In experimental models of multi-component thermal systems, small errors in each submodel can propagate detrimentally through the overall model, resulting in large prediction errors as the prediction time increases. These errors can be problematic when using open-loop or feed-forward control schemes. This paper demonstrates the advantages of a whole-system or integrated parameter estimation approach as opposed to the component-by-component parameter estimation approach that is widespread in the literature. The approach is demonstrated on a combined heat and power system at a laboratory facility, and the resulting model is used to predict the system temperatures up to 20 min in advance. Results show that, when compared to conventional component-by-component parameter estimation, the integrated parameter estimation approach improves the model prediction accuracy significantly.
Copyright © 2012 ISA. Published by Elsevier Ltd. All rights reserved.