Normal factors such as complex control strategies

flexible meteorological parameters

and operation conditions or abnormal factors such as refrigerant charge faults in variable refrigerant flow (VRF) systems can lead to complex and varying fluctuations in the energy consumption. It is difficult to directly diagnose whether normal or abnormal factors cause such fluctuations based on the energy consumption. In this study

an effective energy assessment and diagnosis method is proposed

which combines the support vector regression (SVR) algorithm with the one-class support vector machine (OCSVM) algorithm to diagnose the energy performance of a VRF system. An energy assessment and diagnosis model is constructed based on the normal data set

and is verified by the abnormal energy data set. The results show that the energy assessment and diagnosis model based on SVR-OCSVM has a high accuracy of up to 70%.