A novel remaining useful life prediction method based on CNN-Attention combined with SMA-GPR

Accurately predicting the future capacity and remaining useful life (RUL) of lithium-ion batteries is crucial for ensuring their safety and reducing the maintenance costs of related equipment. However, the aging data of lithium-ion batteries (LIBs) exhibit significant nonlinearity and are also affected by uncertainties such as capacity regeneration. To address this issue, this paper proposes an RUL prediction method based on a Convolutional Neural Networks-Attention Mechanism (CNN-Attention) combined with a Slime Mold Algorithm - Gaussian Process Regression (SMA-GPR) model. Firstly, SVMD is applied to extract capacity regeneration features and capacity decay features. Next, to solve the data dependency of single-model prediction, the SMA-GPR model is applied to improve the CNN-Attention prediction, thus solving the generalization problem of and obtaining an accurate RUL. Next, to verify the accuracy and robustness of the proposed method, the experiments involved long-term aging tests under multiple scenarios including three types of LiFeO4(LFP)batteries including 142Ah square cells、280Ah square cells and 10Ah pouch cells, and varying conditions including temperature, charge-discharge rates and pre-tightening forces. The prediction error based on experimental data applying the three batteries is within 1 %.