Neural Network Based Seismic Inversion for Reservoir Prediction and Pore Pressure Analysis, Balkassar Field, Potwar Sub Basin, Pakistan

Abstract

The Eocene and Paleocene Reservoirs including Chorgalı, Sakesar and Lockhart formations, in the Balkassar Oil field, which is present on the southern flank of the Soan syncline in Central Potwar are considered the main targets for a source of hydrocarbons in the Potwar Sub basin. The Balkassar structure consists of an elongated, fault-bounded, salt-cored anticlinal pop-up structure which trends Northeast- Southwest and mainly formed because of the compressional forces in the area. This study is intended to estimate reservoir properties which is porosity, volume of clay and fluid content prediction using the Artificial Neural Network. Artificial Neural Networks have gained a substantial amount of attention over the past few years, among different linear and nonlinear prediction tools such as curve fitting, regression etc. In this study, 3D seismic data and well logs of the Balkassar Oil field are used for the prediction of reservoir properties. To achieve this, the 3D seismic data was inverted through Probabilistic neural networks to obtain acoustic impedance volume which was then used as part of seismic attribute study applied to the data set. Multi-attribute analysis was performed in order to analyze the effectiveness of specific attributes for training the PNN. A total of seven attributes were found to provide the best training results, after stepwise regression and validation testing. These attributes proved to show a substantial amount of correlation and thus porosity, volume of clay and fluid content volumes were predicted. Horizon maps for three potentially prospective formations of Chorgali, Sakesar and Lockhart were extracted from these volumes to analyze the spatial extent of these attributes, on the basis of which, potentially prospective zones were defined by a probability analysis.