Data-driven simple thermal models: The radiator - gas consumption model

A simple data-driven Lumped Parameter thermal model is developed linking the radiator surface temperature to the whole house gas consumption using operational data collected from a real house. The indoor room air temperatures, the radiator surface temperatures and the whole house gas consumption are used as input data. The model parameters are estimated using the Ordinary Least Squares technique. Results show that an improved fit can be achieved by excluding data points for which gas consumption is being used for purposes other than space heating using a lower threshold (Tr,pred-Tr,meas) of 1oC during the model calibration. Finally, the model is applied to disaggregate the gas consumption data that are linked to space heating from the original whole house gas consumption at half -hourly intervals.