INVERSION OF HIGH-RESOLUTION HIGH-QUALITY SONIC COMPRESSIONAL AND SHEAR LOGS FOR UNCONVENTIONAL RESERVOIRS

摘要

Interpretation of elastic properties honoring fine heterogeneity has garnered recognition recently in petrophysical analysis, bedding failure prediction, and hydraulic fracture job design for unconventional reservoirs. Traditional sonic processing assumes homogeneity of the formation over a specific sonic tool receiver aperture length (e.g., at least 2 ft). This assumption may not be appropriate for highly laminated reservoirs where mechanical properties of interest could vary on a significantly finer scale. Additionally, shear slownesses extracted from low- and high-frequency processing are associated with different wavelengths and different rock volumes. For instance, shear slowness logs from a high-frequency monopole transmitter and a low-frequency dipole flexural mode can exhibit different axial resolutions even when using the same receiver aperture length. This apparent inconsistency and the lack of adequate vertical resolution control may render conventional sonic answer products inadequate for properly addressing the high-resolution challenges of unconventional reservoirs. We have developed a new interpretation algorithm to improve the layer slowness contrast for thinly laminated formations in vertical wells using borehole sonic data from advanced array-based logging tools. This novel interpretation method can yield high quality and high-resolution sonic compressional and shear (P and S) logs. It is based on a robust downscaling technique that jointly combines all logs processed at different array resolutions. An overdetermined matrix is formulated by taking all convolutional relationships among the different-resolution sonic logs. The high-resolution sonic logs (both P and S) are estimated using the Moore-Penrose pseudoinverse method. This method yields the sonic log with an optimal apparent resolution better than that is estimated from the conventional 1-ft single resolution sub-array method. Finally, the residual is formulated to serve as a log quality control flag and is used to automatically switch to more reliable low-resolution logs such as the dipole flexural shear slowness in depth intervals of poor-quality hole data or slow formations. The algorithm was validated with synthetic logs from finite-difference modeling and was then tested on a field data set collected in a vertical well traversing a thinly laminated formation. The inverted high-resolution P and S logs from sonic field measurements have higher depth resolutions than what the maximum resolution conventional processing can achieve and are consistent with a higher resolution ultrasonic log from an ultrasonic imaging tool logged in the same well. The field data application suggests that this downscaling algorithm enhances the spatial resolution and more accurately captures the layer slowness contrast while removing outliers thereby improving the log quality. The application of this method results in a superior characterization of the acoustic properties of thinly layered rocks than what is obtained with conventional processing. The elastic moduli honor the highly heterogeneous nature of the rock and thus could improve stress profiling and rock strength correlations for geomechanical modeling. Operational decisions such as landing laterals or staging stimulation intervals to avoid weak or strong interfaces will also be better informed.