PORE 2-3 Machine Learning application in Porous Media: from image processing to multiscale models

Course Dates & Time: June 24, June 26, July 1 and July 3 from 18:00-21:00 British Summer Time

Please note that this is an online course. The Zoom link and all other course-related information can be found within the individual lessons, which registered participants will be able to access through their user account.

Lecturer: Gege Wen

Gege is an Assistant Professor at Imperial College London, co-appointed by Earth Science Engineering and I-X (Imperial AI). 

Her research interest is developing computational methods for Earth and environmental science problems to help fulfill society’s energy needs and transition toward a low-carbon future. She specializes in (1) multiphase flow and transport for CO2 geological storage, (2) sustainable subsurface energy storage, and (3) ML for scientific computing. 

Gege obtained her Ph.D. from the Energy Sciences & Engineering Department at Stanford Doerr School of Sustainability with Professor Sally M. Benson. She received her Master’s degree in Fluid Mechanics and Hydrology from Civil and Environmental Engineering at Stanford University. Prior to her Ph.D. at Stanford, she received her Bachelor’s degree with honors from Lassonde Mineral Engineering at the University of Toronto.

Gege is the creator of CCSNet.ai, an AI-based physics simulation tool for CO2 geological storage modeling.

• Google Scholar: https://scholar.google.com/citations?hl=en&user=WTEXROwAAAAJ
• LinkedIn: https://www.linkedin.com/in/gege-wen-6234a963

Course Description:

This course explores the intersection of machine learning and porous media research, providing participants with theoretical foundations and practical applications. Students will learn how AI models can accelerate porous media analysis through advanced image processing techniques, including segmentation, super-resolution, and synthetic image generation. The course covers parameter estimation from imaging data and dynamic modeling of transport phenomena in complex porous structures.

Key machine learning architectures covered include Convolutional Neural Networks (CNN) for image analysis, Graph Neural Networks (GNN) for structural representation, Physics-Informed Neural Networks (PINN) for incorporating physical constraints, and generative AI architectures.

Dates: The course will take place on the following days and times:

• Tuesday, June 24th, 2025                  (18:00-21:00 British Summer Time)
• Thursday, June 26th, 2025                (18:00-21:00 British Summer Time)
• Tuesday, July 1st, 2025                      (18:00-21:00 British Summer Time)
• Thursday, July 3rd, 2025                   (18:00-21:00 British Summer Time)

Important note: All lectures will be offered live, and participants are expected to attend all sessions in order to be granted a certificate of attendance. Recordings of the lectures will be provided in most cases within 24 hours after each session. These recordings will be available for 1 month following course completion. Please note that sharing the recordings with others is not permitted.

When you register for a course, your name and email address will be used by the InterPore Office and shared with the course lecturer(s) to facilitate direct communication and provide course updates before it begins. Course access information and further details will be sent to all participants the week before the course starts. For questions, please contact Margaret Dieter at margaret.dieter@interpore.org.