A two-branch convolutional neural network for pre-stack elastic parameter inversion of coal-bearing gas reservoirs

Abstract

Accurate estimation of Poisson’s ratio is essential for the characterization of coal-bearing gas reservoirs, particularly in thin-bed and low signal-to-noise ratio environments where conventional elastic impedance (EI) inversion suffers from wavelet interference, limited resolution, and reliance on linearization assumptions. To address these limitations, we develop a physics-guided two-branch convolutional neural network (TB-CNN) that directly predicts Poisson’s ratio by jointly integrating EI-inverted P-wave velocity, S-wave velocity, and density with small-, medium-, and large-angle seismic stacks. The first branch provides geologically consistent, physics-informed background trends, while the second branch captures thin-bed-sensitive reflectivity features and amplitude tuning effects. The fused latent representation is explicitly regularized using empirical rock-physics relationships to ensure physical plausibility and enhanced generalization. Field validation on the 8# coal seam of the Ordos Basin demonstrates that the proposed TB-CNN improves vertical resolution, sharpens seam boundary delineation, and better preserves thickness variations compared with EI inversion and a single-branch CNN. Near-well comparisons show higher correlation with log-derived Poisson’s ratio, while lateral slices reveal improved continuity and thin-layer detectability. These results confirm that combining physics-guided stability with data-driven resolution provides a robust and interpretable framework for Poisson’s ratio inversion in thin coal seams and holds promise for broader applications in unconventional gas reservoir prediction.

Keywords

Coalbed methane / Elastic parameter inversion / Pre-stack seismic data / Convolutional neural network

Funding

This study was supported by the commissioned project “Development of Seismic Prediction Methods for Coal and Rock Gas Reservoirs Based on Wave Theory” (Contract No. KFKT2024-29), funded by the Exploration and Development Research Institute of Changqing Oilfield Branch, CNPC, and Chengdu University of Technology.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Author contributions

Conceptualization: Fei Li, Mengbo Zhang, Qiang Liang Formal analysis: Yongheng Zhang, Mengbo Zhang, Qingzhou Zhang Investigation: Fei Li, Xiaojie Cui, Na Ni Methodology: Mengbo Zhang, Qingzhou Zhang Visualization: Xiaojie Cui, Yongheng Zhang Writing–original draft: Yongheng Zhang, Qiang Liang, Na Ni Writing–review & editing: Fei Li, Mengbo Zhang, Qiang Liang

Availability of data

Not applicable.

References

[1] Shuey RT. A simplification of the Zoeppritz equations. Geophysics. 1985; 50(4): 609-614. doi: 10.1190/1.1441936

[2] Widess MB. How thin is a thin bed? Geophysics . 1973; 38(6): 1176-1180. doi: 10.1190/1.1440403

[3] Ristow D , Rühl T. Fourier finite—difference migration. Geophysics. 1994; 59(12): 1882-1893. doi: 10.1190/1.1443575

[4] Buland A , Omre H. Bayesian linearized AVO inversion. Geophysics. 2003; 68(1): 185-198. doi: 10.1190/1.1543206

[5] Jardani A , Revil A , Slob E , Söllner W. Stochastic joint inversion of 2D seismic and seismoelectric signals in linear poroelastic materials: A numerical investigation. Geophysics. 2010; 75(1): N19-N31. doi: 10.1190/1.3279833

[6] Hampson DP , Schuelke JS , Quirein JA. Use of multiattribute transforms to predict log properties from seismic data. Geophysics. 2001; 66(1): 220-236. doi: 10.1190/1.1444899

[7] Ge Q , Cao H , Yang Z , et al. High—resolution seismic impedance inversion integrating the closed—loop convolutional neural network and geostatistics: an application to the thin interbedded reservoir. J Geophys Eng. 2022; 19(3): 550-561. doi: 10.1093/jge/gxac035

[8] Zhang B , Pu Y , Dai R , Cao D. Seismic poststack impedance inversion using geophysics—informed deep—learning neural network. Interpretation . 2025; 13(2): T219-T232. doi: 10.1190/int—2024—0041.1

[9] Chen Y , Fomel S. Random noise attenuation using local signal—and—noise orthogonalization. Geophysics. 2015; 80(6): WD1-WD9. doi: 10.1190/geo2014—0227.1

[10] Xue YJ , Zhang H , Zhang JQ , Wang XJ , Cao JX , Liu ZG. Local quantum filtering and denoising of seismic data. Geophysics. 2025; 90(5): V455-V472. doi: 10.1190/geo2024—0723.1

[11] Wu X , Yan S , Bi Z , Zhang S , Si H. Deep learning for multidimensional seismic impedance inversion. Geophysics. 2021; 86(5): R735-R745. doi: 10.1190/geo2020—0564.1

[12] Fang W , Fu L , Zhang M , Li Z. Seismic data interpolation based on U—net with texture loss. Geophysics. 2021; 86(1): V41-V54. doi: 10.1190/geo2019—0615.1

[13] Zhu G , Chen X , Li J , Guo K. Data—Driven Seismic Impedance Inversion Based on Multi—Scale Strategy. Remote Sens. 2022; 14(23): 6056. doi: 10.3390/rs14236056

[14] Jiang X , Cao J , You C , Wang X , Du Z. Deep Neural Networks for Direct Hydrocarbon Detection in Prestack Seismic Data Based on AVO Analysis. IEEE Trans Geosci Remote Sens. 2023; 61: 1-9. doi: 10.1109/tgrs.2023.3333910

[15] Yang N , Li G , Zhao P , Zhang J , Zhao D. Porosity prediction from pre—stack seismic data via a data—driven approach. J Appl Geophys. 2023; 211: 104947. doi: 10.1016/j.jappgeo.2023.104947

[16] Rasht—Behesht M , Huber C , Shukla K , Karniadakis GE. Physics—Informed Neural Networks (PINNs) for Wave Propagation and Full Waveform Inversions. J Geophys Res Solid Earth. 2022; 127(5): e2021JB023120. doi: 10.1029/2021jb023120

[17] Sun J , Innanen K , Zhang T , Trad D. Implicit Seismic Full Waveform Inversion With Deep Neural Representation. J Geophys Res Solid Earth. 2023; 128(3): e2022JB025964. doi: 10.1029/2022jb025964

[18] Jiang X , Cao J , Yang J , Liu J , Zhou P. AVO Analysis Combined With Teager—Kaiser Energy Methods for Hydrocarbon Detection. IEEE Geosci Remote Sens Lett. 2022; 19: 1. doi: 10.1109/LGRS.2020.3029170

[19] Xudong J , Junxing C , Jiangtao H , Xuanchen X , Jie L. Pre—stack gather optimization technology based on an improved bidimensional empirical mode decomposition method. J Appl Geophys. 2020; 177: 104026. doi: 10.1016/j.jappgeo.2020.104026

[20] Jiang X , Cao J , Zu S , Xu H , Wang J. Detection of hidden reservoirs under strong shielding based on bi—dimensional empirical mode decomposition and the Teager—Kaiser operator. Geophys Prospect. 2021; 69(5): 1086-1101. doi: 10.1111/1365—2478.13073

[21] Li P , Wang Y , Liu J , Liang M. Evaluation of Carbon Emission Efficiency and Analysis of Influencing Factors of Chinese Oil and Gas Enterprises. Energy Sci Eng. 2025; 13(3): 1156-1170. doi: 10.1002/ese3.2055

[22] Castagna JP , Sun S , Siegfried RW. Instantaneous spectral analysis: Detection of low—frequency shadows associated with hydrocarbons. Lead Edge. 2003; 22(2): 120-127. doi: 10.1190/1.1559038