非线性二次规划贝叶斯叠前反演

Non-linear quadratic programming Bayesian prestack inversion

叠前反演的目的是基于弹性波理论从地震数据中获得地层参数的可靠估计,进而用于描述地层的流体和岩性特征.然而叠前反演问题都是高维的和非适定的,并且容易受各种噪声和采集过程中不确定因素的影响,因此,为了获得稳定可靠的解必需对反演过程加以合理的约束.本文提出了一种基于非线性二次规划的叠前三参数反演方法.首先基于贝叶斯参数估计理论,假设似然函数服从高斯分布,并使待反演的参数服从于改进的Cauchy分布,从而提高了反演结果的分辨率;其次用协方差矩阵来描述参数间的相关程度,进一步提高了反演结果的稳定性;最后将问题转化为一个非线性二次规划的求解问题,并在多种约束下得到问题的解.仿真实验和实际应用皆已表明,本文提出的反演方法运算速度快捷,既使在信噪比很低的情况下也可获得较好的反演结果,为储层的进一步识别提供更多的物性参数.

The goal of prestaek AVO inversion is to obtain reliable estimation of subsurface parameters from which to describe the properties of subsurface fluid and lithology based on elastic wave knowledge and a set of geophysical measurements. However, prestack inversion problems are multidimensional and ill posed, and they are often strongly affected by noise and measurement uncertainty, so it is necessary to add constraints in order to obtain steady and rational inversion results. A new non-linear quadratic programming Bayesian prestack three-terms inversion method is developed; Firstly, this method is based on Bayesian parameter estimation theory, Gaussian distribution is used for likelihood function and modified Cauchy distribution is used for prior distribution; Secondly, covariance matrix is used to describe the degree of correlation between the parameters; rock physics relations are used to constrain the inversion results; at last, this method is transformed into non-linear quadratic programming problem, and inversion results are acquired under several constraints. Tests on synthetic data and practical application show that all parameters were almost perfectly retrieved for further analysis even the SNR was low.