Intelligent Gas-bearing Prediction of Tight Sandstone Reservoirs Guided by Geological Accumulation Characteristics and Well-Seismic Information

Prediction of gas-bearing in tight sandstone reservoirs is a significant challenge in seismic exploration due to their complex geological conditions. The application of traditional seismic inversion methods in high-water-saturation tight sandstone reservoirs has shown limited effectiveness. The seismic response characteristics of gas-bearing and water-bearing layers are often similar. To address these issues, this study proposes an intelligent prediction workflow for gas-bearing in tight sandstone reservoirs, guided by geological accumulation characteristics and well-seismic information. We focus on investigating how geological accumulation characteristics (e.g., thickness of hydrocarbon source rocks and dip attributes) can guide well-seismic information (e.g., frequency gradient attributes, impedance, gamma ray, low-frequency gas-bearing trends, and relative geological time attributes) to improve the accuracy of gas-bearing prediction. These features are used as inputs to the Transformer model, with well log gas-bearing interpretation results serving as labels. By incorporating the constraints from lithology inversion results, we achieve gas-bearing property prediction for high-water-saturation tight sandstone reservoirs. The field data test shows that geological accumulation characteristics, as a new source of information, can compensate for the limitations of deep learning models caused by insufficient training datasets and a lack of diversity in gas-sensitive attributes. The gas prediction results demonstrate high accuracy and geological consistency.