Development of a Proton Exchange Membrane Fuel Cell Aging Model for Predicting Degradation Under Transient Operating Conditions

Kurzfassung

This thesis investigates the development of a scalable and transferable aging modeling framework for predicting the voltage degradation of polymer electrolyte membrane fuel cells (PEMFCs) under transient operating conditions. A literature review on PEMFC degradation mechanisms and existing aging models was conducted, followed by the selection and extension of the open-source physics based AlphaPEM simulator. Three semi-empirical aging laws describing the degradation of key electrochemical parameters were implemented as a modular aging framework and calibrated using the G20 durability dataset containing 1008 hours of dynamic load cycling. The model was validated by comparing simulated and experimental polarization curves and long-term voltage degradation. To further improve prediction accuracy, a machine learning correction layer based on XGBoost was integrated with the physics-based model to capture transient and cycle-specific degradation effects that could not be represented by the aging laws alone. The hybrid physics-machine learning approach significantly improved the agreement with experimental measurements compared to the standalone physics-based model. The resulting framework provides an extensible and transferable tool for long-term PEMFC durability assessment, requiring only parameter re-identification to adapt to different fuel cell systems without modifying the underlying model structure, thereby supporting the development and commercialization of fuel cell technology for heavy-duty automotive applications.