EDFM-AI for Calibration of Hydraulic and Natural Fracture Geometry

ABSTRACT Due to highly uncertain nature of unconventional reservoirs and difficult monitoring of the engineering effectiveness, a reliable, accurate, and efficient reservoir model calibration workflow is essential to help operators understand/plan their assets. In this paper, a collection of the non-intrusive EDFM (Embedded Discrete Fracture Model), AI (artificial intelligence), and an unconventional-targeting in-house reservoir simulator URSim is built to perform automatic history matching and uncertainty calibration of hydraulic and natural fractures (abbreviated as EDFM-AI). By implementing the proposed workflow, highly uncertain parameters, such as hydraulic fracture half-length, height, conductivity, closure coefficient, etc. are easily characterized. To validate the robustness of the proposed workflow, especially on the level of simultaneous multi-well calibration, a field-scale shale gas reservoir model is prepared with 3 horizontal wells and 200 hydraulic fractures for each well. Known set of hydraulic fracture parameters are inputted, constant gas flow rates are the well constraints, and the simulation outputs are defined as benchmark data. The history matching results showed high accuracy matches. More importantly, the maximum error between calibrated P50 and true value is only 6.8%. This novel EDFM-AI workflow sheds light on post-frac evaluation and completion optimization in unconventional resources. INTRODUCTION Current unconventional reservoir developments are characterized as fast-paced, real-time, and multi-well (pad-wise drilling). Many operators can drill and completion more than 100 unconventional horizontal wells per year. Under this scenario, it is crucial for operators to understand the effectiveness of hydraulic fracturing. Multitude methodologies/studies have been dedicated to study the man-made hydraulic fractures. These include fracturing simulation models (Wan et al., 2020; Weijermars et al., 2020; Leem et al., 2022), experimental methods (Magsipoc et al., 2020; Wei et al., 2021), and diagnostic technologies (Gutierrez et al., 2010; Ugueto et al., 2021; Wang et al., 2022). However, uncertainties associated with fracture geometries still remain as the most challenging problem in the unconventional oil/gas industry.