Field Study: Embedded Discrete Fracture Modeling with Artificial Intelligence in Permian Basin for Shale Formation

Abstract

Abstract Even with the current low oil price environment, oil and gas production from shale formations becomes increasingly important. The drilling and exploration activities in the Permian Basin are the most active unconventional play in the U.S. Therefore, engineers need effective and efficient full field planning and development strategies for an economically successful project. Since in the coming years, thousands of horizontal well drilling activities are expected, field planning, such as the well landing design, spacing study, and operational sequence, could have huge impacts on economic field development. Full field models using unstructured grids can capture detailed geometry information like fracture distribution. It is, however, computationally expensive and often numerically unstable (convergence issues). We investigated the embedded discrete fracture modeling (EDFM) with artificial intelligence (AI) to overcome challenges associated with the unstructured modeling. We applied the concept of EDFM, in which a fracture is embedded in the structured matrix grid that does not conform to the discrete fractures. EDFM uses relatively low matrix resolution but honors fracture geometry and connectivity: fracture orientation, shapes, and intersections. As a result, the flow behavior in fractures is well preserved. Furthermore, EDFM has a large advantage of computational efficiency, compared to unstructured full-field models. We also proposed to employ AI techniques to optimize the hydraulic fracture networks by identifying the most important fracture geometries and eliminating the unpropped structures. Our study combines EDFM and AI techniques to make numerous full field simulation runs making the uncertainty analysis affordable for better planning decisions. In this study, we verified the effectiveness and efficiency of EDFM and AI techniques with various models. Then we applied the technique for a full field well interference study to guide the development sequence. In the full-field Permian Basin case study, the new method accurately captured the influence among four neighboring horizontal wells. The results are very close to the reference solution from full field unstructured simulations, but with a huge improvement in efficiency.