Remaining useful life prediction of lithium-ion battery based on real-time decomposition and tightly coupled convolutional informer

Accurate remaining useful life (RUL) prediction of lithium-ion batteries (LIBs) is crucial for ensuring safety and optimizing the performance of battery-powered systems. However, existing methods often suffer from capacity regeneration and cannot effectively learn the local fluctuations and global nonlinear degradation of the LIBs' capacity. To address these challenges, a novel hybrid framework that combines real-time variational mode decomposition (RTVMD) with a tightly coupled convolutional Informer model (TCCI) is proposed. Firstly, RTVMD introduces a fully stepwise real-time decomposition sampling technique to decompose the capacity series of LIBs in real time. Then, the TCCI module integrates the CSP attention, dilated causal convolution, and passthrough mechanism in a unified architecture to capture the capacity series’ local and long-term dependencies. Finally, several comparative experiments are conducted on three LIBs degradation datasets to demonstrate the effectiveness of the proposed model. The experimental results illustrate that the maximum mean absolute percentage error (MAPE), root mean squared error (RMSE), and mean absolute error (MAE) of the proposed method are merely 0.015, 0.008, and 0.006, respectively. Compared with other methods, the prediction accuracy of the proposed method is improved by at least 20 %. Efficient prediction outcomes can offer sufficient backing for electrical equipment predictive maintenance.