Descripción completa

Views 184 Downloads 12 File size 245KB

Report DMCA / Copyright


Recommend stories

Citation preview


Lars Erik Øi, Associate Professor, Telemark University College, Porsgrunn, Norway Elisabeth Tyvand Selstø, Process Engineer, Kvaerner Stavanger, Norway

for presentation at GPA Europe's meeting in Bergen 13th - 14th May 2002

ABSTRACT Natural gas dehydration is normally performed by absorption in triethylene glycol. Regeneration by traditional distillation achieves about 1 weight % water in regenerated glycol. Use of stripping gas with or without an extra column, Coldfinger, Drizo and the Read Cycle are processes that achieve less than 1 % water in regenerated glycol. These processes are difficult to model because of special details, and convergence is a traditional problem in regeneration calculations. Models for all the glycol regeneration processes have been developed in the process simulation program HYSYS with the Peng Robinson equation of state. The models can also be implemented in other process simulation programs and with other vapour/liquid equilibrium models. Parameters in the models have been adjusted to simulate actual regeneration plant performance. The models have also been used to evaluate alternative regeneration processes.


INTRODUCTION Natural gas is dehydrated to avoid corrosion or pipeline plugging due to ice or hydrate formation. The traditional way to dehydrate natural gas is by absorption in triethylene glycol (TEG). Normally, the water is removed from the (rich) glycol by distillation, and then the regenerated glycol is recycled back to the absorption column. Figure 1 shows a typical process. After absorption, soluble gases like CO2 are first released from the glycol in a flash tank, and then the glycol is heated before it flows to a distillation (regeneration) column. The regeneration column has a reboiler and a condenser. Regeneration of the glycol by atmospheric distillation at about 200 °C achieves about 1 weight % water in regenerated glycol. The regenerated glycol is pumped back to the absorption column and heat exchanged against the rich glycol and cooling water.

Dry gas


Absorption column

Regeneration column

Flash tank

Wet gas

Reboiler Glycol pump

Rich glycol

Figure 1: Traditional TEG absorption and stripping process

To achieve less than about 1 weight % water in TEG, there are several special regeneration processes available. Examples are use of stripping gas with or without an extra column, Coldfinger [1], Drizo [2] and the Read Cycle [3]. There is not much available information in literature about simulation of such processes. GPSA [4] and Kohl [5] are general references for dehydration processes. The aim of this paper is to show how different glycol regeneration processes can be modelled in a simulation program. Then it is showed how these models can be used to evaluate alternative regeneration processes. This paper is based on a M.Sc. thesis from 2001 [6] by Elisabeth Tyvand Selstø at Telemark University College with Lars Erik Øi as responsible supervisor.


SIMULATION OF A TRADITIONAL TEG DEHYDRATION PROCESS Available models Models for traditional TEG dehydration processes are given from the vendors of the process simulation programs HYSYS, PROSIM, PRO/Vision and ASPEN PLUS. Except for this, there is not much information about simulation of dehydration processes in the literature. A paper about dehydration simulation in PROSIM was presented at a GPA conference in 1992 [7]. At Hyprotech 2000, two papers about simulation of glycol dehydration were presented, with emphasis on steady state simulation [8] and dynamic simulation [9]. An example of a HYSYS simulation of a traditional TEG dehydration process is given in figure 2.

Figure 2: HYSYS simulation of a traditional process

Compared to the general process in figure 1, the flowsheet includes a pre-cooler and a pre-scrubber to remove liquid water in the simulation. In the simulation, the pre-scrubber ensures that the gas to the absorption column is exactly saturated with water. The storage tank is simulated as a flash unit to ensure that the flow through the pump is pure liquid (in the calculations). One recycle block is used to control that the recycled glycol flow equals the glycol flow to the absorption column.

Convergence problems The calculation of a traditional TEG dehydration process in HYSYS, PRO/Vision or ASPEN PLUS is quite straightforward, but simulation calculations involving absorption columns, distillation columns and recycles often have a tendency to diverge. This is discussed in reference [8]. The absorption column normally converges if the feed streams (natural gas in feed and recycled glycol) are specified, and the number of theoretical stages is limited (2 or 3 is typical).


The overall process can be difficult to converge because of the recycle from regeneration to the absorption column. The normal way to start the calculation, is to guess the glycol liquid stream to the absorber. When the rest of the process is calculated, the resulting liquid to the absorber is not exactly equal to the guessed stream. This should be solved by some sort of iteration. However, this is often a problem, due to the very small glycol loss that leaves the process in the degassing and the stripper. There are two common ways to solve this. One way is to stop the iteration after two cycles and accept the negligible difference in glycol flow. The other way is to calculate the glycol loss and add this to the recycled glycol stream. The regeneration (stripping) column including reboiler and condenser often diverges. Very few theoretical stages are necessary (2 is typical), and a very low reflux is needed. It is recommended to use reboiler temperature (e.g. 204 °C) and a low reflux ratio (e.g. 0.5) as the specifications.

