AI Battery Lifecycle (v3)
Estimates current-cycle state of health from the 4.0–3.6 V part of a discharge, validated on held-out batteries from NASA, CALCE and Oxford aging data.
Model details
- Architecture
- Ensemble of tabular models (20 evaluated)
- Input
- 18 features from the 4.0–3.6 V portion of a discharge
- Output
- SOH
- Training data
- Experimental
- Trained on
- NASA PCoE; CALCE; Oxford Battery Degradation Dataset 1
- Uncertainty
- Not reported
- Physics constraints
- None
- License
- mit
Limitations
- For research on lab aging data; not certified for battery management, charging control, safety decisions or field use.
- Doesn't provide a validated RUL forecast.
- Check cross-dataset errors before applying it to a new chemistry, manufacturer, sensor set or duty cycle.