.png){kind=icon}
.png){kind=icon}
Return
A Reliable and Explainable Ensemble Framework for Lithium-Ion Battery Charge Estimation
.png){kind=icon}
PRE
AI
B
S
A
F
İ
E
DOI：10.1002/est2.70450.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Li-ion batteries are used in many electronic devices due to their advantages such as high charge capacities, convenient charging times, and long life. Furthermore, Li-ion batteries are especially preferred for long range in electric vehicles. With the widespread use of electric vehicles, State of Charge (SoC) estimation, which is one of the most important components of battery management systems, is becoming increasingly important. This study proposes an interpretable and robust framework for SoC estimation of Li-ion batteries using ensemble learning and explainable artificial intelligence techniques. A Voting Regression (VR) approach is developed by combining Linear Regression (LR), Random Forest (RF), and Gradient Boosting (GB) models under both uniform and weighted configurations. The weighted VR model is particularly designed to leverage the strengths of high-performing estimators by assigning greater importance to stronger models. Experiments are conducted on a discharge subset of the NASA Battery Dataset, using voltage, current, temperature, and load-related features. The results show that the Random Forest model achieves the highest prediction accuracy, with an RMSE of 0.874, MAE of 0.605, and R2 of 0.999, while the proposed weighted VR model provides competitive performance with improved robustness and flexibility. To enhance interpretability, SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) are also utilized to analyze feature contributions. The analysis reveals that voltage-related features are the dominant factors influencing SoC estimation, while temperature plays a secondary role and current-related features have minimal impact. Overall, the study highlights the effectiveness of ensemble learning combined with explainable AI for providing accurate and interpretable SoC estimating for battery management applications.
Keywords:
explainable AI (XAI)
Gradient Boosting
LIME
Linear Regression
Random Forest
SHAP
state of charge (SoC) prediction
Voting Regression
Journal
E
IF:
4
Papers:
984
Citations:
2.2K
Organization
Cited Papers