Glycol dehydration of high-acid

gas streams
A model based on mass and heat transfer closely predicts plant performance data
for glycol dehydration of high-acid gas streams

John Carroll Gas Liquids Engineering

Nathan Hatcher and Ralph Weiland Optimized Gas Treating

G
lycol dehydration is a proc- There are other facets of glycol two topics are inter-related because
ess that presents some dehydration that are interesting just both require the accurate assess-
unique challenges from both from an applied science viewpoint. ment of the interactions between
technical and computational stand- One of them is the heat transfer situ- components in the vapour phase;
points. In the first place, modern ation that ensues in a regenerator however, the solubility calculation
designs almost invariably use tower using both stripping gas and a is more demanding.
internals consisting of structured reboiler (Stahl column). When the Normally, to model the solubility
packing rather than the more tradi- hot gas hits the bottom of the pack- of a gas in a liquid, one uses the
tional bubble cap trays. Structured ing in the wash section atop the Henry’s Law approach. A thermo-
packing offers a lower pressure column, it finds itself suddenly dynamically complete version of
drop and considerably higher going from an environment in which Henry’s Law2,3 is:
capacity than trays, and it is well it is saturated with water in equilib-
suited to handling the very low L/ rium with a predominantly TEG >v–i∞ (P – Pjo)H (1)
a~ x H exp = y Pφ̂ v
G ratios common in dehydration. stream into an environment where it i i ij RT i i
However, until now, the height of is grossly under-saturated with
the packing was estimated using respect to the pure water stream in where
rules of thumb, not good engineer- the wash section. This causes γi: Activity coefficient
ing science. Mass transfer rate-based extremely rapid humidification, and xi: Mole fraction of component i in
modelling, on the other hand, uses the humidification process extracts the solvent (its solubility)
science and therefore offers greater the necessary heat of vapourisation Hij: Henry’s constant for solute i in
reliability of design. The other chal- as sensible heat from the liquid solvent j, kPa/mol frac
-
lenge of dehydration using any water phase. Sudden humidification vi∞ Partial molar volume of
glycol is thermodynamic: water is can drop the wash water tempera- insolvent j at infinite dilution,
the component of interest but it is ture by 30–40°F or even more. m3/kmol
probably nature’s most perversely This article addresses the efficacy Pjo: Vapour pressure of the solvent,
non-ideal chemical, which makes it of a new mass and heat transfer kPa
challenging to devise a truly accu- rate-based model in terms of how P: Total pressure, kPa
rate model for the vapour-liquid well it reflects known phase behav- R: Universal gas constant,
equilibrium, especially in systems iour and how closely it predicts 8.314 kJ/kmol·K
with a high-acid gas content. known plant performance data T: Absolute temperature, K
Traditionally, sweetened gases using both bubble cap trays and yi: Mole fraction of component i in
have been the main candidates for packed columns without recourse the vapour
dehydration, mostly using triethyl- to height equivalent to a theoretical φ̂iv Fugacity coefficient for
ene glycol (TEG) prior to entering plate (HETP) or height of transfer component i in the vapour,
the transmission pipelines. More units (HTU) estimates and other unitless
recently, there has been much inter- rules of thumb.
est in sour gas injection as a means Most people believe that Henry’s
of disposing of gas streams of too Phase equilibrium Law is only applicable to dilute
low quality for sulphur recovery. In Our phase equilibrium concerns are solutions, but the form in Equation
the context of carbon capture, the twofold: accurate calculation of the 1 can be applied without restriction.
recovered CO2 is compressed, lique- equilibrium water content of high- It would be unusual to do so, but
fied and injected into underground and low-pressure gases containing this equation can even be applied
formations. Handling these sour very high levels of CO2 and/or H2S; to mixtures that are not typically
and high CO2 streams requires and calculation of the solubility of considered to refer to the solubility
dehydration prior to compression all gases, including water, in the of a gas in a liquid at all; for exam-
and/or liquefaction. dehydration solvent itself. These ple, methanol and water.

