| Title |
Battery State-of-health Estimation using a Dual-branch Model |
| Authors |
심소현(Sohyeon Sim) ; 박용주(Yongju Park) ; 박진욱(Jinuk Park) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.2.43 |
| Keywords |
State-of-health; Battery states; Deep learning; Temporal features; Global features |
| Abstract |
Battery degradation and malfunction pose significant safety concerns, increasing the demand for accurate state?of?health (SOH) prediction and early anomaly detection. Most existing studies, however, focus on cycle-based aging prediction rather than estimating battery degradation during charging. In this work, we propose a Dual-Branch framework that leverages real-time voltage, current, and temperature measurements obtained during the charging process. The proposed model jointly exploits temporal features captured by a sequence-based branch and statistical summary features extracted over the entire charging interval. Experimental results on the NASA and MICH public datasets demonstrate that the model achieves high prediction accuracy. In particular, cross-dataset validation shows that our method improves MAE by 21% compared to the baseline Random Forest model, indicating strong generalization capability across heterogeneous data distributions. |