Vapour/liquid equilibrium In the HYSYS calculations, the Peng Robinson (PR) equation of state has been used. This model is recommended by Hyprotech [10] for TEG dehydration. The PR equation is simple, reliable and effective. But it is not very accurate, and it also has some limitations. The PR equation has only one adjustable binary parameter for each binary component pair. This is not enough to fit vapour/liquid equilibrium data accurately. An activity model with parameters fitted to new TEG/water equilibrium data is more accurate [11]. This is important in absorber design calculations. For flowsheet development and comparison calculations, the PR equation is satisfactory. The PR equation in some cases calculates significant amounts of TEG in top of the regeneration column. According to the very low vapour pressure of TEG, this is not correct according to equilibrium data. The reason for this error is that the TEG/hydrocarbon binary parameter has been used to fit hydrocarbon solubility, and this results in very wrong TEG solubilities in hydrocarbons. However, the error in calculating loss of glycol is of minor practical importance. The actual loss of glycol from a dehydration plant is due to mechanical losses (droplets or foam) from the absorption column, the flash tank or the regeneration column.

Base case A dehydration case from the North Sea with 36500 kmol/h natural gas at 8550 kPa and temperature 26 °C is used as a base case. The water specification in dried gas is less than 37 ppm (parts pr. million on a mol basis). In the simulation calculation, the regeneration pressure was set to 110 kPa (slight overpressure) and the reboiler temperature was set to 204 °C. The glycol circulation rate was set to 24 kg TEG pr. kg absorbed water [4]. This results in a glycol purity of 98.9 weight % or 1.1 weight % water. The water concentration in the dried gas was calculated to 58 ppm. This is far from the specified 37 ppm. It is not recommended to increase the reboiler temperature, because the glycol will degenerate above this temperature. It is normally not accepted to operate the regeneration column below atmospheric pressure because oxygen leakage into the system can cause an explosion hazard. To increase the number of theoretical stages will only reduce the water content slightly. The conclusion is that regeneration by traditional distillation will only achieve about 1 weight % water in regenerated glycol, and the water specification in the dried gas will not be achieved in the actual base case.


SIMULATION OF REGENERATION PROCESSES Stripping gas to reboiler The simplest way to reduce the water content in the glycol, is to introduce stripping gas to the regeneration column. The effect of this is shown by GPSA [4]. The stripping gas can be nitrogen, recycled natural gas or flash gas (from the flash tank after the absorption column). Traditionally, the stripping gas is added to the liquid in the reboiler. The additional feed of stripping gas compared to the HYSYS simulation in figure 2 is shown in figure 3. (The apparent feed of stripping gas to the middle of the column is due to the HYSYS graphics. The actual stripping gas feed is to the reboiler.)

Figure 3: HYSYS simulation of stripping gas to reboiler

The implementation in HYSYS is easy, and the regeneration column will converge almost as easy as without stripping gas with suitable specifications. Recommended column specifications for easy convergence is reboiler temperature and reflux ratio. When convergence is achieved, it is easy to vary e.g. the stripping gas rate to achieve the specified water concentration in the product gas.

Stripping gas to extra stripping column More efficient use of the stripping gas can be achieved by adding an extra stripping column below the reboiler. (This column is often called a Stahl column.) This result in a reduced amount of stripping gas needed to achieve a specified glycol concentration [4]. The implementation of such a column in the HYSYS simulation is shown in figure 4. Compared to the simulation in figure 2, an extra recirculation stream is included. A new recycle block is needed to control that the stripping gas flow from the extra stripping column equals the stripping gas to the regeneration column.


Figure 4: HYSYS simulation of extra stripping column

The suggested calculation method to calculate the new recycle, is to start with a dry stripping gas stream from the extra stripping column, then calculate the distillation column and then calculate the extra stripping column. This can be repeated until convergence. Another possibility is to simulate the regeneration column and stripping gas column as one column [8]. The reboiler is then implemented as a heat input some trays above the bottom. The stripping gas is fed to the bottom tray.

READ Cycle The principle of the Read cycle [3] is that the hydrocarbon gases from the regeneration column is recycled (and used as stripping gas). The gas from top of the distillation column is cooled, the water is condensed, (some of the) hydrocarbons are pressurised with a blower and used as stripping gas in the regeneration process. The advantage is that there is no need for additional stripping gas. A READ cycle process implemented in HYSYS is shown in figure 5. Compared to the HYSYS simulation of the processes in figure 2 and 4, some heat exchangers are added. A purge gas flow from the Read cycle is also included. Except for an extra recycle, the introduction of a Read cycle does not add much simulation complexity to the calculations. A relevant simulation is to calculate the amount of stripping gas and necessary compressing effect to achieve the specified water in the dried gas.


