AVA inversion with adaptive selection of regularization parameters based on generalized Stein's unbiased risk estimation
Received date: 2023-11-22
Online published: 2025-01-14
Copyright
Appropriate regularization parameters are a key point in the geophysical inverse problems. It is related to the quality of inversion results. In AVA inversion of pre-stack seismic data, the frequently-used method is quality control to select the regularization parameters. This method adjusts the value of regularization parameters, and performs the inversion with near-well seismic traces, and selects the one whose inversion results has the best match with the actual well log. The selected regularization parameters are used to the inversion for whole seismic traces. Hence, it is based on trails which are very empirical and circumscribed. Due to heterogeneity of underground medium, the selected regularization parameters from quality control with near-well seismic traces may be not appropriate for the whole seismic traces. To slove this problem, this paper adopts a new method to select regularization parameters. The new method solves the generalized Stein's unbiased risk estimation function and adaptive estimates the regularization parameters in the process of inversion. From the results of model numerical test and real seismic data application, one can see that the AVA inversion with adaptive selection of regularization parameters increase the quality of inversion results.
Ying HU , RongHuo DAI , Cheng YIN , Jun YANG . AVA inversion with adaptive selection of regularization parameters based on generalized Stein's unbiased risk estimation[J]. Progress in Geophysics, 2024 , 39(6) : 2265 -2274 . DOI: 10.6038/pg2024HH0424
图2 真实模型参数与两种AVA反演结果对比(a)纵波速度;(b)泊松比;(c)密度.蓝色曲线为真实模型参数,红色曲线为自适应正则化参数AVA反演结果,黑色曲线为质量控制AVA反演结果,绿色曲线为先验低频模型参数. Fig 2 Comparison between real model parameters and two inversion results (a)P-wave velocity; (b)S-wave velocity; (c)Density. Blue curves are real model parameters, red curves are inversion results from AVA inversion with adaptive selection of regularization parameters, black curves are inversion results from AVA inversion with quality control, green curves are the a priori model parameters. |
表1 不同正则化参数选择方法反演结果与真实模型参数之间的相对误差Table 1 Relative errors of inversion results with different regularization parameter selection methods compared to the real model parameters |
| 正则化参数选择方法 | RE |
| 自适应 | 0.1405 |
| 质量控制 | 0.1977 |
图5 自适应正则化参数AVA反演结果(a)纵波速度;(b)泊松比;(c)密度. Fig 5 Results of AVA inversion with adaptive selection of regularization parameters (a)P-wave velocity; (b)Poisson's ratio; (c)Density. |
图7 井旁地震道反演结果与测井曲线对比(a)纵波速度;(b)泊松比;(c)密度.黑色曲线为真实模型参数,蓝色曲线为自适应正则化参数AVA反演结果,红色曲线为常规AVA反演结果. Fig 7 Comparison between well-log curves and inversion results from near-well seismic trace (a)P-wave velocity; (b)Poisson's ratio; (c)Density. Red curves are real model parameters, blue curves are inversion results from AVA inversion with adaptive selection of regularization parameters, black curves are inversion results from conventional AVA inversion. |
表2 不同正则化参数选择方法井旁地震道反演结果与测井曲线之间的相对误差Table 2 Relative errors of inversion results from near-well seismic trace with different regularization parameter selection methods compared to the well log curves |
| 正则化参数选择方法 | RE |
| 自适应 | 0.1671 |
| 质量控制 | 0.1753 |
感谢审稿专家提出的修改意见和编辑部的大力支持!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/
| 〈 |
|
〉 |