Probabilistic assessment of pore pressure prediction at predrill and while-drilling stages

Josue Fonseca
Pore pressure prediction is vital to prevent disasters when drilling the subsurface and for designing cost-effective drilling plans for ventures that depend on subsurface resources. This property depends on the geological and physical process occurring through the geo-history of an area of interest. However, its prediction is generally simplified by the use of empirical models. Thus, we propose novel methods to include process-based modeling when predicting and quantifying the uncertainty related to these estimations. Basin modeling simulations are performed at predrill stages and then rapidly conditioned to offset well and seismic data with surrogate calibration (https://pubs.geoscienceworld.org/geophysics/article/88/6/M239/628018/Ba…). Afterward, the retrieved posterior Earth models are utilized to design probabilistic models to reproduce the identified predrill pore pressure uncertainty. Bayesian networks represent the uncertainty as this probabilistic graphical model allows fast updates in pore pressure beliefs at deep targets when while-drilling data from top layers are observed.