Pioneering Big Data Fracture Calibration Framework: EDFM-AI For Field-Wise Hydraulic Fracture Network Quantification

This paper presents a scalable EDFM-AI framework for automated calibration of effective hydraulic fracture geometries across multiple shale gas well pads, marking the industry’s first integration of big data fracture modeling with machine learning. The proposed workflow combines geological and petrophysical inputs with a star-boundary fracture generator and EDFM-coupled GPU-accelerated reservoir simulation. It further integrates an ESMDA-assisted proxy model to address high-dimensional uncertainties in fracture geometry, enabling rapid and accurate history matching. The methodology is applied to 23 horizontal wells across four shale gas pads, resulting in calibrated fracture half-lengths and heights ranging from 70–220 meters and 10–25 meters, respectively. Bottomhole pressure matches exceed 90% accuracy, and the 20-year gas EUR forecast quantifies the long term production potential range to be 0.09–0.15 billion cubic meters, showcasing the model's predictive capability. This novel workflow transforms unconventional reservoir characterization by enabling data-driven, field-scale fracture calibration and production forecasting at previously impractical scales. Its automation, scalability, and high fidelity make it a powerful tool for optimizing shale gas development across large well populations.