[Link] GAS 2011 43

 it is rarely greater than two.
Saturated water content of gases
Therefore, it is very important to
include this contribution in the
GPSA ref. Mole % (dry basis) Temp, °F Pres, psia H2O lb/MMscf solubility model.
CH4 CO2 H2S Meas’d ProTreat
Ex 20-1 100 0 0 150 1000 220 216 The fugacity coefficient in the
Ex 20-2 80 20 0 160 2000 172 188 vapour phase can be calculated
Fig 20-9 0 100 0 100 500 132 125.3 using an equation of state such as
750 110 102.5 the Soave-Redlich-Kwong (SRK) or
1000 125 100.7
2000 215 215.1 Peng-Robinson (PR) equations.
3000 238 247.8 These cubic equations of state have
5.31 94.69 0 100 850 88 96.9 been used for many years for
1125 81 99.2 modelling the fluid properties and
1500 128 148.6
2000 139 184.2 phase equilibrium in petroleum
Fig 20-16 89 11 0 100 2000 40.6 41.1 systems containing light hydrocar-
89 11 0 160 1000 286 283.9 bons and a few associated
80 20 0 100 2000 40.6 45.1 non-hydrocarbons such as carbon
80 20 0 160 1000 282 292.5
80 20 0 160 2000 172 188.5 dioxide, hydrogen sulphide and
92 0 8 130 1500 111 103.5 nitrogen. However, they are less
72.5 0 27.5 160 1367 247 252.6 successful with water and other
83 0 17 160 1000 292 293.4 polar substances such as alcohols
30 60 10 100 1100 81 81.2
9 10 81 100 1900 442 264.4 and glycols. Typically, the equa-
5.31 94.69 0 77 1500 109.2 95 tions of state must be modified to
5.31 94.69 0 122 2000 164.6 234.5 predict accurately the pure compo-
nent properties, and the mixing
Table 1 rules must account for the complex
interactions.
The value of the Henry’s constant altogether at these conditions means At low pressure, both the
is a function of both the solute and the solubility is under predicted by Poynting correction and the fugac-
the solvent. Thus, the Henry’s Law approximately 10%. ity coefficients are approximately
constant, Hij, is different for meth- A common definition used in unity. Furthermore, if the solubility
ane in water than it is for methane describing the solubility of a gas in is low, Equation 1 reduces to the
in methanol, and Hij is different still a liquid is partial pressure. The more familiar form of Henry’s Law:
for ethane in water. Furthermore, partial pressure, Pi, of a component
for every solute-solvent pair, the Hij in a gas mixture is defined as: xiHij = yiP (4)
is a function of the temperature.
When comparing different solutes Pi = yiP (2) or equivalently:
in the same solvent, the larger the
Henry’s constant the lower the The non-idealities in the vapour xiHij = P i
(5)
solubility of the solute. phase are embodied in the fugacity
The activity coefficient accounts coefficient. As an approximation, Physically, this equation says that
for the effect of concentration on the fugacity is to the partial pres- the solubility of a gas increases
the activity of the component in the sure as the compressibility factor is directly as the partial pressure of
liquid phase. The infinite dilution to the molar volume. Included in the component in the gas increases.
definition of the activity coefficient this quantity is the effect of pres- However, this idealised version is
is used here; its significance is that sure, temperature and composition limited to about 200–300 kPa and
at infinite dilution, the activity coef- on the gas phase non-ideality. For gases with solubilities less than
ficient defined this way is unity. an ideal gas mixture, the fugacity about 0.1 mol%.
The exponential term is the coefficient of all components, i, in The ProTreat simulation tool’s
Poynting correction. It accounts for the mixture is unity: dehydration model uses the Peng-
the effect of pressure on the refer- Robinson equation of state (EOS)
ence fugacity, and it is important φ̂iv = 1 (3) for the vapour phase and currently
(significantly different from unity) offers a four-suffix Margules equa-
only at high pressure. For a small However, for a non-ideal mixture, tion activity coefficient model based
molecule such as methane, hydro- this quantity can be less than, on the data of Bestani & Shing1 for
gen sulphide or carbon dioxide in greater than or equal to unity. At the liquid phase, as reported by
water, the partial molar volume at relatively low pressure, it is gener- Clinton et al.4 A similar model
infinite dilution is approximately ally less than unity and it can be as based on the less conservative data
0.032 m3/kmol. Thus, at 7000 kPa small as 0.1 (and even smaller for of Parrish et al6 is planned for a
and 20°C (293°K), the Poynting liquids). At high pressure and high future release.
correction for these systems is about temperature, the fugacity coefficient There are two important aspects
1.1. Neglecting the Poynting effect can be greater than unity; however, to thermodynamic modelling of

