.png){kind=icon}
.png){kind=icon}
Return
Battery smart sensing via a virtual reference electrode
.png){kind=icon}
OA
AI
DOI：10.1038/s41467-026-76535-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Battery safety is regulated by management systems that use current and cell voltage to define operating limits, but these signals can miss lithium metal plating. Negative electrode potential can reveal this failure mode, yet direct measurement requires an additional reference electrode that is difficult to implement in practical cells. Here, we introduce a virtual reference electrode that estimates negative electrode potential during battery operation without a physical reference electrode. Trained on measured negative electrode potentials from three-electrode cells, the model predicts this internal state using only signals available from two-electrode cells, with root-mean-squared and mean absolute errors of 0.023 and 0.018 volts, respectively. Electron microscopy validates the predicted transition between intercalation and lithium metal plating near the thermodynamic threshold. Integrated into adaptive charging, the virtual reference electrode adjusts current in response to predicted failure risk and extends cycle life by 8.25 times relative to a constant-current constant-voltage baseline under the tested low negative-to-positive capacity ratio. More broadly, this framework may enable virtual sensing of internal battery states without changing battery chemistry or architecture. Battery safety is limited by hidden lithium plating. Here, authors use machine learning to build a virtual reference electrode that estimates negative electrode potential from standard cell signals, detects plating risk and enables an 8.25-fold extension in cycle life.
Journal
IF:
15.7
Papers:
9.3W
Citations:
91.2W
View more organizations