Collective AHU Anomaly Detection for Building Energy Optimization

As for the national agenda, carbon neutral 2050, the efficient use of energy is very important along with the development of new and renewable energy. In the case of building energy, energy efficiency largely depends on the operating efficiency of the HVAC system. The AHU is a main part of the HVAC system. Therefore, the normal operation of the AHU device is very important for energy efficiency. This paper summarizes the study results about AHU device anomaly detection. It is very difficult to obtain AHU device fault data at the actual site, so we generated simulation data by using domain knowledge and a simulation tool. The building energy simulator energyPlus™ was used, and the data that was the basis for the simulator was measured from the actual building in KIER. Ensemble algorithms were used for multiclass classification and obtained over 90% accuracy for a cooling season and 99% accuracy for a heating season.