44 GAS 2011 [Link]

phase equilibrium in glycol dehy- be simulated without this column if
Water content vs tray count
dration systems: the water content desired. Two condenser outlet
of the treated gas; and the hydro- streams permit the withdrawal of
carbon, acid gas and especially the Number Water wet stripping gas and/or water
of trays lb/MMscf
benzene, toluene, ethylvenzene and 5 8.5 vapour from the system (Stream 19)
xylene (BTEX) content of the water- 6 6.7 and the removal of a specifiable
laden glycol. Table 1 compares 7 5.7 portion of condensed water (Stream
ProTreat model results with Gas 8 5.1 20), with the remainder returned as
Processors Suppliers Association reflux.
(GPSA) Data Book entries for satu- Table 2 Table 2 shows the effect of the
rated water content. Generally, actual tray count on the water
ProTreat reproduces measured absorbed in the TEG solution. content of the dehydrated gas.
values of water content to within ProTreat results conform closely to ProTreat simulation indicates six
the accuracy of the data. The Peng- the conclusions of RR-131 as they trays are adequate to reduce the
Robinson EOS that performs these should, because ProTreat’s solubil- water content from 88.7 lb/MMscf
saturated water content calculations ity model has been regressed to the to the target level of <7 lb/MMscf
applies a large number of interac- actual measured BTEX solubilities. (32°F dew point). Tower diameter
tion parameters (kijs) for the for 70% flood is 3ft. These values
interactions between water and the Process simulation are in line with GPSA Data Book
various gases, as well as between The GPSA Data Book contains a nice results. In summary, the available
the gases themselves. example of dehydration with TEG data indicate that the model is accu-
Other components whose solubil- (Example 20–11). The gas is water rately reflecting literature data on
ity in TEG is pertinent are the acid saturated at 600 psia, with other the vapour-liquid equilibrium
gases and hydrocarbons, especially particulars noted in Figure 1. Two (VLE) and general TEG dehydra-
the BTEX components. Vapour- cases are detailed, both requiring tion experience, as reported by
liquid equilibrium constants two theoretical stages. One uses GPSA.
(K-values) for BTEX and o-xylene bubble cap trays, which, at a tray
are available in GPA RR-131, and efficiency of 25–30%, translates into Dehydration column performance
the data there have been used to fit 6–8 actual trays. The other case uses Until now, only an equilibrium-
the ProTreat solubility model for 10ft of an unspecified structured stage model has been available for
these species. The data indicate that packing. ProTreat has provision for calculations involving the perform-
at typical contactor conditions, a separate Stahl column, shown ance of structured packing.
approximately 10–30% of the immediately below the stripper in However, the HETP of the particu-
aromatics in the gas stream may be Figure 1, but the stripper can also lar packing is surely related to

<7 lb H20/MMSCF
(<0.01474 mole% H20) Dehydrated gas

350,000 reboiler duty

Outlet
3
99%wt TEG 12 19
3 gal TEG/lb H20 removed = 5.1 gpm
14
Contactor
2 18 CB 17 13 20
Outlet
Recycle Controller Cooler Stripper
Wet 10
gas Two theoretical stages
25-30% bubble cap efficiency = 6.8 bubble cap trays
1 16 9
Diameter = 3.4 ft
or
10 ft of structured packing
30 MMSCFD gas Flash gas
100F 8
4 6
600 psia
Water saturated HeatX
5 Flash tank 15 11

