Improving energy efficiency through forecast-driven control in hybrid heat pumps

Abstract

Hybrid heat pumps (HHPs) are increasingly used for residential space heating, especially where stand-alone heat pumps (HPs) are inefficient. Typically, HHPs and HVAC systems controls rely on simple rule-based approaches. Smart controllers that employ building-system modeling can improve energy efficiency by determining which heat generation unit to activate and setting the supply water temperature according to actual building heat demand. Data-driven models are particularly suitable for widespread use, as they can self-learn building thermal characteristics and optimize system operation. In this study, we employed an autoregressive model to forecast short-term hourly energy demand and the corresponding water supply temperature to the heat emitters. These predictions helped to estimate generators performance and select the optimal unit to minimize energy costs while meeting heat demand. The predictive control procedure was tested on various case studies, both simulated and field-monitored, representative of the Italian housing stock. Results showed that in non-renovated buildings with radiators, the predictive control strategy can reduce operating costs by up to 20% compared to current commercial HHP controls. This improvement was mainly due to better supply temperature set-point evaluation and increased HP use. Similar benefits were observed in environmental and primary energy metrics. Conversely, in newer, well-insulated houses with low-temperature emitters, current controls are already efficient. Finally, we showed that the proposed control strategy deviates less than 3% from an ideal prediction and control in realistic on-field monitored test cases, representing a valuable trade-off between achievable benefits, data requirements, computational efforts, and implementation feasibility in real industrial HHP devices.