基于线性预测倒谱系数的地震相分析  

Seismic facies analysis based on linear prediction cepstrum coefficients

作　　者：解滔[1,2] 郑晓东[1] 张? XIE Tao ZHENG Xiao-Dong ZHANG Yan(Research Institute of Petroleum Exploration ~ Development, Beijing 100083, China China Earthquake Networks Center, China Earthquake Administration, Beijing 100045, China)

机构地区：[1]中国石油勘探开发研究院,北京100083 [2]中国地震台网中心,北京100045

出　　处：《地球物理学报》2016年第11期4266-4277,共12页Chinese Journal of Geophysics

基　　金：国家重点研发计划重点专项(2016YFC060110701);国家自然科学基金(40504110)联合资助

摘　　要：本文借鉴语音识别技术中的线性预测倒谱系数(LPCC系数)特征参数提取方法对地震数据进行分解,这种方法的优点是:可以获得将子波和反射系数信息分离的地震语音特征参数,对地质现象边界具有较好的描述能力,使我们可以从不同维度更细致地观察隐藏在地震数据中的地质特征.理论模型分析表明,基于LPCC系数的地震分析具有较高的地震相划分能力.实际地震资料应用表明,LPCC系数对储层特征的描述比常规三瞬属性更为细致,不同阶次LPCC系数在描述储层不同特征时也保持了内在的联系.采用K均值聚类方法对提取的12阶和24阶LPCC系数进行聚类分析,聚类结果与目的层段古地形较为吻合,较好地反映了研究区的断裂、礁滩相带、深水扇和储层的分布特征,说明在地震相分析中采用LPCC系数作为特征参数是可行和有效的.In this paper,the linear prediction cepstrum coefficient（LPCC）,which is widely and successfully used in speech recognition,is introduced to extract multi-dimensional feature parameters for seismic facies analysis.The merit of seismic speech features is that the wavelet and reflection coefficients information in LPCC are well separated,which allows interpreters to more effectively detect geologic characteristics hidden in seismic data from different dimensions,especially the recognition of geologic boundaries.The results from theoretical modeling indicate that seismic analysis using LPCC can achieve good seismic facies division.Through analysis on real seismic data,LPCC displays its advantages in more detailed description of reservoir characteristics compared with the conventional instantaneous amplitude,frequency and phase attributes.LPCC of different orders keeps inner relations while they describe different aspects of the reservoir.In order to display the robustness of LPCC,K-means algorithm,the simplest and widely used method,is employed to group the LPCC of 12 and 24orders extracted from the target interval,respectively.The clustering results show the good accordance between seismic facies distribution and paleogeomorphology of the target interval in analysis block.The characteristics of faults,reef facies belts,deep-water fans and reservoirs in the analysis block are well displayed.Theresults also demonstrate the feasibility and effectiveness of LPCC in seismic facies analysis.

关 键 词：线性预测倒谱系数 地震相分析 储层预测 K均值聚类 语音识别 

分 类 号：P631[天文地球—地质矿产勘探]