Scholarlink

<p>Accurate estimation of State of Charge (SoC) and State of Health (SoH) values in battery management systems is crucial for maintaining energy efficiency, ensuring operational safety, and sustaining battery life. This study proposes a hybrid deep learning (DL) architecture that combines multi-scale convolutional neural networks (CNNs), residual blocks, and long-short-term memory (LSTM) units. The proposed model is designed to learn both local spatial patterns and time-dependent relationships found in battery data. The developed model was trained using real battery datasets, and its performance was evaluated comparatively against traditional machine learning (ML) algorithms such as Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GBM). Model performance was analyzed using common regression metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Coefficient of Determination (R²). The results show that the proposed model achieves lower prediction errors compared to the methods evaluated (MAE = 0.1507, MSE = 0.2030, R² = 0.9998). Furthermore, scatter plots and residual distribution analyses reveal that the model predictions are quite close to the actual values and that the error distribution does not contain any significant systematic bias. Overall, the results demonstrate that the proposed hybrid architecture can produce accurate and stable results for SoC estimation and offers an approach that could contribute to the development of reliable battery management systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid multi-scale CNN-Residual-LSTM approach for robust state-of-charge estimation in lithium-ion batteries

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月22日

查看全文
  • Saadin Oyucu,
  • Onur Polat,
  • Muammer Türkoğlu,
  • Faruk Kürker,
  • Ahmet Aksöz,
  • Emre Biçer

摘要

Accurate estimation of State of Charge (SoC) and State of Health (SoH) values in battery management systems is crucial for maintaining energy efficiency, ensuring operational safety, and sustaining battery life. This study proposes a hybrid deep learning (DL) architecture that combines multi-scale convolutional neural networks (CNNs), residual blocks, and long-short-term memory (LSTM) units. The proposed model is designed to learn both local spatial patterns and time-dependent relationships found in battery data. The developed model was trained using real battery datasets, and its performance was evaluated comparatively against traditional machine learning (ML) algorithms such as Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GBM). Model performance was analyzed using common regression metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Coefficient of Determination (R²). The results show that the proposed model achieves lower prediction errors compared to the methods evaluated (MAE = 0.1507, MSE = 0.2030, R² = 0.9998). Furthermore, scatter plots and residual distribution analyses reveal that the model predictions are quite close to the actual values and that the error distribution does not contain any significant systematic bias. Overall, the results demonstrate that the proposed hybrid architecture can produce accurate and stable results for SoC estimation and offers an approach that could contribute to the development of reliable battery management systems.