Ensemble Learning for Precise State-of-Charge Estimation in Electric Vehicles Lithium-Ion Batteries Considering Uncertainty
Accurate state-of-charge (SoC) estimation is crucial for enhancing the performance, longevity, safety, and reliability of lithium-ion batteries (LiBs) in electric vehicles (EVs). This study presents a comprehensive machine learning (ML)-based approach for SoC estimation of EV LiBs, addressing the ch...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10904247/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate state-of-charge (SoC) estimation is crucial for enhancing the performance, longevity, safety, and reliability of lithium-ion batteries (LiBs) in electric vehicles (EVs). This study presents a comprehensive machine learning (ML)-based approach for SoC estimation of EV LiBs, addressing the challenges of model reliability, uncertainty, and real-world data variability. To ensure the model’s robustness and generalizability, preprocessing techniques, including normalization and scaling, were employed alongside rigorous cross-validation methods. A well-structured ML pipeline was developed to integrate these processes, optimizing the entire model development cycle for efficiency and practical implementation. In the ML pipeline, we utilized Extra Trees Regressor (ETR) and Light Gradient Boosting Machine (LightGBM) and proposed an ensemble model, combining the strengths of ETR and LightGBM, namely ETR-GBM. We benchmarked the model’s performance against other ML models, such as CatBoost and Random Forest (RF). Under uncertain conditions, the proposed model emphasized its reliability and robustness, and its conclusions underscored the efficacy of the SoC estimation approach. The ETR-GBM consistently outperforms the individual models (ETR, LightGBM, XGBoost, CatBoost, Support Vector Regression (SVR), Random Forest (RF), and Bayesian) when noise is added to the training dataset. With a noise standard deviation of 0.1, the ETR-GBM demonstrated superior performance, achieving a Root Mean Square Error (RMSE) of 0.41%, surpassing the individual models, which exhibited RMSE values ranging from 0.85% to 0.91%. |
|---|---|
| ISSN: | 2169-3536 |