Development of a Digital Twin Model for a Hydrogen Production Plant Based on Artificial Intelligence Techniques

Abstract

The global energy transition has driven the adoption of green hydrogen as a key energy vector for industrial decarbonization and the integration of renewable energy sources. This work presents the development of a Digital Twin model applied to a Proton Exchange Membrane (PEM) hydrogen production plant, integrating artificial intelligence techniques for simulation, prediction, and adaptive updating. The proposed model integrates multi-domain variables —specifically electrical, thermal, and hydraulic data acquired from the National University of Engineering’s experimental facility—with artificial neural network (ANN) algorithms to predict hydrogen yield and overall system efficiency. The methodology includes data acquisition, preprocessing, normalization, and incremental learning stages, allowing the digital twin to adapt to new operating conditions and account for data drift. The results demonstrate a high correlation between real and estimated values, confirming the accuracy and robustness of the model under different operating scenarios. Therefore, the results confirm that the proposed digital twin represents a reliable tool for monitoring, prediction, and performance analysis of hydrogen production plants.

Aurelio Morales Villanueva

Professor of Reconfigurable Computing

My research interests include reconfigurable computing, computer architecture and embedded systems.