Capturing natural fracture topology in DFNs for energy and storage applications

Many renewable subsurface energy systems rely on understanding fracture networks. In geothermal systems, fractures often provide the primary pathways for fluid flow, while in underground hydrogen storage (UHS) and carbon capture and storage (CCS), fracture networks can strongly influence pressure communication, injectivity or caprock integrity. Similarly, in nuclear waste repositories, fractures can compromise barrier integrity and limit containment. This study investigates the representativity of discrete fracture networks (DFNs) generated from high-resolution photogrammetric outcrop data to support robust models and simulations.

Observed data such as fracture orientations, lengths, intensities, and topological node classifications (X-, Y-, and I-nodes) are used to construct synthetic DFNs via (1) geometric modelling, (2) fracture-growth algorithms, and (3) tracemap extrusion. These DFNs are then meshed and integrated into single-phase flow simulations. Pressure gradients are applied to quantify the influence of fracture intensity and topology on flow behaviour across above fracture generation methods.

Results show systematic topological deviations between natural and synthetic networks. Geometric and growth-based methods overestimate X- and I-nodes while underrepresenting Y-nodes, affecting connectivity and predicted flow paths. Tracemap extrusion reproduces geometry more accurately but requires significantly higher computational resources. Flow simulations reveal that fracture intensity and node topology strongly influence pressure evolution and steady-state attainment. Both parameters are central to injectivity forecasting, (thermal) breakthrough prediction, and storage containment assessment.

Overall, the results demonstrate that current DFN generation methods reproduce fracture geometry reasonably well but struggle to match natural network topology, introducing systematic biases into models and simulations. Improving the representation of Y-node-dominated branching structures is therefore essential for developing more reliable models and simulations of fractured reservoirs and repositories.