基于特定反射系数压制与最大似然属性的断层识别方法

The fault identification method based on specific reflection coefficient suppression and maximum likelihood attribute

甄宗玉 陈华靖 张鹏志 郑江峰

中海石油（中国）有限公司天津分公司，天津 300459

Tianjin Branch of CNOOC Ltd., Tianjin 300459, China

方差、相干等方法在断层识别中的应用已较成熟，但这些方法应用于复杂断裂发育区存在断面表征模糊、平面组合不清晰等缺点。基于相似系数改进的最大似然属性，利用断层倾角和方位角扫描搜索似然属性的最大值进行断层识别，可以更清晰地表征断面和断层组合。在河流相储层广泛发育的复杂断块油田中，河道砂体的地震响应多表现为强振幅反射，砂体边界引起的同相轴尖灭会导致最大似然属性将其错误地识别为断层响应。针对上述问题，文中提出了基于特定反射系数压制与最大似然属性的组合方法进行断层识别：首先分析最大似然属性的基本原理，并发现了该属性在河流相储层干扰下的缺点；随后利用谱反演和层位自动解释的特定反射系数压制方法，压制河道砂体的响应；最后基于压制河道响应后地震数据，利用最大似然属性进行断层识别。模型数据和实际地震资料应用表明，该组合方法具有良好的应用效果。

Variance and coherence methods have been widely used in fault identification, but in complex fault development areas, these methods show some shortcomings such as fuzzy description in fault surface and unclear fault combination. The modified maximum likelihood attribute based on similarity coefficient can describe the fault surface and fault combination clearly by fault dip and azimuth scanning. In the complex fault-block oil fields with fluvial reservoirs widely developed, the seismic response of channel sand body shows strong amplitude reflection, and the maximum likelihood also may recognize the channel boundary as fault response due to the boundary of sand also caused the pinch-out of seismic event which is similar to fault response. In order to solve the problem, we proposed a combination method based on specific reflection coefficient suppression and maximum likelihood attribute for fault identification. Firstly, this paper analyzed the basic principle of the maximum likelihood and found its disadvantages under the interference of fluvial reservoir. Then, the method of specific reflection coefficient suppression was used to suppress the response of channel sands based on spectral inversion and horizon automatic interpretation. Finally, based on the seismic data of suppressed channel response, the fault identification was carried out by using the maximum likelihood attribute. The application of model data and actual data shows that the combination method has good application effect.

10.6056/dkyqt202103009