Pore Pressure Prediction from Conventional Well Logging Data Using Machine Learning Algorithms

Abstract

Accurate determination of pore pressure is fundamental to defining the in-situ stress tensor and plays a critical role in hydraulic fracturing design, sand production management, safe mud weight window determination, reservoir performance evaluation, and wellbore stability analysis. Despite its importance, reliable pore pressure estimation remains a significant challenge in the oil and gas industry. This study aims to develop a continuous pore pressure profile along the wellbore depth directly from conventional well logging data. Two machine learning techniques—Artificial Neural Networks (ANNs) and Support Vector Machines (SVM)—were implemented using conventional well log measurements, including compressional transit time (DTCO), shear transit time (DTSM), bulk density (RHOZ), total porosity (PHIT), and gamma-ray (GR), as input variables. Directly measured pore pressure data were used as the target output for model training and validation. The dataset comprised measurements from 385 wells drilled across multiple oil fields in southern Iraq. The results demonstrate that integrating multiple logging parameters significantly enhances pore pressure prediction accuracy. Both ANN and SVM models exhibited strong predictive capabilities for computational pore pressure estimation. However, comparative performance analysis indicates that the SVM model outperformed the ANN in terms of prediction accuracy and robustness. Field-scale application confirms that the proposed models can predict continuous pore pressure profiles with accuracy exceeding 90%. The developed models are suitable for integration with real-time drilling data systems, enabling visualization through mud logging platforms. This approach provides drilling engineers with a practical and efficient tool to estimate pore pressure using widely available conventional logging data, without requiring detailed reservoir elastic parameters. Consequently, the proposed methodology supports both pre-drilling well planning and real-time pore pressure evaluation during drilling operations.