3D structural modeling for seismic exploration based on knowledge graphs

Abstract

Three-dimensional structural models provide valuable references for studying exploration challenges and often are used for survey design and seismic method validation as well as to assist with reservoir interpretation. The current structural modeling strategy primarily relies on seismic images, which can suffer from degraded image quality in areas of complex structures. This leads to less-reliable structural interpretations. It is difficult to build appropriate models with unreliable inputs by data-driven modeling methods when the input data contain no other constraints. To mitigate this problem, we have adopted a knowledge-data-driven structural modeling method for seismic exploration based on knowledge graphs that contain certain digitized prior information related to the subsurface structures. The knowledge graph of the structural models formalizes the knowledge of geoscientists by describing the structural model entities and their interrelations in the form of a graph. We first establish the knowledge graph of the structural model to constrain the model topology, then estimate the geologic intersections (lines) between the geologic surfaces under the guidance of the knowledge graph, and finally reconstruct the surfaces under the constraints of the intersections and generate closed geologic bodies. The knowledge graph defines the overall structure of the model, which sets up a priori constraint for the modeling and plays a role in transferring expertise to computers. We use two field models to demonstrate our structural modeling process and prove the effectiveness and feasibility of the method.