Reflux coil
7

Figure 1 GPSA Data Book example 20-11

[Link] GAS 2011 45

 sizes shown in Figure 2 indicate
0 that there can be a significant
M125.X temperature bulge in dehydration
5 M170.X
Distance from top of bed, ft columns caused by the phase
M2.X
change (condensation) of the water
10 M250.X
being absorbed. Water absorption
M350.X
generates sensible heating equiva-
15
lent to its latent heat of
condensation. Also, when small-
20
crimp packings are used, the
25 temperature bulge is closer to the
bottom of the column because the
30 larger surface area allows water to
absorb so much faster. Figure 3
35 shows the water content of the gas
at various positions along the
40 height of the column. It is evident
1000 110 120 130 140 that, after traversing the bottom
Temperature, ºF 20ft of M350.X packing, the gas is
about as dry as 99.95 wt% TEG at
Figure 2 Temperature profiles and the effect of packing size in the Mellapak X Series 100°F can get it. With M125.X pack-
ing, on the other hand, water is still
packing size. For structured pack- degree crimp angle) packings being removed even after the gas
ing, size can be expressed in terms affects dehydration performance. has passed through 40ft of packing.
of specific surface area and crimp For Sulzer Mellapak structured So the required bed height is very
size, characteristics that are geomet- packings, the packing designation much a function of the packing size.
rically related. Under otherwise (eg, M250.X) is a close indicator of It is not that rules of thumb cannot
identical process conditions, one the specific area, in this case be made to work; rather, it is that
should expect that large crimp 250 m2/m3. Simulations were all the right rules of thumb, at least for
packing will require a deeper bed run with 40ft of packing and the packing, depend on too many
to give the same performance as a absorber was sized for 70% flood parameters (not just packing type
relatively short bed of small crimp (9–11ft diameter depending on and size but also on the gas and
packing simply if for no other crimp size). The absorber was set solvent fluxes through the column)
reason than because the surface up to dehydrate 49 000 lbmol/h of and this makes them rough at best
area of the small crimp material is wet sweet methane (trace CO2 and and unreliable at worst. Until now,
so much larger. H2S) using 250 gpm of 99.95 wt% the answer to this dilemma has
Figures 2 and 3 are simulations of TEG. been to over-build the columns;
how packing size within the The gas-phase temperature however, in a competitive environ-
Mellapak X Series (roughly 60- profiles for the various packing ment, surely being able to avoid
over-design gives the knowledgea-
ble contractor and the astute
0 internals vendor a commercial
M125.X advantage. Mass and heat transfer
5 M170.X rate-based simulation is the preci-
Distance from top of bed, ft

M2.X sion tool that allows this to be

10 M250.X done.
M350.X Figure 4 shows that the blanket
15 rule of thumb of 6–8 trays for
dehydration can be a gross over-
20
simplification. The number of trays
depends at least on the dryness to
25
be achieved; that is, the dryness of
the solvent. If the target dew point
30
is not too stringent (say, 10 or 20lb
35 H2O/MMscf), 6 or 8 trays seem
adequate. But in very low dew
40 point applications, such as LNG
0 1 10 100 1000 plants, two or three times that
Water in gas, Lb/MMscf number of trays may be required to
get to the desired dryness. With
Figure 3 How water removal depends on packing size and packed bed depth 99.97% TEG, for example, it is

46 GAS 2011 [Link]

possible to get to below 1 lb/MMscf
water content, but even after 16 100
trays, water is still being removed.
For the particular conditions

