It might be a valuable addition to have a simple physical model for heating demand. One option would be, that you give insulating surfaces (areas with corresponding widths and u-values for the insulation) and two time series for temperatures. In the most simple case, we could just take the transmission as demands.
What do you think?