Scholarlink

<p>Reliable prediction of lithium-ion battery capacity is important for evaluating battery degradation and overall performance. However, many existing methods rely on direct capacity measurements, make limited use of multi-source health information, and do not effectively capture both local characteristics and long-term degradation trends. The model employs multi-scale convolutional layers to extract local features, a transformer block with trend-aware nonlinear positional encoding to model long-term temporal dependencies, and a dual-output prediction strategy to improve prediction stability. Experimental results on the NASA battery dataset show that the proposed method achieves higher accuracy and robustness compared to several comparison models. Ablation studies and sensitivity analysis further confirm the contribution of each model component.</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

An improved CNN-transformer for lithium-ion battery capacity prediction with multi-source indirect health indicators

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Hairong Wang,
  • Ying Chen,
  • Yuxuan Lin,
  • Dequn Zhou

摘要

Reliable prediction of lithium-ion battery capacity is important for evaluating battery degradation and overall performance. However, many existing methods rely on direct capacity measurements, make limited use of multi-source health information, and do not effectively capture both local characteristics and long-term degradation trends. The model employs multi-scale convolutional layers to extract local features, a transformer block with trend-aware nonlinear positional encoding to model long-term temporal dependencies, and a dual-output prediction strategy to improve prediction stability. Experimental results on the NASA battery dataset show that the proposed method achieves higher accuracy and robustness compared to several comparison models. Ablation studies and sensitivity analysis further confirm the contribution of each model component.