Key points are not available for this paper at this time.
With the development of seismic acquisition methods, research theories, and application of various interactive interpretation software, the application of seismic technology has developed from structural interpretation to reservoir evaluation 1, lithology interpretation, reservoir description, fluid identification and other aspects. Multi-scale fracture identification driven by seismic data has been widely used in the development of shale oil and gas reservoirs. Due to the introduction of coherent data, artificial intelligence and visual interpretation software in recent years, the ability and accuracy of fault structure identification based on seismic data have been greatly improved. The methods used to predict fractures with seismic data include shear wave splitting detection 2, 3, conversion wave fracture detection 4, longitudinal wave anisotropy detection and post-stack seismic attribute method. Post-stack seismic attributes contain rich lithology and physical property information, which is the most widely used fracture prediction method at present. However, its structural interpretation is mainly based on the seismic waveform characteristics of Inline and Crossline profiles extracted from seismic data. Due to the small fault distance, the characteristics of in-phase axial dislocation in time profiles are not obvious. In addition, part of the time section is missing in the seismic interpretation grid line, which determines that the interpretation precision of micro-structures such as small faults is low. Based on the 3D post-stack seismic data, well logging and 5 horizons interpretation data a shale gas reservoir in southwest China, we carried out the multi-scale fracture identification research. The whole working area is about 200km2, and the fracture development of the Longmaxi Formation buried to a depth of 3700 ~ 4200m is the focus of the study. In order to achieve a more accurate and effective interpretation of small and medium-sized fractures and other micro-amplitude structures, the research on seismic multi-scale fracture identification of small and medium-sized fractures is carried out on the three-dimensional seismic data in the study area. Combined with the seismic and geological data, the fracture display of the whole area is enhanced by structure-oriented filtering. The machine learning method has a good application effect on large fracture identification. The fracture identification algorithm based on CNN (Convolutional Neural Network) image segmentation transforms the problem of image classification into the problem of image segmentation. The CNN network based on U-net is used to identify large-scale fractures. And then d-LoG (Laplacian of Gaussian Operator) coherent sensitive to medium-small scale fractures are extracted. Finally, the identification results of large, medium-small scale fractures are fused to evaluate the development of fractures in the working area.
Qin et al. (Mon,) studied this question.