An intelligent data-driven model for Dean–Stark water saturation prediction in carbonate rocks

Abstract

Carbonate rocks have a very complex pore system due to the presence of interparticle and intraparticle porosities. This makes the acquisition and analysis of the petrophysical data and the characterization of carbonate rocks a big challenge. In this study, a functional network (FN) tool is used to develop a model to predict water saturation using petrophysical well logs as input parameters and the Dean–Stark measured water saturation as an output parameter. The dataset is comprised of 150 well logs points with the available core data. The developed FN model was optimized by using several optimization algorithms such as differential evolution, particle swarm optimization (PSO), and covariance matrix adaptation evolution strategy. FN model optimized with PSO was found to be the most robust artificial intelligence tool to predict water saturation in carbonate rocks. The results showed that the proposed model can predict the water saturation with an accuracy of 97%. In addition to the development of optimized model, an explicit FN-based empirical correlation is also presented for a quick use. To validate the proposed correlation, three most commonly used water saturation models, namely Simandoux, Bardon and Pied, and Fertl and Hammack, were tested on the blind dataset. The results showed that FN model predicted the water saturation with an error of less than 5%, while the other saturation models predicted water saturation with an error up to 50%. This work clearly shows that machine learning techniques can determine water saturation with high accuracy.