利用平滑模型约束的频率域多尺度地震反演

Seismic multi-scale inversion in the frequency domain based on smooth model constraint

频率域地震反演是利用带限地震信号的频率域响应信息预测出地层模型参数。针对频率域反演的高分辨率特性和地震信号的带限特征,本文综合探究频率域反演和贝叶斯反演的思路,提出了基于平滑模型约束的频率域多尺度贝叶斯反演方法。通过对目标函数加入平滑模型约束信息,有效地解决了带限地震信号的低频分量缺失问题,并进一步提高了频率域反演过程的抗噪性和反演结果的空间连续性。此外,由于地震信号在频率域中实现了多频率分量的自动解耦,从而可通过多频率分量逐级迭代方式搜索到反问题的最优解,进而提高反演结果的收敛精度和分辨率,且减弱了反演算法对初始模型精度的过度依赖。最后,通过平稳和非平稳理论模型测试及实际资料的处理,验证了该方法的高分辨率特性和稳健性。

Using partial frequency components of band-limited seismic data, an inversion in the frequency domain yields stratigraphic impedance information. In view of high resolution in the frequency domain inversion and band-limited characteristics of seismic signals, we consider jointly the frequency domain inversion theory and Bayesian framework, and propose a robust seismic multi-scale inversion method with smooth model constraint in the frequency domain. With smooth model constraint, this method not only enhances low frequency compensation, but also improves noise removal and lateral continuity of the inversion. Since decoupling different frequency component decoupling of seismic gathers can be realized in the frequency domain, successive iterations of different frequency components can obtain the optimal solution of the inverse problem. Tests on stationary, non-stationary models, and seismic data demonstrate the feasibility and stability of our proposed method. © 2016, Editorial Department OIL GEOPHYSICAL PROSPECTING. All right reserved.