Study Of Multiple Models Stair-like Generalized Predictive Control Strategy

Model predictive control (MPC) is the combined outcome of optimal control theory and industrial practice along with development of computer technology. Model predictive control has been successful applied in the petroleum, chemical, power and other processes. It is also successfully extended to robots, aircrafts, network systems and other field. The study of linear model predictive control is becoming more and more mature. The controlled object often have characteristics such as uncertain, nonlinear, strong coupling, and delay time-varying in the real industrial process. With improvement of product quality and maximize pursuit of economic profit, industrial processes become more and more complex, control research of complex industrial process is particularly important.It is difficult to obtain satisfactory control performance for the complex process. As an effective way for complex control system, the stratege of multiple models has achieved many achievements in theoretical research and industrial application. The main innovations of this dissertation are as follows:(1)To deal with a system with jumping parameters, a stair-like generalized predictive control algorithm based on multiple models switching is posed. Multiple fixed models and two adaptive models are established to identify the dynamic characteristic of the system in paiallel. Multiple fixed models are used to improve the transient performance, while the re-initialized adaptive model can eliminate the steady-state error. The stability of the system is ensured by the conventional adaptive model at the same time. At each sampling time, the stair-like generalized predictive controller is designed based on the best model which is selected according to the switching index. Then, the global stability control is achieved.It has been proved that this strategy can guarantee the bounded stability for input and output system.The convergence analysis and simulation results are also been given.(2) A stair-like nonlinear generalized predictive adaptive control algorithm based on neural networks and multiple models is proposed for a class of uncertain nonlinear discrete time dynamical system. The algothm is composed of a linear stair-like generalized predictive adaptive controller,a neural network nonlinear stair-like generalized predictive adaptive controller and a switching mechanism.The linear stair-like generalized predictive adaptive controller can ensure the bounded stability of the input and output signals in the closed-loop system and the neural network nonlinear generalized predictive adaptive controller can improve the dynamic performance of the system.The purpose of using switching mechanism is to obtain stability and improve system performance. Finally,Stability and convergence analysis of the proposed adaptive control system is given. Simulation examples are also included.