Statistical modeling for real time pore pressure prediction from pre-drill analysis and well logs

Abstract

The challenge of pore pressure prediction in an overpressured area near a well is studied. Predrill understanding of pore pressure is available from a 3D geologic model for pressure buildup and release using a basin modeling approach. The pore pressure distribution is updated when well logs are gathered while drilling. Sequential Bayesian methods are used to conduct real-time pore pressure prediction, meaning that every time new well logs are available, the pore pressure distribution is automatically updated ahead of the bit and in every spatial direction (north, east, and depth), with associated uncertainty quantification. Spatial modeling of pore pressure variables means that the data at one well depth location will also be informative of the pore pressure variables at other depths and lateral locations. A workflow is exemplified using real data. The prior model is based on a Gaussian process fitted from geologic modeling of this field, whereas the likelihood model of well-log data is assessed from data in an exploration well in the same area. Results are presented by replaying a drilling situation in this context.