Prediction of gas hydrate saturation throughout the seismic section in Krishna Godavari basin using multivariate linear regression and multi-layer feed forward neural network approach

Abstract

Stepwise linear regression, multi-layer feed forward neural (MLFN) network method was used to predict the 2D distribution of P-wave velocity, resistivity, porosity, and gas hydrate saturation throughout seismic section NGHP-01 in the Krishna-Godavari basin. Log prediction process, with uncertainties based on root mean square error properties, was implemented by way of a multi-layer feed forward neural network. The log properties were merged with seismic data by applying a non-linear transform to the seismic attributes. Gas hydrate saturation estimates show an average saturation of 41 % between common depth point (CDP) 600 and 700 and an average saturation of 35 % for CDP 300–400 and 700–800, respectively. High gas hydrate saturation corresponds to high P-wave velocity and high resistivity except in a few spots, which could be due to local variation of permeability, temperature, fractures, etc.