Milestone-Präsentation von Ivan Buntic am 23. März 2023

7. März 2023 /

Ivan Buntic hält am Donnerstag, den 23. März 2023 seine Milestone-Präsentation mit dem Titel "Data-driven optimisation algorithms for local dynamic model adaptivity"

Ivan Buntic ist akademischer Mitarbeiter am Lehrstuhl für Hydromechanik und Hydrosystemmodellierung und Mitglied im SC SimTech. Am Donnerstag, den 23. März 2023 um 11.00 Uhr wird er seine Milestone-Präsentation mit dem Titel "Data-driven optimisation algorithms for local dynamic model adaptivity" halten.

Datum: Donnerstag, 23. März 2023
Zeit: 11.00 Uhr
Titel: "Data-driven optimisation algorithms for local dynamic model adaptivity"
Ort: Pfaffenwaldring 61, U1.003 (MML)

Abstract

Hydrogen promises to be an emerging option for a climate friendly energy carrier of the near future. Therefore, it is important to think about and investigate storage scenarios and the interactions of hydrogen with its surroundings while being stored. Possible storing options include depleted gas fields, salt caverns and aquifers though underground aquifers offer vast amounts of volume for storage and are therefore of central interest to us. Additionally, hydrogen interacts with hydrogen-reducing microbials which can be found in these aquifers. In order to obtain a better understanding of this interaction, together with partners from the TU Delft, we conduct experiments for measuring contact angles in a hydrogen-stone-"living brine" system.

While these experiments constitute a part of the doctoral study, its main goal is to investigate the spreading of hydrogen when being injected into and extracted from an aquifer. Although the Darcy equations were established almost two centuries ago to describe flow through porous media, we intend to couple Darcy models of different dimensionalities in order to create a multi-scale model which balances accuracy and efficiency. Specifically, a full-dimensional Darcy model is coupled to the so-called Vertical Equilibrium (VE) model, the system is then solved in an implicit manner. Besides that, local criteria are introduced to automate the optimal distribution of the models in the domain. To improve the speedup even more, data-driven machine learning approaches are utilized to learn analytic functions for the VE model instead of solving local non-linear equations.

Kontakt

Dieses Bild zeigt Lehrstuhl für Hydromechanik und Hydrosystem-modellierung

Lehrstuhl für Hydromechanik und Hydrosystem-modellierung

 
 

Anfahrt

Pfaffenwaldring 61, 70569 Stuttgart

Zum Seitenanfang