Data-Driven Optimization of Recyclable Battery Cells for Electrified Systems Using Machine Learning Regression Models

This work presents a data-driven framework for optimizing and evaluating recyclable battery cells for electrified systems using machine learning (ML) techniques. A custom battery tester circuit is designed to measure critical parameters that are voltage, current, capacity, internal resistance, and temperature. This work explores the prediction of the remaining useful life (RUL) of battery cells. Several regression algorithms are evaluated for RUL, where the Random Forest model achieved the best performance with the lowest Mean Absolute Error (MAE) and strong generalization across test sets, showing effectiveness in forecasting battery lifespan. The measured parameters of the recyclable battery are used to train multiple supervised ML models to estimate the battery’s State of Charge (SoC) and State of Health (SoH). This study compares several regression algorithms for the dataset of the recyclable battery cells. Results showed that Random Forest and Gradient Boosting achieved the lowest MAE, confirming their robustness and accuracy for SoC estimation under variable thermal conditions. A hybrid numerical approach is used to estimate the SoH, which combines capacity fade and resistance growth and provides an effective unsupervised degradation analysis. This ML study helps in forecasting the performance of the recyclable battery cells that are configured for the battery packages in the electrified system applications.