Artificial intelligence for reservoir modeling and property estimation in petroleum engineering

Abstract

This study aims to advance the application of artificial intelligence (AI) in reservoir modeling by developing and evaluating machine learning (ML) techniques for estimating key subsurface properties, including permeability, porosity, and relative permeability. While prior research has applied AI methods in this domain, most approaches focus on narrow datasets or lack integration with physical reservoir behavior. In contrast, this work combines multiple AI techniques—artificial neural networks (ANNs), support vector machines (SVMs), fuzzy logic (FL), and evolutionary algorithms—into a hybrid modeling framework that captures complex, nonlinear rock–fluid interactions in multiphase flow environments. The study introduces novel preprocessing and training strategies to address data quality issues and improve generalizability. Case studies using real field data demonstrate that the proposed AI models outperform conventional empirical and statistical methods in prediction accuracy, robustness, and computational efficiency. The findings suggest that AI can significantly reduce reliance on expensive laboratory measurements, support faster reservoir characterization, and enhance decision-making under geological uncertainty. This work contributes a scalable and interpretable ML-based workflow that bridges data-driven insights with engineering principles, offering a step forward in the digital transformation of reservoir management.