Abstract - Rawlings - 9708497 The goal of this research project is to further develop the theory of model predictive control (MPC) and further its application as a means to improve real-time operation of chemical processes. The research focuses on four areas: (1) Developing new theory for the linear plant case. This new theory is needed to support the next generation of MPC technology for plants described by linear models, which are the MPC models for most current industrial implementations. (2) Handling large-scale applications through better optimization methods. The industrial trend in several of the process industries is towards larger scale MPC applications. The industrially implemented optimization approaches are reaching their limits on these large problems. This research develops the optimization theory and provides new optimization algorithms to enable practitioners to tackle these large-scale applications. (3) Developing improved methods for controlling nonlinear plants. Although the straightforward extension of linear MPC to nonlinear plants works from a control theory perspective, the optimization difficulties are intractable from the implementation perspective. Building on the research of several groups during the last several years, the PI plans to implement several new approaches for the nonlinear plant, which do not require the solution to non-convex optimization problems. (4) Application of MPC in collaboration with industrial partners. The applications serve both as a vehicle to transfer technology and as a test-bed for evaluating the strengths and weaknesses of these new theoretical results. The PI's approaches are made freely available in high-level computing languages to aid practitioners in testing and evaluation via computer simulation.
