Ensemble Learning for Precise State-of-Charge Estimation in Electric Vehicles Lithium-Ion Batteries Considering Uncertainty

Estimating the State of Charge (SoC) of lithium-ion batteries with high accuracy is essential for electric vehicle battery management systems. Due to various internal and external uncertainties such as temperature, load variations, and non-linear battery behavior, precise SoC estimation remains a challenging task. This project focuses on developing a robust solution using ensemble learning techniques to enhance SoC prediction accuracy. Multiple regression algorithms are explored, including Extra Trees Regressor (ETR), XGBoost, CatBoost, and Support Vector Machine (SVM). The core contribution is the proposed ensemble-based Stacking Regressor model that combines Random Forest, Decision Tree, and Linear Regression as base learners. The integration of these models allows for capturing complex patterns and reducing prediction error by leveraging the strengths of each individual model. The dataset includes features such as voltage, current, temperature, and time, which are vital for SoC estimation. The aim is to create a learning-based framework that improves accuracy and reduces model uncertainty using machine learning.

Keywords: State of Charge (SoC), Lithium-Ion Battery, Ensemble Learning, Stacking Regressor, Machine Learning, SoC Estimation, Regression Models, Battery Management System (BMS), Model Uncertainty, Feature Engineering.