Stochastic discrete fracture network modeling in shale reservoirs via integration of seismic attributes and petrophysical data

Abstract

The prediction of natural fracture networks and their geomechanical properties remains a challenge for unconventional reservoir characterization. Since natural fractures are highly heterogeneous and sub-seismic scale, integrating petrophysical data (i.e., cores, well logs) with seismic data is important for building a reliable natural fracture model. Therefore, I introduce an integrated and stochastic approach for discrete fracture network modeling with field data demonstration. In the proposed method, I first perform a seismic attribute analysis to highlight the discontinuity in the seismic data. Then, I extrapolate the well log data which includes localized but high-confidence information. By using the fracture intensity model including both seismic and well logs, I build the final natural fracture model which can be used as a background model for the subsequent geomechanical analysis such as simulation of hydraulic fractures propagation. As a result, the proposed workflow combining multiscale data in a stochastic approach constructs a reliable natural fracture model. I validate the constructed fracture distribution by its good agreement with the well log data.

Review of discrete fracture network characterization for geothermal energy extraction

“…The use of seismic attributes is used to generated 3D DFN models due to their capability to detect the principal trends, and presence of faults that can represent preferential flow pathways for the geothermal fluids (Bense et al, 2013;Schneider et al, 2016;Cho et al, 2019;Marchesini et al, 2019;Cho, 2021;Fadel et al, 2022;Huang et al, 2022;Xing et al, 2022;Zhang E. Y. et al, 2022;Chiodini et al, 2023).…”

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

“…Various volume attributes, including semblance [1], variance [2], and coherency [3] are commonly employed for fault detection. However, seismic attributes are vulnerable to both seismic noise and complex stratigraphic features, introducing a potential for significant misinterpretation [4]. Unlike the single volume attributes, the ant-tracking algorithm initiates with a fault attribute and utilizes "ant agents" that follow assumed fault features.…”