Dry gas water content,

simulated in this study, it should be
mentioned that high TEG viscosity 10

lb/MMscf
is a consideration and will nega-
tively affect internals performance
compared with light hydrocarbons,
for example. At the solvent mois- 1 99.97%
ture levels encountered in these 99.95%
simulations, viscosity is not signifi- 99.90%
cantly affected by water content, 99.50%
0.1
and at the lowest temperature (feed 0 8 12 16
solvent at 100°F) the viscosity is Number of trays
about 19 cP (for reference, corn
syrup is 50–100 cP). This does not
necessarily rule out using trays, Figure 4 Water removal to low dew points requires deeper beds and more trays
although it does tend to make
structured packing more attractive. Observations from a mass transfer 1 psig. The dehydrator is treating
rate model 440 MMscfd of wet, sweet gas with
Stahl columns and stripping gas The mass and heat transfer rate a TEG flow of 1000 gpm, and the
A Stahl column gives an extra stage model uses real trays, both in bone-dry nitrogen stripping gas
of regeneration by taking the number and mechanical detail, and flow to the Stahl column is 1.9 scf/
solvent from the reboiler and real packing in terms of actual bed gal. The bed starts with a 2ft-deep
contacting it with a flow of dry depths of a specific packing with wash section for TEG recovery
inert gas. Stahl columns are essen- defined geometry, including crimp followed by an 8ft-deep stripping
tial when the dried gas must have a angle, crimp size, surface treatment, section for water removal. The
very low dew point. Figure 5 shows specific surface area and vendor. stripper was simulated by dividing
the effect of stripping gas rate (SCF The simulation of a packed column it into 40 segments, each having a
per gallon of TEG solvent) on the begins by finely segmenting the 3in depth. Finer segmentation is,
TEG purity and on the water packed height into a large number of course, possible, but it
content of the treated gas for of thin cross-sectional slices to adds nothing to the detail and
Example 20-11 from the GPSA Data approximate the continuous nature very little more to the accuracy of
Book scaled up by a factor of five of contacting in packing. As a the simulation. We will traverse
and using six bubble cap trays in result, one gets to observe mass and the regeneration column, starting
the absorber and a 10ft bed of heat transfer effects on a fairly with the condensate (essentially
FlexiPac 1.6Y in the regenerator detailed scale. pure water), which enters the
(includes a 2ft reflux wash section). Figure 6 shows the vapour and column from the condenser at
In terms of scale, the absorber is liquid temperature profiles through 180°F and is heated by the gas
nearly 7ft in diameter and is drying a 10ft-deep bed of 2in metal Nutter stream (stripping gas and water),
gas that is water saturated at 600 Rings in a regenerator operating at which is at 198°F.
psia and 100°F. The regenerator is
only 15in in diameter.
The simulation results in Figure 5  
show that using even a modest flow
of stripping gas can reduce the  

D r y
ga s,
lb/ M M sc G

water content in the treated gas by

Glyc ol
pur iU y w t %

 
more than a factor of two. Stripping
gas permits more water removal  
from the solvent and increases the
dry TEG from 98.8 wt% to 99.8  
wt%. This particular treated gas is  
dry enough by transmission line
standards, but it is a long way from  5&(
XU 
dry enough for an LNG plant, for MC..TD'
 
example. However, this demon-       
strates the principle that a Stahl Stripping gas rate, SCF/gallon
column can serve a useful purpose
— it should be used when treated
gas needs to be drier than usual. Figure 5 Effect of stripping gas flow on solvent dryness and water content of gas

