Scholarlink

<p>Accurate estimation of the State of Health (SOH) of lithium-ion batteries enables battery management systems to effectively monitor battery status, thereby preventing the occurrence of battery safety accidents. To address the challenges of difficult acquisition of complete charge–discharge data and low estimation accuracy under actual operating conditions, this study proposes an SOH estimation method based on time–frequency analysis and charging voltage segments. Health-related features are extracted within the voltage interval with the highest frequency in the charging data, and Discrete Wavelet Transform (DWT) is utilised to perform time–frequency decomposition on the input features. Each decomposed component is transmitted to the Temporal Convolutional Network (TCN) and Bidirectional Long Short-Term Memory (BiLSTM) branches respectively. Concurrently, the Transformer is used to capture global information, and finally, the SOH estimation value is output through the fully connected layer. Relatively accurate estimation of battery SOH can be achieved using only charging voltage data with a length of 0.1&#xa0;V. Validation and analysis on the CACLE dataset demonstrate that the mean absolute error is within 1%. Generalisation verification is completed on the Oxford dataset, indicating that the proposed model exhibits excellent generalisation performance.</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

A method for estimating the SOH of lithium batteries based on DWT-fused neural network and charging voltage segments

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • Hai Tian,
  • Jing Peng,
  • Wei Duan,
  • Wenjie Zhu,
  • Haixin Yu,
  • Luping Dong

摘要

Accurate estimation of the State of Health (SOH) of lithium-ion batteries enables battery management systems to effectively monitor battery status, thereby preventing the occurrence of battery safety accidents. To address the challenges of difficult acquisition of complete charge–discharge data and low estimation accuracy under actual operating conditions, this study proposes an SOH estimation method based on time–frequency analysis and charging voltage segments. Health-related features are extracted within the voltage interval with the highest frequency in the charging data, and Discrete Wavelet Transform (DWT) is utilised to perform time–frequency decomposition on the input features. Each decomposed component is transmitted to the Temporal Convolutional Network (TCN) and Bidirectional Long Short-Term Memory (BiLSTM) branches respectively. Concurrently, the Transformer is used to capture global information, and finally, the SOH estimation value is output through the fully connected layer. Relatively accurate estimation of battery SOH can be achieved using only charging voltage data with a length of 0.1 V. Validation and analysis on the CACLE dataset demonstrate that the mean absolute error is within 1%. Generalisation verification is completed on the Oxford dataset, indicating that the proposed model exhibits excellent generalisation performance.