L1范数约束的快速多道叠前反演  

Fast multichannel pre-stack inversion constrained by L1 norm

作　　者：秦思萍 贺锡雷[1] 谌洪平 张陈强 张雨强 雷扬 陈明春 QIN Siping;HE Xilei;CHEN Hongping;ZHANG Chenqiang;ZHANG Yuqiang;LEI Yang;CHEN Mingchun(College of Geophysices,Chengdu University of Technology,Chengdu 610059,China;Nanfang Branch,Sinopec Geophysical Corporation,Chengdu 610041,China)

机构地区：[1]成都理工大学地球物理学院,成都610059 [2]中石化石油工程地球物理有限公司南方分公司,成都610041

出　　处：《物探化探计算技术》2025年第2期159-168,共10页Computing Techniques For Geophysical and Geochemical Exploration

摘　　要：叠前反演在求取储层弹性参数和油气储层预测方面具有重要作用。传统的叠前反演忽略了相邻道之间的相关性,多道叠前反演方法充分考虑了道间的关联性,增强了反演结果的横向连续性。现有的多道反演算法倾向于将整个地震数据重新排列成一列进行反演,占用计算机大量内存且计算量大。基于地震反射系数的强稀疏性,笔者将相对于普通稀疏正则化方法精度更高、抗噪音性更好的全变分正则化技术引入多道叠前弹性反演,采用基于L1范数约束的全变分正则化目标函数对反演参数进行横纵向约束,避免了重新排列地震道,提高了计算效率和反演精度,同时也降低了对内存的需求。通过理论模型测试和实际运用,从反演的精度、横向连续性和抗噪能力分析,验证了所提方法的正确性和可靠性。Pre-stack inversion plays an important role in obtaining reservoir elastic parameters and predicting oil and gas reservoirs.The traditional pre-stack inversion neglects the correlation between adjacent channels,In contrast,the multi-channel pre-stack inversion method fully considers the correlation between channels and enhances the horizontal continuity of inversion results.The existing multichannel inversion algorithms tend to rearrange all seismic data into a column for inversion,which occupies a large amount of computer memory and requires a large amount of computation.Based on the strong sparsity of the seismic reflection coefficient,this paper introduces multi-channel pre-stack elastic inversion using the total variational regularization technique,which has higher accuracy and better noise resistance than the ordinary sparse regularization method.It uses the total variational regularization objective function based on L1 norm constraint to constrain the inversion parameters horizontally and horizontally,which avoids the rearrangement of seismic tracks and improves computational efficiency and inversion accuracy.It also reduces the memory requirements.The correctness and reliability of the proposed method are verified by the analysis of inversion accuracy,transverse continuity and anti-noise ability through theoretical model tests and practical application.

关 键 词：叠前反演 稀疏正则化 L1范数 全变分正则化 多道反演 

分 类 号：P618.13[天文地球—矿床学]