Geometrical characterization of the multiscale pore structure in clay-rich shale: A machine learning approach

AbstractThe pore structure of shale plays a critical role in hydrocarbon storage and flow; however, the microscale complexity of Gulong shale remains poorly characterized, limiting efficient resource exploitation. To address this, argon-ion polishing and field emission scanning electron microscopy were applied, combined with random forest-based image segmentation, to quantify the pore structure. Results show that inorganic pores dominate the shale, primarily as nanopores (∼0.18 μm), acting as the main pathways for fluid migration but prone to collapse, highlighting the importance of clay-fluid interaction studies and optimized fracturing designs. Organic pores are largely isolated (∼0.16 μm) with limited connectivity, suggesting the need for stimulation to enhance hydrocarbon flow. Micro-fractures are mostly interlayer bedding fractures (∼1.06 μm in width), emphasizing the role of advanced stimulation strategies. By integrating high-resolution imaging with machine learning-enhanced segmentation, this study provides a comprehensive geometric dataset for Gulong shale. These findings advance understanding of the microstructure of clay-rich shales and provide practical metrics for reservoir modeling in shale oil exploitation.