Water saturation estimation using Support Vector Machine

Abstract

Low saturation gas and commercial gas have similar amplitude responses at shallow water reservoir (less than 2000 m), because a small amount of gas can dramatically reduce the bulk modulus of the gas‐water mixture. Increasing the depth of reservoir, the bulk modulus of gas increase gradually. Gas saturation can be significant to affect bulk modulus of gas‐water mixture. This improves the chances to discriminate low saturation gas from commercial gas with seismic data. Using a water saturation curve calculated from density and resistivity logs of a gas well at the Gulf of Mexico, we trained an ε‐insensitive Support Vector Machine (SVM) to regress the water saturation from seismic data. Validations of SVM regression on the gas reservoir and a prospect of low saturation gas suggest that it is possible to distinguish between commercial gas and low saturation gas at deep water reservoirs.