occurs.3.For a class of discrete-time linear systems subject to networked packet loss,a new RMPC is proposed.The networked packet loss is described as Bernoulli probability distribution,and the information of networked packet loss is considered in designed controller,so that the controller has greatly reduced conservativeness.At each sampling time,the complex "min-max" optimization problem is transformed into a convex optimization problem with LMIs,which greatly reduces the calculation burden.In addition,in order to consider the problem more comprehensively,the cases of the packet loss rate with uncertainty and the unknown packet loss rate are also reaserched based on the idea of the free weight matrix.4.For a class of discrete time-varying and time delay linear systems subject to networked packet loss and actuator saturation,a mixed H2/H? RMPC with active compensation mechanism is studied.The networked packet loss is described as Bernoulli probability distribution.Furthermore,a sector region method is introduced in order to deal with the actuator saturation.The mixed H2/H? performance index can guarantee both the optimal performance and robustness.The H2 performance index is used as the optimization index of system and the H? performance index is treat as the constraints of system.Moreover,for the optimization problem,the proof of recursive feasibility and robust stability is also given.5.In the process industry,due to some sudden changes such as faults and disturbances,the process can randomly switch among system modes.A Markov chain can be used to describe the process of random switching,which can be treat as Markovian jump systems.However,due to collision,delay and jamming in the network,the problem of operation mode disordering can often occur in the Markovian jump systems.For this problem,a new RMPC is proposed.In order to deal with the problem of operation mode disordering,on the one hand,a bijective mapping strategy is introduced between the original Markov process and the Markov process of mode disordering so that a new Markov process can be obtained.On the other hand,in order to deal with the uncertainties between the original Markov process and the new Markov process,a non-fragile method is used.In addition,based on the idea of the free weight matrix,the cases of transition probability with uncertainty and partial information are also studied.
Keywords/Search Tags:Robust model predictive control, Disturbance rejection, Fault detection, Networked packet loss, Actuator saturation, Markovian jump systems, Linear matrix inequalities
PDF Full Text Request
Related items