多道随机稀疏反射系数反演

常规随机稀疏反射系数反演方法采用单道反演策略,实现过程中需要依据经验预设非零反射系数个数,不恰当的预设参数会引起诸多问题,另外,方法的收敛准则从本质上放大了以振幅较大为特征的异常值的影响。为此,引入相邻地震道数据作为约束条件,提出了一种多道的随机稀疏反射系数反演方法。通过对目标地震道数据的非零反射系数的位置和倾角进行同时非线性搜索,有效提升了反演预设参数的选择稳定性,同时减轻了异常值对反演结果的影响。作为单道随机稀疏反射系数反演方法的拓展,该方法保留了可以对反演结果进行不确定性分析的优势,可以提高地震数据高分辨率处理结果的可信性。利用该方法获得的高分辨率反射系数和地震剖面能够得到更多的构造和界面信息,有利于后续的解释工作。

Conventional stochastic sparse reflectivity inversion adopts a single-trace strategy,and the value of non-zero reflectivity needs to be pre-set according to previous experiences,since inadequate pre-set parameters result in a variety of problems.In addition,the convergence criterion of the method essentially enlarges the influence of outliers characterized by large amplitude values.To solve these two problems,a multi-trace inversion method was proposed by introducing adjacent seismic trace data as constraints.By simultaneously conducting a non-linear search for locations while identifying slopes of non-zero reflectivity of the target trace,the selection stability of the preset parameters was effectively improved,and the influence of the outliers on the inversion results was reduced.As an extension of the single-trace method,the proposed method retains the uncertainty analysis of the inversion results,which can improve the credibility of high resolution seismic data processing results.The proposed method can help obtain high resolution reflectivity and seismic sections,which is conducive to subsequent interpretation.

中国博士后科学基金资助项目(2019M662005)和中国石油化工股份有限公司科研项目“川西侏罗系致密碎屑岩甜点预测技术研究”(P20055-7)共同资助。