Applications of Data Science to Electrochemistry

Abstract

Electrochemical Energy storage is a promising, yet deceptively complex method of powering our world in a sustainable manner. The intricacies of the chemical reactions, kinetics, diffusion, and transport mechanisms involved have made it an area of research where there is a lot to discover and understand about creating powerful and efficient batteries. As the need for moving on from traditional energy sources becomes more pressing, so does the demand for higher performance batteries as replacements.

Discerning which combinations of materials and their properties lead to which characteristics is difficult given all the mechanical and electrochemical factors involved, but the growth of data science and statistics as a tool for all disciplines hopes to shed some more light on this problem from a different perspective. Some performance characteristics such as rate retention, capacity retention, and coulombic efficiency can characterize a battery’s performance, but these values are not known until after extensive testing. The goal of this project is to use machine learning algorithms and other data analysis tools to contribute to this ongoing effort of relating different battery chemistries directly to their properties without costly experimentation and testing. Some of the possible applications of machine learning to this field include estimating state of charge, predicting performance characteristics, and identifying different regimes and electrochemical behaviors from battery cycling data.