Water saturation plays a vital role in calculating the volume of hydrocarbon in reservoirs and defining the net pay. It is also essential for designing the well completion. Innacurate water saturation calculation can lead to poor decision-making, significantly affecting the reservoir's development and production, potentially resulting in reduced hydrocarbon oil recovery. Various techniques to estimate the water saturation in both clean and shaly formations. However, the most widely used approaches in the petroleum industry rely on petrophysical models, including Archie's equation, Waxman-Smits, Simandoux, Indonesia, and dual-water models. Most of these methods are only valid for clean sands or carbonate, while the presence of clay significantly limits the accuracy of these models. On the other hand, the estimation of the water saturation through core analysis does not usually cover a large interval of the well, is highly costly, and requires much time. In this study, an empirical equation for predicting water saturation based on the weight and biases of the artificial neural networks (ANN) was developed. 334 data points of the shale volume, formation deep resistivity, porosity, and permeability and their corresponding water saturation collected from the Epsilon Field in Greece were considered for optimizing the ANN model. The ANN model was trained on 252 data sets, where the water saturation was predicted with an average absolute percentage error (AAPE) of 0.90%. Then, an empirical equation was developed based on the optimized ANN model and its weights and biases. The developed equation predicted the water saturation for the remaining 82 data sets (testing data) with an AAPE of 1.08%. The newly established empirical correlation enhances the precision of water saturation prediction and provides a cost-effective means to acquire a continuous water saturation profile, a critical asset for oilfield management and hydrocarbon exploration.
© 2025 The Authors. Published by American Chemical Society.