Research Of Nonlinear Predictive Functional Control Based On Block-structured Models

Model predictive control (MPC) is one of the successful applications of advanced control methods in industrial process control. As one of the newest research domain of MPC, predictive functional control (PFC) has made great development about the theory and application. However, most processes in industrial control are actually nonlinear systems. Thus research on nonlinear predictive functional control (NPFC) has become an important issue in control field. Some problems of nonlinear predictive functional control are researched in this thesis about some nonlinear processes with special structures. The main contents are as follows:(1) A survey on the development and status of the model predictive control is given.(2) According to the special structures and features of block-structured models, a N-L two steps identification method is given. Different forms of multivariable systems' block-structured models are shown and their advantages and disadvantages are discussed. The SISO NPFC algorithm is introduced based on block-structured models.(3) A nonlinear adaptive predictive functional control (NAFPC) strategy is developed based on the block-structured models. The nonlinear component of the predictive model can be determined from the steady-state data of the process, and the ARMAX model of the linear component can be estimated from the online input-output data using the recursive least square method with forgetting factor. A nonlinear dynamic optimization problem is solved to obtain the nonlinear adaptive functional predictive controller, which is suitable for time-varying process. The online computational cost is greatly reduced by the efficient solution of the first order derivatives of the output of the block-structured models. In the heat exchanger and pH process examples, it can be seen that the nonlinear adaptive functional predictive control strategy gives better performance than PID control and the nonlinear predictive control.(4) A nonlinear predictive functional control (NFPC) of multivariable processes is developed using the block-structured models. The nonlinear system optimal control problem is transformed to the combination of a linear control problem and the solution of nonlinear equations. The explicit solution of manipulated variables of the control system can be obtained if the model is known. The computational speed is improved greatly. A Hammerstein model example is given to show that the presented algorithm is feasible.At last, a summary and perspective is given.