Model Predictive Control

B. Wayne Bequette

Model Predictive Control

Model predictive control is the class of advanced control techniques most widely applied in the process industries. A primary advantage to the approach is the explicit handling of constraints. In addition, the formulation for multivariable systems with time-delays is straightforward.

MPC was developed in the process industries in the 1960's and 70's, based primarily on heuristic ideas and input-output step and impulse response models. The basic principle is to solve an open-loop optimal control problem at each time step. The decision variables are a set of future manipulated variable moves and the objective function is to minimize deviations from a desired trajectory; constraints on manipulated, state and output variables are naturally handled in this formulation. Feedback is handled by providing a model update at each time step (often the "additive disturbance correction"), and performing the optimization again.

Since students from a number of disciplines will be taking this course, we first provide an overview of important issues in process control, and a review of continuous and discrete models. A historical perspective of various MPC approaches is provided, then we derive the analytical solution for the unconstrained problem. The quadratic programming solution for linear models and constraints and a quadratic objective function is presented.

Recent results on stability, robustness and state estimation will be covered. Also, extensions to nonlinear systems using different types of nonlinear models will be covered. MATLAB and SIMULINK will be used for the simulation of MPC applied to different problems. In addition to homework assignments students will complete a major research/design MPC project.

Course taught: Fall 1998, Fall 2000, Fall 2002, Fall 2004, Fall 2007

MPC Course Syllabus (.pdf)

MPC References (.pdf)

A Personal Retrospective on Nonlinear Model Predictive Control (.pdf) Bequette, B.W. "Nonlinear Model Predictive Control: A Personal Retrospective," Canadian Journal of Chemical Engineering, 85(4), 408-415 (2007)

The Mathworks home page is Mathworks