Data-driven Simple Thermal Models: The Importance of the Parameter Estimates

Abstract

A simple 1st order data-driven lumped parameter model of a domestic building is developed to explore the effect of using different model parameter values in the model outputs. The adequacy of the Ordinary Least Square estimation technique is explored. Results show that an improved fit to the measured data can be achieved by varying the initial model parameter values of capacitance (up to 78%), resistance (-46%) and effective window area (-59%). This highlights the importance of having a reference set of parameters based on the known physical characteristics of the building. Finally, the model residuals are deemed appropriate to inform the decision making process for further model development.