Full text loading...
-
A Machine Learning Technique to Rank the Realizations of Seismic Geostatistical Inversion
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 81st EAGE Conference and Exhibition 2019, Jun 2019, Volume 2019, p.1 - 5
Abstract
Seismic inversion is aimed to model the elastic properties of subsurface layers. Traditional methods of inversion provide a smooth map of elastic properties which is most often far from the actual ones. On the other hand, reservoir characterization methods such as fluid flow simulation need an input model which describes the heterogeneities of the reservoir. To overcome these difficulties, geostatistical tools have been added into the inversion process to increase the resolution of the inversion results, but an important issue with such methods is a generation of many realizations with the same probability. In this study, a Bayesian geostatistical inversion was applied on an Iranian oilfield and 250 realizations of acoustic impedance and shear impedance were generated. A machine learning technique was employed to classify the lithofacies obtained from well data. The result of the cross-validation of well log data showed an accuracy of 99.6%. Furthermore, the classification was generalized to all realizations and P10, P50 and P90 scenarios of resulted oil facies volume were determined for further characterization studies.