Predicting the Lifespan of Lithium-Ion Batteries: Integrating Initial Data with Machine Learning Approaches

Accurate lithium-ion battery cycle life prediction from early operational data is essential for optimizing manufacturing processes, ensuring product reliability, and enabling proactive maintenance. This study presents a machine learning-based approach that predicts the remaining cycle life using features derived from the first 100 operation cycles, including discharge capacity trends, internal resistance changes, and voltage characteristics. A linear regression model with elastic net regularization was trained on data from 124 cells, with hyperparameters optimized through cross-validation. The model achieved a root mean square error (RMSE) of 210 cycles and an average percentage error of 9.96% on test data, demonstrating strong predictive performance for cycle lives between 500 and 1200 cycles despite underestimations for longer-lived cells due to data distribution imbalances. This work highlights the promise of data-driven methods for early-stage battery life cycle prediction, offering a scalable and interpretable framework to support advancements in battery technology and maintenance strategies.