Modeling of Air-Source Heat Pumps in Smart Residential Applications

This paper presents a dual-framework for modeling and forecasting the power and energy consumption of a central air-source Heat Pump (HP) at a Canadian home during winter, when HP electricity demand is significantly higher, based on actual measurements. The first approach is a physics-based model that combines a Thermal Equivalent Circuit (TEC), calibrated based on measurements, with a temperature-dependent Coefficient of Performance (COP) derived from manufacturer specifications. The second approach is a data-driven method that employs supervised Machine Learning (ML) techniques, specifically Artificial Neural Networks (ANN) and Long Short-Term Memory (LSTM) networks, to forecast the short-term electricity consumption of the HP, based on diverse inputs such as power demand measurements, outdoor temperature, temporal variables, and solar radiation. The results demonstrate the superior performance of the proposed LSTM model, outperforming ANN-and physics-based models. The proposed dualmodeling framework enables accurate residential HP power and energy forecasting that facilitates the deployment of demand-side flexibility in smart grids.