Using Machine Learning to Predict Permeability from Well Logs: A Comparative Study of Different Models

Abstract

Abstract Permeability is a crucial parameter in reservoir engineering, as it determines how easily fluids can flow through subsurface rocks. Traditional methods of measuring permeability are expensive, error-prone and time-consuming, which has led to the development of machine learning models that can predict permeability based on well-log data. This study aims to determine the most effective machine learning models for predicting permeability based on available well-log data. The study covers a detailed explanation of the data-gathering and pre-processing techniques used. Four input features were used in the models, namely gamma ray (GR), bulk density (RHOB), neutron porosity (NPHI), and resistivity (RT). Six different models were trained and evaluated based on their performance, namely linear regression, support vector machine, decision trees, gradient boosting, multi-layer perceptron, and random forest models. The models were evaluated based on their Mean Squared Error and RSquare values. Our results showed that the random forest model outperformed the other models, achieving a Mean Squared Error (MSE) value of 0.88. The multi-layer perceptron and gradient boosting models also showed promising results, achieving MSE values of 0.81 each. These findings suggest that machine learning models can be a powerful tool for predicting permeability from well-log data, and the random forest model, in particular, holds great promise for future modelling efforts in this area.