The Design Of The Robustness Of Predictive Control System Based On The Normal Matrix Approach

In the application of predictive control technology, state feedback predictive control has strong ability to inhibit disturbances and better robustness, while normal matrix approach in the design of robust control system has the advantages of simple controller, strong system robustness and so on. It's very important how to combine the advantages of both and to apply in the predictive control design.In this paper, the relationship between the parameters of predictive control system and the robust stability of the closed-loop system and the gradient of normal matrix and the robust stability of the closed-loop system are analyzed based on multi-variable state feedback predictive control system with the state-space model. Robust design based on normal matrix approach for the predictive control system is given. The main content can be summarized as follows:Ⅰ. Aanalysis and contrast about main the advantages and disadvantages in the respect of the robust stability of the closed-loop system and the simplicity of controller between the normal matrix and the H∞design approach are given.Ⅱ. Analysis of the robustness for single-value prediction control system based on the equation of state is proposed, the method to determine the parameters of the control algorithm is given, the relationship between the parameters (prediction time-length, time delay mismatch, control output attenuation coefficient, state feedback factor, etc.) for predictive control system and its robustness of closed-loop system is discussed as well.Ⅲ. The relationship between the robustness of the closed-loop system and the gradient of normal matrix is discussed based on the concept of the eigenvalue, Singular value and gradient of matrix. The condition of robust stability of multi-variable state feedback predictive control system is given. Ⅳ. The robust design of single-value predictive control system based on the normal matrix approach is given, including the prediction time-length selection, the adjustment of attenuation coefficients for controller output while model mismatch, and corresponding simulation is made.