Linear Parameter-Varying Models for Predictive Control Design: Application to Nonlinear Chemical Reactors

Thiago V. da Costa, Lívia M. Tizzo, Flávio V. da Silva, and Ana Maria F. Fileti


model predictive control, system identification, local model networks, process control


This paper presents the application of an identification algorithm based on local model networks able to split the full model dynamics in linear parameter-varying (LPV) models for different regions on the process operating range. It is shown that a model based controller equipped with an efficient LPV model performs better than when a single linear time-invariant (LTI) model is used. Results demonstrated that model adaptation over several regions provides better system representation leading to more efficient and consistent control in already implemented control loops.

Important Links:

Go Back