基于自适应空间反射结构正则化的多道同时波阻抗反演

Compared to the trace-by-trace impedance inversion, multi-trace simultaneous impedance inversion can effectively decrease the ambiguity of inversion result due to the introduction of spatial structure regularization, and is attracting more research attentions. However, the present methods for multi-trace inversion mainly make use of the geometrical constraints, such as dip and strike, which are estimated from the spatial gradient of reflection amplitude. In presence of noise and other disturbance, the estimate accuracy of geometrical parameter is greatly decreased. In addition, the geometrical parameters are not able to describe the spatial variation of dynamic attributes resulted from heterogeneous reservoir. We propose simultaneous multi-trace impedance inversion based on self-adaptive regularization of spatial reflection structure. Base on the nonstationary autoregression theory, reflection structure operator can be estimated from the seismic data. The operator not only contains the spatial continuity and structural relation of seismic reflection, but also describes the spatial variation of dynamic seismic attributes. Then, the operator is introduced to the objection function of multi-trace inversion based on the Bayes theory as a spatial structure regularization for decreasing the ambiguity of inversion result. The research work has theoretical value and practical potential for improving the reliability of reservoir prediction using seismic inversion.

由于引入了不同空间位置的约束关系，多道波阻抗反演较单道波阻抗反演能够有效地减小反演结果多解性，是近年来波阻抗反演的重要发展方向。然而，现有的多道反演方法主要是利用地质结构的倾角和倾向等几何约束关系，这些几何信息来源于相邻地震道振幅变化的空间梯度，很容易受随机噪声等各类干扰的影响，另外，这类几何参数也不能反映和约束储层结构横向变化引起的反射振幅等波动力学信息的变化。为此，我们提出了一种基于空间反射结构正则化自适应约束的多道波阻抗反演方法。该方法基于非稳态自回归理论，从地震数据本身提取反射结构表征算子，该算子描述了反射结构在空间方向的连续性和延展关系，不仅隐含有地质结构的几何信息，也包含了反射振幅等动力学信息的空间分布关系。然后，在贝叶斯理论框架下，将该算子引入到多道反演目标函数的正则化条件，减小反演结果的多解性。本课题的研究对于较大幅度地提高地震反演储层预测精度具有重要的理论价值和突破潜力。

地震频带和观测孔径的有限性导致了地震反演多解性。在贝叶斯反演理论框架下，将先验信息引入到地震反演的正则化条件是减小多解性的根本途径。地震信号空间结构具有鲜明的个性化特征，可以对反演结果的可靠性进行判识和甄别，将其作为反演系统的正则化约束，能够减小反演结果的多解性。首先，本课题基于非稳态自回归理论，从地震数据本身构建了一种地震反射结构表征算子。该算子不仅包含地层产状等几何信息，也描述了反射振幅等动力学信息的空间变化。然后，将反射系数稀疏结构和地震反射空间结构同时引入到多道地震反演的正则化条件，在纵向稀疏结构和横向反射结构的引导下进行多道同时反演，大幅度降低了反演结果高频分量的多解性。基于本课题研究成果开发了高分辨率地震反演软件系统，并在东部油田薄层油气藏地震预测中进行了实际应用，反演结果提高了薄层结构的表征能力。