Anomaly Detection of Storage Battery Based on Isolation Forest and Hyperparameter Tuning

The safety of an uninterruptible power supply (UPS) unit is very important in the operation of a telecommunication room. It is necessary to identify and replace abnormal electrical batteries of the UPS to ensure the normal operation of the equipment. In this paper, a single-model method based on isolation forest and hyperparameter tuning is proposed for detecting abnormal batteries. Experimental results show that the proposed method is efficient in offline situations. A multi-model method is also proposed to deal with the online anomaly detection problem, which is found performing well.