Model correction mechanishm for nonlinear time variant systems as support to predictive control strategies

Abstract This work presents a strategy to estimate and to correct dynamics variations in nonlinear time variant systems. This correction is carried out by estimating the internal parameters of the process and determining the differences with an available nonlinear model of the system. The proposed approach has a double functionality; on the one hand, it allows a better performance of using nonlinear models for control purposes, like for nonlinear predictive controllers; and on the other hand, it can be used as a diagnosis mechanism since it provides relevant information about the current state of the system. Thus, in order to use this technique with nonlinear time variant systems, a nonlinear model predictive control strategy has been used. The estimator proposed within the framework of this work is similar to the Moving Horizon Estimation strategy. Experimental results on a real tank process are presented to show the main properties of the proposed architecture.