.png){kind=icon}
.png){kind=icon}
Return
A method for reconstructing the charging voltage curve of lithium batteries based on a deep time-series prediction framework
.png){kind=icon}
PRE
AI
X
Y
S
DOI：10.1016/j.tca.2026.180384.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Battery charging curves contain substantial information regarding critical internal states of batteries. However, it remains challenging to collect complete charging profiles in practical battery applications. This study proposes a battery charging curve reconstruction method based on a deep time series forecasting framework, which decomposes the reconstruction process into forward and backward reconstruction components, implemented through a multi-step iterative mechanism. Additionally, a multi-phase reconstruction strategy integrating long-horizon and short-horizon models is introduced to address two critical challenges: training blind zone and error accumulation. Evaluations on two publicly available datasets (Oxford dataset and Aachen dataset) validate the effectiveness of the proposed approach. Comparative evaluations on the Oxford dataset demonstrate that the proposed method outperforms existing state-of-the-art reconstruction methods, achieving superior performance with a narrower average input voltage window (0.235 V versus 0.3 V). Specifically, it reduced the average RMSE of reconstructed voltage curves by 16.9%, while decreasing estimation errors for maximum capacity and maximum energy by 27.0% and 15.7%, respectively. Reconstruction assessments on the Aachen dataset further validate that DTSF-CVCR maintain robust performance across input segments with varying voltage ranges and exhibit significant advantages over mainstream time-series prediction models across multiple error metrics.
Journal
IF:
3.5
Papers:
9.5K
Citations:
1.9W
Organization
No organization information available
Cited Papers