基于压缩感知和稀疏反演的表面相关多次波估计与成像方法研究

Estimation and imaging of surface-related multiples (i.e., all seismic waves reflected at the free surface at least once) are challenging topics in exploration seismology. Multiples always destructively interfere with primaries. Conventional processing techniques require input data that consist of primaries only. Failure to eliminate multiples may lead to erroneous estimates for migration velocity or result in strong coherent artefacts that interfere with the migrated reflectors. Recently an estimation of primaries by sparse inversion (EPSI) method was proposed to avoid limitations of prediction-subtraction technique. Then a robust EPSI (REPSI) method outperforms EPSI in term of flexibility. However, REPSI essentially relies on a rather complicated spectral-projected gradient (SPG) method, which hinders REPSI’s adoption by industry. Using the feedback model and a linearized Bregman method, this study proposed a simple and effective sparse inversion method for estimation of primaries, surface multiples, and source signatures. Multiples assist in overcoming amplitude-scaling and phase-shifting ambiguities of conventional problem in which multiples are not considered. The shot-to-shot constant/variant source wavelets are obtained via an exact line-search scaling. No assumption on the source signature need be made (zero- or minimum-phase). .Multiples are normally discarded as noise, but they carry complementary information compared to primaries, as they interact with the free surface and are therefore exposed more to the subsurface. Smaller reflection angles and longer travel paths make multiples supplement additional information and illumination otherwise absent in primaries. To utilize information carried by multiples, based on the theory of compressive sensing and sparse inversion, this study also proposed an efficient approach for imaging with surface-related multiples. Modelling of surface-related multiples in the described method derives from the feedback model (i.e., the measured seismic data themselves act as a secondary source to predict surface-related multiples) instead of putting a free surface boundary condition in the wave-equation modelling kernel. Different from the work by previous researchers, no monopole sources are injected at each physical source location (with a positive sign) and the mirrored location w.r.t the free surface (with a negative sign). Besides, the source wavelets, which are extremely crucial for imaging with primaries and surface-related multiples, are accurately estimated in company with multiples in the data space instead of during the imaging process, because in this way, multiple prediction need not know subsurface properties (e.g., velocity, density, elasticity, anelasticity, anisotropy) explicitly. This study represents a new step in estimation of surface-related multiples and provides a new perspective of imaging with primaries and surface-related multiples.

多次波估计与成像是勘探地震学研究热点。常规地震处理技术假设数据仅由一次波构成，多次波对一次波处理和成像造成强烈干扰。最近提出的稳健稀疏反演法通过迭代估计地下脉冲响应与子波，进而重构一次波与多次波，克服了预测相减法存在的问题，但其依赖极为复杂的谱映射梯度法，难以工业应用。本研究拟采用压缩感知领域简单高效的线性Bregman稀疏反演法，估计一次波、多次波和随炮集变化的子波，便于工业应用，验证多次波有助于克服传统子波估计存在的振幅和相位非唯一问题。.相比一次波，多次波传播路径更长、覆盖范围更广，本研究拟基于压缩感知和稀疏反演理论，利用多次波高效成像。子波对一次波多次波联合成像极为关键，其决定一次波和多次波相对关系。不同于前人在成像时估计子波，本研究在多次波估计时获得高精度子波，此时多次波预测不必考虑地层参数（速度、密度、粘弹性等）的影响。本研究发展了多次波估计新方法，提供了一种多次波成像新思路。

自由表面相关多次波估计与成像是油气勘探地震学研究的热点和难点。实际油气勘探采集的地震数据中自由表面相关多次波经常与一次波强烈干扰在一起。常规地震数据处理技术通常假设数据仅由一次波构成，自由表面相关多次波对一次波处理和成像造成强烈干扰，降低了地震数据处理与解释精度。近年来提出的稳健稀疏反演一次波估计方法（Robust estimation of primaries by sparse inversion，REPSI）通过迭代估计地下脉冲响应与地震子波，进而重构出一次波与自由表面相关多次波，克服了传统预测相减法存在的问题，但REPSI依赖极为复杂的谱映射梯度L1算法（Spectral-projected gradient L1，SPGL1），难以在实际油气工业界推广应用。由于实际数据采集时，每炮所激发的地震子波不尽相同，而目前几乎所有的自由表面相关多次波估计方法均假设所有地震子波相同，因此本研究开展了随炮集或空间位置变化的地震子波估计研究。本研究基于压缩感知理论，采用较为简单高效的加速线性Bregman稀疏反演法，估计一次波、自由表面相关多次波和随炮集变化的地震子波，便于在实际油气工业界推广应用。本研究验证了自由表面相关多次波有助于克服传统地震子波估计过程中存在的振幅和相位非唯一性问题。本研究研发了精确的线性搜索方法来估计随炮集变化的地震子波。相比一次波，多次波传播路径更长、覆盖范围更广，传统方法将估计得到的多次波看作噪声，而本研究基于压缩感知和稀疏反演理论，研发利用自由表面相关多次波高效成像方法。本研究将采集的地震数据作为二次震源来预测表面相关多次波。地震子波对一次波多次波联合成像极为关键，其决定一次波和多次波相对关系。不同于前人在偏移成像时估计子波，本研究在多次波估计过程中获得高精度子波，此时多次波预测不必考虑地层参数（如速度、密度、粘弹性等）的影响。本研究利用岩丘模型数据和开源的Pluto1.5数据验证了所提出方法的有效性。本研究发展了自由表面相关多次波估计新方法，研发了一种自由表面相关多次波成像新技术，在油气地震勘探工业界具有较好的市场需求和良好的应用前景。