Optimised reservoir characterisation workflow using multi attributes classification ? a case study on the Wandoo field, NW Shelf, Australia

Abstract

Grid-guided volume attributes and 3D volume seismic transforms integrated through multi-dimensional seismic classification algorithms provide new inputs for the seismic interpreter to rapidly converge towards a 3D seismic facies hypothesis fitting a solid structural framework. We illustrate the approach with the 75 million barrels Wandoo field from the Australian North West Shelf. At the Barremian target sands level (600 m), the structural framework of the field is complex and the seismic data of low fold, with high frequency noise and multiples. From the onset, seismic interpretation was performed in a multi-attribute 3D visualisation environment. Model-based noise attenuation techniques were used to pre-process the input seismic volume to facilitate the time structural interpretation of regional reflectors. New grid-guided volume attributes based on orthonormal polynomial trace reconstruction captured subtle lateral differences in seismic facies that were not represented on the observed wiggle trace. Multi-dimensional geostatistical (Fisher, Contextual Bayesian) and Neural Network classification algorithms were used to produce a set of class and probability maps from all volume attributes generated on the reservoir interval. Calibration of the derived classes for facies and fluids was performed using a priori information from well measurements. Simultaneously with the grid-guided approach, the entire seismic volume was transformed into a set of 3D seismic facies attributes (texture, edge enhancing and termination cubes) to enhance both lateral and horizontal discontinuities. The 3D classification algorithms were used to integrate the information content of multiple attribute cubes into a single probability volume. The latter was displayed in 3D visualisation space to track voxels of given seismic facies and automatically extract 3D fault surfaces. These results were validated by additional exploration wells not included in the initial study. Further well trajectories could then be planned in 3D visualisation space to reach potential targets with a higher degree of confidence. The entire procedure has been captured and is presented as an optimum workflow.