Modulus of Elasticity Prediction with Machine Learning Techniques in Sandstone Gas Reservoir, Bengal Basin

An accurate data-driven predictive model is crucial to obtain the continuous profile of formation modulus of elasticity which is also a promising study in the area of rock modeling for gas reservoirs by scholars. The objectives of this study are i) to assess the predictive models with machine learning for obtaining modulus of elasticity (E), and ii) to rank the most effective variables to predict the rock strength parameter, E of clastic sedimentary rocks in the Bengal Basin. The supervised machine learning (ML) techniques, namely, random forest (RF) and least square support vector machine (LSSVM) are applied to develop the data-driven connectionist models for the prediction of E, respectively. The data sets of formation log data are processed and utilized to construct predictive models to obtain the output. For the evaluation of model performance, statistical metrics have been utilized in the study. Considering the simulated results, it is found that the LSSVM outer performed than the RF approach with minimal least error and high correlation coefficient to predict E. Based on the model’s parametric sensitivity analysis, compressional and shear wave velocities are mostly significant than the other predictor variables such as formation gamma-ray and bulk density to predict the dynamic E. These systematic study steps for model development and feature ranking can be applied to the application of petroleum engineering fields such as prediction of reservoir rock properties, drilling rate of penetration, and wellbore failure potentiality of geoenergy reservoir rocks.