| Title |
LSTM-based State-of-health estimation using Lithium-ion battery DC and AC characteristic factors |
| Authors |
Jiwoong Kim ; Jaehyeong Lee ; Dongcheol Lee ; Jaea Lee ; Jonghoon Kim |
| DOI |
https://doi.org/10.6113/TKPE.2026.31.2.135 |
| ISSN |
1229-2214(pISSN), 2288-6281(eISSN) |
| Keywords |
Lithium-ion Battery; State-of-health; State-of-charge; Electrochemical impedance spectroscopy; Long short-term memory |
| Abstract |
Lithium-ion batteries have become central to power and transport systems with the growth of electric vehicles and renewable energy, underscoring the importance of accurate state-of-health (SOH) estimation. The high state-of-charge (SOC) is a particularly important operating region. In the analysis of battery behavior, power capability is closely linked to safety, but the diagnostic signals are weak and strongly influenced by operating conditions. This study takes the high-SOC region as the basis for SOH estimation and compares DC indicators derived from the terminal voltage and current, which can be easily obtained from battery management system logs. However, these indicators are sensitive to external conditions. AC indicators obtained from electrochemical impedance spectroscopy (EIS) provide reproducible and mechanism-aware information when the temperature and SOC are controlled. In this study, a long short-term memory network is employed to capture long-term degradation and short-term fluctuations, and single-indicator inputs are evaluated within a unified framework. The results show that AC features, particularly the charge transfer resistance (Rct), yield the most consistent SOH predictions and outperform DC metrics. These findings confirm the diagnostic value of the high-SOC region for practical SOH estimation. |