sefraniBattery SoH LSTM
LSTM that predicts the state of health of EV batteries from 8 per-cycle features, using cycle-level encoding after CyclePatch (BatteryLife).
Model details
- Architecture
- Per-cycle FFN encoder, 2-layer LSTM and MLP head
- Input
- 8 per-cycle features (discharge voltage, charge capacity, temperature, internal resistance, ...)
- Output
- SoH
- Training data
- Not stated
- Trained on
- 150 batteries; split by battery ID
- Uncertainty
- Not reported
- Physics constraints
- None
- License
- mit