Scholarlink

<p>This study proposes a novel stepped island flow field structure for proton exchange membrane fuel cells (PEMFCs). The optimal island geometry and number of stepped layers were first determined using the control variable method. Subsequently, a Long Short-Term Memory (LSTM) network was employed as an artificial neural network (ANN)-based surrogate model to establish the nonlinear mapping relationship between the geometric parameters and the net power density. The LSTM model was trained using datasets generated from a computational fluid dynamics (CFD) model. A two-level genetic algorithm (GA) strategy was adopted: at the first level, GA was used to optimize the hyperparameters of the LSTM to improve its predictive accuracy; then, the optimized LSTM served as a fast surrogate model for the second-level GA to search for the optimal structural parameters. This two-level GA–LSTM hybrid optimization framework was applied to further refine the stepped island flow field configuration. The optimal geometric parameters were determined as follows: island width 0.4&#xa0;mm, length 2&#xa0;mm, elliptical shape, and three stepped layers. The optimal areas and heights for the first, second, and third steps are 0.92 mm<sup>2</sup>and 0.15&#xa0;mm, 1.413 mm<sup>2</sup> and<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:\:0.32mm,\)</EquationSource> </InlineEquation> and 1.6328 mm<sup>2</sup>and 0.53&#xa0;mm, respectively. Compared with the base case, the optimized stepped elliptical island PEMFC achieves a 5.9% enhancement in net power density. This research demonstrates the reliability of this optimization strategy in improving flow field performance for PEMFCs.</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

Optimization of stepped island flow field for proton exchange membrane fuel cells using GA-LSTM hybrid optimization framework

  • 期刊论文

  • 发表日期：

    2026年6月22日

查看全文
  • Kuan Yang,
  • Shanwu Xu,
  • Ziquan Li,
  • Pengfei Feng

摘要

This study proposes a novel stepped island flow field structure for proton exchange membrane fuel cells (PEMFCs). The optimal island geometry and number of stepped layers were first determined using the control variable method. Subsequently, a Long Short-Term Memory (LSTM) network was employed as an artificial neural network (ANN)-based surrogate model to establish the nonlinear mapping relationship between the geometric parameters and the net power density. The LSTM model was trained using datasets generated from a computational fluid dynamics (CFD) model. A two-level genetic algorithm (GA) strategy was adopted: at the first level, GA was used to optimize the hyperparameters of the LSTM to improve its predictive accuracy; then, the optimized LSTM served as a fast surrogate model for the second-level GA to search for the optimal structural parameters. This two-level GA–LSTM hybrid optimization framework was applied to further refine the stepped island flow field configuration. The optimal geometric parameters were determined as follows: island width 0.4 mm, length 2 mm, elliptical shape, and three stepped layers. The optimal areas and heights for the first, second, and third steps are 0.92 mm2and 0.15 mm, 1.413 mm2 and \(\:\:0.32mm,\) and 1.6328 mm2and 0.53 mm, respectively. Compared with the base case, the optimized stepped elliptical island PEMFC achieves a 5.9% enhancement in net power density. This research demonstrates the reliability of this optimization strategy in improving flow field performance for PEMFCs.