Scholarlink

<p>Proton Exchange Membrane Fuel Cells (PEMFCs) offer clean power genera-tion with high efficiency and low emissions. However, their large-scale use is limited by performance degradation, parameter uncertainty, and the difficulty of adapting traditional models to changing operating conditions. This study proposes a hybrid modelling framework that combines physics-based equations with an intelligent parameter adaptation layer and uncertainty analysis. The complete system is developed and tested in MATLAB/Simulink. The core physical model includes the Nernst potential, activation, ohmic, and concentration losses, as well as dynamic mass and heat balances. Degradation is represented through catalyst activity decay and membrane resistance growth, while uncertainty is assessed using Monte Carlo simulation. The hybrid framework improves predictive accuracy and operational robustness by adjusting key parameters online as conditions change. Compared with a conventional physics-only PEMFC model, the proposed approach reduces the root-mean-square prediction error by 65%. It also maintains stable performance when integrated into a hybrid renewable energy system. The results show that this framework is a reliable and adaptable tool for energy management, predictive maintenance, and real-time control of PEMFC applications.</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

Hybrid physics AI enhanced PEM fuel cell modelling with real-time degradation and uncertainty quantification

  • 期刊论文

  • 发表日期：

    2026年2月4日

查看全文
  • Avinash Mahajan,
  • Nischal Mungle,
  • Archana Mungle,
  • Rajesh Bodkhe

摘要

Proton Exchange Membrane Fuel Cells (PEMFCs) offer clean power genera-tion with high efficiency and low emissions. However, their large-scale use is limited by performance degradation, parameter uncertainty, and the difficulty of adapting traditional models to changing operating conditions. This study proposes a hybrid modelling framework that combines physics-based equations with an intelligent parameter adaptation layer and uncertainty analysis. The complete system is developed and tested in MATLAB/Simulink. The core physical model includes the Nernst potential, activation, ohmic, and concentration losses, as well as dynamic mass and heat balances. Degradation is represented through catalyst activity decay and membrane resistance growth, while uncertainty is assessed using Monte Carlo simulation. The hybrid framework improves predictive accuracy and operational robustness by adjusting key parameters online as conditions change. Compared with a conventional physics-only PEMFC model, the proposed approach reduces the root-mean-square prediction error by 65%. It also maintains stable performance when integrated into a hybrid renewable energy system. The results show that this framework is a reliable and adaptable tool for energy management, predictive maintenance, and real-time control of PEMFC applications.