基于高精度地层倾角线性反演的地震全体积拉平方法

Volumetric flattening through linear inversion scheme with accurate seismic dips

地震数据的拉平分析是三维地震资料解释中的常见方法,一般通过拾取某一标准层位并将其校正到一个基准面的方式拉平地震数据,这样会导致距离标准层越远的层位拉平效果越差.近几年发展起来的三维地震数据全体积拉平方法是避免这种拉平缺陷的有效途径.本文提出一种基于高精度地层倾角的全体积自动拉平方法,该方法将一种自适应加权的向量滤波法获得高精度的局部地层倾角作为观测数据,根据拉平变换对每个数据样点产生的垂向偏移量与地层倾角之间的近似线性关系,建立最小二乘反演机制.在优化反演模型时,引入由梯度结构张量法得到的误差控制模板,抑制断层等复杂地质构造对拉平效果的影响;此外,增加模型参数的平滑度量项,避免拉平后的数据出现较大的畸变,并采用模型重新参数化方法,加快反演算法的收敛速度.三维合成数据和实际地震资料的试算结果表明这种三维全体积拉平方法是有效和可行的.

Flattening seismic data is one commonly used interpretation technique that helps extracting geologic and reservoir features. Traditional approaches need to track a horizon throughout the data volume and then flatten data on the picked horizon. These approaches may produce unsatisfactory results for layers far from the referred horizon which can be avoided by volumetric flattening approaches. This paper proposes a volumetric flattening approach to automatically flatten seismic data with local seismic dips. The approach utilizes the accurate seismic dips as observation data for the flattening algorithm which are calculated by a weighted vector directional filter scheme. By considering the approximate linearity relationship between vertical flattening shifts and local dips, a least-squares inversion model is established. Several optimization strategies are taken into account for this inversion problem. An error mask calculated by gradient structure tensor is included to reduce influences of structural complexities such as faults. And a regularization term measuring vertical smoothness of the model is added to diminish waveform distortion of flattened data. Besides, model reparameterization is also adopted to accelerate the convergence of the conjugate-gradient algorithm. Both the synthetic and real data examples illustrate that the proposed approach is feasible and effective.