Abstract: In Huanjiang–Hongde–Yanwu area in the western margin of Ordos Basin, oil reservoirs were widely developed in the eighth member of the Mesozoic Yanchang Formation (Chang 8 member), with huge reserve amount, which is a favorable replacement resource for further exploration and increasing reserves in the region. However, the electrical resistivity of Chang 8 member oil reservoir generally ranges in 3.0–15.0 Ω·m, and the resistivity ratio between oil layer and water layer is lower than 2, showing low contrast ratio, and leading to great difficulty in oil layer identification. The conventional logging interpretation method has a low accuracy, which is unable to meet the needs of high-efficiency reserve increase and capacity construction. Based on a large amount of well drilling, core data, 3D seismic data, and comparative analysis of logging curves, the genesis of Chang 8 member low-resistivity oil reservoir has been studied from the perspectives of structural evolution and mineralization characteristics. The study results indicate that Chang 8 member low-resistivity oil reservoir in the western margin of Ordos Basin was mainly due to high salinity of formation water and high saturation of bound water. By using an artificial neural network model and variables of six key logging parameters, two sensitive parameters have been introduced, i.e., QRw and PC1, and a machine learning model has been established to prepare a cross plot between them, obtaining good distinguishment results among oil layer, oil–water layer, and water layer, with an accuracy of reservoir fluid type recognition improved to 88.9%.

Key words: Ordos Basin; low-resistivity oil layer; high salinity of formation water; artificial neural network