Prediction of Rock Mechanical Parameters in Deep Coal-Bearing Strata and Fracturability Classification Evaluation of Coalbed Methane Reservoirs

The Ordos Basin is a core area for deep coalbed methane (CBM) accumulation and production in China. The No. 8 coal seam at the Benxi Formation top is a primary target due to its wide distribution, stable thickness, and good gas-bearing capacity. However, complex roof/floor lithology and coal heterogeneity lead to intricate petrophysical responses of mechanical parameters, affecting fracturing efficiency. Thus, an integrated system incorporating rock properties for mechanical parameter prediction, in-situ stress calculation, and fracturability classification is critical for deep CBM sweet spot identification.

Sixty core samples of major lithologies (limestone, mudstone, sandstone, coal rock) were collected from the central-eastern Ordos Basin. Comprehensive laboratory tests were performed, including X-ray diffraction (XRD), uniaxial/triaxial compression, Brazilian splitting, and basic physical tests (P-/S-wave velocity, density). For coal rock, an extended Gassmann-based petrophysical model was established via Biot’s porous medium theory, incorporating pore-fracture extrusion-spray flow effect to accurately predict cleat density. Pearson correlation analysis identified key factors governing mechanical parameters: acoustic velocity, density, cleat density for coal; acoustic velocity, density, shale content for roof/floor rocks. Multiple nonlinear regression models were built for their mechanical parameters, with R²>0.75 between predicted and measured values, ensuring high accuracy.

Using the predicted rock mechanical parameters, the combined spring model was employed to calculate the in-situ stress of the coal-bearing strata. The prediction results demonstrated excellent consistency with field measured data, with an average relative error of less than 10%. Focusing on CBM reservoirs, relevant parameters were extracted, including rock mechanical properties (USC, E, V……) and in-situ stress components (σ_H, σ_h……). The correlation between these parameters and single-well daily gas production was systematically analyzed. The XGBoost ensemble learning algorithm was utilized to screen key influencing parameters from high-dimensional data, identifying four critical factors: minimum horizontal stress difference between reservoir and roof, reservoir horizontal stress difference, reservoir tensile strength, and reservoir elastic modulus. A fracturability evaluation model (FI) was constructed based on these key factors, and clustering analysis was applied to classify reservoir fracturability through iterative updating of cluster centers. The classification results yielded three reservoir grades: Class Ⅰ (FI > 0.32) with excellent fracturability, facilitating the formation of complex fracture networks; Class Ⅱ (0.21 < FI ≤ 0.32) with moderate fracturability, tending to form relatively simple fracture networks; and Class Ⅲ (FI ≤ 0.21) with poor fracturability, for which fracturing stimulation is not recommended.

The results of this study show great potential in evaluating deep CBM in the basin. It significantly improves the accuracy of parameter prediction (R² > 0.75) and in-situ stress calculation (error < 10%). Meanwhile, the combination of the FI model and classification standard effectively enhances evaluation precision and decision-making efficiency, providing strong support for targeted fracturing and sustainable deep CBM development.