Application of Machine Learning to Battery Technologies

The goal of the project is to utilise big data and machine learning approaches to assess the current state of a battery and predict its future capacity directly from electrochemical and other analytical data. It involves building systems for experimentalists to record and store their data accurately, and metadata regarding cell construction and operation, in order to be machine readable. The aim is then to use regression and other machine learning approaches to link features in the electrochemical data to physical processes occurring in the battery, in turn using these to predict its future performance. Another strand is combining data from multiple different sources (electrochemistry, diffraction, spectroscopy and imaging) to learn new correlations between diverse datasets that are indicative of performance and current battery state.