Determination of Methanol Loss Due to Vaporization in Gas Hydrate Inhibition Process Using Intelligent Connectionist Paradigms

Abstract

The clathrate hydrate formation in pipelines and treatment systems of gas and natural gas liquid (NGL) is an undesirable operating phenomenon. It interrupts the gas flow continuity, reduces safety level, and imposes substantial costs on both gas and NGL processing plants. Therefore, it is necessary to prevent or at least postpone this undesired and high-risk phenomenon. The application of inhibitor agents to shift phase equilibrium of gas hydrate formation to lower temperatures and higher pressures is a mature technology. Methanol (MeOH) is a well-known thermodynamic agent in the hydrate inhibition process that cyclically injects to the gas phase and then recovers and reuses. Significant amounts of methanol vaporize/loss during its recovery in a three-phase separator. An accurate determination of this loss is necessary to estimate the amount of methanol make-up. Therefore, this study tries to determine the methanol loss using six intelligent connectionist approaches, i.e., least-squares support vector machines (LS-SVM), adaptive neuro-fuzzy inference systems, and artificial neural networks. The LS-SVM was finally detected as the most accurate paradigm for the considered purpose. Methanol loss estimation by the LS-SVM model is in excellent agreement with 196 real-field datasets in the literature, i.e., AARD = 0.295% and R2 = 0.9999. An economic study shows that methanol loss may impose more than 132 million US Dollars per year to a gas plant that processes 674 million standard cubic meters of gas per day.