Abstract
As a high-resolution geophysical method employed by the oil and gas industry, well logging can be used to accurately investigate reservoirs. Challenges associated with shale gas reservoir exploration increase the importance of applying elastic parameters or velocity at the logging scale. An efficient shale rock physics model is the foundation for the successful application of this method. We propose a procedure for modelling shale rock physics in which an appropriate modelling method is applied for different compositions of shale rock. The stiffnesses of the kerogen and fluid (oil, gas or water) mixture are obtained with the Kuster-Toksöz model, which assumes that the fluid is included in the kerogen matrix. A self-consistent approximation method is used to model clay, where the clay pores are filled with formation water. The Backus averaging model is then used to simulate the influence of laminated clay and laminated kerogen. Elastic parameter simulations using well logging data show the importance of treating the volume fractions of laminated clay and kerogen carefully. A comparison of the measured compressional slowness and modelled compressional slowness shows the efficiency of the proposed modelling procedure.
Introduction
Shale gas reservoirs present significant characterization challenges due to their complex composition of minerals, organic matter (kerogen), and intricate pore networks. For geophysicists, predicting how sound waves travel through these rocks is essential for interpreting well logs. However, shale is inherently anisotropic—its properties vary by direction because of its fine, layered structure.
Most existing models either oversimplify shale as isotropic or focus on low-frequency seismic data, leaving a gap for high-frequency (~10 kHz) logging scales. Without accounting for the specific distribution of clay and kerogen, predicted velocities often deviate from reality, leading to inaccurate reservoir assessments. This research proposes a multi-stage modeling workflow to bridge the gap between mineral composition and field-measured elastic parameters.
Methods
The proposed modeling procedure applies different mathematical theories to specific components of the shale to build a comprehensive "bottom-up" model.
- Kuster-Toksöz (KT) Model: This model calculates the stiffness of the kerogen and fluid mixture. KT is particularly effective for modeling various pore geometries, such as "penny cracks," as inclusions within the kerogen matrix.
- Self-Consistent Approximation (SCA): SCA is used to model the clay and bound water mixture. Unlike other models that assume a "host" material, SCA treats all constituents equally, making it suitable for the high-frequency environment of sonic logging.
- Matrix Calculation: The rock's mineral "skeleton" (quartz, calcite, dolomite) is determined using a linear average of their moduli.
- Backus Averaging: In the final step, the isotropic mixtures are "stacked" into layers. This simulates the lamination of clay and kerogen, transforming the model into a transversely isotropic (VTI) representation.
Key Results
The model was validated using a dataset from a shale gas reservoir in China, including mineralogy logs, organic content, and sonic logs (compressional slowness, or DTC).
Initial attempts using a purely isotropic model resulted in significant discrepancies, with predicted DTC values around 210 µs/m compared to measured values of 250 µs/m. However, after incorporating Backus averaging to account for lamination, the modeled DTC (blue line) achieved a near-perfect match with the measured DTC (red line) across the entire depth interval.
Discussion and Conclusion
A critical insight from this work is that the ratio of layered to randomly distributed clay and kerogen is the primary driver of elastic behavior. Even with accurate total volume measurements, models fail if they do not account for lamination. Furthermore, the choice of SCA over standard Gassmann theory is vital at logging frequencies (~10 kHz), where pore pressures do not have time to equilibrate.
While the model shows high efficiency, a current limitation is that the volume fractions of layered versus random components are not directly measured by standard tools and must be determined through best-fit optimization. Future work should focus on developing petrophysical correlations to quantify lamination directly and testing the workflow across diverse global shale plays.
Summary updated