Optimization approaches for large-scale model predictive control

This work deals with the analysis of optimization algorithms for large scale Model Predictive Control (MPC). Different solution strategies are developed, analyzed and implemented on MPC case studies.; Model Predictive control has rapidly found significant application in the process industry. Its ability to handle constraints and multi-variable processes and its intuitive way of posing the control problem in the time domain are the reasons for its popularity. MPC problems with constraints are typically solved via quadratic programming using active set methods but these methods are known to scale badly with problem size. Developments in optimization, and especially in primal-dual interior point methods, have produced a new set of algorithms that are more efficient for large problems.; This thesis report the first known application of Krylov subspace methods to the sparse linear system in MPC. The results reported in this thesis show that the interior point method with Krylov subspace technique scales linearly with the prediction horizon N. Also, the implementation of the Krylov subspace iterative technique to active set method reduces the cost of computation from O(N3) to O(N½).