A new approach for estimating water saturation in low-resistivity hydrocarbon-bearing reservoirs using artificial neural network (ANN)

The low-resistivity pay zone phenomenon is among the main challenges in the oil industry, where the hydrocarbon-bearing pay zone shows low-resistivity values instead of declaring high-resistivity values. It was first noticed in the Algerian Southern Field, and several studies have been conducted to correct the underestimation of the hydrocarbon volume obtained from the petrophysical logs, specifically resistivity logs, whereas it has been established that heavy and electrically conductive minerals are the primary cause of the underestimated results. Furthermore, in previous literature, some researchers suggested using a modified Simandoux equation with compensating terms. Nonetheless, these terms are added empirically with no practical background, making generalizations on similar cases in the same field complicated and of low reliability. This study’s main goal is to design a modified model for estimating water saturation (Sw) in low-resistivity reservoirs based on a trained artificial intelligence algorithm. This main contribution of this paper is that the developed algorithm is based on combining the artificial neural network-derived model with a support vector machine technique (ANNSVM). The estimated water saturation values using this algorithm are in full agreement and better matched with the field measurements derived using the drill stem tests and the modular dynamic tester, indicating its applicability and accuracy to estimate the Sw in this kind of challenging reservoir. Moreover, in comparison to the conventional water saturation models, the developed ANNSVM model enables predicting more reliable data corresponding to real data. Therefore, this proposed algorithm could be successfully applied to accurately estimate the Sw in other low-resistivity reservoirs in the Algerian basins and worldwide.