Comparing the capability of artificial neural network (ANN) and CSMHYD program for predicting of hydrate formation pressure in binary mixtures

In the present study, we investigated forecasting of hydrate formation pressure of binary mixtures including at least one hydrocarbon using a feed-forward multi-layer artificial neural network. For this purpose, 895 experimental data which cover a wide range of temperatures and compositions were collected from different studies cited in the literatures. In order to find the best model, different ANN types are tested through the absolute average relative deviation percent (AARD), mean square error (MSE) and the regression coefficient (R2) and the optimal configuration is selected. It is found that the selected ANN model is based on the statistical analysis has an excellent agreement (AARD=1.02, MSE=1.27×10−5 and R2=0.9938) with the collected experimental data. The obtained results reveal that the developed MLPNN model is an applicable and feasible tool to predict hydrate formation pressure with high accuracy with respect to Colorado School of Mines Hydrate (CSMHYD) Program.