Search

High pressure subsea natural gas dehydration (NGD) units using ethylene glycol (MEG) absorption have been proposed. To expand the experimental database and assist design qualification, new vapor–liquid equilibrium (VLE) experimental data have been measured for a 20-component glycol–water–natural gas mixture at T = (288–323) K, p = (6.0, 12.5) MPa, and wMEG,feed = (90, >99.8) %. MEG, H2O, CO2, N2, and alkane (methane to n- and i-pentane) phase distributions have been quantified. Experimental uncertainty ranges from ±2–42%, with the greatest uncertainty for the quantification of trace components. Experimental results are modeled using the Cubic-Plus-Association (CPA) equation of state. Overpredictions (∼9%) are observed for the water content of the vapor phase. CO2 is shown to have a large effect on yMEG, leading to modeling deviations in the order of 65%. A relatively accurate prediction of the natural gas partition coefficients was observed for major components C1–C3 and CO2, with modeling errors ranging from 5% for methane to 10% for CO2. More significant deviations were observed for trace components, with the largest deviation of 73% N2. The CPA model provides both satisfactory and conservative results suitable for use in NGD process designs. On the basis of this work, operation at subsea conditions would significantly improve dehydration capability.

Novel technologies in the field of subsea gasprocessing include the development of natural gas dehydration facilities, which may operate at high pressure due to their proximity to reservoirs. For the qualification and design of these processing units, ternary vapor−liquid equilibrium data are required to validate the thermodynamic models used in the design process. For this purpose, 16 new ternary data points were measured for ethylene glycol (1) + water (2) + methane (3) at 6.0 and 12.5 MPa with temperatures ranging from 288to 323 K and glycol content above 90 wt %. Glycol in gas (y1),water in gas (y2), and methane solubility (x3) were measured with relative experimental uncertainties (ur(x) = u(x)/|x|)below 12%, depending on the type of data. The Cubic-Plus-Association (CPA) equation of state was used to model the data. Literature pure component and binary interaction parameters were used. It was found that the model provides a good qualitative description of the experimental data for y1 and y2, while a significant over-prediction occurs for x3. The modeling errors for CPA ranged between 5−40% average absolute relativedeviation.