Remaining useful life prediction of lithium-ion battery based on fusion model considering capacity regeneration phenomenon

Remaining useful life (RUL) holds significant importance in battery management system, and accurately predicting RUL is incredibly important to guarantee the safety and stability of the battery operation. However, capacity regeneration phenomenon during the deterioration of lithium-ion battery is usually unavoidable, which affects the precision of RUL prediction. To address above issue, this paper proposes a RUL prediction method utilizing particle filter and autoregressive integrated moving average (PF-ARIMA) model considering capacity regeneration phenomenon. Firstly, capacity regeneration point (CRP) is detected by using Wasserstein-distance (W-distance) to capture the discrepancy between the prior and posterior probability distributions in the particle filter (PF). Secondly, the absence of real capacity during the predictive stage prevents the online updating of degradation model parameters in the PF algorithm, so the autoregressive integrated moving average (ARIMA) model is introduced to recursively obtain the capacity as the observation of the PF to further realize the online estimation of the battery capacity. Finally, the validity of the RUL prediction method in this paper is examined using the NASA lithium-ion battery dataset and compared with existing methods, and the experimental results reveal that the errors of the proposed method are basically within 5 %, which improves about 70 % in comparison with other methods, demonstrating the highly stable and reliable performance of the proposed method.