Research On Process Control Systems Based On Input Nonlinear MPC

Predictive control has emerged and developed as a powerful computer control technique since the 1970's, and is one of the typical process control algorithms. This thesis starts from the history of the control theory, surveys the basic ideas and current status of MPC, including the transition of linear model predictive control to input nonlinear model predictive control, and the connection between general nonlinear model predictive control and input nonlinear model predictive control. It suggests that input nonlinear model predictive control compromises between linearity and nonlinearity.Firstly, it considers the simple linear model predictive control algorithms. The level and temperature systems are simulated via MATLAB Simulink, and the level system is experimented by applying generalized predictive control.Then, it focuses on input nonlinear system, including Hammerstein model and input saturation. Two-step model predictive control is applied, which decomposes the MPC problem into a dynamic optimization problem upon linear model and a static rooting problem of nonlinear algebraic equation. The first step calculates a desired intermediate variable without considering constraint, nonlinearity and uncertainty. The second step deals with nonlinearity by solving a nonlinear algebraic equation (group) and satisfies constraint by desaturation. Based upon the existing results in the literature, two-step model predictive controller with uncertainty is considered. Polytopic description is applied for the uncertainty. Lyapunov method is applied to obtain the exponential stability conditions, with the approaches for calculating and tuning the domain of attraction given. The stability results are validated with a simulation example.