Speaker
Description
Capillary pressure plays a crucial role in multiphase transport and has applications in carbon dioxide sequestration and underground hydrogen storage. Characterizing it is challenging when rock samples are unavailable; thus, it is often estimated using the J function, but the scaled results are scattered. This presentation discusses a new approach for estimating capillary pressure using the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) to capture the complex relationships between capillary pressure, pore structure, permeability, and porosity. First, the study used Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to cluster 118 rock samples recovered from depths ranging from 161 to 16,678 ft below the surface. Next, it converted their capillary pressure, permeability, and porosity measurements into images by introducing the concept of a constrained capillary pressure image. Then, it augmented the constrained images to increase their number to 1,166. Later, it designed a conditional WGAN-GP, with its hyperparameters tuned by analyzing the quality of generated images and loss values. The images generated in the best scenario showed no mode collapse; thus, they were converted back into capillary pressure measurements. The capillary pressures generated in the best scenario exhibit key characteristics of tight gas sandstones, such as partial percolation and nonzero entry pressure. The results are interesting and have applications in characterizing capillary pressure far from the wellbore.
| Country | USA |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
