基于包络反演的高低波数同步反演方法

全波形反演方法存在过于依赖初始模型及局部极值等问题。为此,利用坡印廷矢量进行梯度分解并构建高低波数同步反演方法,同时将该方法与包络反演相结合,提出了一种提高全波形反演方法稳定性的分步多尺度反演策略,即首先将线性速度模型作为初始速度模型进行包络反演,以构建浅层背景速度场;再将反演结果作为初始速度,利用坡印廷矢量实现偏移分量及层析分量的分解和同步迭代反演;最终构建扰动速度场及中、深层背景速度场。迭代反演过程中,将利用层析分量得到的梯度更新量补偿到常规反演梯度中,从而恢复中、深层低波数速度模型,同时避免了偏移/反偏移计算,减少了计算量。将该方法应用于Marmousi2模型数据的反演结果表明,基于包络反演的高低波数同步反演方法对中、深层背景速度恢复能力强;误差曲线表明,基于包络反演的高低波数同步反演方法的收敛误差小、收敛速度快且稳定性强,反演得到的速度模型为油气预测奠定了基础。

The Full-Waveform Inversion (FWI) presents several issues,such as the strong dependence on the initial model and local extremum.A synchronous inversion of the high and low wavenumber is proposed in this study,which is based on the gradient decomposition by means of the Poynting vector.By combining synchronous and envelope inversions,a multi-scale inversion strategy is proposed to improve the stability of the FWI.Firstly,the background velocity field of the shallow layers is established by envelope inversion,using the linear model as the initial model.Subsequently,the Poynting vector is utilized to decompose the migration and tomography components,and a simultaneous iterative inversion is implemented to obtain the perturbation and mid-deep background velocity field.During the iterative process of inversion,the updated gradient calculated from the tomographic component is added to the conventional gradient,so that the mid-deep low wavenumber velocity can be restored.Meanwhile,the processes of migration and de-migration are avoided,thus reducing the computational cost.A model test showed that the high and low wavenumber synchronous inversion can successfully update the background velocity in the mid-deep layer.The error curve indicates that the convergence error is small,while the convergence speed is fast,which confirms the effectiveness and stability of the method to construct the velocity model.

国家自然科学基金(41574125、41774139)、国家科技重大专项(2017ZX05018-005)、中国石油天然气集团有限公司科研项目(2016A-33)和国家重点研究发展计划(2017YFB0202903)共同资助。