A comprehensive variable refrigerant flow heat recovery model for building performance simulation.

A comprehensive, physics-based, and modular Variable Refrigerant Flow with Heat Recovery (VRF-HR) model is developed for multi-year simulations of large-scale VRF systems. The model is designed to simulate various operational modes, including single-mode (cooling-only, heating-only) and heat recovery mode, across any number of indoor units (IUs), outdoor units (OUs), and compressors. A parameter-estimation procedure leveraging manufacturer data is implemented to calibrate the model, ensuring accurate system representation. A machine learning-based control strategy is introduced to emulate real-world compressor selection for partial load operation. The model is validated using two years of operational data from a large-scale VRF system serving the first floor of the former ASHRAE Headquarters Building in Atlanta, USA, which consists of 22 indoor units, 2 outdoor units, and 8 compressors. Results demonstrate that the manufacturer-tuned model accurately predicts total energy consumption, achieving a relative error of 9.5 %, an NMBE of 6.2 %, and a CVRMSE of 27.2 % over the first year. For the second year, the model achieves a CVRMSE of 25.3 %, an NMBE of 5 %, and a relative error of 7 %, meeting ASHRAE calibration criteria.