The Prediction of Hydrate Formation Rate in the Presence of Inhibitors

Abstract

The main objective of this study was to present a novel approach to access more accurate hydrate formation rate predicting models based on the combination of flow-loop experimental data with learning power of artificial neural networks. Therefore, more than 2,300 data of C 1 , C 3 , i-C 4 , and CO 2 hydrate formation rate in the presence of two kinetic inhibitors (PVP and L-Tyrosine) and two inhibitor intensifying additives (PPO and PEO) was used. It was found that such models can be used as powerful tools, with total errors less than 2% for the developed models, in predicting hydrate formation rate in these cases.

Comparison of intelligent systems, artificial neural networks and neural fuzzy model for prediction of gas hydrate formation rate

“…The number of neurons in the input and output layers are equal to the number of variables that are being presented to the network as inputs and targets, respectively. Determination of the appropriate number of the existing neurons in the hidden layer(s) which are principally responsible for feature extraction is difficult and time-consuming and is often done by trial and error (Graupe, 2007, Blusari,1995, Shadravanan, 2010.…”