沙漠地震勘探四维非均匀波特性随机噪声数理表征及其非线性降维消噪系统

Seismic prospecting is an important means for oil and gas resources exploration. Seismic random noise source in desert region is formed by wind, near-field and far-field humanity activities. The wind as the main contributing factor can be characterized by the superposition of the point source effect. The change of dune, wind presssure and surrounding geographical environment makes seemingly simple desert region produce random noise with complex characteristics, including the low frequency, changing nonstationarity, nongaussianity, quasilinear, instantaneity, inhomogeneous phase transition undulatory property etc. This kind of noise limits the effect of the existing excellent denoising technology. This project applied modern statistical theory to complete qualitative and quantitative analysis of the spatiotemporal properties of this noise; Using Sommerfeld integral, the simulation of four dimensional inhomogeneous wave of seismic random noise in desert were gradually obtained by inhomogeneous surface wave decomposition of point source and the solution of inhomogeneous wave equation; The mathematical characterization of low-frequency random noise in desert region were established. This project further constructed nonlinear dimensionality reduction model with sub-block feature based nonlinear dimensionality reduction theory and established the block matching criterion. Finally block sparse compressed sensing theory was applied to high precision-pick up and reconstruct effective events. The completion of this project will be contributed to establish the mathematical characterization and reduction system of desert low-frequency random noise systematically. Meanwhile, the theory of seismic random noise is improved and the digital processing level of random noise attenuation is enhanced.

地震勘探是油气资源探查的重要手段。沙漠地带地震随机噪声源由风作用、近场和远场人文活动组构，作为主要成因的风作用可由点源效应的叠加所表征。由于沙丘变化、风压的时空变化、周边地理环境的不同使看似简单的沙漠地带产生了具有复杂特性的随机噪声，包括低频性、变化的非平稳性、非高斯性、非时限性、非均匀变相波动特性等。这类噪声限制了已有的优秀的消噪技术的效果。本项目应用现代信号统计理论对其时空特性进行定性定量分析；并利用Sommerfeld积分，经过点源的非均匀相变波分解，求解非均匀波动方程，逐步得到沙漠地震随机噪声的四维非均匀波特性的模拟；建立沙漠地带低频随机噪声数理表征。基于非线性降维理论构建具有子块特征的非线性降维模型，建立块匹配准则，应用块稀疏压缩感知理论拾取并重构有效信号同相轴。此项目将全面系统地完成沙漠地震随机噪声的数理表征并建立对应的消减系统。完善随机噪声理论，提升地震勘探消噪处理水平。

当今油气能源供需关系日渐紧张，迫使油气勘探开发向“深、低、非、海”领域延伸。塔里木沙漠地区作为“超深地，非常规勘探”的典型区域，是油气勘探的重点目标区域。面对沙漠环境复杂的随机噪声干扰，建立沙漠随机噪声数理表征模型及有效的应对方案至关重要。本项目基于上述背景开展了沙漠地带四维非均匀波特性随机噪声数理表征及其消减系统的研究，主要内容及成果如下：构建“源-场-库”一体化的四维非均匀噪声场的系统数理描述；采用现代统计学理论对噪声关键特性进行精准建模，结合信噪特征差异建立稀疏分解模块化导向下网络特征提取及新型自适应阈值优化匹配去噪框架；应用地震数据几何纹理特征指导噪声水平估计，以稀疏形态分解方式同步实现矩阵稀疏和低秩先验信息增强，全面提升数据多层次细节丰富度，建立系统化、智能化、可实现同频带信噪有效分离的沙漠地震噪声消减系统；以深度学习智能模型深入挖掘沙漠地震勘探数据中的深层次地质结构信息，开发“模拟增广和真实数据并重”的训练策略，同步扩建完备沙漠地震资料消噪训练集，确立CNN架构下沙漠地震勘探智能消噪体系，满足智能消噪模型处理三维地震不同测线接收数据的高标准、高效率、高精度、普适性及泛化性的消噪要求。项目中所有研究方案及实验结果均经过大量不同参数仿真数据实验验证，同时应用于塔里木盆地沙漠测区（但不限于）的野外实际采集地震数据。结果表明本研究项目的消噪方法可很好提升地震勘探记录质量，提升勘探资料信噪比。项目成果以第一标注在国内外期刊发表SCI检索论文38篇，国际会议论文及汇报6篇，获得授权发明专利11项。该项目成果可为获取沙漠地带地震勘探高质量地震资料及分析地球内部精细结构提供强有力的技术保障。