| One of the advanced control methodologies which has made a significant impact on industrial control engineering is Model Predictive Control (MPC). This technique is a design philosophy—not just a specific controller design.; Despite its advantages, MPC is deficient because it does not offer a clear numeric relation between design parameters and control performance. In that respect, it would be very useful to develop simple predictive control algorithms which guarantee closed-loop stability and offer clear meanings of parameters.; This thesis presents a modification to the Generalized Predictive Control algorithm which guarantees closed-loop stability. The new controller Principal Component GPC (PC-GPC) is designed using a terminal equality constraint with a move suppression factor λ = 0. The available degrees of freedom are presented to the designer as a new parameter—called number of principal components (NPC). These components can be left, or removed, from the solution to obtain different performances.; Several methods to select the new parameter NPC are presented. These can be based on percentage of minimized index and control effort applied to the process, among others. This methods offer an alternative to the empirical selection of the weighting control factor λ. |
