Liquid electrolytes govern the performance — efficiency, energy density, life — of solution-based processing and battery technologies. The vision is to integrate AI and ML tools into our experimental workflow and accelerate the discovery, design, and optimization of electrolyte materials.
What we’re working on
- Bayesian optimization–accelerated electrolyte design. The design space for liquid electrolytes (salts, solvents, diluents) is high-dimensional, and experimental resources are limited. By coupling high-throughput electrochemical measurements with ML-based surrogate modeling, we are identifying optimal aqueous electrolyte compositions that maximize metal deposition kinetics and efficiency. The iterative loop provides mechanistic insights into how ionic speciation and interfacial chemistry govern performance.
- Physics-AI hybrid models. Accurate prediction of thermodynamics, kinetics, and transport across complex electrolytes is limited by missing parameters and nonlinear composition effects. We are developing a physics-informed AI framework that integrates thermodynamic modeling with data-driven learning to predict metal deposition potentials, charge-transfer kinetics, and ionic transport — extending predictive accuracy to multicomponent, highly concentrated electrolytes.
Research highlights
Early-stage lifetime prediction for Li-ion batteries
A two-stage deep-learning framework predicts Li-ion battery lifetime from limited early-cycle data, combining features extracted across multiple cycles to extend the reliability of early lifetime forecasts — reducing the time and cost of battery qualification. Read the paper in Frontiers in Energy Research →
Two-level optimization for grid-scale battery energy storage
A two-level optimization framework for battery energy storage systems improves economics while minimizing long-term capacity fading, bridging operational dispatch decisions with degradation-aware scheduling. Read the paper in Journal of Energy Storage →
Physics-guided ML for capacity-fading mechanism detection
A physics-guided ML framework classifies the dominant degradation mechanism using physics-informed features and regresses fading rates from early-cycle data — delivering both accurate prediction and mechanistic insight into why a cell is aging. Read the paper in Batteries →
Selected publications
- Frontiers in Energy Research, 2022, 10, 1059126 — Two-stage deep learning for early-stage Li-ion lifetime prediction.
- Journal of Energy Storage, 2023, 63, 106943 — Two-level optimization framework for battery energy storage.
- Batteries, 2024, 10(8), 283 — Physics-guided ML for capacity-fading mechanism detection.