Automatic Calibration of Complex 3D Discrete Fracture Networks using EDFM-AI

Abstract

Abstract In the realm of shale reservoir development and modelling, the problems of completion evaluations and long-term well performances are particularly important but highly challenging. Without suitable measurement metrics of completion designs and production potential, the operators have limited ability to quantify and optimize the quality of completion. Presence of natural fractures within shale reservoir adds higher level of complexity to an already challenging problem. Ability to integrate natural fractures is critical for reliable history matching and long-term production forecast. These challenges limit the accuracy of the traditional numerical methods and needs to be addressed. The objective of this study is to evaluate a novel numerical method called embedded discrete fracture model (EDFM) to non-intrusively model the contribution of hydraulic fractures and natural fractures. In this study, we tested and incorporated cutting edge artificial intelligence with our EDFM method (EDFM-AI) to perform automatic history matching. These improved numerical methods are utilized to handle the above challenges in a systematic manner. The new EDFM-AI workflow is applied to a naturally fractured shale reservoir. Key untertainty features for hydraulic fractures such as half-length, height, water saturation, conductivity, and cluster efficiency are defined, with various confidence intervals. Highly uncertain parameters for natural fractures, such as total number of fractures and their conductivity are also included. Through EDFM-AI workflow, it is found out that the total effective fracture flowing area is about 2.21 million cubic feet for this Eagle Ford horizontal well case study. Furthermore, the conductivity of natural fractures are 1/100 times that of hydraulic fractures, but these natural fractures can contribute 25% of the 30-year oil EUR. Finally, the visualization shows that the drainage radius increased from 200 ft at end of the history period to roughly 500 ft after 10 years of production, which differentiate the optimal well spacing at different producing time. By implementing this novel workflow, the vast uncertainties can be quantitatively analyzed. The successful application of EDFM-AI workflow demonstrates its benefits compared to other unconventional modelling workflows in the oil/gas industry.