Prediction of Equilibrium Conditions for Hydrate Formation in Binary Gaseous Systems Using Artificial Neural Networks

Abstract

AbstractThe present work attempts to indicate the potential of artificial neural networks (ANN) for the fast and reliable estimation of the equilibrium conditions of single, binary, and multiple hydrocarbon gas hydrates. The ANN used in this study was a network with the tangent‐sigmoid (tansig) propagation transfer function in the hidden layer and a final layer with the linear (purelin) transfer function. The number of hidden neurons has been determined by minimizing the error of the calculation. The obtained results and the ANN model reliability were compared with other predictive methods. Results showed that the ANN method is able to reliably predict the hydrate equilibrium conditions of hydrocarbons, particularly for binary gas systems, using simple input parameters such as the weight fractions and normal boiling points of the components.