Experimental Investigation of Thermally Induced Stress Shadow Effects in Multiwell Hydraulic Fracturing and Neural-Network-Based Fracture Identification.

Efficient stimulation of deep high-temperature, low-permeability reservoirs critically depends on the creation of complex fracture networks via multiwell hydraulic fracturing, where thermally induced stress shadow effects modulate fracture propagation and connectivity. In this study, we present a comprehensive experimental investigation using a self-developed high-temperature true triaxial multiwell fracturing apparatus and an artificial dual-well model to systematically characterize the coupled influence of the temperature field, perforation spacing, and fracturing sequence on fracture interaction and stress perturbation. A series of fracturing experiments were conducted under varying thermal conditions and fracturing strategies (sequential and synchronous). A postfracturing quantitative analysis of fracture geometry was conducted using CT-derived three-dimensional reconstructions. Furthermore, we developed an improved U-Net neural-network framework to automatically identify fracture connectivity and spatial patterns from low-contrast CT images, achieving improved prediction accuracy over conventional linear models. The integrated "physical experiment-quantitative fracture characterization-neural-network recognition" framework elucidates the competitive and synergistic mechanisms governing the coupled interactions among thermal, geometric, and stress shadow effects in deep reservoirs. The findings demonstrate that increasing perforation spacing weakens stress interference, promoting independent fracture growth. High temperatures reduce peak injection pressure and intensify complex fracture morphologies through thermal microcrack development. The machine-learning-based identification approach reliably discriminates fracture connectivity states and morphological types, offering a robust basis for optimizing multiwell fracturing design and evaluating post-treatment effectiveness in deep high-temperature tight reservoirs.