Scholarlink

<p>Precisely forecasting the State of Health (SOH) of lithium-ion batteries is essential to enhance vehicle safety and refine battery management systems. This study seeks to address the shortcomings of conventional approaches in feature representation, temporal dependency modeling, and forecasting precision by introducing an attention-enhanced bidirectional long short-term memory (Bi LSTM) model for battery SOH estimation. First, the method selects five key features highly correlated with aging from the battery’s charge–discharge cycles as model inputs and validates their effectiveness using Pearson correlation coefficients to ensure the reliability and representativeness of the input data. During model construction, Bi LSTM is employed to fully exploit the temporal dependencies between features, and the attention mechanism dynamically adjusts feature weights, enabling the model to concentrate on more predictive information. Finally, evaluation and validation using NASA’s publicly accessible lithium-ion battery dataset demonstrate the method’s superior effectiveness in SOH prediction for both individual and similar batteries. Compared to a standard LSTM model, it achieves an average decrease of approximately 1.27% in root mean square error (RMSE) and around 1.26% in mean absolute error (MAE). This improvement enhances the model’s ability to generalize while effectively capturing the battery capacity degradation trend and reducing prediction errors.</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

Attention-based bidirectional LSTM model construction and application for lithium-ion battery state-of-health prediction

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Zhoujian An,
  • Jin Ma,
  • Xiaoze Du,
  • Yong Ding,
  • Dong Zhang,
  • Jian Fu

摘要

Precisely forecasting the State of Health (SOH) of lithium-ion batteries is essential to enhance vehicle safety and refine battery management systems. This study seeks to address the shortcomings of conventional approaches in feature representation, temporal dependency modeling, and forecasting precision by introducing an attention-enhanced bidirectional long short-term memory (Bi LSTM) model for battery SOH estimation. First, the method selects five key features highly correlated with aging from the battery’s charge–discharge cycles as model inputs and validates their effectiveness using Pearson correlation coefficients to ensure the reliability and representativeness of the input data. During model construction, Bi LSTM is employed to fully exploit the temporal dependencies between features, and the attention mechanism dynamically adjusts feature weights, enabling the model to concentrate on more predictive information. Finally, evaluation and validation using NASA’s publicly accessible lithium-ion battery dataset demonstrate the method’s superior effectiveness in SOH prediction for both individual and similar batteries. Compared to a standard LSTM model, it achieves an average decrease of approximately 1.27% in root mean square error (RMSE) and around 1.26% in mean absolute error (MAE). This improvement enhances the model’s ability to generalize while effectively capturing the battery capacity degradation trend and reducing prediction errors.