Lithium-ion batteries (LIBs) have emerged as the dominant energy storage medium in electric vehicles and other applications, owing to their high energy density, low self-discharge rate, and superior performance characteristics. Accurate end-of-life (EOL) prediction is thus critical for system safety and reliability. Deep learning methods have become the primary solution for EOL estimation in big data environments. However, LIBs' aging involves complex interactions among internal and external factors, introducing significant uncertainty into the aging process, making the quantification of the uncertainty in EOL estimation a core challenge. This paper proposes a Wiener-LSTM fusion framework that combines stochastic degradation models with deep learning to capture the randomness inherent in LIB degradation mechanisms systematically. The framework not only provides confidence interval estimation for EOL, but also enables dynamic remaining-useful-life interval estimation with incomplete degradation data. Experiments on the NASA lithium battery dataset demonstrate that our method achieves superior accuracy in estimating the EOL interval, with consistent performance across diverse conditions and scenarios. By leveraging these precise estimations to control decision risks, our work offers a viable technical pathway for predictive maintenance in battery management systems.
SHI et al. (Thu,) studied this question.