An efficient method of effective porosity prediction using an unconventional attribute through multi‐attribute regression and probabilistic neural network: A case study in a deep‐water gas field, East Coast of India

Abstract

Effective porosity prediction in 3D space away from the well is essential to characterize reservoir effectively. Precise prediction of effective porosity is a challenging task because of the non-uniqueness in its relationship with conventional seismic attributes. The challenge further intensifies incase of complex geological setup having heterogeneous reservoir properties. In the present case, major channel-levee complexes associated with smaller episodes of channel-cut-fill and migration has made the study area a geologically complex one. To overcome this challenge, the authors have adopted an approach which combines multi-attribute linear regression with Probabilistic Neural Network (PNN) technology. However, choice of unconventional attribute, “Gas Volume”, along with conventional attributes and parameter selection has resulted in much improved prediction of effective porosity volume. This effective porosity volume has been found to contain finer detail amenable for further quantitative reservoir characterization efficiently.