Accurate Prediction of Electrochemical Degradation Trajectory for Lithium-Ion Battery Using Self-Discharge

Accurate prediction of battery performance is crucial for timely battery health management. However, it is challenging to forecast precisely battery's performance due to its intricate degradation mechanisms and variability in the degradation rates. To overcome the challenges, we employed an artificial intelligence (AI)-driven approach. By training an indicator that reflects well structural degradation of a battery (self-discharge; SD), a model could predict battery performance metrics with a high accuracy compared to models using direct indicators (e.g., capacity). In a comparative analysis, the self-discharge model with the couple of bidirectional-long-short term memory demonstrates outstanding prediction accuracy with a mean absolute percentage error of 0.39% for capacity retention (CR) prediction and a root mean square error of 13.2 cycles for remaining useful life prediction. These findings underscore the significance of incorporating indicators reflecting the internal electrode health of batteries for accurate lifespan prediction.