Volume 63

Forecasting Building Thermal Demand Using Machine Learning Lloyd Corcoran, Daniel Carr, Carlos E. Ugalde-Loo

Abstract

Rising summer temperatures in the UK, alongside stricter building regulations on thermal efficiency, are increasing the importance of residential cooling demand. While very few households currently use active cooling, adoption is expected to grow, placing additional pressure on electricity networks. This paper presents a study combining physics-based modelling with machine learning to forecast building thermal demand using basic dwelling characteristics and weather data. Results show that the method achieves high accuracy in predicting the thermal demand of previously unseen dwellings—showcasing the potential extreme gradient boosting may have in forecasting cooling demand in a warming world.

Keywords building thermal demand, machine learning, extreme gradient boosting, IES VE