A method to evaluate the energy performance of inverter-driven heat pumps in real operating conditions

Inverter-driven heat pumps, owing to their modulation capability, represent a promising alternative to gas boilers. As these systems frequently operate under transient and part-load conditions, their actual performance can differ significantly from the nominal values reported in manufacturers catalogues, despite remaining a valuable source of information. This paper presents a methodology that leverages manufacturer data to predict the performance of inverterdriven heat pumps. The objective of the work is to propose a simplified and reliable procedure to simulate heat pumps performance in partial and full load operation in real operating conditions whenever the available data are obtained by manufacturers' datasheets. A classifier was developed to identify the system's operating modes based on heat and power load patterns, as well as temperature gradients. A regression model was then trained using manufacturer data and validated against field measurements from 21 residential installations in the United Kingdom (20 air source and 1 ground source heat pumps). Two simulation approaches were compared: one that considered only stable operation and another that included all operations (both transient and stable). The results show strong agreement with measured data (R2 between 0.64 and 0.95, rMAE between 7%-28%, CVRMSE between 12%-36%), with only a moderate reduction in accuracy when transient behaviour is included. Although discrepancies may occur at the hourly scale, they tend to compensate over time, leading to a mean relative error of the SCOP of 9% for both the simulations. Furthermore, the analysis shows that comparable predictive performance can be achieved with a reduced yet well-distributed training dataset.