Development and validation of a Python-based model for an AEM electrolyser: performance and degradation analysis

The transition to renewable energy sources requires efficient energy storage and conversion systems. Among these, AEM (Anion Exchange Membrane) electrolyzers have emerged as a promising technology for green hydrogen production due to their potential to combine the advantages of both alkaline and PEM electrolyzers. However, their long-term performance and degradation behaviour under real operating conditions remain crucial aspects for industrial-scale deployment. This thesis project, developed in collaboration with Edison S.p.A., presents the modelling of a realistic AEM electrolyzer test bench. In particular, the aim of this Python-based simulation model is to simulate the dynamic behaviour of AEM electrolyzers, integrating both their electrical response and degradation over time. The model simulates hydrogen production based on variable energy inputs from renewable sources, making it a valuable tool for assessing future operational scenarios integration with RES. The degradation mechanisms are evaluated based on experimental tests performed during this thesis in accordance with the Joint Research Center (JRC) protocols. Various modelling techniques, including empirical, semi-empirical, and physics-based approaches, were examined in from the scientific literature. Based on this review, the most suitable formulations and equations were selected to accurately describe the polarization curves, thermal management, and degradation phenomena of the AEM electrolyzer. The model was fine-tuned to reproduce the dynamics of a real AEM electrolyzer operating at the Officine Edison laboratories. The model takes into account the electrical configuration, layout and operational constraints of the test bench, allowing for a faithful reproduction of the system’s behaviour within the simulation. By integrating these practical aspects, the model bridges the gap between theoretical formulations and real-world implementation. The Python model receives as input the energy powering the AEM and calculates the corresponding hydrogen output over time, considering key electrochemical and thermal parameters. The results of the model are then validated by comparing them with experimental data, ensuring its reliability in real-world applications. The validation results confirmed the model effectively predicts the electrolyzer’s behaviour, with a low root mean squared error (RMSE) and mean absolute percentage error (MAPE), particularly for voltage and hydrogen production. Initially, a discrepancy in the temperature sensitivity was identified, attributed to the assumption of a constant polarization curve temperature in the model. Additionally, the analysis revealed a higher-than-expected current prediction, linked to an increased membrane humidification degree, which in turn affected conductivity and efficiency. This behaviour led to the identification of an operational fault in the real electrolyzer, demonstrating the model’s potential even as a diagnostic tool. Afterwards, the fault has been solved by substituting the stack of the electrolyser (a critical component) and the model has been validated. The model successfully approximated hydrogen production, with the cumulative output aligning closely with experimental values. Its accuracy is supported by the statistical metrics results: RMSE of 0.1410 and MPE of 0.1685. This highlights the model’s reliability in predicting AEM electrolyzer performance, making it a valuable tool for scenario analysis, operational optimization, and long-term performance assessment.