Application of Multi-attribute Clustering Technology to Description of Fault-Controlled Fracture-Cavity Reservoirs in Changxing Formation, Longgang Area

The Changxing Formation in Longgang area has developed subtle faults and developed fracture-cavity reservoirs in the platform facies. It has great exploration potential. In order to accurately describe the development of fault-controlled fracture-cavity reservoirs, the deep learning algorithm, a kind of machine learning method based on convolutional neural networks, is used to describe the faults. Secondly, the abnormal seismic reflection of fractured and vuggy body is highlighted by using the average value subtraction with principal component analysis method. The gradient structure tensor is selected to identify the boundary of the fractured vuggy body, after the pre-processing with principal component analysis technique. The texture attribute is used to describe the internal structure of fractured vuggy body. The final step is to get the fuzzy C-means clustering result of gradient structure tensor and texture attribute. Eventually, two types of seismic facies closely related to reservoir development are characterized: bead-like and irregular strong amplitude seismic reflection. The fusion display of cluster and deep learning result reflects the relationship between fracture-cavity reservoirs and faults. The analysis shows that the karst cave represented by the beaded reflection is mostly related to the main fault, and the irregular strong amplitude is mostly caused by the secondary fault. The automatic identification method of fracture-cavity reservoirs based on seismic multi-attribute clustering plays an important role in fine description, and lays a foundation for finding favorable oil and gas reservoirs.