Organic-rich shale lithofacies, primarily defined by mineral composition and organic matter richness, reflects the features of the two critical factors for unconventional shale reservoirs. The research of shale lithofacies can aid in identifying shale gas productive zones and designing horizontal well and hydraulic fracturing. Seven shale lithofacies in Marcellus Shale have been defined by mineral composition and TOC content. Prediction of shale lithofacies by conventional logs is the key step to define the distribution of shale lithofacies laterally and vertically, but the relationship between lithofacies and logs is non-linear and complex. The effectiveness of conventional mathematical methods is limited. Artificial intelligence (AI) classifiers, such as artificial neural network (ANN), and support vector machine (SVM), can solve complex nonlinear problems. In addition, learning algorithms based on AI could also work together with AI classifiers to recognize shale lithofacies. Meanwhile, an innovative decomposition method, hierarchical decomposition, has been proposed and used to enhance the performance of ANN and SVM classifiers in predict Marcellus Shale lithofacies. In this paper, we devoted ourselves to comprehensively discuss the strength and weakness of these AI algorithms in pattern recognition and present an integrated workflow for organic-rich shale lithofacies prediction. This methodology should be helpful for recognizing shale lithofacies in other shale-gas plays, which aids in identifying high productive shale gas sweet spots.

This content is only available via PDF.
You can access this article if you purchase or spend a download.