3D Pore Pressure and Geomechanics: Work Smarter and Faster Integrating Geoscience with Machine Learning

In unconventional plays, given the comparatively short drilling times and the likelihood that operators have multiple active rigs, wells are drilled and data are acquired at an unprecedented rate whereby a new well is completed every 1–2 days at a cost of $6–9M per well. Therefore, performing manual workflows for petrophysics, pore pressure and geomechanics prediction can be impractical due to turnaround considerations and the multiple personnel required. This, together with technical challenges of complex stratigraphy, multiple facies, variable rock properties, and the interaction of pore pressure and geomechanics, calls for more consistent, sophisticated, and faster analytical tools. A supervised deep neural network approach is presented as an innovative tool to devise solutions which simultaneously integrate myriad data types. Furthermore, an algorithm was developed to predict a certain number of attributes solely from a facies-based seismic inversion, namely Vp, Vs, and Rho. The application of these algorithms on various blind wells from a Permian Basin case study, both within and outside the seismic survey, shows a reasonable accuracy when compared to manually interpreted counterparts but were obtained in a fraction of the time, hence, provide a promising outlook for the application of deep learning in integrated studies.