Application of interactive convolutional neural network micro-fracture prediction technology based on prestack depth migration data in deep shale gas reservoirs Available to Purchase

Micro-fractures in shale gas reservoirs are crucial for structural interpretation and exploration. Accurate prediction relies on high-resolution seismic data, but conventional methods struggle to identify fractures < 20m. This study integrates high-resolution 3D pre-stack depth migration (PSDM) with an interactive convolutional neural network (CNN) to enhance micro-fracture prediction. PSDM improves imaging quality through iterative approximation and well-seismic constraints. The CNN is trained using real drilling data and geological knowledge, enabling automatic micro-fracture prediction. Results show 100% accuracy for fractures > 10m and 78% for those < 10m. The integration of PSDM and interactive CNN reliably identifies micro-fracture spatial distribution, significantly improving efficiency and accuracy compared to manual methods. Future work will expand training data to enhance prediction. This study provides new insights for applying high-resolution seismic imaging and AI in micro-fracture identification.