Scholarlink

<p>Accurate health estimation is critical for the safe and reliable operation of proton exchange membrane fuel cells (PEMFCs). However, achieving precise diagnostics is impeded by the complex, non-linear coupling between macroscopic performance decay and microscopic degradation mechanisms. To address these limitations, this study proposes a physics-informed data-driven framework. Leveraging a validated coupled multi-physics platform, we constructed a multiscale dataset bridging external macroscopic characteristics and internal microscopic degradation parameters, thereby enabling precise internal health characterization. A transformer model was then employed to capture the long-range dependencies of these polarization features and accurately predict the internal metrics. The novel multiscale health index was then constructed using the geodesic distance to quantify any deviation of the current state from the healthy baseline. As such, quantitative mapping between the stack voltage and internal indicators was established, enabling the internal states to be inferred from voltage inputs. The model achieved a normalized root mean squared error (NRMSE) &lt; 0.020 and a coefficient of determination (R²) &gt; 0.990 for degradation-indicator prediction, and the health state assessment attained an NRMSE of 0.009 and an R² of 0.998 against voltage-based benchmarks. Overall, the proposed method provides a solution for conducting a non-invasive assessment of the degradation of internal components within PEMFC 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

State of health prediction for proton exchange membrane fuel cells using multi-scale indicators

  • 期刊论文

  • 发表日期：

    2026年1月17日

查看全文
  • Haitao Min,
  • Xiubing Liu,
  • Xia Sheng,
  • Weiyi Sun,
  • Zhaopu Zhang,
  • Yipeng Lin

摘要

Accurate health estimation is critical for the safe and reliable operation of proton exchange membrane fuel cells (PEMFCs). However, achieving precise diagnostics is impeded by the complex, non-linear coupling between macroscopic performance decay and microscopic degradation mechanisms. To address these limitations, this study proposes a physics-informed data-driven framework. Leveraging a validated coupled multi-physics platform, we constructed a multiscale dataset bridging external macroscopic characteristics and internal microscopic degradation parameters, thereby enabling precise internal health characterization. A transformer model was then employed to capture the long-range dependencies of these polarization features and accurately predict the internal metrics. The novel multiscale health index was then constructed using the geodesic distance to quantify any deviation of the current state from the healthy baseline. As such, quantitative mapping between the stack voltage and internal indicators was established, enabling the internal states to be inferred from voltage inputs. The model achieved a normalized root mean squared error (NRMSE) < 0.020 and a coefficient of determination (R²) > 0.990 for degradation-indicator prediction, and the health state assessment attained an NRMSE of 0.009 and an R² of 0.998 against voltage-based benchmarks. Overall, the proposed method provides a solution for conducting a non-invasive assessment of the degradation of internal components within PEMFC systems.