RESERVOIR CHARACTERIZATION OF SAWAN FIELD USING PETROPHYSICAL AND NEURAL NETWORK ANALYSIS, MIDDLE INDUS BASIN, PAKISTAN

Abstract

The agenda of this dissertation is to draw a comparison between the reservoir properties generated using conventional petrophysics and neural networks. Based on the geographical location, the study area lies in Khairpur, Sindh Province, Pakistan. Stratigraphic cum structural traps are found in this area. Wireline logs and a 3D seismic cube has been used to conduct this research. Target horizon is sand of C Interval of Lower Goru Formation. Probabilistic Neural Network (PNN) and Multilayer Feed Forward Network (MLFN) has been used to estimate porosities along with conventional petrophysical analysis. Results generated by both the methods are similar. However, PNN produces more reliable results whereas the MLFN overestimates values at some points. Average hydrocarbon saturation within the wells lies in between 50- 70%. Effective porosity ranges between 4-11% which is ideal in the case of sandstones.