battery-hub

Hybrid ML-MCDM battery end-of-life framework

RISHABH KUMAR

State-of-health estimationRUL predictionNMCLFPNCALCONa-ionZn-ionTrained on synthetic + experimental data

Hybrid ML-MCDM battery end-of-life framework

Ensemble for routing EV batteries at end of life in India, with anomaly gates, global and per-chemistry XGBoost SoH models, and XGBoost and TCN RUL models.

Model details

Architecture
XGBoost, TCN, Isolation Forest and VAE ensemble
Input
32–38 audited cell features; 60-cycle sequences for the TCN
Output
SoH, RUL, out-of-distribution flag
Training data
Synthetic + experimental
Other training data
NASA PCoE; CALCE; Stanford; 130 synthetic cells (PyBaMM and BLAST-Lite)
Uncertainty
Not reported
Physics constraints
None
License
mit

Limitations

  • For research and demonstration; not ready for live BMS data, fleet management or regulatory submissions.