LAUSR

Abstract In variable refrigerant flow (VRF) systems with multiple indoor units, individual energy metering (IEM) not only contributes to a fair charge for the tenants, thus avoiding their complaints on the charging fee, but also shows the actual allocation of capacity for IUs. However, it is extremely challenging to measure the accurate capacity of each IU directly according to current researches though total capacity can be measured. In this paper, methods of individual energy metering for VRF systems were reviewed, and three new methods, based on the electronic expansion valve (EEV), machine learning model (MLM), and throttling model (TM), were proposed. The accuracy and applicability of the three methods were investigated and analyzed according to experiments on a water-cooled VRF system. In the experiments, the maximum deviation factors of the EEV-, MLM-, and TM-based methods were ±7.8%, ±6.0%, and ±6.9%, respectively. In addition, coefficients of variation of the root-mean-square error of the above methods were ±4.4%, ±3.3%, and ±4.4%, respectively. Considering the practicability and feasibility in real projects, the TM-based method was applied to an air-cooled VRF system, showing superior accuracy in cooling conditions.