Reducing experimental effort in the certification process through model-based performance prediction of variable-speed heat pumps

Heat pumps are essential for decarbonizing the heating sector, particularly air–water heat pumps, which are favored for their low costs and ease of installation. In case of unrefurbished existing buildings, hybrid heat pumps – combining heat pump and gas boilers – offer the possibility to decouple the replacement of the heating system from further renovation measures to the building envelope. For tailoring the heat pump capacity to the building demand, manufacturers offer different variants with different nominal capacities that are based on similar hardware, but certification costs quickly add up if each different combination of heat pump and boiler needs to be measured. To reduce certification cost, we evaluate the possibility of using standard test data for monoenergetic electric heat pumps to derive operating points required by the EN14825 standard for different hybrid heat pump capacities. To achieve this, we present and evaluate a semi-analytical heat pump model with focus on reducing the number of required parameters. The model predicts efficiency (COP) and heat flow rates across various operating points. As a result, the model requires a small experimental dataset of 9 laboratory measurements and only 5 input parameters to fit 5 unknown parameters. Using 115 performance measurements of five commercial heat pumps tested in a certified laboratory, the prediction quality is evaluated and additionally benchmarked against an innovative AI method based on the tabular foundation model TabPFN. The semi-analytical model can predict COP with an average error margin of 6.6% for norm measurements, and SCOP with a maximum error of 9.2%. The efficiency class of the machines according to the Ecodesign regulation (EU 813/2013) is correctly predicted in 8 out of 10 cases using the semi-analytical model, and up to 10 out of 10 cases using the tabular foundation model.