深度学习与边缘增强相结合的断裂综合检测技术——顺北地区超深走滑断裂检测应用实例

塔里木盆地顺北地区超深断控缝洞型储集体发育，近年来不断有高产井涌现； 同时，大量研究证实高角度走滑断裂对油气藏的运移和聚集起决定性作用。受“断控储集体”埋藏深，且走滑断裂断距小、难闭合等因素的影响，顺北地区地震资料信噪比低、断面特征不清晰，导致走滑断裂检测及空间解释难度较大。针对上述超深走滑断裂检测研究面临的难点，文中提出深度学习与边缘增强相结合的多尺度断裂综合检测技术：首先将走滑断裂按规模划分为主干断裂、伴生次级断裂、小尺度裂缝；通过正演主干断裂、次级断裂、裂缝等不同断裂模式的地震响应特征并进行方法实验，认为可应用U-Net卷积神经网络深度学习技术识别主干断裂、振幅梯度矢量凌乱性检测技术识别伴生次级断裂、Aberrance增强属性识别小尺度裂缝。将该套技术应用于顺北地区走滑断裂实际检测中，取得了显著效果。

Application of comprehensive fault detection technology combining deep learning with edge enhancement in detecting ultra-deep strike-slip faults in Shunbei block

Karst fracture-cave reservoirs are well-deve-loped in the Shunbei block of the Tarim Basin，and high-productivity wells have emerged in recent years. A large number of studies have confirmed that the development of strike-slip faults with high angles plays a decisive role in the migration and accumulation of oil and gas reservoirs. Due to the deep burial of fault-controlled reservoir small fault throws，and hard closure，the SNR of seismic data in the Shunbei block is low，and the characteristic of fault planes is not clear，which make the detection and spatial interpretation of strike-slip faults difficult. Given the difficulties faced by studies on ultra-deep strike-slip fault detection，this paper proposes a comprehensive detection technology combining deep learning with edge enhancement for multi-scale faults. Specifically，the paper divides the strike-slip faults into main faults，associated secondary faults，and small-scale fractures by scale and carries out targeted studies. According to the seismic response characteristics of different fracture modes including forward main faults，associated secondary faults，and small-scale fractures and method experimental tests，it is believed that U-Net convolutional neural network deep learning technology can be used to identify main faults，and amplitude gradient vector disordered detection technology can be applied to identify associated secondary faults. In addition，the Aberrance enhancement attribute can be adopted to identify small-scale fractures. The proposed technology has been applied to detect strike-slip faults in the Shunbei block and achieved remarkable effects.