The influence of frosting on typical day selection for the evaluation of heat pumps
This study investigates the influence of frosting on the selection of typical days for heat pump performance testing. The objective is to reduce experimental time by using a typical-day method, where annual performance is extrapolated from measurements on a few typical days selected via time-series clustering of weather data. We analyze how the weighting of weather characteristics, ambient temperature, relative humidity, and solar irradiation, in the clustering algorithm impacts the accuracy of predicting the Seasonal Coefficient of Performance (SCOP). Simulations of a building energy system with heat pump were conducted for the three German locations Potsdam, Garmisch and Sylt using a dynamic simulation model that includes a frosting and defrosting algorithm. The results demonstrate that ambient temperature is the dominant factor to accurately approximating the SCOP. For the building energy system model used in this study, high weights on the ambient temperature and low weights on the relative humidity and solar radiation yield SCOP approximation errors below 5% across the three locations. Furthermore, using four typical days proved to be an effective compromise, significantly reducing testing time while maintaining sufficient accuracy.