A centrifugal heat pump model identifiable from catalogue data: unified calculation of cooling, heating, and heat-recovery operation for annual building energy simulation

Annual building energy simulation needs heat-source models that are identifiable from publicly available information, physically sound in extrapolation, and computationally cheap; empirical models fail outside their underlying data, while rigorous physics-based models require proprietary design information. This paper proposes a hybrid model of centrifugal heat pumps that combines a machine-independent theoretical refrigeration cycle with a single manufacturer-specific efficiency characteristic, linear in six coefficients over two similarity-based dimensionless variables. It is identified from catalog data alone by inequality-constrained least squares, yielding a unique solution without iterative optimization. Cooling, heating, two heat-recovery modes, and operation below the capacity-control range are handled uniformly. The model reproduced the catalog of a 3341 kW inverter-driven HFO-1234yf machine with 1.7% mean absolute percentage error (maximum 6.9%), and the cooling-mode calibration alone predicted the heat-recovery rated point, outside the catalog map, within 2.3%. An annual simulation of 8760 hourly states runs in under one second.