Drilling risk predictions are crucial in pre-drill planning. Drilling risks such as kicks, stuck pipes, lost circulation and hole collapses will make borehole instable and risky for drilling. Accurate pore pressure and wellbore stability prediction are required in well planning and drilling. Traditionally, these predictions are manually executed by geomechanics experts. The procedures are usually complicated and take time, the results also highly depend on the executor's expertise. In this paper, we utilized machine learning methods to perform prediction in a simpler manner. The digital models were trained with existing geology data, drilling data, well logs etc with physics-based hybrid algorithms. The trained models are then used to predict pore pressure and wellbore instability with mud weight window along planned well trajectory, to identify drilling risks and recommend solutions. This novel method was tested and validated in an offshore field in South China Sea, China. The target reservoir is a shale formation known as abnormally high-pressure and hole instability. An extended reach well was planned to drill and require identifying potential drilling risks with an optimal well trajectory and platform location. Such prediction in planning phase usually takes weeks with traditional manual methodology. Applying our machine learning digital model to this field, the initial digital geomechanics model was trained using existing exploration wells data and interpreted field geology surfaces data. It was then used to predict pore pressure for the others exploration wells. The prediction results were compared to actual downhole pressure tests and confirmed the average accuracy of about 97%. The hole instability risks were predicted with machine learning together with mud weight window, including formation collapse pressure, mud loss pressure and breakdown pressure. The comparison with manual results was accurate, the optimal well location and well plan trajectory were decided. The extended reach well was then drilled and completed without any drilling complexities. This digital machine learning method reduced the pre-drill analysis time from weeks to hours, this case study confirms its efficiency and reliability.
Gao et al. (Sat,) studied this question.