PhD Forum Abstract: Towards Comfort-Aware and Energy-Efficient Control of Domestic Heat Pumps

Heating decarbonisation demands control strategies that are both effective and transparent. However, in domestic air-source heat pumps, thermostat controller heuristics and system physics often intertwine, making it difficult to reason about energy use or to test alternative control policies. We adopt a clear control boundary: the controller selects a flow-water temperature set-point. A separate system model predicts next-minute flow and return temperatures and instantaneous electrical power from the current state and weather. Using household telemetry, we fit a compact model with a single-step temperature update towards the set-point, thermal power from the flow-return difference, and a coefficient of performance (COP) linked to temperature lift which is the difference between flow and outdoor temperature. Model coefficients are estimated by constrained least squares. Validation uses rolling-origin splits and physics checks (energy balance, sensible COP), with day replays around set-point changes. This ongoing work contributes a clear controller-system separation at the set-point boundary, a small set of physically motivated equations calibrated from household telemetry, and a reproducible protocol that supports counterfactual policy testing and an offline reinforcement learning environment providing a practical route towards improved heat pump control.