Hybrid heat pump and district heating (HP–DH) systems are increasingly used to support building electrification, but their supervisory supply water temperature setpoints are often fixed and do not respond to partial-load hydraulics, occupancy, or energy-price signals. This paper proposes a BMS-ready, demand-responsive framework that predicts the absolute supply water temperature setpoint from routinely available signals (hydronic temperature difference Δ T , AHU valve position/state, outdoor temperature and occupancy schedule) using supervised learning, and ensures safe operation using a transparent rule-based supervisor. The supervisor enforces Δ T limits, keeps critical-valve opening within a hysteresis band, adds occupancy and room-feedback biasing, and allows price-based setpoint offsets while preserving comfort constraints. The method is trained and tested on two commercial buildings: Case 1 in Portugal and Case 2 in Estonia. System-level relevance is further shown at a Norwegian commercial site operating a hybrid HP–DH plant. In the field deployment, the yearly average COP increased from 6.26 to 7.25 and the heat pump share rose from 7.1% to 14.4%, indicating higher electrification without loss of operational stability. Overall, combining data-driven setpoint prediction with explicit safety constraints provides a practical way to increase heat pump operation while using district heating for peaks and maintaining acceptable return temperatures. • Supervised learning predicts absolute supply-water temperature setpoints from BMS signals. • A rule-based supervisor ensures safe and stable operation. • Models are validated in 2 buildings in different climates and system architectures. • Field deployment in a hybrid HP–DH plant increased COP by 15.8% and doubled heat-pump share. • Plug-and-play integration via baseline-plus-correction requires no new hardware.
Sukhanov et al. (Wed,) studied this question.