Three-dimensional Geological Modeling of X Block in Songliao Basin

Abstract

To address the bottlenecks in the Songliao Basin—namely, the high multi-solution nature of deep tight source-reservoir systems and the difficulty of cross-scale characterization—this paper proposes a three-dimensional, multi-scale modeling approach that integrates geologic facies-controlled constraints with a depth-generating network. At the macroscale, this approach overcomes the limitations of single-well data by strongly integrating inorganic geochemistry, sedimentary facies, and 3D seismic data. Using a sedimentary facies grid as a constraint medium and applying sequential indicator and sequential Gaussian algorithms for mutual property inference, it overcomes the bottleneck of sparse sampling to accurately reconstruct the 3D spatial distribution of effective hydrocarbon source rocks in the target intervals; At the microscopic level, to address the challenge of topological connectivity loss under sparse sampling, a three-way multi-discriminator adversarial network incorporating multi-scale channel attention was introduced. By applying rigid geometric constraints via independent discriminators along the X, Y, and Z axes, the model faithfully reconstructed the directionality and connectivity characteristics of the microscopic pore skeleton. Applications in the Songliao Basin work area demonstrate that this model bridges the information gap between macroscopic facies-controlled evolution patterns and microscopic intelligent reconstruction, providing a feasible, high-precision approach for the three-dimensional quantitative prediction of deep source rocks and reservoirs.