Robust Predictive Control Theory And Method For Uncertain Constrained Systems Under The Networked Environment

Traditional process industries,such as petroleum,chemical industry,metallurgy and so on,are the important industry areas which should be reformed and upgraded on the new road of industrialization,and realization of the whole production process of networked control is one of the main directions of transformation and upgrading.Predictive control is the most widely used and effective advanced control algorithm in the process industry,and it has become a recognized standard for dealing with multivariable problems with complex constraints.However,the conventionally designed predictive control,which is applied in the process industry,can not deal with the problem of data loss and quantization error which may occur in the data transmission process.If the industrial process which adopts the conventionally designed predictive control is directly placed in the networked control mode,the control quality,even the production safety will be under great threat.This thesis takes the networked control systems(NCSs)and predictive control as the research object and basic scheme,respectively.The approaches of robust predictive control are systematically investigated under the networked environment with data loss and quantization error.The main contents of this paper are as follows:1.A robust constrained predictive control for T-S fuzzy uncertain quantized system with data loss is investigated.Bernoulli process and sector bound approach are employed to settle data loss and quantization error,respectively.In the process of designing controller,non-parallel distributed compensation control is used.For the purpose of getting the new stability conditions with less conservatism,this thesis combine slack matrix with extended non-quadratic Lyapunov function.In the infinite horizon,min-max optimization problem can be transformed into minimizing the performance objective function by linear matrix inequality(LMI).The effectiveness of proposed method is proved by the example of continuous stirred tank reactor(CSTR)in industry.2.The off-line algorithm of robust constrained predictive control for uncertain system under the networked environment with data loss is studied.Bernoulli process is employed to model the data loss of NCSs.Due to the urgent demands of computational burden,one off-line algorithm is proposed based on invariant ellipsoid.By a group of states,a nested set of ellipsoid with constant control laws can be created.When solving the optimization problem on-line,it is needed to find the smallest ellipsoid which contain the current state and apply the corresponding control law.The effectiveness of proposed method is proved and the asymptotic stability is guaranteed by the example of CSTR in industry.3.The output feedback predictive control for uncertain linear time-varying system under the networked environment with quantization is researched.Firstly,logarithm quantizer is introduced to characterize the process of stochastic data loss,thus the mathematical model of linear time-varying NCSs with data loss is established and the state observer is also designd off-line.Further,based on estimated state,the output feedback predictive control algorithm which parameterizes the infinite horizon control move into one free control move and a linear feedback law is provided.In addition,an essential formula for on-line refreshing ellipsoidal bound of estimation error is introduced,which meets the requirement of guaranteeing the stability of output feedback predictive control with constrainrt.Finally,the effectiveness of proposed method is proved by the example of coupled spring-mass system in industry.