基于粘滞声波方程的局域相空间反Q偏移方法研究

Seismic wave propagation in real media is different in many aspects from the ideal solid media. The anelasticity of the medium causes dissipation of seismic energy, thus decreasing the amplitude of high frequencies of the propagating wavelet and causing high-frequency wave components to travel faster than low-frequency waves. The two effects, dissipation and dispersion, are simply referred to as the attenuation effect. These effects must be considered to extract more detailed information about the subsurface from seismic data or to construct images with high resolution. The attenuation effect can be described by the earth quality factor, the Q factor. In China, the lithologic oil and gas reservoirs are developed in thin beds structure with high spatial heterogeneities. To solve the problem of attenuation compensation and strong lateral inhomogeneity, we propose the local phase space wave field downward continuation attenuation compensation imaging method, that is, viscoacoustic beamlet domain inverse Q migration. The localization of the phase space wavefield downward continuation operator has the ability to adapt to lateral variation of velocity and Q field, improves the wave propagation accuracy, produces high resolution migration and imaging results. In the attenuation compensation algorithm, the amplitude of the back propagation is greater than 1.0 to amplify the energy loss. So the stability is the key to the inverse Q migration. In this project, we study the stability issue in the beamlet decomposition (local wavenumber domain filter or decomposition with attenuation compensation) and local wavefield downward continuation process, aiming to develop a local phase space attenuation compensation migration method. Filed data will be tested to verify the application on target area.

由于实际地下介质具有粘滞效应，地震波在传播时能量会被介质吸收而发生衰减和频散，表现为频带变窄、主频降低、能量变弱。发展高精度成像或储层预测技术，必须对地层吸收效应进行补偿。针对我国岩性油气藏呈薄互层结构、油气储层横向非均质强、地层倾角小的特点，本项目提出基于局域相空间波场延拓的吸收补偿成像方法，即粘声波介质小波束反Q偏移成像。局域相空间波场延拓算子空间上具有局域化性质，可适应速度和品质因子的横向变化，提高波的传播精度，实现高精度偏移成像。反Q偏移成像中，波场延拓的稳定性是核心问题，本项目拟通过小波束分解（数据局域波数域滤波、带吸收衰减补偿的小波束分解）和局域波场延拓等环节研究使波场延拓稳定的技术方法，发展出完善的局域相空间吸收补偿偏移成像方法，并在实际数据中测试其应用效果。

本项目首先完善了Dreamlet偏移成像的理论和算法.Dreamlet偏移成像目的是探索一类能够对地震波场和单程波传播算子同时分解和压缩的理论和方法，也即实现在压缩域的传播与成像．本项目研发了Dreamlet传播算子与观测系统沉降法偏移相结合的理论与方法．Dreamlet观测系统沉降过程只保留用于成像观测系统下部地质结构的有效数据，自动丢弃已经用于成像观测系统上部而对下部成像没有贡献的信号．通过二维SEG/EAGE叠后和Marmousi叠前数据算例展示了Dreamlet传播算子应用于观测系统沉降法偏移的这一特点．.其次，本项目提出了一种变网格有限差分地震波场模拟理论与算法.由于重力引起的岩石压实效应，一般来说，地震波传播速度由浅入深整体逐渐增大.梯形坐标系设计可耦合速度由浅入深逐渐增大的变化，该坐标系中均匀网格采样所对应的物理直角坐标系网格由浅入深逐渐增大，也即浅部低速区对应细网格，深部高速区对应粗网格．在Marmousi模型上的测试结果表明，在模拟精度可比的条件下，本项目提出的方法是传统均匀网格有限差分算法效率的2.9倍，而内存效率仅为其75%..再次，本项目提出了三维共偏移距叠前地震数据反假频射线束形成偏移成像技术. 对稀疏／非规则采样或者低信噪比数据，射线束提取困难并伴随有假频产生，对叠加剖面和道集造成严重干扰．为了提升射线束偏移在稀疏和低信噪比地震数据采集中的成像效果，本项目提出基于三角滤波的局部倾斜叠加波束形成偏移假频压制方法．以某海上三维实际数据为例，文中展示了反假频射线束形成对偏移叠加剖面和共成像点偏移距道集中的噪声进行了有效压制..最后，根据国家战略以及本学科的发展特点，本项目探索了机器学习和人工智能方法在地学大数据中的应用. 本项目以甘肃省平凉市灵台县为目标研究区域，基于地理空间和历史滑坡数据，利用四种机器学习模型构建滑坡易发性评价分析模型.本项目还探索了基于残差卷积神经网络的波阻抗反演方法，并引入迁移学习提高其在不同特征空间模型上的应用效果.