CPs ee New gas-water-TEG equilibria Gas dehydrators give improved dew point depression when using higher concentrations of triethylene glycol RL. Hicks and E. A. Senules, ProComp, Inc and Hicks & Associates, Inc., New Orleans, La, NATURAL GAS DEHYDRATION can be required for produet specifications well below the conventional 4 to 7 Ib/MMsef Limit. Most existing triethylene glycol (TEC) design and rating programs use outdated equilibrium data and are not well- suited for getting high dew point depression. With the improvement in analytical tools aveilable for ‘measurement at low concentrations, considerably improved equilibrium daca are now available. This article ciscusses the ‘equilibrium data used in previous programs along with the evolution of an updated, more accurate program. New predictions are compared with actual operating data. The results show that tools are now available to the hydrocarbon processing industry to predict the interaction between TEG and water vapor in natural gas over a broader range, thereby allowing designs for ultrarlow dew point applications to be made with confidence. PREVIOUS PROGRAM Gas dehydratars commonly wte TEG, In these cehydrators, the natural gas is contacted with lean TEG, typically in countercurrent absorbers with bubble cap trays or packing. Calculations to determine the number of trays or fect of ucking, the required glycol concentration, oF the glycol Circulation rate, are performed as other typical sbsorption processes—requiring vaporiquid equilibrium data In absorption calculations, the equilibrium datais expressed usually in terms of component K-values as follows: K a yolte a where K = equilibrium ratio of water dy = mol fraction of water in gas phase Xe = mol fraction of waver in lean glycol The K-values can be calculated directly from equilibrium data or from various thermodynamic models such as liquid activity correlations or equations of state. ‘The most commonly used data available in the literature are the equilibrium dew poim curves for the natural gas-water- TEG system reported by Worley.) These curves were devel- oped in 1966 and are published in the latest edition of the GPSA Engineering Data Book (10th Edition, 1987, Fig, 15-15, represented in Fig. 1 as solid lines), The curves were derived by extrapolating data from fieldesize pilot plants to an infinite number of trays and to infinite glycol rates. To use agli water dow pins, ¢F 8 B08 Contac temperature, 73000 Fig, 1—Werley's curves (solid ines) used forlater comparison with Perish ‘sala (éashed Ines). these curves in dehydrator calculations, K-values are deter= mined by correlating the equilibrium dew point curves with natural gas water content data such as that of McCarthy, et al.,? McKetta and Wehe! or Bukacek.t ‘Using activity coefficients, an alternative method of caleu- lating K-values for TEG absorbers was developed in 1961 by Scauzillo.$ Scauzillo used the experimental cquilibrium daca of Townsend and of Wise, et al.,? for the natural gas-water- TEG system to calculate activity coefficients of water in the liquid phase, Townsend's data revealed that the ratio of the vwater content in equilibrium with a glycol solution to that of the saturated gas was essentially constant for any given glycol concentration, Furthermore, this ratio was equal to the activity of water in the TEG solution, independent of pressure and not significantly affected by temperature in the range of interest. Thus, using the following Eq. 2, Scauzillo derived the expression for the K-values shown in Eq. 3 as follows: ee) & X = (2.105¥10-5) (44) (41) ° where 1 = activity coefficient of water in liquid phase aeRBs i i i i = er o EG concoration, wr Fig. 2—Dew penis culled by Sonualle’s method (P = 7.000 pos) p08 wl wh Why's curves, 18 cocoa we Fig. Scale coeticiens fox TEG verve war compved wih Pars deta. Wg = water content in gas phase in equilibrium with lean TEG, WMMef Wy m water content in gan phase at saturation, Ib/MMcf ‘Ae = moi fraction of water in liquid phase ‘Using water content deta of McCarthy, et al., and K-values calculated by the ineihod just described, Scausillo presented a graph showing equilibrium dew points available with glycol concentrations of GO 10 99.9 wie and in the temperature range af 40 to 120°F. Scausillu's dew point curves closely Tt the curves presented by Worley in 1906 as shown in Fig. 2 Fig. 2 represents the caine equilibrium data as in Fig. 1, presented diferently to show the ininimal disparity between the Scauzilo ealeulations and the curves presented by Worley. ‘The author's original prograin used the Seauzillo method of calculating K-values asite equilibrium basis. Eq 3 provided a simple means co calculate the equilibrium K-value by use ofthe activity coefficients which were obtain by curve fiting. Seauaillo's original activity coefficient plot shown as a solid ine in Fig. 3. Absorption calculations. The major