Numerical study of fracture dynamics in different shale fabric facies by integrating machine learning and 3-D lattice method: A case from Cangdong Sag, Bohai Bay basin, China

Shale oil, primarily existing in organic-rich shales from lacustrine basin, would amount to a huge resource in China. Currently, this kind of resource is pervasively distributed in basins like Songliao, Ordos, Jungar, Qaidam, and Bohai Bay. However, accompanied by the great potential, difficulties of exploitation dramatically increase, as a result of strong heterogeneity and complex lithology of shale formation. Taking the continental shale formation in Bohai Bay basin as an example, the abundance of organic matter differs greatly with the varying lithofacies. Moreover, diversified mineral constitutions and well-developed laminas are found in deeply buried shale coring by adopting X-Ray diffraction technique. All these characteristics have squeezed the in-depth understanding of mechanisms behind the generation of complex fracture network during the hydraulic fracturing treatment. In this paper, a workflow of integrating machine learning algorithm and 3-D lattice method is established to investigate the fracture propagation process in four predominant shale fabric facies. Sensitivity analyses are also conducted to quantify impacts of different controlling parameters on the fracture propagation. Our study provides a guideline for improving the stimulation performance of volumetric fracturing in continental shale formation, and further helps to create more periodic secondary fractures and thus enhance the well productivity. •A workflow of integrating machine learning algorithm and 3-D lattice method is established.•On-site core photography and statistical methods are used to detect diversified lithology and distinct mineral compositions.•XGBoost is adopted to estimate rock mechanical parameters.•Fracture dynamics in four predominant shale fabric facies are investigated.