Data-Driven Model for Lithium-Ion Battery Electrode Property Estimation

• Developed machine learning model to estimate 3D electrode effective properties using it’s physical descriptors.
• Generated and characterized a large dataset of ∼ 17000 unique stochastic porous electrode microstructure with varied active material composition and binder phase, ellipsoidal particle shapes, sizes, and orientation.
• Developed various ML models using scikit-learn library with top performing gradient boost model having R2≥0.9
• Performed feature ranking analysis to highlight relative importance of input descriptors at output predictability.