Lithium-ion battery health state and remaining useful life prediction based on hybrid model MFE-GRU-TCA

Accurate prediction of battery state of health (SOH) and remaining useful life (RUL) is crucial for reducing the risk of energy storage battery failures and intelligent management of energy storage power stations. Currently, most existing research methods only consider capacity as the input for their models, disregarding the interconnectedness of internal battery feature data. A model based on MFE-GRU-TCA (Multi-Feature Extraction and Temporal Convolutional Attention Gated Recurrent Units) is proposed to improve the accuracy of lithium-ion battery SOH and RUL prediction. The MFE module is used to extract data features from multiple charge/discharge cycles of lithium-ion batteries that have undergone data selection and data scaling, and then concatenating them with overall features from the cycles. The GRU module is then used to capture the longterm dependencies in the sequential data. The TCA module is used to better represent the decay trend of the capacity series and mitigate the influence of capacity regeneration phenomenon. Moreover, the TCA module leverages temporal convolutional attention to focus on relevant temporal states and produce more accurate predictions. Extensive experiments were conducted on the NASA and CALCE datasets, and comparisons were made with existing methods. The experimental results demonstrate that the proposed model achieves more accurate predictions of lithium-ion battery SOH and RUL. The Root Mean Square Errors (RMSE) on the NASA dataset and CALCE dataset are below 0.832% and 0.614% respectively.