Exploring Structural Uncertainty using a Flow Proxy in the Depositional Domain

Abstract

Abstract Among various sources of uncertainty in reservoir modeling, structural uncertainty is often underrepresented. Evaluating different faulting scenarios requires gridding each of these scenarios. Preserving the structural geological realism requires careful consideration of structural hierarchy and abutting relationships. Enforcing such relationships often requires manual editing results. In this paper, a proxy method to efficiently represent uncertainty in structural modeling by building fault scenarios in the depositional grid is proposed, which is often a simple Cartesian grid. In this way accuracy (building actual complex fault models) is traded off with the ability to study uncertainty. A bottleneck in properly modeling structural uncertainty is caused by representation of faults as explicit surfaces. This is resolved in this paper with a proxy method by representing a faulting scenario as a fault block map which represents the compartments created in a reservoir by faults as a property on a Cartesian grid. These compartments defined on a Cartesian grid are easier to manipulate, allowing for fast generation of geologically consistent fault network realizations. Faults bounding these compartments are represented as transmissibility multipliers on the Cartesian grid taking into account their throw. Another bottleneck associated with structural uncertainty evaluation comes at the stage of simulating flow through a faulted reservoir. Running simulations in the physical domain where the grid is deformed with respect to fault throws may be cumbersome due to grid cell geometries around faults. Utilizing state-of the-art flow diagnostics on the Cartesian domain instead of a relatively slow finite element reservoir simulator on the physical domain allows the rapid evaluation and selection of fault networks for uncertainty quantification of forecasts. To showcase the main functionalities of the proxy method, a small field example is used with small number of faults, where the full (and time-consuming) exploration of uncertainty is compared with uncertainty derived from the proxy method.