THE APPLICATION OF NEURAL NETWORK METHOD IN PETROPHYSICAL EVALUATION OF ASMARI FORMATION IN A PRODUCING WELL IN SOUTHWEST OF IRAN

Determination of petrophysical parameters and their distribution in the reservoir can lead to new zonation and change of production thickness zone.clay minerals exist in most of oil reservoirs and reduce important parameters such as porosity, permeability and production potential.The purpose of this study was to investigate the petrophysical properties of Asmari Formation by combination of different traditional petrophysical methods for volume of clay estimation and reservoir evaluation studies. The traditional calibration of gamma ray log such as Bhuyan – Passey, Larionov-1, Steiber, Clavier and Jozanikohan relationships were applied which resulted to 45% relative error for estimation of clay minerals in compare to the 15 known laboratory values of this parameter. In the next step, neural network modeling was performed to reduce relative error. 259 data were estimated from laboratory values and trained with Tangent Sigmoid Activation Function, Levenberg-Marquardt training algorithm, 6 neurons and 1 hidden layer in a MLP neural network. The clay volume outputs of the neural network were classified and the body of the reservoir determined to be sandstone-clay. By investigating the density-porosity cross-plots, formation lithology and good quality reservoir intervals were introduced for Perforation operation. The gamma-ray data and neutron porosity data were also categorized to give the low to high quality intervals. Finally, by combination all the results in this study, the quality of Asmari formation were estimated to be "good".