Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 20: Data Driven Material Science: Big Data and Workflows III
MM 20.4: Vortrag
Dienstag, 19. März 2024, 11:00–11:15, C 243
Predicting Equilibrium Pressure for Hydrogen Storage: A Cheminformatics Approach Using Deep Neural Networks — •Sinan S. Faouri1,2, Kai Sellschopp2, Paul Jerabek2, and Claudio Pistidda2 — 1Applied Science private University — 2Helmholtz-Zentrum hereon
Hydrogen storage is a critical aspect of hydrogen-based energy systems, and predicting the equilibrium pressure during storage processes is essential for optimizing storage conditions. In this study, we employ a cheminformatics approach by extracting a comprehensive set of descriptors, including electronegativity, electron affinity, atomic radius, thermal conductivity, and more, to characterize the hydrogen storage process. These descriptors serve as inputs for deep neural networks (DNNs) to predict the equilibrium pressure. We compare the performance of the DNN model against three other machine learning models to assess its predictive capabilities. The evaluation metrics of all four models are thoroughly examined and compared, providing insights into their respective strengths and weaknesses. This comparative analysis aims to elucidate the effectiveness of the cheminformatics-driven DNN approach in predicting equilibrium pressure for hydrogen storage, contributing to the advancement of efficient and reliable hydrogen storage technologies. The findings of this study have broader implications for the development and optimization of hydrogen-based energy systems.
Keywords: Hydrogen storage; Deep neural networks; Cheminformatics descriptors; Machine learning; Enthalpy