Physic-guided multi-azimuth multi-type seismic attributes fusion for multiscale fault characterization

Accurate characterization of the fault system is crucial for the exploration and development of fractured reservoirs. The fault characterization technique based on multi-azimuth and multi-attribute fusion is a hotspot. In this way, the fault structures of different scales can be identified and the characterization details of complex fault systems can be enriched by analyzing and fusing the fault-induced responses in multi-azimuth and multi-type seismic attributes. However, the current fusion methods are still in the stage of violent information stacking in utilizing fault information of multi-azimuth and multi-type seismic attributes, and the fault or fracture semantics in multi-type attributes are not fully considered and utilized. In this work, we propose a physic-guided multi-azimuth multi-type seismic attributes intelligent fusion method, which can mine fracture semantics from multi-azimuth seismic data and realize the effective fusion of fault-induced abnormal responses in multi-azimuth seismic coherence and curvature with the cooperation of the deep learning model and physical knowledge. The fused result can be used for multi-azimuth comprehensive characterization for multi-scale faults. The proposed method is successfully applied to an ultra-deep carbonate field survey. The results indicate the proposed method is superior to self-supervised-based, principal-component-analysis-based, and weighted-average-based fusion methods in fault characterization accuracy, and some medium-scale and microscale fault illusions in multi-azimuth seismic coherence and curvature can be removed in the fused result.