Use of artificial neural networks for estimating water content of natural gas mixtures taking into account heavy hydrocarbons contribution

When natural gas (NG) is extracted from the reservoir, it brings water in a vapor phase. Water present in natural gas needs to be removed to prevent disastrous consequences such as: corrosion, problems in pipelines, hydrate formation, difficulties in compression and transport processes. Accurate estimation of the water content in natural gas mixtures is the basis of the dehydration process design. In this regard, many water content methods have been developed (rigorous and semi-empirical methods). Some methods are based on pressure and temperature data. Other ones include the concentration of hydrogen sulfide (H2S), carbon dioxide (CO2), and methane in natural gas. There are no reported methods for calculating water content that considers the hydrocarbons heavier than methane present in mixtures of rich gas, gas condensate, or liquefied petroleum gas (LPG). Furthermore, some available methods for estimating water content in natural gas have low accuracy, and not all apply to typical operating conditions for natural gas processing. With artificial intelligence development, alternative estimation methods such as artificial neural networks (ANN) have proven to be accurate in estimating data for engineering applications. ANN base their structure on the biological neural networks functioning and can learn from a set of previous data to predict new data. Hence, this graduation project aims to develop a predictive model based on artificial neural networks to precisely estimate water content in natural gas mixtures, taking into account the composition of the heavy hydrocarbons present in the mixture for typical gas processing conditions. For this purpose, experimental data from open literature of water content in NG was required. The experimental data were processed, and the ANN was designed. Then, validation stages with other predictive methods were necessary to check the proposed method's applicability. As Ecuador is a hydrocarbon-producing country, the engineering tools for calculating the water content in natural gas mixtures are of interest to guarantee operational continuity in gaseous hydrocarbon production and processing facilities.

• Váldodahkki: Montenegro Madroñero, Jhon Fabián (author)
• Materiálatiipa: bachelorThesis
• Giella: eng
• Almmustuhtton: 2021
• Fáttát: Gas natural; Contenido de agua; Red neuronal artificial; Hidrocarburos pesados; Procesos de deshidratación; Natural gas; Water content; Aartificial neural network; Heavy hydrocarbons; Dehydration process
• Liŋkkat: http://repositorio.yachaytech.edu.ec/handle/123456789/376