A Mechanistic Approach to the Management of Uncertainty in Reservoir Characterization

This paper deals with developing a rational framework for quantifying and controlling the uncertainty in predictions associated with poro-elastic models of formations used in reservoir simulators for oil recovery. It aims at integrating data collection with model-based reservoir simulation for the purpose of reliable decision making. In particular the paper develops an analytical and computational framework for answering the following two questions: 1) Given sparse information about the hydrological and geomechanical properties of geophysical formations, along with observations from corresponding flow patterns, compute the level of confidence to be associated with the corresponding numerical predictions, and 2) prioritize subsequent modeling and data acquisition efforts. Poro-elastic models are essential for correctly predicting the mechanisms of oil mobility in fractured compressible rock masses such as chalk and unconsolidated sands. For other types of formations, poroelastic models typically provide a small, but sometime significant, correction to estimates obtained from other, non-coupled, models, and they clearly provide an improvement, in terms of accounting for the physics, over those models.