Abstract

One of the key objectives of seismic reservoir characterization is to predict and map the distribution of reservoir properties in a 3D sense. Appropriate use of supervised artificial Intelligence (AI)/machine learning techniques can do a considerably good job in this aspect. However, to make the property estimation more robust, suitable attributes need to be fed into the AI training process. Pre-stack seismic inversion provides relevant elastic attributes, which are better correlated with the desired reservoir properties. Therefore, an integrated workflow consisting of prestack seismic inversion and a robust supervised machine learning technique, when implemented judiciously, will help in accurately predicting the petrophysical and rock-mechanical reservoir properties away from the drilled well locations.