Onboard in-situ warning and detection of Li plating for fast-charging batteries with deep learning

Abstract

Accurate lithium plating detection and warning are essential for developing safer, longer cycle life, and faster charging batteries. However, it is difficult to in-situ detect lithium plating from the electrochemical signals without the introduction of sensors or reference electrodes. Here, we proposed an online lithium plating detection and warning method based on anode potential construction. By establishing a precise mapping relationship between battery voltage and three-electrode potential through deep learning, we can reconstruct the three-electrode curve of batteries accurately without introducing a reference electrode. So that lithium plating can be detected over the full life cycle of batteries with a positive rate of 99.9%. Furthermore, with the combination of the voltage prediction module, the future anode potential can be predicted and the lithium plating can be warned with a positive rate of 98.7%. Our approach provides a new possibility for the development of fast-charging technology and life extension strategies.