Scholarlink

<p>Efficient design of proton exchange membrane fuel cells (PEMFCs) requires balancing high electrical output with thermal stability, yet the complex interactions among operating parameters make this a challenging task. Addressing this gap, this study develops an integrated predictive–optimization–decision framework that systematically models PEMFC performance, explores trade-offs, and guides application-specific design choices. The primary innovation lies in combining multi-layer perceptron neural networks (MLPNN) with metaheuristic optimization, particle swarm optimization (PSO), modified particle swarm optimization (MPSO), multi-objective Harris hawks optimization (MOHHO), and multi-objective PSO (MOPSO), followed by decision-making using the additive ratio assessment (ARAS) method. Predictive modeling results demonstrate variable-specific advantages of optimization strategies: PSO-MLPNN yielded superior accuracy for electrical power output prediction (MAPE = 0.233%), while MPSO-MLPNN achieved marginally better accuracy for cell temperature prediction (MAPE = 0.301%). Multi-objective optimization revealed the inherent trade-off between power and temperature, with MOHHO providing broader Pareto fronts and greater diversity than MOPSO. Optimal operating conditions (ST<sub>An</sub> ≈ 2.0–2.15, ST<sub>Ca</sub> ≈ 2.1–2.3, RH<sub>Ca</sub> ≈ 60–66%, T<sub>in</sub> ≈ 26&#xa0;°C) enabled peak power outputs near 5300 mW while maintaining stable cell temperatures around 39.5&#xa0;°C. Finally, ARAS-based decision analysis identified seven design scenarios. The scenario with balanced weights yielded a cell power output of 5205.9 mW, representing an increase of approximately 6.94% compared to the mean cell power of 4867.9 mW in the dataset. The corresponding cell temperature was 39.53&#xa0;°C, which is about 20.3% lower than the mean cell temperature of 49.61&#xa0;°C. These results demonstrate the proposed framework’s ability to provide flexible and application-specific design strategies, simultaneously enhancing electrical performance and maintaining thermal stability and safety.</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

Integrated neural network and metaheuristic algorithms for balancing electrical performance and thermal safety in PEMFC design

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月28日

查看全文
  • Naim Ben Ali,
  • Borhen Louhichi,
  • As’ad Alizadeh,
  • Mohamed Shaban,
  • Narinderjit Singh Sawaran Singh,
  • Abdellatif M. Sadeq,
  • Walid Aich,
  • Khalil Hajlaoui

摘要

Efficient design of proton exchange membrane fuel cells (PEMFCs) requires balancing high electrical output with thermal stability, yet the complex interactions among operating parameters make this a challenging task. Addressing this gap, this study develops an integrated predictive–optimization–decision framework that systematically models PEMFC performance, explores trade-offs, and guides application-specific design choices. The primary innovation lies in combining multi-layer perceptron neural networks (MLPNN) with metaheuristic optimization, particle swarm optimization (PSO), modified particle swarm optimization (MPSO), multi-objective Harris hawks optimization (MOHHO), and multi-objective PSO (MOPSO), followed by decision-making using the additive ratio assessment (ARAS) method. Predictive modeling results demonstrate variable-specific advantages of optimization strategies: PSO-MLPNN yielded superior accuracy for electrical power output prediction (MAPE = 0.233%), while MPSO-MLPNN achieved marginally better accuracy for cell temperature prediction (MAPE = 0.301%). Multi-objective optimization revealed the inherent trade-off between power and temperature, with MOHHO providing broader Pareto fronts and greater diversity than MOPSO. Optimal operating conditions (STAn ≈ 2.0–2.15, STCa ≈ 2.1–2.3, RHCa ≈ 60–66%, Tin ≈ 26 °C) enabled peak power outputs near 5300 mW while maintaining stable cell temperatures around 39.5 °C. Finally, ARAS-based decision analysis identified seven design scenarios. The scenario with balanced weights yielded a cell power output of 5205.9 mW, representing an increase of approximately 6.94% compared to the mean cell power of 4867.9 mW in the dataset. The corresponding cell temperature was 39.53 °C, which is about 20.3% lower than the mean cell temperature of 49.61 °C. These results demonstrate the proposed framework’s ability to provide flexible and application-specific design strategies, simultaneously enhancing electrical performance and maintaining thermal stability and safety.