Fractional long-range dependence in the capacity degradation trend of Li-ion batteries for RUL estimation

The capacity degradation trend of lithium-ion batteries is not monotonic. After a certain number of charge-discharge cycles, the battery capacity can show a significant increase, accompanied by substantial fluctuations. Research has found that the current degradation state of a lithium-ion battery is influenced by its historical data, demonstrating long-range dependence (LRD). However, the existing prediction methods often ignore these large fluctuations and LRD. To address these limitations, this paper proposes a fractional Poisson process (fPp) difference prediction model. The model's ability to capture both jumps and long-range dependence is validated using the Hurst exponent. The model parameters are then identified using maximum likelihood estimation. The proposed model is evaluated on the NASA lithium-ion battery degradation dataset, which consists of constant current-constant voltage (CC-CV) cycling data. The results demonstrate that the fPp difference prediction model offers significant advantages over existing approaches.