Real-time knowledge- and data-driven reliability analysis for lithium-ion battery energy storage system by Bayesian fault propagation network

With the wide application of Lithium-ion Battery (LIB) Electrochemical Energy Storage System (BESS) under complex working conditions, relevant disasters increase and cause huge losses. The safety of BESS becomes a vital concern, and the reliability analysis plays an important role in ensuring BESS safety. However, most of the existing methods are unable to update the failure probability and track the failure propagation path in real time, so that the potential risks in BESS are difficult to capture. In addition, firefighting systems and thermal insulation materials are widely applied in BESS as fault blocking conditions to prevent failure propagation, but the existing methods do not consider their influence on calculating failure probability in BESS reliability analysis. To fill in the gaps, a novel reliability analysis framework named Bayesian Fault Propagation Network (BFPN) is proposed in this paper. This framework applies the Bayesian network to determine the failure probability in BESS and analyze the path of failure propagation. In order to dynamically update the failure probability in real time, a systematic knowledge- and data-driven algorithm is proposed by integrating Prognostics and Health Management (PHM), FIDES, and fuzzy logic. Moreover, the fault blocking conditions are considered in the framework to approach quantitative and realistic fault propagation, such that the proposed method suits the real scenarios. The main contribution is that the proposed method offers an applicable tool for the dynamic reliability analysis of entire BESS, taking fault blocking conditions into consideration. The experiment is conducted in a BESS that has twelve battery packs and 216 cells (eighteen cells for each pack). The results show that the proposed framework can evaluate the reliability of BESS in real time and provide the failure probability for BESS safety. The experiments validate the proposed framework as an effective tool for evaluating failure risks in BESS, offering comprehensive insights into fault coupling and propagation mechanisms within the system. • Establish a Bayesian network-based framework for fault propagation in BESS. • Develop a hybrid knowledge-data driven method for failure probability estimation. • Implement real-time failure probability updating. • Integrate fault blocking conditions for realistic failure propagation in BESS.