A hybrid artificial intelligence method to predict gas solubility and bubble point pressure

Bubble point pressure (Pb) and gas solubility (Rs) are reservoir fluid properties that have significant effect on oil reservoir characterization and reservoir engineering computations. Better knowledge of these PVT properties improve the reliability of reservoir simulation. Perfectly, these properties should be obtained from actual laboratory measurements. However, these measurements are either not available or very costly to be obtained. For these reasons, there is a need for a quick and reliable method for predicting these PVT properties. There are numerous approaches for predicting PVT properties, namely, empirical correlations and computational intelligence schemes. The objective of this research is to develop reliable model by utilizing a novel comprehensive approach using hybrid Self-adaptive Differential Evolution and Artificial Neural Network (SaDE-ANN) that could be used as a robust and effective model to predict Pb and Rs of crude oils based only on oil and gas gravities in addition to the reservoir temperature. This model was developed using data sets collected from published sources. Statistical error analysis was also used to check the validation of the two proposed empirical correlations introduced based on the weights and biases of the developed SaDE-ANN model. Based on the results generated by the proposed model compared with other conventional methods, that SaDE-ANN model is a reliable and accurate approach for predicting Pb and Rs of crude oils with a high accuracy and correlation coefficient of 99%. The developed approach is inexpensive with no additional required equipment or tools. This paper, for the first time, introduced a novel approach that utilizing only on three input parameters; oil gravity, gas gravity and reservoir temperature to predict Pb and Rs of crude oils. This approach has not been applied or investigated before this research. Therefore, it is a step forward to eliminate the required PVT lab experiments and discover the capabilities of hybrid optimization and artificial intelligence models as well as their application in predicting PVT properties. Outcomes of this study could help reservoir engineers to have better understanding of reservoir fluid behavior when experimental data samples are not available.