Elastic impedance inversion incorporating fusion initial model and kernel Fisher discriminant analysis approach

Abstract

Seismic inversion is a significant technique for estimating petroleum reservoir parameters. The low frequency component of the initial model represents the geological background information, which plays an important role in the seismic inversion. It is challenging to precisely depict the actual geological model in seismic inversion because of the inherent velocity-depth ambiguity. Therefore, the initial model which is closer to genuine geological backdrop is essential. We propose a workflow which estimates a fusion initial model based on data fusion algorithms. It is well known that seismic facies analysis can provide more low-frequency information about the geological background. For example, the boundaries of sedimentary bodies can be represented by seismic facies classification data. We utilize a combination of the seismic facies classification data and well curves interpolation initial models to accurately invert the special geological body with the support of a feature-level fusion algorithm. Then, a practical pre-stack seismic inversion method is implemented, and a field data example further demonstrates its applicability and steadiness in seismic inversion.