各种速度分析与反演方法的对比研究

Comparison among velocity analysis and inversion methods

速度估计与建模是勘探地震学的核心技术。速度估计问题是一个标准的反演问题。原则上,反演问题应该在贝叶斯框架下进行,但是石油工业界根据生产实际形成了一套速度分析与建模的技术系列。针对目前速度估计与速度建模技术研究及应用现状,试图把各种速度估计方法都纳入贝叶斯估计框架下进行审视。该框架的基本逻辑是建立2种目标泛函:成像空间中的相关最佳泛函(或聚焦最佳泛函)和数据空间中的逼近误差的方差最小泛函。在此目标泛函的基础上,利用梯度导引类的优化算法或Monte Carlo类的全局寻优算法,甚至扫描(枚举)算法实现各种尺度下的速度估计及模型建立。在上述理论框架下,系统地分析目前典型方法技术的共同特征,可以指出新的速度估计方法的发展方向。

Velocity estimation and modeling technique is a key issue in exploration seismic. As a typical inversion problem, velocity estimation should be carried out under the framework of Bayesian estimation. However, a series of techniques for velocity analysis and modeling are developed by the petroleum industry. In view of the present situation of research and application of velocity estimation and modeling technique, we try to analyze the current velocity estimation methods under the framework of Bayesian estimation. All the existing velocity estimation methods can be incorporated into the following two objective functions： the best correlation criterion （the best focusing criterion） in the image-domain or the least error criterion in the data-domain. The gradient-guided local optimization method and Monte Claro method can be used to estimate the seismic velocity in different scales. Under this framework, we can analyze the common features of all the current methods, and develop new methods and techniques for velocity estimation.