Evaluating feasibility of batteries for second-life applications using machine learning

TL;DR

This paper introduces a machine learning-based method to quickly assess retired electric vehicle batteries for second-life use or recycling, using simple features from voltage and current data.

Contribution

It develops a Gaussian Process Regression approach with feature selection for fast, accurate battery evaluation across diverse aging datasets, improving second-life decision-making.

Findings

Mean RMSE less than 1.48% in evaluations

Effective feature ranking via correlation analysis

Validated on diverse battery aging datasets

Abstract

This paper presents a combination of machine learning techniques to enable prompt evaluation of retired electric vehicle batteries as to either retain those batteries for a second-life application and extend their operation beyond the original and first intent or send them to recycle facilities. The proposed algorithm generates features from available battery current and voltage measurements with simple statistics, selects and ranks the features using correlation analysis, and employs Gaussian Process Regression enhanced with bagging. This approach is validated over publicly available aging datasets of more than 200 cells with slow and fast charging, with different cathode chemistries, and for diverse operating conditions. Promising results are observed based on multiple training-test partitions, wherein the mean of Root Mean Squared Percent Error and Mean Percent Error performance errors are found to be less than 1.48% and 1.29%, respectively, in the worst-case scenarios.