21 - System Identification and Model Predictive Control

Summary

Uncertainty is ubiquitous in engineering practice and models.Parameters that are estimated via online measurement or by experiments always carry a certain level of uncertainty with them – which can in fact be significant for difficult-to-measure systems.Other sources of uncertainty are fluctuations in process inputs, e.g. concentrations, flow rates, temperatures, etc.And finally, one may not be certain of the structure of models, e.g. the actual chemical reaction mechanism(s) may be uncertain.All these necessitate special handling of such models, and where the uncertainty can be quantified by probabilistic measures this allows special formulations and solution procedures to be employed so as to derive robust solutions with respect to the uncertainty involved.All these, along with the necessary theoretical concepts, are presented in this chapter, with subsequent emphasis for practical application to the multiple scenario approach for the handling of parametric uncertainty.