Multi-level anomaly detection of lithium battery energy storage system

Abstract

Abstract The energy storage technology route represented by lithium battery energy storage strongly supports China’s energy structure transformation. The widespread use of lithium batteries also poses a significant safety risk that is often overlooked. Energy storage system security is facing severe challenges. It is very beneficial for the safety of energy storage systems to predict the potential faults of lithium batteries before they are used and find anomaly batteries in time. This paper proposes a three-stage anomaly detection method based on statistics and density concepts to provide real-time potential fault prediction of lithium battery energy storage systems. Considering both efficiency and accuracy, this method can detect anomaly samples from shallow to deep. Finally, the battery condition of a large energy storage plant was studied, and 11 anomaly batteries were detected. The proposed method is significant for safely operating large energy storage power stations.