反射波层析反演速度建模方法

Velocity model building using reflection tomography

我国油气地震勘探目标逐渐转向小尺度(薄互层)岩性油气藏,油气地震勘探技术也逐渐向单点高密度、宽带、宽方位(“两宽一高”)地震勘探方向转变。“两宽一高”的地震数据采集为宽带波阻抗成像提供了数据基础,精确的背景(偏移)速度建模是宽带波阻抗成像必不可少的环节。在实际应用中,反射波层析反演是主要的宏观速度建模技术,典型的方法如成像道集层析或立体层析等,但它们在反演精度及计算效率方面仍存在不足之处。在数据域中,反射波时差的精确测量仍然是一个颇具挑战性的问题。为解决这一难题,从叠前地震数据的稀疏表达与特征分解出发,利用特征波场分解和包络走时定义准则,提取叠前地震数据的运动学信息,并构造了特征数据体(包含地震波的斜率、走时及子波波形)。通过分析特征波在地下的聚焦特性,提出了一种反射波残差(时差或者地下偏移距)的自动测量方法,避免了地震波振幅对反射波运动学信息拟合的影响。基于反射波残差,提出了一种特征波反射层析反演(characteristicreflectioninversion,简记为CRI)方法,通过极小化反射波时差或地下偏移距实现背景速度反演。理论分析可知:反射波时差测量方法仅适用于二维模型,而地下偏移距测量方法可以适用于二维和三维模型,因而其适用性更广。在数值计算方面,反射波残差目标泛函的梯度可以用波动理论计算,也可以用射线理论近似。出于计算效率的考虑,利用射线理论计算泛函梯度并结合梯度去噪方法,实现低波数的速度更新。数值实验表明:该方法无需长偏移距观测数据或低频信息,对初始模型依赖性低、计算效率高,且整个反演流程可以实现全自动化,可以为后续的宽波数带速度建模提供较为可靠的低波数背景速度模型。

As the target of oil exploration in China is changing toward the small-scale (thin-interbed) lithologic reservoirs,the seismic acquisition methods are being gradually upgraded to the broadband,wide-azimuth and high-density (BWH) techniques.Such new techniques permit broadband impedance imaging,which require an accurate background model (in terms of velocity,anisotropy,even Q).In exploration seismology,the macro-velocity model is often estimated using reflection tomography.Typical inversion methods,such as the Ray-based tomographic migration velocity analysis or the stereo-tomography,still suffer from low resolution and high computational costs.In the data domain,accurately measuring the difference in the reflection travel time between the observed and synthetic data remains a challenge.To address this problem,we begin with a sparse presentation and characterization of pre-stack seismic data,from which we extract kinematic information (slope,travel time,and waveform of the reflection event) using the characteristic wavefield decomposition method.Furthermore,by analyzing the focus property of a wavefield in the subsurface space,two reflection-misfit calculation strategies are proposed:the reflection travel time-residual and the subsurface-offset measurements.Consequently,the new reflection inversion method,which is referred to as the characteristic reflection inversion (CRI),is based on the minimization of reflection travel-time residuals or subsurface offsets.According to a theoretical analysis,the strategy for measuring the reflection travel time residual should be valid for any 2D media,while the subsurface offset measurement method should be valid for both 2D and 3D media.The functional gradient can be calculated either by wave-equation linearization or by the ray theory.To increase the computational efficiency,the ray theory version of the gradient is calculated,followed by a denoising method in such a way that the low-wavenumber components of the velocity model can be updated.Numerical examples demonstrate that the CRI neither requires low-frequency seismic data nor long-offset acquisition.In addition,the CRI is computationally efficient,since it eliminates the need for calculating and saving the common image gathers and travel time picking.Furthermore,it is a fully automatic procedure,and is insensitive to the accuracy of the staring model,making it quite promising for automatic macro-velocity model building.