Electrochemical Energy Storage (Batteries) 

Research Contents

• Data curation & feature engineering: build standardized, cross-chemistry datasets and robust health indicators from noisy field logs. 

• Physics-informed ML & transfer: embed electrochemical constraints and transfer models across cells, packs, and use cases.

• Uncertainty & prognostics: quantify prediction intervals for RUL/SOH and propagate to risk-aware maintenance.

• Operations co-optimization: link diagnostics to charging, warranty, and asset-management policies.

Predicting battery lifetime and assuring reliable operation remain central challenges because degradation emerges from tightly coupled electrochemical–thermal phenomena, manufacturing variability, and usage diversity that are rarely captured in controlled cycling tests. Field data are heterogeneous, sparse, and often event-driven, which complicates model training and validation and leads to poorly calibrated uncertainty in remaining-useful-life estimates. Our approach combines physics-informed modeling with rigorous feature engineering and transfer learning so that models honor conservation laws and material limits while still adapting across chemistries, formats, and duty cycles. Calibrated uncertainty quantification translates predictions into risk-aware maintenance, warranty, and safety decisions that can be executed at cell, pack, and fleet scales. By fusing laboratory evidence with operational telemetry, we construct non-invasive diagnostics and digital twins that reveal near-wall thermal behavior, detect precursors to accelerated aging, and support optimal charging strategies under real constraints. 

Associated members: Sangjun Jeon, Yooil Son