The faults in reservoir models have long been afflicted by ‘non-geological’ reservoir simulation practices. Many traditional production history matching workflows treat geological faults as ‘flood-gates’ that can be completely open, closed, or partially sealing as needed to match the production behavior observed at the wells. The predictive power of such models remains questionable. More realistic dynamic models consume a detailed fault seal analysis that utilizes fault displacement, juxtaposition, and clay content predictions to estimate properties that control flow across the faults. However, even when adopting this more geologically-sound approach, the mechanism of fault transmissibility multipliers (TMs) is used to incorporate the dynamic characteristics of faults in reservoir simulators. These TMs are not physical properties and instead represent a relationship between fault rock properties and host cell properties. Therefore, they only remain valid if the grid properties are not modified during history matching. Also, since a TM is not a physical property, rather a numerical simulation artefact, it is difficult to assign meaning and a reasonable range of uncertainty to it during history matching.
This paper presents a new workflow for the assessment of uncertainty in fault flow dynamics. The workflow has been developed for a next-generation unstructured grid that uses cut cells to accurately honor structural and stratigraphic complexities. We describe the methods used to calculate side-dependent fault rock permeability and thickness estimates that are then directly provided as inputs to a next-generation reservoir simulator. The simulator incorporates the fault properties into the transmissibility calculation for cell-cell pairs separated by the fault. A scaling factor can be used to consistently scale fault rock properties if needed, and this is applied independently of the grid permeability.
Example cases are presented to contrast the traditional approach with this improved approach. The potential shortcomings of the constant TM approach are exposed by examining the calibrated fault TMs for a history-matched case. We highlight the potential inconsistencies introduced in reservoir models with the use and subsequent manipulation of variable fault TMs. This inevitably leads to unphysical modifications to the spatial distribution of TMs during model calibration and a loss of geological insight. The implications on the reliability of production forecasts are discussed. A sensitivity study is conducted using the improved approach to quantify the relationship of the fault properties on flow performance. The benefits of the new methodology for the scaling of fault transmissibility during history match are demonstrated.
This new approach provides the reservoir engineer with direct visibility of and control over geological fault rock properties in simulation models. This constrains the calibration of fault rock properties within a sensible uncertainty range, allows the grid permeability to be modified independently of the fault rock properties during history matching, and increases the reliability of production forecasts by preserving geological fidelity.