Remaining useful life prediction of lithium-ion batteries: A decomposition-based multi-model prediction approach

Accurate prediction of the Remaining Useful Life (RUL) of lithium-ion batteries is critical to ensuring the reliability and safety of battery-powered systems. During the degradation process, lithium-ion batteries are usually accompanied by the capacity regeneration phenomenon, resulting in a non-linear and non-smooth capacity degradation curve. To address this challenge, this paper proposes a novel hybrid prediction method. First, Catch Fish Optimization Algorithm (CFOA)-Variational Mode Decomposition is used to adaptively decompose the original battery capacity series to reduce the complexity of the battery capacity series. Then, Gaussian Process Regression (GPR) and Convolutional Neural Network (CNN)-Echo State Network (ESN) are used to predict the aging trend signals and oscillatory signals, respectively. Finally, the prediction results are integrated. The proposed method is validated on two publicly available battery aging datasets (NASA and CALCE), achieving a root mean square error of no more than 0.038 56 and 0.010 52, which significantly outperforms traditional single models such as CNN, GPR, and ESN. The effectiveness of the proposed method is validated on two publicly available battery aging datasets, which provide higher prediction accuracy and generalization capability compared to the traditional single-model approach. The results confirm the good effectiveness and generalizability of the proposed hybrid method. This work provides a promising data-driven solution for the prediction of lithium-ion batteries in practical applications.