Prediction of the Gas Compressibility Factor Using Coefficient-Matrices Based on ANN

Abstract Gas compressibility factor (Z-factor) is an important parameter that is widely used in petroleum and chemical engineering. An accurate and fast calculation of this parameter is of crucial importance and is used as an essential input in petroleum reservoir simulation. The Standing-Katz chart was published in 1942 and it has been considered an industry standard for the gas compressibility factor. Several methods have been developed to calculate z-factor by correlations such as the one proposed by Brill-Beggs. Other methods such as Dranchuk and Abou-Kassem (DAK) use iterative-based solutions. The DAK correlation gives most reliable fit to the experimental data along with a low average percentage error compared with those from other methods. However, for high values, the DAK correlation fails with very high error compared with the experimental data. In this study, artificial neural networks (ANN) are used to predict Z-factor. Data used for constructing the Standing-Katz charts and experimental data from the literature were used to build the ANN model and evaluate the quality of the new model compared with the other methods. Data used for constructing the Standing-Katz and Katz compressibility charts were used in constructing the neural networks model. 70% of the data was used for training, 15% for validation, and the remaining 15% for testing. The experimental data collected from the literature was given as a new data to test the quality of the model. The ANN model constructed using the DAK correlation gave an average absolute percentage error of 0.197 for testing; this value is better than that given by the DAK correlation. The error evaluated using the experimental data for ANN model is very low compared with the DAK correlation and those presented in the literature. In addition, a mathematical model is generated from the ANN model that can be used easily to determine the Z-factor at any conditions.