Total Differential Capacity Plot Analysis Using Data Science Methods

Abstract

Differential capacity plots can be a powerful tool for uncovering battery performance characteristics buried within large datasets of charge-discharge curves. Due to the difficulty in analyzing these datasets in their entirety, arbitrarily chosen subsets of cycles are typically reported in the literature and used to draw qualitative conclusions describing the electrochemical changes that drive the peak shifts. Herein, open-source software we developed to quantitatively analyze entire cycling datasets is discussed. Peak features, such as peak locations and areas, are extracted by individually fitting each of the differential capacity plots. We implemented a database that allows users to perform this analysis, save model fits, and return to their data at a later point. Further, we demonstrate the ability to differentiate between two battery chemistries using peak characteristics and a support vector classifier, with an accuracy of 77%. This work provides the framework for more in-depth, quantitative analyses of differential capacity data.