Finite horizon predictive control of chemical processes

This dissertation presents a strategy for model-predictive control of processes modelled by nonlinear differential-algebraic equations. This strategy makes use of a repeated optimization of an open-loop performance objective over a finite time horizon. The manipulated variable profile determined from the solution of the open-loop optimization is implemented until a plant measurement becomes available.Two nonlinear programming approaches for solving the open-loop optimal control problem are examined in detail, and example problems are solved using both methods. The first approach eliminates the dynamic constraints by solving them as a subproblem. The second second approach uses orthogonal collocation on finite elements to convert the differential equations to algebraic constraints that are solved directly by standard nonlinear programming software.Feedback is incorporated into the algorithm by using the measurement to update the optimization problem for the next time step. Several methods for updating the optimization problem are discussed, including resetting the model's states to measured or estimated values, estimating model parameters, or inferring output disturbances.