Improved Battery Manufacturing Using Machine Learning

Advanced Research: Clemson University, Claflin University, College of Charleston, and NASA Langley Research Center

Machine Learning driven manufacturing of lithium-sulfur energy storage systems

This project focuses on advancing Lithium-sulfur batteries (LSBs) with capacities up to 500 Wh/kg at the pack level, significantly surpassing current Lithium-ion batteries (LIBs) in both performance and cost-effectiveness. We will achieve this through innovative manufacturing strategies in collaboration with Clemson Nano-bio lab, Claflin University, the College of Charleston, and NASA Langley Research Center.

The first objective involves understanding the fundamental electrochemical mechanisms of semicrystalline sulfocarbons (SC2), particularly sulfurized poly(acrylonitrile), known for its high capacity, cycling stability, and low production cost. The goal is to increase sulfur loading in SC2 up to 40% by weight for practical LSB pouch cells, essential for pilot-scale manufacturing.

In the second objective, we will develop ML models to predict battery performance based on electrode microstructure, including capacity, rate capability, and cyclability. This involves analyzing over 100 samples with extensive physical and electrochemical characterization. Data collection covers a wide range of electrode materials and involves advanced imaging techniques, while preprocessing focuses on enhancing data quality and consistency. Feature extraction will quantify microstructural and electrochemical characteristics, crucial for the predictive model. The model building involves deep learning and gradient boosting methods, with training parameters designed for accuracy and generalizability. Evaluation emphasizes prediction accuracy and generalizability, employing various statistical metrics.

Lastly, we will train two underrepresented minority undergraduates from Claflin and the College of Charleston in a 10-week summer research program.