利用字典学习快速匹配追踪压制煤层强反射  

Strong coal-seam reflection suppression by fast matching pursuit based on dictionary learning

作　　者：韩站一[1] 宋炜[2] 李继光[1] 赵翠霞[1] 赵爱国[1] 马晓义 HAN Zhanyi;SONG Wei;LI Jiguang;ZHAO Cuixia;ZHAO Aiguo;MA Xiaoyi(Geophysical Research Institute,SINOPEC Shengli Oilfield,Dongying,Shandong 257022,China;College of Geophysics,China University of Petroleum(Beijing),Beijing 102249,China)

机构地区：[1]中国石化胜利油田分公司物探研究院,山东东营257022 [2]中国石油大学(北京)地球物理学院,北京昌平102249

出　　处：《石油地球物理勘探》2022年第5期1156-1163,I0008,共9页Oil Geophysical Prospecting

基　　金：中国石化股份公司项目“西部探区高密度地震采集技术应用研究”(P20034-4)资助。

摘　　要：煤层强反射信号上、下和内部通常隐藏着弱有效信号,如何有效分离煤层强反射信号周围的弱有效信号,受到了越来越多的关注。匹配追踪算法是解决该难题的有效算法之一。然而,匹配子波母函数早期用Gabor子波,现在常用Morlet小波和Ricker子波等,而这些子波都是理论子波,和实际地震资料波形特征存在较大差异,因此在匹配过程中会产生残余分量。针对该问题,通过引入了字典学习算法,从地震数据中学习获得字典的有限匹配子波而不是预先定义字典,极大提高了信号分解的匹配度。针对目标区煤层反射不稳定的问题,选择按单个同相轴开展字典学习,从而构建了振幅、相位、主频和极性变化的单峰字典库。通过叠后、叠前的衰减煤层强反射处理分析,验证了新方法在强反射压制方面比常规方法更可靠且有效。应用基于匹配子波的Wigner-Ville高分辨率时频分析方法,揭示了煤层强反射消除前、后时频属性特征,为煤层附近储层的精确描述提供了技术支撑。Weak effective signals are usually hidden in the upper,lower,and internal parts of strong coalseam reflection signals.The research on methods to effectively separate weak effective signals from strong coal-seam reflection signals has attracted more and more attention,and matching pursuit is one of the effective algorithms to solve this problem.In the early days,the Gabor wavelet was used as the generating function of the matching wavelet,some other theoretical wavelets such as Morlet and Ricker wavelets are commonly used now,which are greatly different from the waveform characteristics of actual seismic data,which results in the generation of residual components in the matching process.To solve these problems,this paper introduces the dictionary learning algorithm to learn the atoms of the dictionary in the seismic data instead of predefining the atoms,which greatly improves the matching degree of signal decomposition.Aiming at the problem of unstable coal-seam reflection in the target area,we carry out dictionary learning according to a single event to build a unimodal dictionary with different amplitude,phases,dominant frequency,and polarity.Through the analysis of the attenuation of strong coal-seam reflection of post-stack and prestack data,it is verified that the new method is more reliable and effective than the conventional method in strong reflection suppression.The highresolution time-frequency analysis of Wigner-Ville distribution based on the matching wavelet reveals the time-frequency attribute characteristics before and after the elimination of strong coal-seam reflection,which can provide technical support for accurate description of reservoirs near coal seams.

关 键 词：煤层强反射 匹配追踪 字典学习 WIGNER-VILLE分布 匹配子波 

分 类 号：P631[天文地球—地质矿产勘探]