Petrofacies characterization using best combination of multiple elastic properties

Abstract

Elastic properties such as P-impedance and the ratio of P-to S-wave velocity are key features for linking well log data to seismic data for reservoir characterization. Crossplotting methods from well log data are commonly applied to delineate the distribution of petrofacies on the seismic volume. However, only two or three elastic properties (two-or three-dimensional data) can be used in the crossplotting methods because of data visualization issues. Furthermore, the elastic properties used in the crossplotting method have to be chosen subjectively or empirically by a human interpreter. In this study, we propose a new workflow to overcome the limitations of conventional crossplotting methods. We use more than three elastic properties from well log data and apply non-linear transformations to the properties, to improve separability of the non-linear datasets. Then, we choose the best combination of the non-linear transformed elastic properties using class-scatter-matrix based separability measure. Lastly, Bayesian inference with multivariate probability density function (PDF) of petrofacies is performed to predict petrofacies probability volumes in the seismic area. An application to Vincent oil field data demonstrates the effectiveness of the proposed workflow. The accuracy improved from 80.7% to 86.2% and uncertainty is significantly reduced compared to the conventional workflow with crossplotting methods.