Hierarchical fault diagnosis and severity identification method of building air condition systems

Abstract

Faults are inevitable of building Heating, Ventilation and Air Conditioning (HVAC) system, which causing either excessive energy consumption, bad thermal comfort, or undesirable indoor air quality. Existing fault diagnosis research generally uses algorithm optimization and data processing to improve performance, but not enough consideration is given to process design and the difference of physical properties of parameters during fault. Thus, this paper proposed a two-level hierarchical fault diagnosis and severity identification method for building Variable Refrigerant Flow (VRF) system. The Extreme Gradient Boosting (XGB) was studied to train fault diagnosis model at first level, and the association rule dug out each fault feature with different severities and XGB was used to identify fault severity at second level. Three faults with ten severities were simulated to verify this method. The results indicated that the fault diagnosis accuracy of abnormal refrigerant charge, four-way valve fault and Electronic Expansion Valve (EXV) fault could reach 99.46%, 100% and 99.65% respectively. The association rule analysis found that the discharge temperature, the sub-cooler gas outlet temperature, and sub-cooler EXV opening had the highest support to three faults respectively. Based on adaptive input features, the average fault severity identification rate of three faults could reach 98.23%, 99.94% and 96.16% respectively.