Fault detection and safety risk evaluation of lithium-ion batteries based on confidence interval of Gaussian process regression for real-world application

Electric vehicles (EVs) have emerged as a significant contributor to reducing carbon emissions and promoting sustainable development. Accurately evaluating the safety risk of lithium-ion batteries is the key to ensuring the safe and reliable operation of EVs. However, the occurrence and progression of battery faults are uncertain, especially in complex practical scenarios. To this end, we design a novel fault detection and safety evaluation framework featuring adaptive thresholds based on confidence interval (CI) of Gaussian process regression (GPR). This framework enables battery faults to be detected, located, and tracked in real-world scenarios. The effectiveness of the proposed method is verified on a dataset of 1046 cells from nine faulty EVs, two battery modules and twenty-eight battery systems. The average precision and false positive rate of fault detection results are 0.99 and 0.07, respectively. Compared with the battery management system alarm, the proposed method provides 22.3–240.1 h and 2–28 s early warnings for developmental and sudden faults, respectively. Furthermore, our method accurately tracks the evolution trend of faults and enables the visualization of battery safety risks. This work highlights the potential and generalizability of GPR-based CI as adaptive thresholds for monitoring and diagnosing battery faults.