Research on Fault Diagnosis Method of Lithium Battery Energy Storage System Based on Data Drive

To ensure the safe and reliable operation of lithium battery energy storage systems, it is crucial to study fault diagnosis methods for lithium batteries. Early diagnosis of internal short circuits is particularly significant, as it can prevent subsequent serious accidents. However, in the early stages of voltage change, the internal short circuit produces only a minimal voltage variation, which does not trigger alarms and is highly concealed. Therefore, this paper employs an unsupervised learning algorithm to diagnose faults in the energy storage system. First, the Local Outlier Factor (LOF) algorithm is utilized to calculate the local outlier factor of the battery voltage. Next, the 3 $\sigma$ principle is applied as the fault trigger threshold to identify the faulty batteries. To validate the effectiveness of this method, a comprehensive dataset of battery information from the energy storage system is constructed using a second-order equivalent circuit and a coupling model of internal short circuits. The experimental results demonstrate that the LOF algorithm can effectively identify faulty batteries within the energy storage system.