Abstract: Fractures are the main flow conduits in coal reservoirs,and the accuracy of their prediction directly affects development outcomes.Targeting the issues of low signal-to-noise ratio of seismic data,great difficulty and low accuracy in post-stack fracture identification for the Kizilnur Formation coal reservoirs in Dibei Area of Tarim Basin,a multi-attribute pixel-level fusion technique was applied to comprehensively identify coal reservoir fractures.Considering the fracture-controlling effect of structural deformation intensity,a nonlinear relationship model was established among fault zones under structural control,structural curvature,and fracture density.The research shows that the multi-attribute fusion volume is more sensitive to subtle deformation and small faults,predicting fracture spatial distribution with little noise influence and yielding clear sheet-like features.Fractures in Dibei Area exhibit significant structural zonation:fracture density grows exponentially with structural curvature;high-angle tensile fractures predo minantly develop in the core of structural uplifts,with apertures of 20-35 μm;the deviation between dominant fracture orientations and the direction of maximum curvature is less than 10°.This fracture prediction method has guided subsequent well deployment in Dibei Area;the correspondence between predicted fractures and image log results reaches about 80%,fully demonstrating the method feasibility in guiding fracture prediction and optimizing drilling deployment,providing critical support for efficient development of deep coal-rock gas.

Key words: coal-rock gas, structural control, fracture prediction, Kizilnur Formation, Tarim Basin, multi-pixel