Oblique-incidence reflectivity difference method combined with deep learning for predicting anisotropy of invisible-bedding shale

Abstract

Deep learning methodologies have revolutionized prediction in many fields and is potential to do the same in the petroleum industry because of the complex oil–gas reservoir. A limitation remains for dense shale exploration in that the shales with invisible bedding are difficult to characterize measurably because of the considerable complexity of the geological structures. The oblique-incidence reflectivity difference method (OIRD) is sensitive to the surface features and was used to obtain a layered distribution of dielectric properties in shales. In this paper, we report a combination of OIRD and deep learning method to identify the dielectric anisotropy of an invisible-bedding shale. The model performs well and clearly identifies the bedding of the shale based on the output values associated with the probability. Only a single direction was determined to have laminations with widths of 20–60μm. The anisotropy features detected by OIRD also existed in the invisible-bedding shale and were caused by the smaller cracks and denser particles’ orientation relative to general shales. As current dense reservoirs include rich invisible-bedding shales, we believe that the OIRD method combined with deep learning method can help improve the exploration efficiency of shale reservoirs.

Keywords

• Oblique-incidence reflectivity difference method
• Deep learning method
• Shale
• Anisotropy