GANN: A Hybrid Model for Permeability Prediction of Oil Reservoirs

Abstract

Permeability is an important property of a petroleum reservoir that indicates the amount of oil in the reservoir and its ability to flow. The ability to predict reservoir permeability can significantly improve oil field operations and management. One method to obtain reliable permeability data is to analyze cores in the laboratories, which is very expensive, time consuming and not applicable in all cases. Another method better suited to smart cities is to use log data from oil wells to predict permeability, which is fast, reliable, and very cheap. In this study, we apply multiple artificial intelligence (AI) techniques to well logs to predict oilfield permeability in search of a more powerful hybrid model. In this paper, we propose Genetic Algorithm Neural Network (GANN), a hybrid model for permeability prediction, using the neural network as the primary model to calculate weights for the prediction and the Genetic Algorithm as the secondary model to optimize the results generated by the Neural Network be used. The experimental results show that the GANN model can estimate the permeability of oil reservoirs with higher correlation coefficients and lower mean square errors compared to the individual AI techniques.