基于勘探地震学介质特征高效稳健的地震波保真成像方法研究

Least-square reverse time migration (LSRTM) is now a standard seismic amplitude-preserving imaging method for estimating high-wavenumber parameters in exploration seismology. LSRTM usually tends to coverage very slow or even divergence in field data application, the following are the two factors that are most likely to be ignored when analyzing convergency of LSRTM. First, the commonly used approximation for linearized scattering wavefield calculation in LSRTM is based on volume or surface scattering assumption alone, which is not exactly consistent with the medium characteristic in exploration seismology that there exist widespread reflecting surfaces together with some anomalous geological bodies. Secondly, as one of the key steps in LSRTM, the implementation of preconditioned gradient relies on an iterative approach in a black box manner, the comprehensive consideration of preconditioned gradient’s physical meaning is absent. As result, when removing wave propagation effects in an inversion imaging process, various effects are coupled with each other and the coverage rate of inversion is affected. In this proposal, we start from the forward problem involved in LSRTM. We focus on developing new scattering wavefield calculation method that is more suitable for subsurface medium in exploration seismology. The new linearized scattering wavefield calculation method aims to reduce the complexity of data misfit. Then under certain assumptions, we propose new explicitly expressed preconditioned gradient term with clear physical meaning. The present preconditioned gradient is designed to stably remove the main propagation effect in the early stage of iterations so that the whole coverage rate can be accelerated. Finally, a new stable and efficient seismic amplitude-preserving imaging method with better suitability for imaging subsurface medium in exploration seismology will be proposed. We will also verify the effectiveness our present method with complex field dataset.

最小二乘逆时偏移（LSRTM）是当前勘探地震学中估计高波数参数标志性的地震波保真成像方法。LSRTM在实际应用中易出现收敛慢甚至不收敛的情况，以下两点是易被忽视的影响LSRTM收敛性的重要因素。首先，LSRTM中线性化散射场计算采用单一的体或面散射模式，这与勘探地震学面对的以层状反射界面为主，同时存在若干地质异常体的介质特征不完全吻合。其次，对于LSRTM中核心步骤之一预条件梯度项的实施，通常采用类黑匣子式的迭代求解方式，缺乏对该项物理意义的充分考虑，使得在反演成像过程中消除波传播效应时多种因素相互耦合，影响收敛速度。本项申请拟从LSRTM的正问题出发，探索更适合勘探地震学介质特征的散射场计算方法，减少数据匹配难度。然后，在一定的假设下，给出物理意义明确的预条件梯度项的显式表达，用以在迭代中首先以稳定的方式消除主要波传播因素，加快收敛速度。最终本项申请拟形成一套针对勘探地震学介质特征高效稳健的地震波保真成像方法，并在实际数据中取得效果。

随着油气勘探目标的不断精细，对高分辨率、高保真地震波成像方法提出了客观需求。随着“两宽一高”地震数据采集技术的进步，为实现高精度地震波保真成像奠定了良好的数据基础。.最小二乘偏移是目前最具有代表性的高精度地震波保真成像技术，其试图通过匹配模拟数据和观测数据的振幅值来估计参数的高波数扰动。然而，最小二乘偏移的理想实现需将Hessian逆算子作用到常规成像结果上，对于大规模实际数据的处理，这在计算和存储量上都较难承受。此外，Hessian逆算子的估计精度很容易受到叠前数据观测不规则、正算子对物理波现象的描述不精确以及地震子波未知等多种因素影响，从而出现迭代收敛慢甚至不收敛情况，阻碍了最小二乘偏移成像在大规模实际资料中实用化进程。.因此，针对目前地震波保真成像实际应用中存在的计算不稳定、计算效率低、成像结果分辨率及保真度提升不明显等问题，有必要发展一套计算稳健高效、成像结果分辨率高、保真度好的地震波保真成像方法。为实现以上目标，本项目在Hessian 算子的近似显式表达，超级分辨率成像以及密度效应补偿等方面开展了研究, 取得了如下重要结果：.（1）.构造了包含最主要波场传播效应的显式Hessian算子及其相应逆算子对这些波场传播效应进行消除，较大程度上提高了成像质量；.（2）.构造了快速算法计算局部角度域点扩散函数，然后将成像结果与角度域点扩散函数的高维反褶积过程构造为稀疏约束的超级分辨率反演成像问题，较大程度上提高成像结果的分辨率；.（3）.通过分析波场在常密度和变背景密度介质中的传播规律，构造变背景密度介质中振幅补偿算子，对由于变密度导致的成像振幅异常进行校正，为后续宽频带波阻抗反演提供良好的输入数据。.以上相关结果本项研究发展了一套反问题凸性更好、计算高效稳健的地震波保真成像方法在理论和实际应用方面都具有较为重要的意义。