Enhanced Reservoir Characterization of Thin Coal Seams: Application of 2D CNN-Based Seismic Impedance Inversion in Deep CBM Exploration

Abstract Deep coalbed methane (CBM) exploration presents unique challenges in reservoir characterization, particularly when dealing with thin coal seams. This study introduces a novel application of two-dimensional Convolutional Neural Network (2D CNN) based seismic impedance inversion for characterizing the No. 8 coal seam in China's Ordos Basin. By integrating post-stack 3D seismic data with limited well log information, we achieve high-resolution impedance inversion across an 80 km² deep CBM block in the Daji area. The results demonstrate superior accuracy in delineating thin coal seams compared to conventional inversion methods, validated through blind well tests. The resulting high-resolution impedance model enables well data-driven reservoir characterization, including porosity, ash content, and gas content estimation, providing crucial insights for well placement optimization and horizontal well trajectory design.