calculation required fir TEG unis isto determine the circulation rate required to achieve a yiven dew point depression with a given eoncentrs~ ion ind number of trays, There are various methods for making this absorprion ealeulation, Although trsy-to-tray caleulations could have been per formed, the cyuilibrium daia was. not believed tn he of sufficient accuraty to warrane the effort. Additionally, per ‘onal computers at that time were not fast enough to slaw he tray-to-tray ealculations. Other simplified methods were aca TE (ae uth wale gat Fig, ¢—xample with § actual rays and TOO°F coat temperature tbs chest of ogher TE conusttahan Fig, §—Sxample wit 6 acual aye, 100°F contact iompovature end selanded to igher TES concent are available, such as the MeCabe-Thiel graphical method and the Sherwood method, However, 10 the essemtially constant abicxjion Cactors over the entire column, The aceompanying box, ‘The Kremser- Brown method,” provides more details on the reais for this constant value as well as how the equation is used in dhe calculations, Comparison to actual date, To determine the accuracy (or indeed the applicability) nfa calculation method, one needs {o be able to compare the computed rents with actual operating data. In our literature search and in discussions withBB iso] Tenorialae,F TE rt, gle wate ngs ig. Su-Water coment Below hyorate point adjueled by exrepolt sie ala. aan speraing engineers at majo cil eompeinies, we were unable 9 cbiain any mote sompiete data tha that of Worley-the mame data from which the GPSA equiilrim curves were generated, Worley's data, expressed as operating curves for serual TEG eontactre, are reproduced in Figs. 4 ad 5 (GPSA Figures 20-40 and 20-41) for 6 and 8 actual ays, respectively. Note that thete curves are applicable for 100°F contact rempererures only. Although the curves ale 25 yeais nid (1966), they are considered accurate up to lean TEG concentrations of 99.9%. However, extapelations te higher rencentrations have much wnecrtainty ‘Numerous calculations were made withthe original glycol srogeam, using the Senszillo squibrium data, to provide ezults for comparison with the Worley operating curves and ‘dh mare init data from glyecl unit maaufoeturers. Using he Kremer Brown methead with = wonsant try efficiency, poor agreement wat inally obtained between the program sults and the operating dala. Mowever afer incorporating coreation to adjust the overall column efficiency based an centration driving forces and other operating parameters cellent agreement was obtained between the program results ind nctunl operating da ‘Many glyenl units were designed with his earlier version oh which lowed more skernatives 10 be susideted widh iuiinnal engineering efor. Increasing con- idence in the design program allowed engincers 10 design ants with less excess capacity, hereby recucing capital and sperating coms, ‘srage REVISED PROGRAM Altiough the eatlier program provided exceptional seius x compared 10 the serual field data, there was some concern as wits accuracy at higher glyeol eoneentrations. With publication Fig. 7—Sxame wy 10 actual aye, 00°F conta iemporatr ‘nln to haher TEG coneentaiona na of new, broader ranged equilibrium data by Parvsh, ot al.? there arose much interest in modifying the program to include the new deta which would conceivably alow mere accurate ‘alelations for low dew paint applicsions Earlier Figs. and 3 show comparisons forthe equilriam values and sctivity cociiconts, respectively, between the ‘newer eeilibrium data (dashed lines) and chat of Worley (slid Tines). The Parrish dew point curves were generated for lov coven syserns hy using their experimental acing coef Sienis wilh a nvthon of expoating he vapur ihe ancl the iquid phase fagaitien (Apprraln B of the Parti per). Tn this method, the water vapor fugacity eoeffiient (G*™) and the Poynting correction faor forthe liquid (the expose term shown later in Eq, §) were assumed to be unity. “To determine what effect uring the activity coefficient siven by Parrih might have, equilibrium dew point curves were generated using the Parrish activity coefficients with Ea, Sand water content data of Bukacck. ‘The ealeylated dew point values were considerably divergent from the equilibrium ‘dew point curves presented by Parrish, particularly ar increas ing prescure. Thins, it was conchided that the Parrish dew point curves were not applicable to the high pressure natural Haswater-TEG syste, although the experimental activity coeficlents for water wore voriered 10 be valid and neces sary for use in upgrading the equilibrium basis ofthe progracn, New data for watar content. The iniial atempe to we the activity coefficients in the alsin calculations involved theealeulation of K-values using the calculation ofthe fugacity coefficient of water vapor (x) and the Poynting convection {ator a3 fellows: Sle = Pate P O) Foam taf Uy exp (—PVedRT) © STU