基于HHT与PCA及K-maxmin聚类的地层划分方法及其应用  

Appication for sequence stratigraphy division based on Hilbert transform,principal component analysis and K-maxim clustering

作　　者：王彪 杨巍[1,2] 朱仕军[1,2] 王震 修金磊[4] 朱鹏宇 WANG Biao;YANG Wei;ZHU Shi-jun;WANG Zhen;XIU Jin-lei;ZHU Peng-yu(State key Laboratory of Oil and Gas Reservoir Geology and Exploitation,Southwest Petroleum University,Chengdu 610500,China;School of Geoscience and Technology,Southwest Petroleum University,Chengdu 610500,China;Exploration and Development Research Institute Northwest Branch of Sinopec,ürümqi 830011,China;Shengli Oilfield Branch of Sinopec,Dongying 257000,China;Southwest Geophysical Company of Dongfang Geophysical Prospecting Incorporated,Chengdu 610213,China)

机构地区：[1]油气藏地质及开发工程国家重点实验室,西南石油大学,成都610500 [2]西南石油大学地球学与技术学院,成都610500 [3]中石化西北分公司勘探开发研究院,乌鲁木齐830011 [4]中石化胜利油田分公司,东营257000 [5]东方地球物理勘探公司西南物探分公司,成都610213

出　　处：《地球物理学进展》2020年第5期1861-1869,共9页Progress in Geophysics

基　　金：国家重大科技专项“塔里木盆地碳酸盐岩油气田提高采收率关键技术示范工程”子课题“缝洞型油藏流体识别研究”(2016ZX05053-001-01-03);“中石化胜利油田分公司科技攻关项目”(YKK1807);“西南石油大学油气藏地震采集与反演青年科技创新团队”(2017CXTD08)联合资助.

摘　　要：划分不同级次的层序界面、识别地层内部的沉积旋回类型是层序地层分析的重要基础.利用主成分分析、希尔伯特黄等方法,有效地提取了隐藏在测井信号深度域和频率域中的丰富的地质信息,最终利用提取的信息进行地层层序的划分.结果表明:经过K-maxim聚类方法、主成分分析后的深度域信号能更加准确的识别薄层,而经过Hilbert-Huang变换后的频率域信号能更加准确识别突变点的位置,适用于确定层序界面.通过机器学习的思想以及信息融合技术,将2种方法结合.以测井数据的深度域信息为参考,以测井数据在频率域的突变点在沉积界面处反映灵敏为依据,将玛湖凹陷区NX1井处的三叠系百口泉组划分成1个长周期基准面旋回、1个短周期基准面旋回.利用该方法划分的结果不仅与已有的划分成果具有一致性而且其中的百口泉组的地层划分更加精细,为利用测井资料进行高分辨率层序地层划分提供了新思路.Dividing sequence boundary of different levels and identifying the type of sedimentary cycle within the stratum is an important basis for sequence stratigraphy analysis.Using principal component analysis,Hilbert Huang and other methods,the rich geological information hidden in the depth and frequency domains of the logging signal is effectively extracted,and finally the information is used to classify the stratigraphic sequence.The results show that the signal in the depth domain after K-maxim clustering method and principal component analysis can identify the thin layer more accurately,and the signal in the frequency domain after Hilbert-Huang transform can identify the location of the mutation point more accurately,which can determine the sequence stratigraphy.Through the idea of machine learning and information fusion technology,the above two methods are combined.Taking the depth domain information of the logging data as a reference and based on the mutation point in the frequency domain of the logging data sensitively reflecting sediment interface,The Triassic Baikouquan formation at the NX1 well in the Mahu depression area is divided into a long-period base-level cycle and a short-period base-level cycle.The results of this method are not only consistent with the existing classification results,but also the stratigraphy division of the Baikouquan formation is more elaborate,which provides a new idea for using logging data for high-resolution sequence stratigraphy division.

关 键 词：层序划分 主成分分析 K-maxim聚类 经验模态分解 HHT变换 

分 类 号：P631[天文地球—地质矿产勘探]