Combining Machine Learning And Seismic Inversiontechnique For Reservoir Characterization Of Dhodak Field, Central Indus Basin, Pakistan

Abstract

The main aim of studying this area is to evaluate the hydrocarbon potential of Dhodak area, Central Indus Basin, Pakistan. Major structures present in this block are anticlinical pop-up structures and thrusted fault blocks produced due to the decollement by the basement. Generally, target Formations were Pab Sandstone, Lower Ranikot Formation and Dungan Limestone. Lower Ranikot Formation was our main target formation specifically for analysis to predict its reservoir properties. Synthetic seismogram was developed with the help of Density log(RhoB) and Sonic log(DT) for the confirmation of geological section made by Seismic interpretation. For the conformation of these structures Time, Velocity and Depth contour maps of these Formations are also generated which are delineated through seismic data. Hydrocarbon potential of Lower Ranikot Formation in Dhodak-01 & 05 well was evaluated by Petro-physical analysis by using borehole logs. It consists for the calculation of Petrophysical properties like shale’s Volume, sand’s volume, Porosity calculations, Density calculations, Saturation of Water Calculations, Gas effect calculations and Hydrocarbon Saturation(Shc). Different sort of logs that are present were used like Gamma Ray, Compensated Neutron LLD, LLS, MSFL (Resistivity logs), Spontaneous Potential log (SP Log). Rock physics modeling have used for the prediction of P and S waves variations and how Poisson’s ratio occurs. Rock physics depicted the missing log prediction like S-wave specifically and fluid substitution also occurred with different conditions to confirm reservoir availability. Model Based Inversion (MBI) have been used to predict and confirm more better results came from seismic interpretation. Wavelet was extracted from the control line and the correlation occurred. LFM (Low Frequency Models) have been generated as well. Afterward, Quality control of data was occurred by inverted techniques. Lastly to confirm results with seismic inversion technique, Machine learning (PNN) based on Bayesian classifier method which gives reliable prediction of petro physical properties so it is used predict the Volume of clay prediction, Porosity prediction and Saturation of water was predicted by Actual and Predicted values.