Study on the Adaptability of Geological Models for Complex Carbonate Gas Reservoirs

There are many technical difficulties in constructing a three-dimensional geological model for the Feixianguan Formation gas reservoir in the Tieshanpo area of the Sichuan Basin, which is characterized by its high and steep structure, large reservoir thickness and strong heterogeneity, and complex gas water relationship. This has affected the formulation of technical strategies for gas reservoir development. Using data such as rock cores, logging, geophysics, well testing, and production performance, the traditional 3D geological modeling approach involves: depositional facies modeling to determine reservoir environments and frameworks, integrating local and global modeling for fluid classification, detailed characterization of spatial heterogeneity in different reservoir types, and combining deterministic and stochastic methods to establish the geological model. The deep learning geological modeling method based on intelligent gas fields in this study is based on the basic idea of deep learning geological modeling, considering the principle of generative adversarial network geological modeling, and establishing a conditional generative adversarial network geological model: ① Achieved 3D spatial prediction and quantitative characterization of geological features, with deep learning-based interwell predictions better conforming to geological patterns; ② Established a depositional microfacies model identifying major reservoir depositional environments; ③ Main reservoirs are distributed in Cycles I, II, and III, evolving from two sets to one multi-layered stack from east to west, with the P5 well area showing optimal quality; ④ The gas reservoir lacks a unified gas-water contact, with local water bodies and bottom water developed in the P2 and P1 well areas respectively; ⑤ The P5 well area of the gas reservoir has the largest reserve scale and abundance, followed by the P2 and P1 well areas respectively. Subsequent actual drilling and model reserves have confirmed the reliability of the established model, which can be used for numerical simulation prediction and provides support for optimizing gas reservoir development methods. This method and research results have reference significance for establishing 3D geological models and adaptability evaluation of similar complex carbonate gas reservoirs.