Identification of Diagenetic Facies in Low-Permeability Sandstone Reservoirs Based on Self-Organizing-Map Neural Network Algorithm

Quantitative identification of diagenetic facies is critical for favorable reservoir prediction. In this study, the diagenetic facies of the Chang-8 reservoir in the Zhenbei area of the Ordos Basin was investigated using an integrated analysis of casting thin sections, scanning electron microscopy (SEM) and X-ray diffraction (XRD). The Chang-8 reservoirs can be subdivided into five major diagenetic facies categories: 1) weakly-dissolved chlorite cemented facies, 2) moderately-compacted mineral dissolution-susceptive facies, 3) moderately-dissolved kaolinite-bearing facies, 4) moderately-compacted carbonate cemented facies, and 5) strongly-compacted tight sandstone facies. On the basis of the above analyses, the diagenetic facies were identified from well logs by involving the supervised-mode self-organizing-map neural network (SSOM) algorithm. Six wireline logs sensitive to the diagenetic facies characteristics were used as the model input, the diagenetic facies prediction model was built using SSOM. The prediction results of the diagenetic facies are in good agreement with the core analysis types, with a matching of 83.87%. Our work also sheds light on reservoir typing by linking the diagenetic facies with reservoir quality and oil testing data.