Fluid flow modelling with seismic cluster analysis

Laurence R. Bentley, Xuri Huang, Claude Laflamme

Cluster analysis is used to construct fluid flow zones from seismic attributes. The steps are: 1. Remove grid points that contain outliers in any seismic attribute; 2. Scale each attribute to zero mean and unit variance; 3. Use principal component analysis to transform the scaled attributes to uncorrelated principal component attributes; 4. Principal component attributes are grouped into categories of similar seismic response using cluster analysis; 5. Upscale the seismic grid to the computational grid scale using a weighted voting procedure (morphing); 6. Spatially filter cluster assignments to remove small, isolated spots using a weighted voting scheme; 7. Assign a seismic zone to spatially connected elements with the same cluster category. Using a Gulf of Mexico test case, the zoning procedure produced useful computational zones and reduced the time required for history matching.