结合CEEMDAN和主成分分析的低信噪比微地震初至信号检测  

A first arrival detection method in low SNR microseismic signals based on CEEMDAN-PCA

作　　者：胡瑞卿 王彦春[2] 尹志恒 王伟 HU Ruiqing;WANG Yanchun;YIN Zhiheng;WANG Wei(Geodetection Laboratory, Ministry of Education, China University of Geosciences ( Beijing), Beijing 100089 China;School of Geophysics and Information Technology, China University of Geosciences (Beijing), Beijing, 100089 China;SINOPEC Geophysical Corporation, Beijing 100020, China;BGP Training Center, CNPC, Zhuozhou, Hebei 072750, China)

机构地区：[1]中国地质大学(北京)地下信息探测技术与仪器教育部重点实验室,北京100089 [2]中国地质大学(北京)地球物理与信息技术学院,北京100089 [3]中石化石油工程地球物理有限公司,北京100020 [4]东方地球物理公司培训中心,河北涿州072750

出　　处：《石油地球物理勘探》2019年第1期45-53,I0002,共10页Oil Geophysical Prospecting

基　　金：国家科技重大专项"前陆冲断带及复杂构造区地震成像关键技术与构造圈闭刻画"(2016ZX05003-003)和地下信息探测技术与仪器教育部重点实验室开放课题"Bayes框架下优化遗传退火反演方法研究"(GDL1609)联合资助

摘　　要：微地震资料信噪比过低,传统方法的初至拾取精度与稳定性大多不理想。将基于自适应噪声完备经验模态分解(CEEMDAN)与主成分分析(PCA)相结合,有效地实现了低信噪比资料中的初至特征检测。针对低信噪比微地震资料进行CEEMDAN处理后,对各阶本征模态函数(IMF)进行PCA,再对各阶IMF的主成分进行加权重构,同时对次要成分进行压制与剔除,使三分量信号中具有较强一致性的初至信息得以保留。设计多组不同信噪比的测试信号,对方法的可行性进行测试,并最终应用于三分量实测信号。结果表明,该方法在极低信噪比条件下仍可实现对微地震信号初至的有效识别与检测。In most cases, conventional methods cannot extract precise and stable first arrivals from microseismic signals due to its low signal-to-noise ratio (SNR). To overcome this difficulty, we propose here a new method based on comprehensive ensemble empirical mode decomposition ( CEEMDAN ) with adoptive noise and principal component analysis (PCA). First PCA of each order intrinsic mode function ( IMF) is carried out after CEEMDAN processing. Then the main components of each order of the IMF are reconstructed and the secondary components are suppressed or removed so that the initial information with strong consistency in threecomponent signals can be retained. Tests on multiset synthetic data with different SNR and real data for this method were carried out. The test results provide a solid evidence for the feasibility of the proposed method in first arrival extraction from low SNR microseismic signals.

关 键 词：微地震事件 初至信号检测 自适应噪声完备经验模态分解 主成分分析 

分 类 号：P631[天文地球—地质矿产勘探]