The Governing Council of the University of Toronto

Published purpose

This project will develop a physics informed machine learning approach to predicting battery durability to be implemented into a self-driving laboratory being developed by the NRC. It leverages University of Toronto AI-assisted tools for fitting electrochemical impedance spectroscopy (EIS), a central tool in measuring electrochemical systems, and generating statistically significant and unbiased models of physical processes. The Recipient will refine and apply this tool to support the NRC’s development of novel battery cathodes by developing (1) an automated sensitivity analysis and out of distribution detection algorithm to enable model updating during active learning studies, (2) a robust modeling framework for generating physical insights into battery performance and degradation, which will permit scientifically informed adjustments to battery formulations, and (3) an active learning tool that combines these tools to predict battery longevity without long term cycling studies. All data and code generated will be released publicly to benefit all Canadians.