Find in Library

Understanding porosity is crucial in various industries, especially those involved in resource exploration and production, such as oil and gas, mining, and geology. Since porosity prediction in tight sandstone reservoir is inherently uncertain due to the complex pore structure, geological heterogeneity, and noisy readings, achieving the necessary precision with robust uncertainty quantification plugin remains a significant challenge. While Machine Learning algorithms have proven effective in making predictions across various domains, they are often limited to the deterministic prediction only. This study aims to evaluate the reliability of porosity prediction with high accuracy while quantifying uncertainty through the utilization of a stochastic Extreme Gradient Boosting (XGB) model. The proposed model is compared with a plain XGB and Gaussian Process and the research reveals that the proposed model can achieve the highest accuracy (.63) among them. Additionally, analysis of 2D sections reveals that internal regions exhibit lower uncertainty compared to their boundaries, as determined by various realizations of the stochastic XGB model. This remarks on the potential of the model to serve as a robust tool for porosity prediction in tight reservoir zones.