Quantitative characterization of pore types in laminated and massive oil-bearing shales in Qintong Sag and Jiyang Depression using traditional and deep-learning methods
Abstract
Quantitative characterization of pore types in shale oil formations is crucial for the evaluation of shale oil content and mobility but is challenging. This study presents a novel integration of traditional petrophysical analyses with a U-Net deep-learning approach to quantitatively characterize pore types in laminated and massive oil-bearing shales from the Qintong Sag and Jiyang Depression, China. The methodology overcame the limitations of conventional techniques by enabling simultaneous quantification and classification of pore types (interparticle, intraparticle, organic, and organic-inorganic pores) from scanning electron microscope (SEM) images. Results revealed that massive shales exhibited higher surface porosity than laminated shales. This finding could be attributed to undercompaction effects. For shale samples from the Qintong Sag, the dissolution-induced intraparticle pores were mainly contributed by dolomite and feldspar. However, for shale samples from the Jiyang Depression, the dissolution-induced intraparticle pores were mainly contributed by calcite. Intraparticle pores were more circular than interparticle pores. This study provides helpful support for the qualitative and quantitative evaluation of pores in unconventional reservoirs.
Document Type
Journal Article
Date of Publication
2-1-2026
Volume
152
Issue
1
Publication Title
Journal of Energy Engineering
Publisher
American Society of Civil Engineers
School
School of Engineering
Funders
National Natural Science Foundation of China (42272202, 52264001) / Yunnan Fundamental Research Projects (202501AS070128, 202501CF070116, 202401BE070001-035)
Copyright
subscription content
Comments
Xiong, W., Yuan, Y., Xie, Y., Liu, X., Zhang, D., & Zou, J. (2025). Quantitative characterization of pore types in laminated and massive oil-bearing shales in Qintong Sag and Jiyang Depression using traditional and deep-learning methods. Journal of Energy Engineering, 152(1). https://doi.org/10.1061/JLEED9.EYENG-6421