An AI-Driven Integrated Geology and Engineering Workflow for Hydraulic Fracturing Optimization Design in Deep Coalbed Methane Multiwell Pads

Summary The extraction of methane from deep coal seams (more than 1,500 m) has gained increasing attention recently. The deep coal seams present distinct characteristics, including lower reservoir permeability, higher in-situ stress, and stronger plastic deformation, when compared with shallower coal seams. An effective hydraulic fracturing design is crucial, taking into account sophisticated geological information and realistic, 3D fracture networks. However, conventional fracturing design methods have challenges of timeliness and adaptability, particularly regarding the fracture-driven interactions (FDIs) among multiple wells. Here, an artificial intelligence (AI)-driven integrated geology and engineering workflow is proposed and implemented for deep coalbed methane (CBM) multiwell pads. The workflow links geological model, fracture-propagation model, reservoir simulator, and optimization model in a unified framework. Special considerations in balancing the stimulated reservoir volume (SRV) against FDIs are incorporated into the multiwell-pad fracturing optimization. This workflow has been applied to two multiwell pads in the Linxing and Shenfu deep CBM fields, Ordos Basin. In the Linxing field, the vertical height separation between the laterals of two adjacent wells in a multiwell pad is 10~20 m (in a cross-sectional view). Special fracturing treatment design and parameters were considered to control the fracture height in case of fracture hit. In the Shenfu field, a nine-well pad was investigated. The synergistic optimization of well spacing, fracturing parameters, and operational sequence enables effective reservoir stimulation, resulting in high gas productivity. The framework explicitly predicts the fracture networks and determines optimal pumping schedules based on the effective permeability area, SRV, estimated ultimate recovery (EUR), and FDIs. The AI-assisted automated history matching, combined with graphics processing unit (GPU) parallel computing, facilitates rapid and reliable post-fracturing analysis. Furthermore, our findings suggest that multiround fracturing combined with diverters is suitable for exploiting deep CBM resources. The results provide valuable insights into the selection of optimal completion parameters and the development of effective strategies. The key insights from this research are expected to serve as a theoretical guide for the effective development of deep unconventional resources.