[Link] GAS 2011 47

 When the vapour leaves the top of
0 the stripping section, its water
1 content is very nearly in equilibrium
Distance from top of bed, ft 1
with the liquid in the stripping
section. The liquid there is better
3
than 95 wt% TEG with only 5 wt%
4 (17–18 mol%) water. So, although
5 saturated with water, it is saturated
6 with only an 18 mol% water phase.
7 When it hits the first reflux segment,
this vapour is suddenly exposed to
8
a pure, hot water stream (flowing
9 Vapour
Liquid
from the wash section), so it is seri-
10 ously under saturated against pure
150 175 200 225 250 275 300 325 350 water. The consequence then is for
Temperature, ºF water to evaporate as fast as possi-
ble so as to resaturate the hot gas,
Figure 6 Temperature profiles in regenerator with stripping gas via Stahl column now against pure hot water. This is
a typical humidification operation
As the condensate trickles down this way? The answer for the gas and has some interesting character-
through the bed, it continues to be phase is relatively simple: feed, istics. The evaporation process is not
heated by the wet stripping gas, but preheated to 300°F, enters the mass transfer rate limited. Rather, it
within the first 18in its temperature column at the 2ft level and flashes is limited by the rate at which the
reaches a peak (194.5°F) and then into its vapour and liquid parts. latent heat of vapourisation
suddenly plummets to about 160°F. The fact that the vapour tempera- demanded by the humidification
The gas becomes hotter as we ture changes radically at the feed process can be drawn from the bulk
descend through the column, and point is simply the result of the hot liquid to the gas-liquid interface.
eventually it experiences a rapid feed meeting a cold reflux water This process is one of heat transfer
climb to 294°F. The question is: why stream. But why does the reflux and, indeed, the rate of humidifica-
do the phase temperatures trend water cool so much? tion right above the feed point is
controlled by essentially conductive
heat transfer across the liquid film
 running over the packing. The proc-
ess is heat transfer rate limited. The

D ist a nc e
f rom
t op
of


mass transfer driving force for

humidification is so high that

enough heat is drawn from the
be d,
f t

 reflux water to cool it, in this case by

 approximately 33°F.
 As we traverse lower into the
7BQPVS column and approach the bottom of

-JRVJE the stripping section, liquid starts

          to meet much hotter vapour coming
Temperature, ºF from the reboiler. (The bubble point
of TEG-water mixtures is quite
sensitive to water content when the

water content is low.) This cools the
 vapour and heats the liquid, and

D ist a nc e
f rom
t op

 again some of the heat transfer is

the result of water transferring from
of
be d,
f t


 the vapour back into the liquid in
the bottom couple of segments of

the stripping section.

7BQPVS Figures 7a and b are intended to
 show how the use of stripping gas
-JRVJE
 affects these temperature profiles.
           The plots are for the same 10ft bed
Temperature, ºF of Flexipac 1.6Y structured packing
referenced in Figure 5. Even when
Figure 7 Stripping gas is not necessary for anomalous temperature profiles stripping gas is not used, the
(a) No stripping gas (b) Stripping gas at 3 scf/gal condensed water leaving the reflux

44 GAS 2011 [Link]

section meets a much hotter gas,
and a good part of the temperature 100
Sweet
equilibration takes place by water

