--> Estimation of 3-D Confidence Index for Consistent Integration of Seismic Data Into Reservoir Models

2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Estimation of 3-D Confidence Index for Consistent Integration of Seismic Data Into Reservoir Models

Abstract

3D seismic data is routinely used to constrain structural and stratigraphic models as well as for property modelling. Seismic amplitudes, or more commonly facies or elastic parameters from seismic inversion, act as local average trends in a version of co-simulation. Conventional workflows may fail to integrate data properly for two main reasons: (1) well data, especially from horizontal wells, have not been properly tied to seismic; and (2) the simulation method does not account for spatial variations of seismic data uncertainty. In this paper, we propose a novel methodology to correct for the former. We also address the second point by estimating a 3D confidence cube that can be used subsequently to properly weight the seismic information into the reservoir model. The first step of the workflow consists of running supervised electrofacies analysis using core facies description at well positions. The resulting facies logs are transformed to synthetic elastic logs based on statistical analysis of well data, which are in turn interpolated on the reservoir grid, and upscaled to the seismic scale. Second, seismic facies analysis is performed on the seismic inversion results, before being converted to local facies proportions. These volumes are then mapped to synthetic elastic parameters using well statistical analysis results. Well trajectories are then shifted locally on the seismic cube to ensure an optimal correlation between seismic and well elastic parameters. At the tying position, a local correlation coefficient is estimated on a sliding window along the well trajectories and interpolated in the reservoir grid. The methodology has been tested and compared to existing models built on a given carbonate reservoir. The quality of the reservoir model was globally improved due to two elements: First, all the available well data could be tied on seismic with a global matching index (correlation coefficient) generally above 80%, although locally, the confidence index varies significantly from well to well and across stratigraphic units. Second, after interpolation by nonstationary kriging on the reservoir grid, the resulting confidence index has been used to weight optimally seismic data into the reservoir model.