基于聚类和多元地质统计学的电—震联合建模约束反演技术及应用  

Technique and application of joint magnetotelluric and seismic modeling and constrained inversion based on clustering and multivariate geostatistics

作　　者：杨博 张祥国 刘展[1] 徐凯军[1] YANG Bo;ZHANG Xiangguo;LIU Zhan;XU Kaijun(Department of Geophysics,China University of Petroleum(East China),Qingdao,Shandong 266580,China;Research Institute of Exploration and Development,Daqing Oilfield Co Ltd,PetroChina,Daqing,Heilongjiang 163453,China)

机构地区：[1]中国石油大学(华东)地球科学与技术学院,山东青岛266580 [2]中国石油大庆油田公司勘探事业部,黑龙江大庆163453

出　　处：《石油地球物理勘探》2021年第3期670-677,I0016,共9页Oil Geophysical Prospecting

基　　金：国家重点研发计划项目"地下及井中地球物理勘探技术与装备"(2018YFC0603300)资助。

摘　　要：随着待勘探目标地质体越来越复杂且埋深增大,单一地球物理勘探方法的片面性和局限性日益突显,综合利用多种地球物理技术及相应数据已成为现今必然趋势。为此,提出基于多元地质统计学的交叉—变差函数建立速度与电阻率之间的岩石物理关系,并在此基础上利用机器学习中的引导模糊C均值聚类算法进行基于岩石物理关系的多重约束反演,实现电—震联合建模。大杨树盆地南部坳陷实际资料的应用结果表明,该电—震联合建模约束反演可逐步降低单一地球物理方法的多解性,提高对目标地质体的识别能力。非地震与地震方法所得结果相互印证,展示了该联合建模约束反演技术具有良好应用潜力。With more and more complex and deeper geological bodies,the one-sidedness and limitation of a single geophysical method have become more and more prominent,and comprehensive utilization of multiple geophysical technologies and data has become an inevitable trend.This paper proposes cross-variogram in multivariate geostatistics to establish the petrophysical relationship between velocity and resistivity.Based on the relationship,the guided fuzzy C-means clustering algorithm in machine learning is used to conduct multi-constraint inversion and realize MT-seismic joint modeling.Application to real data recorded from the southern depression of the Dayangshu Basin has proved that the MT-seismic joint modeling and multi-constraint inversion can gradually reduce the ambiguity of a single geophysical method and improve the ability to identify target geological bodies.The results of MT and seismic methods are verified each other,demonstrating that the joint modeling and multi-constraint inversion technology is potential in field application.

关 键 词：聚类分析 多元地质统计学 联合建模约束反演 

分 类 号：P631[天文地球—地质矿产勘探]