PREDICTION OF POROSITY AND PERMEABILITY FOR RESERVOIR CHARACTERIZATION USING MACHINE LEARNING

Abstract

Reservoir characterization is the process of determining the petrophysical properties of the subsurface, including porosity, permeability, relative permeability, water saturation, fluid saturation, capillary pressure, shale volume and sand volume. These petrophysical properties of the subsurface can be determined using experimental, simulation and machine learning method. Reservoir characterization using experimental method provides accurate results but it is costly and time consuming. The accuracy of reservoir characterization using simulation method depends on the amount of data. Moreover, the use of initial conditions and assumptions before the simulation might lead to uncertainty. The use of complex algorithms, mathematical models and modeling techniques in the Petrel software makes simulations computationally intensive and time consuming. So, to make reservoir characterization accurate and cost-effective machine learning is applied to save both cost and time. Three machine learning algorithms Linear Regression (LR), Support Vector Regression (SVR) and Artificial Neural Network (ANN) have been used to predict porosity and permeability for reservoir characterization. LR and ANN have performed better in the prediction of porosity and SVR and ANN have performed better in the prediction of permeability. In both the prediction of porosity and permeability, ANN has provided the most accurate results.