Hybrid heat pump demand response potential estimation in UK households using physics-informed machine learning
Hybrid heat pumps (HHPs) have arisen as a pragmatic variant of heat pumps for domestic space and hot water heating. As a heating device that combines a gas boiler and an air-source heat pump, HHPs can electrify a significant amount of the domestic heat demand, without requiring thermal upgrade or storage. One of the most important features of HHPs is flexibility provision, i.e., providing demand response by leveraging building thermal mass and switching between fuels. There has been limited large-scale field trial data of HHP flexibility in the UK, and few studies that realistically represent the short-run thermal dynamics of buildings. This study addresses this gap in the modelling by constructing a machine learning model based on the Lumped Capacitance Model to represent domestic building thermal dynamics, extracting information from monitoring data from over 700 UK dwellings. Optimized heating operation is then simulated to estimate the potential for demand response within thermally acceptable bounds under a tariff-based flexibility mechanism. The impact of building thermal properties and the tariff structure on HHP demand response potential is examined. It is found that even during peak wintertime, HHPs with an output capacity of 4kW can electrify 86% percent of the heating demand, while providing 77% of monovalent heat pump (MHP) peak time demand response in addition to reduced baseline peak demand, without compromising on thermal comfort. This study constructs a portfolio of empirically grounded building energy flexibility models, which can be easily integrated into a distribution-level power network model, to investigate the impact of HHP domestic heating flexibility in a wider system.