Fault diagnosis model of refrigeration system combining Boruta and LightGBM methods for energy saving

Abstract Energy consumption of HVAC system accounts for a large proportion of building energy consumption, and it is necessary to improve the energy efficiency of the system to achieve the purpose of energy saving. Once the HVAC system malfunctions, the energy efficiency of the system will decrease, resulting in unnecessary energy waste. Aiming at common failures of refrigeration systems, this research proposes a fault diagnosis model combining Boruta and LightGBM methods to diagnose faults in time. The results show that compared with other three tree-based models, this optimized LightGBM model can not only efficiently diagnose the refrigerant charge amount faults of the VRF system, achieving a fault diagnostic accuracy of 97.52%, but also can be transplanted to other systems for fault diagnosis, thus effectively avoiding unnecessary energy waste caused by those failures.