粘声介质下的互相关最小二乘逆时偏移方法研究

Least-squares reverse-time migration (LSRTM) has more advantages in some aspects such as amplitude fidelity and high-resolution imaging compared with conventional RTM. The viscosity for seismic wave is ubiquitous in the subsurface media. Therefore, the consideration of media viscosity has more practical significance in the process of LSRTM. Given the effect of viscous media, we develop the viscous acoustic LSRTM in this project. In the implementation of LSRTM, we first derive the optimized staggered-grid finite-difference coefficients based on the theory of the numerical dispersion relationship and least-squares method. And combining with the strategy of adaptive variable-length spatial operators, we solve the viscous wave equation, in order to improve the computing accuracy and efficiency from the aspect of algorithm. Then, to overcome the strong dependency of conventional least-squares migration on initial model and source information and the instability problems of objective function, we explore a stable cross-correlation objective function by matching the phase information of seismic data, and finally achieve the true amplitude high resolution seismic migration. In addition, this project application starts out from the main steps of visco-acoustic cross-correlation LSRTM, studies corresponding graphic processing unit (GPU) fast algorithms and strategies to further improve the computing efficiency, and then promotes the practical application of this imaging technology to oil and gas exploration.

最小二乘逆时偏移相对于传统逆时偏移而言，在保真振幅和高分辨率成像等方面更具优势。实际地下介质对地震波的粘滞效应是普遍存在的，所以在最小二乘逆时偏移成像过程中，考虑介质的粘滞性更具有实际意义。本项目考虑介质的粘滞性影响，发展粘滞声波最小二乘逆时偏移方法。在研究粘滞声波最小二乘逆时偏移实现过程中，首先以数值频散关系和最小二乘理论为基础，推导出优化的交错网格差分系数，并将其与自适应变空间算子长度的策略相结合来求解粘滞声波方程，以从算法上改进计算精度和提高计算效率；然后为克服常规最小二乘偏移对初始模型和震源信息依赖性强及目标函数容易不稳定的问题，探讨一种基于地震数据相位信息匹配的稳定互相关目标函数，最终实现真振幅、高分辨率的地震成像。此外，本项目拟在粘滞声波互相关最小二乘逆时偏移的具体实现过程中，研究相应的GPU加速算法和策略，以进一步提高成像计算效率，从而推动该成像技术在油气勘探中的实际应用。

由于实际地下介质对地震波的粘滞效应是普遍存在的，并直接影响着地震资料的成像质量，所以本项目考虑介质的粘滞性影响，发展了高精度高效率的粘滞声波最小二乘逆时偏移成像方法。在研究粘滞声波最小二乘逆时偏移的具体实现过程中：首先，基于数值频散关系和最小二乘理论，推导了优化的交错网格有限差分系数，并将其与自适应变空间算子长度的策略相结合来求解粘滞声波方程，改进了计算精度和提高了计算效率。而且，还通过结合Taylor级数展开和样点逼近提出了新的优化交错网格有限差分方法。然后，为了克服传统最小二乘逆时偏移的不稳定性问题，本项目通过采用地震数据的相位信息进行匹配来构建目标函数，最终实现了稳定的粘滞声波最小二乘逆时偏移成像。此外，本项目在粘滞声波互相关最小二乘逆时偏移的编程过程中，还采用了相应的GPU加速策略，进一步提高了成像计算效率，并表明研究成果具有较好的应用前景。