Water in treated gas,

CO2
evaporation from the reflux stream. 120 H2S
This is, again, a humidification

lb/MMscf
process and the reflux water cools
80
as a result of the demand for latent
heat for vapourisation. In Figure 7b,
the presence of stripping vapour 60
dilutes the gas leaving the stripping
section, which results in slightly 40
more driving force for humidifica- 5 10 15 20
tion together with the small Number of trays
additional capacity of the gas for
water. (How much the stripping Figure 8 Effect of tray count on water content of methane, CO2 and H2S dehydrated
gas adds to the driving force and with 99.5 wt% TEG at 15 psig and 100°F using bubble trays
capacity depends on the relative
flows of stripping gas versus an order of magnitude, from 2.0 to possible; for example, trading
reboiler vapour.) Again, the reflux 0.2 lb/MMscf. Twelve trays gave expensive reboiler duty for possibly
water temperature peaks and then no improvement. So much for rules cheaper stripping gas, or optimising
drops from 197°F to about 141°F, of thumb! dehydration unit performance in a
for a total cooling of 56°F, the same Finally, at 12 trays of dehydrator demanding LNG setting. Since this
as for the case without stripping contacting, some of the reboiler kind of tool is so reflective of the
gas. However, with stripping gas duty was traded for stripping gas. real engineering physics, the real
(dry nitrogen), the temperature Lowering the reboiler duty from 2.5 chemistry and the real processing
bulge in the reflux section is a to 2.0 MMBtu/hr and doubling the happening in an amine, dimethyl
consequence of the dilution by stripping gas rate (from 3 to 6 SCF/ ether of polyethyleneglycol
nitrogen. We note in passing that, gallon) resulted in a predicted (DMPEG) or glycol plant, using the
to maintain water balance, the no water content of 0.13 lb/MMscf (a tool as a virtual laboratory, pilot
stripping gas case required a 212°F dew point well below the hydrate plant and full-scale virtual plant is
condenser temperature versus 180°F formation temperature). Simulated very attractive.
for the case with stripping gas. glycol purity was 99.975 wt%. So it
appears that stripping gas is a more Dehydration of nearly pure acid
Factors that affect dehydration of effective way to improve dehydra- gases: H2S
sweet gases tion system performance. Down-well disposal of acid gases
To learn more about the practical There is much that can be learned (so-called acid gas injection)
limits of glycol dehydration, a by playing with a mass transfer rate requires the gas to be compressed
sensitivity analysis was conducted, model, even for such a seemingly to a very high pressure. If the gas is
starting with the 3 SCF/gallon mundane process as glycol dehy- wet, compressing it will cause
stripping gas rate at 1.72 MMBtu/ dration. The mass and heat transfer liquid water to drop out and this
hr reboiler duty referenced in the rate model allows one to probe the liquid will be saturated with acid
preceding section. First, the reboiler limits of what is in practice gas at high pressure. Unless one is
duty was increased to 2.5 MMBtu/
hr, which amounted to an increase
from 1124 to 1634 Btu/gallon. This 130
change allowed the glycol purity to Sweet
120
increase from 99.79 to 99.96 wt% CO2
Water in treated gas,

110
TEG. The water content of the H2S
dehydrated gas was predicted to 100
lb/MMscf

drop from 2.7 to 2.0 lb/MMscf, 90

assuming six bubble cap trays for 80
contacting. This is not much of an 70
improvement for nearly a 50%
60
increase in reboiler energy. Are
there enough absorber trays to take 50
advantage of the greatly improved 40
12 16 20 24 28
TEG dryness? The number of trays
Packing depth, ft
was increased from 6 to 8 in the
glycol contactor, while maintaining
the lean glycol purity of 99.96 wt%. Figure 9 Effect of tray count on water content of methane, CO2 and H2S dehydrated with
This dropped the water content by 99.5 wt% TEG at 15 psig and 100°F using Mellapak M170.X structured packing

