You are currently viewing: Articles


Glycol dehydration as a mass transfer rate process

Glycol dehydration is a process that presents some unique challenges from technical and computational standpoints. In the first place, modern designs almost invariably use tower internals consisting of structured packing rather than the more traditional bubble cap trays.

Nathan A Hatcher, Jaime L Nava and Ralph H Weiland
Optimized Gas Treating
Viewed : 476
Article Summary
Structured packing offers lower pressure drop and considerably higher capacity than trays, and it is well suited to handling the very low L/G ratios common in dehydration. However, until now estimating height of packing used rules of thumb, not science. Mass transfer rate-based modelling, on the other hand, uses science and therefore offers greater reliability of design. The other challenge of dehydration using any glycol is thermodynamic.

The dehydration of streams having very high concentrations of acid gases is hard to model reliably because the thermodynamics of vapour-liquid phase equilibrium involves water, one of nature’s most perversely non-ideal chemical species. Interactions between water and the acid gases CO2 and H2S, as well as with most hydrocarbons in the gas phase must be taken into account for a thermodynamic model to be reliable.

Furthermore, in the liquid phase, aqueous glycol solutions themselves are quite non-ideal because both water and glycol are polar molecules.

There are other facets of glycol dehydration that are interesting just from an applied science viewpoint.  One of them is the heat transfer situation that ensues in a regenerator using both stripping gas and a reboiler (Stahl column). When the hot gas hits the bottom of the packing in the wash section atop the column it finds itself going from an environment in which it is saturated with the water contained in a predominantly TEG stream into an environment where it is grossly under-saturated with respect to the pure water stream in the wash section. This humidification process extracts the necessary heat of vaporisation as sensible heat from the liquid water phase and this can drop the wash water temperature by 30°F, 40°F or even more.

Optimized Gas Treating, Inc has recently released a new glycol dehydration model, currently for TEG, and being extended to MEG and DEG. This paper addresses the efficacy of the model in terms of  how well it reflects known phase behaviour and how closely it predicts known plant performance data using both bubble cap trays and packed columns without recourse to HETP or HTU estimates and other rules of thumb.

Phase equilibrium
The concern here is with the accurate calculation of equilibrium water content of high- and low-pressure gases containing very high levels of CO2 and/or H2S. The ProTreat simulation tool’s dehydration model uses the Peng-Robinson equation of state (EOS) for the vapour phase and currently offers a 4-suffix Margules equation activity coefficient model based on the data of Bestani & Shing (1989) for the liquid phase as reported by Clinton et al. (2008). A similar model based on the less conservative data of Parrish et al (1986) is planned for a future release.

There are two important aspects to thermodynamic modelling: water content of the treated gas and the solubility of hydrocarbon, acid gas, and especially the BTEX in the water-laden glycol. Table 1 compares ProTreat model results with GPSA Data Book entries for saturated water content. Generally, ProTreat reproduces measured values of water content to within the accuracy of the data.  The Peng-Robinson EOS that performs these saturated water content calculations applies a large number of interaction parameters (kijs) for the interactions between water and the various gases as well as between the gases themselves as outlined, for example, by Carroll and Mather (1995).

Other components whose solubility in TEG is pertinent are the acid gases and hydrocarbons, especially the BTEX components. Vapour-liquid equilibrium constants (K-values) for benzene, toluene, ethyl benzene and o-xylene are available in GPA RR-131 and the data there have been used to fit the ProTreat solubility model for these species. The data indicate that at typical contactor conditions approximately 10–30% of the aromatics in the gas stream may be absorbed in the TEG solution. ProTreat results conform closely to the conclusions of RR-131 (as they should, because ProTreat’s solubility model has been regressed to the actual measured BTEX solubilities).

Process simulation
The GPSA Data Book contains a nice example of dehydration with TEG (Example 20-11). The gas is water saturated at 600 psia with other details noted in Figure 1. Two cases are detailed, both requiring two theoretical stages. One uses bubble cap trays which at a tray efficiency of 25 to 30%, translates into 6 to 8 actual trays. The other case uses 10-ft of an unspecified structured packing. ProTreat has provision for a separate Stahl column, shown immediately below the stripper in Figure 1 but the stripper can also be simulated without this column if desired. Two condenser outlet streams allow wet stripping gas withdrawal from the system (Stream 19), and removal of a specifiable portion of condensed water (Stream 20), with the remainder returned as reflux.

Table 2 shows the effect of the actual tray count on the water content of the dehydrated gas. ProTreat simulation indicates 6 trays are adequate to reduce the water content from 88.7 lb/MMscf to the target level of <7 lb/MMscf (32°F dew point). Tower diameter for 70% flood is 3’-0”. These values are in line with GPSA data book results which are annotated in Figure 1. In summary, the available data indicate that the model is accurately reflecting literature data on the VLE and general experience as reported by GPSA.

Dehydration column performance

Until now only an equilibrium stage model has been available for calculations involving the performance of structured packing. However, packing size is surely related to the HETP of the particular packing. Packing size can be expressed in terms of specific surface area and crimp size, characteristics that are geometrically related. Under otherwise identical process conditions, one should expect that large crimp packing will require a much deeper bed to give the same performance as a relatively short bed of small crimp packing simply because the surface area of the small crimp material is considerably higher.

Figures 2 and 3 simulate how packing size within the Mellapak X-series (higher crimp angle) packings affects dehydration performance. For Sulzer Mellapak structured packings, the packing designation, eg, M250.X is an approximate indicator of the specific area, in this case 250 m2/m3.
Current Rating :  2

Add your rating:

Your rate: 1 2 3 4 5