Adaptive Model Predictive Control Scheme Research Articles

A new recursive PLS structure for rapid identification on demand

Regression techniques that are used for online estimation and control generally yield poor models if the data is not rich enough. Further, restrictions on the lower limit of forgetting factor, used to prevent ill conditioning of covariance matrix, confine the application of such regression techniques to slowly changing processes only. In this paper, a modified block-wise recursive PLS (RPLS) technique that is based on selection of rich data has been proposed for online adaptation and control. Due to its ability to accommodate a wider range offorgetting factors, the technique has been shown to quickly capture the dynamics of slow as well as fast changes in processes. The suitability of a data block to estimate and update models in a reliable way has been measured in terms of the condition number of an appropriate input matrix. The potential of the proposed technique has been illustrated with two process examples. i) dynamic model identification of a liquid level system with a sudden change in the process parameters, ii) adaptive model predictive control strategy for a highly non linear system.