Bayesian reservoir characterization

Luiz Lucchesi Loures

This research aims to provide a complete solution for reservoir properties determination, including estimation and uncertainty analysis. The strengths of this reservoir characterization methodology is:

i) uncertainty analysis and

ii) integration of multiple geophysical data-sets, rock physics analysis and prior information in a straightforward way provided by the Bayesian framework.

The inference problem reported on this paper is formulated to solve the problem of porosity estimation. The sources of information are pre-stack seismic data; well log data and core samples. The reservoir is considered a volume composed of cells. The final solution is a probability density function (pdf) for porosity for each cell of the reservoir volume. These pdfs, call posterior pdfs, represent deductions about porosity and incorporate all related uncertainties.

The waveform elastic inversion is incorporated in this methodology to access the porous medium physical property information from pre-stack seismic data. Geostatistical modelling is incorporated to access the porous medium spatial variability information from well-log data.

The methodology is implemented to consider a reservoir composed for block cells. One posterior pdf is computed for each reservoir cell. Two cell volumes present the final result. One is constructed with the modes of the posterior pdfs related to each cell and represents the estimated porosity model, and the other is constructed with the confidence interval of the posteriors pdfs and represents a measure of the related uncertainties.