[Link] GAS 2011 47

 prepared to build compressors and cannot be vented to atmosphere artificial residence times, HETPs,
other equipment from unobtainium, with the water vapour stripped HTUs and other rules of thumb.
it is paramount that the water be from the solvent. The mass transfer rate-based tower
removed from the gas before Its high solubility in TEG is really model does not use rules of thumb
compression. Therefore, it is of the main difficulty with dehydrat- — it uses science and good sound
interest to compare dehydration of ing very high concentration H2S; engineering to predict performance.
otherwise pure but water-saturated otherwise, in terms of just water These results were all obtained
H2S with the equivalent sweet gas absorption, the acid gases are no without any correction factors
volume. As Figure 8 shows, it is more challenging than any other whatsoever. They are true predic-
difficult to lower the water content gas from a purely process stand- tions in every sense of the word.
to the same level as high-pressure point. The difficulty is accurately
pipeline gas, for example, simply representing the gas-phase non- ProTreat is a mark of Optimized Gas Treating
because of the treating pressure. idealities that arise from the highly Inc.
The lines in this plot are simulation polar nature of water and H2S.
results for bubble cap trays in a References
1 Bestani B, Shing K S, Infinite dilution activity
10ft-diameter column treating 30 Conclusions
coefficients of water in TEG, PEG, glycerol and
MMscfd of either pure methane, Rules of thumb are fraught with
their mixtures in the temperature range 50 to
CO2 or H2S saturated with water at danger because often the rules 140ºC, Fluid Phase Equilibria, 50, 1989.
120°F. Lean solvent is 200 gpm of simply do not apply. A blanket tray 2 Carroll J J, What is Henry’s Law?, Chem. Eng.
99.5 wt% TEG at 100°F. Figure 9 efficiency of 25% is close to the Prog., 87, 9, 1991, 48–52.
shows similar results for Mellapak truth most of the time. But in deep 3 Carroll J J, Use Henry’s Law for
M 170.X structured packing. water removal, 25% is optimistic multicomponent mixtures, Chem. Eng. Prog.,
Under otherwise identical condi- 88, 8, 1992, 53–58.
tions, the water content of H2S in 4 Clinton P, Hubbard R A, Shah H, A review
equilibrium with the lean TEG is The mass transfer of TEG-water equilibrium data and its effect
on the design of glycol dehydration units,
somewhat higher than it is for
either CO2 or even methane. This rate-based tower Laurence Reid Gas Conditioning Conference,
Norman, OK, 2008.
has to do with the very polar nature
of H2S, which binds water more model uses science 5 Engineering Data Book, Gas Processors
Suppliers Association, 12th Ed, vol II, 20, 2004.
tightly to itself than CO2 does. But
this is information that can be and good sound 6 Parrish W R, Won K W, Baltatu M E, Phase
behavior of the triethylene glycol-water system
gleaned more easily directly from
VLE calculations. The mass transfer
engineering to and dehydration/regeneration design for
extremely low dew point requirements, 65th
rate-based dehydration model,
however, shows that in this instance
predict performance Annual Convention of the GPA, San Antonio,
TX, 1986.
H2S is pretty much completely
dehydrated after about 12 trays and, unless one adds several addi- John Carroll
(about 20ft of packing), whereas tional safety trays, failure will
CO2 and methane require 15 trays threaten. The situation with pack-
(24–26ft of packing) to reach their ing, be it random or structured, is
ultimate dryness. much worse. Quoting or recom-
In an overall sense, the acid gases mending a single HETP or HTU is
appear to be no harder than hydro- foolhardy. The right value depends Nathan Hatcher is Vice-President,
carbons to dehydrate, although H2S on the packing as well as on the Technology, Optimized Gas Treating, Buda,
is very polar, which leaves it a little operating conditions and the gas Texas. He has spent almost his entire career
wetter than most other gases. dryness sought. Rules of thumb had in amine treating and sulphur recovery, and
However, there is nothing particu- their place when the best one could has extensive experience in the practical
larly challenging about dehydrating do was an equilibrium-stage calcu- application of process simulation to
operations, troubleshooting and training.
the acid gases, except for their high lation, and reliance had to be placed
He holds a BS in chemical engineering from
solubility in the TEG solvent. For on experience as expressed (and
the University of Kansas and is currently a
the conditions corresponding to mis-expressed) in such rules of member of the Amine Best Practices Group.
Figures 8 and 9 (15 psig and 120°F), thumb. Today, we have available Email: [Link]@[Link]
the methane, CO2 and H2S content powerful heat and mass transfer Ralph Weiland is a co-founder of Optimized
of the rich solvent is simulated to rate-based simulation tools, and Gas Treating in Clarita, Oklahoma. He began
be 8.4 ppm, 0.66 mol% and 3.32 reliance on rules of thumb and working in gas treating in 1965 and taught
mol%, respectively. Thus, the water- other approximations and guesti- chemical engineering for 30 years at universities
laden solvent contains quite a bit of mates is no longer warranted. in Canada, Australia and the US, and directed
H2S and this must be removed The beauty of the heat and mass graduate research in gas treating. He holds
before the solvent is regenerated to transfer rate-based approach to BA Sc, MA Sc and PhD degrees in chemical
engineering from the University of Toronto.
a suitable state of dryness, because simulation is that one never has to
Email: [Link]@[Link]
it is a certainty that this much H2S worry about tray efficiencies,

44 GAS 2011 [Link]