Data-Driven Estimation of Internal Resistance of Lithium-IRON Batteries

Abstract

Battery health prediction is crucial for improving efficiency and longevity, thereby enhancing operational effectiveness. Internal resistance serves as a critical parameter indicative of battery health. This study utilizes Hybrid Pulse Power Characterization (HPPC) tests conducted with CALM CAM72 equipment to assess internal resistance. It proposes a data-driven approach for estimation, employing various regression algorithms such as Linear Regression, Ridge Regression, Lasso Regression, ElasticNet Regression, Decision Tree Regression, RandomForest Regression, GradientBoosting Regression, XGBoost Regression, and LightGBM Regression. The performance of these algorithms is compared to identify the most effective model.

Once the best model is selected, the coefficients from the regression are examined to understand the impact of variables such as State of Charge (SOC), temperature, discharge characteristics, and charging rate (C-rate) on battery health prediction. This analysis aims to provide insights into the factors influencing battery performance, thereby optimizing efficiency and extending battery lifespan. Based on the output of the regression model, an optimal operating range is proposed using data-driven optimization algorithms.