Neuro-Bayesian facies inversion of prestack seismic data from a carbonate reservoir in Iran

Abstract

Abstract Facies study is a powerful tool to better understand complexities of carbonate reservoirs. Porosity, frame flexibility factor and bulk modulus of fluid are believed to be the most proper rock physical parameters to define desired facies in carbonate rocks. Bayesian inversion is a natural choice to invert the desired facies from seismic data. The inversion method then often includes (1) Bayesian inversion of elastic parameters from seismic data, (2) Bayesian inversion of rock physical parameters from elastic parameters by considering an appropriate up-scaling method and (3) Bayesian classification of facies from inverted rock physical parameters. Neuro-Bayesian inversion method has been introduced in this study, which is a combination of an artificial neural network (ANN) classifier and Bayesian inversion of rock physical parameters that allows an improved facies prediction. Comparison between Bayesian and neuro-Bayesian methods is performed to illustrate the accuracy of predicting the facies, improved from 67% to 73% in the final results. Moreover, the Bayesian method predicted just two of the three facies where Neuro-Bayesian method predicted all the three facies successfully.