Data-Driven Degradation Modeling in Batteries Using Sparse Feature Selection

Abstract

Battery degradation is a complex, multi-factorial process that significantly influences the performance, safety, and lifespan of modern energy storage systems. In recent years, data-driven approaches have emerged as powerful tools to model and predict battery degradation without the need for complex electrochemical understanding. Among these methods, sparse feature selection techniques provide an efficient pathway to identify the most informative predictors from high-dimensional operational and environmental datasets. This research explores the development of a data-driven battery degradation model using sparse feature selection strategies such as LASSO (Least Absolute Shrinkage and Selection Operator) and Elastic Net regularization. The study examines their ability to enhance model interpretability, predictive performance, and computational efficiency. An experimental evaluation was conducted using real-world cycling data from lithium-ion batteries under varying operational conditions. Results demonstrate that sparse modeling techniques not only reduce the complexity of the model but also maintain high predictive accuracy. These findings highlight the potential of sparse feature-driven approaches in advancing battery health management and lifecycle optimization strategies.