A Topological Approach for Automated Unstructured Meshing of Complex Reservoir

Estimations of petroleum reserves rest on finite volumes computational simulations of the reservoir fluid dynamics. These simulations are operated on 3D meshed reservoir models produced by a complex and poorly automated chain of operations. This paper proposes a mesh building methodology, which uses geological rules for building reservoir meshes in a more automated way. We start from a surface structural model and from a description of the stratigraphy both packed thanks to the industry standard RESQML. We construct a volume structural framework based on generalized map topological structures. These structures include topological boundary relations between the represented geological objects (horizons, faults, units) and some dedicated data attached to the topological cells (vertices, faces, volumes, etc.), such as geometry or geological labels (e.g. names, relative ages, deposit methods). In particular, on a single topological representation, we can attach two different geometric representations respectively describing the present day layer geometry (“folded model”) and the original positions of the various layers in their “deposition space” ("unfolded model"). Thanks to a dedicated rule-based language, we deduce from the geological interpretation, a set of topological and geometric operations that allow an automated building of the structural framework, on which the reservoir meshes will be implemented. This language allows a fast prototyping of complex operations (boolean operations for instance) and it guarantees the geological and topological consistency of the model. Using this consistent and fully informed structural framework, we can create in an automated way various conformal unstructured 3D meshes organized in layers. These meshes agree both with the topology induced by the succession of deposition, erosion and tectonic events that constitute the local geological history and with the peculiarities of the used fluid flow simulators. A use-case is presented to demonstrate the feasibility of our method.