Figure 5: HYSYS simulation of a Read cycle

DRIZO The principle is that a stripping gas medium used in the regeneration columns is recovered as a liquid after the regeneration column. The liquid is separated from the condensed water in a three phase separator and is recycled to the regeneration columns. The stripping medium can be introduced externally (e.g. n-heptane) or it can be absorbed from the natural gas (e.g. BTEX, benzene, toluene, ethylbenzene, xylene). Under some circumstances, surplus of aromatics can be recovered as a condensate product from the Drizo process. In a Drizo process, only a pump is necessary to recirculate the stripping gas. There are several Drizo processes in operation, which achieve very low water dew points in dried gas [2]. Introduction of a stripping medium introduces new complications in the simulation. The calculations will be dependent on the reliability of the equilibrium data between the stripping medium, TEG and water. If the stripping medium is absorbed from the natural gas (as BTEX), there are uncertainties in the absorption equilibrium, in the regeneration equilibrium and in the condensing equilibrium. A suggested simulation is suggested in figure 6. Compared to the HYSYS simulation for the Read cycle in figure 5, a 3 phase separator is introduced, and a solvent is added externally. Also in the Drizo process, a gas purge from the gas circulation is needed.


Figure 6: HYSYS simulation of a Drizo Process

As in the Read cycle simulation, calculation of necessary amount of recycling and necessary recirculation effect is of interest.

COLDFINGER The principle of the Coldfinger process is that the partly regenerated glycol from the distillation column is further concentrated in a two phase tank with a cold spot (the Coldfinger) and a condense collector in the gas phase (figure 7). The condensing of the water results in reduced water pressure in the tank, and more water will evaporate from the glycol [1]. The advantage of the Colfinger process, is that it achieves more concentrated glycol with a very simple unit, using no additional stripping gas.

Coldfinger Feed



Liquid Product Figure 7: The Coldfinger unit


The Coldfinger unit is difficult to simulate in a steady state simulation program because the unit is not in equilibrium. The condensing at the cold spot is at a lower temperature than the glycol. The suggested simulation in HYSYS is performed with two flash units. The first is a flash tank simulating the equilibrium between the glycol and the vapour above. The vapour above is sent to a second flash tank simulating the condensing and removal of the water at a lower temperature. The simulation is showed in figure 8. Three recycle blocks are necessary in the calculations. Two of them are used to control that the recycled flows to the absorber column and the regeneration column equals the flows to the columns as in the previous HYSYS simulations in figure 4, 5 and 6. The third recycle block is used to control that the gas flow recycled from the condensing part of the coldfinger unit is equal to the gas feed to the equilibrium part.

Figure 8: HYSYS simulation of a Coldfinger Process

The extra cost of the Coldfinger unit is due to the extra equipment and the extra heat added to compensate for the cooling. It is possible to calculate the necessary temperature in the condensing process and the extra heat added to achieve the specified water in the dried gas.


EVALUATION OF THE SIMULATION CALCULATIONS Tuning of the models The simulation calculations in HYSYS (or in any simulation program) are in principle dependent on the reliability of the gas/liquid equilibrium calculations. However, it is expected that the most important uncertainties in the calculations are connected to the estimation of the efficiency in the columns. A measure of the efficiency is the number of equilibrium stages in the column. Performance data for similar regeneration plants have been used to tune the models to actual efficiency performance. In a simulation model for a regeneration process in operation, the absorption column efficiency and the extra stripping column efficiency were fitted to represent measured performance data. The fitted column efficiencies were used in the base case models for comparison calculations.

Comparison of the processes The models developed have been used to evaluate the possible alternatives. All the models except the alternative without stripping gas were able to reduce the water content down to the water specification in dried gas. The performance of the processes were compared taking into account - the amount of stripping gas - additional equipment - effects in pumps/blowers - heat consumption The cost of stripping gas is the sum of the stripping gas itself and the need for processing (often recycling) of the stripping gas. In general, the cost differences between the simulated processes were quite small. Care must be taken in evaluating such comparisons. There are uncertainties in the modelling, in the equilibrium calculations and in the equipment efficiency estimation (especially the column efficiencies). There are also differences in cost and reliability that can not be taken into a simulation calculation. A comparison analysis based on simulation calculations can be valuable for comparing the efficiency of different process principles. But the traditional problem normally brakes down to a comparison of investment cost, operation cost and reliability of available offers.

Other simulation tools The models in this work have been implemented in HYSYS. HYSYS is an equation based simulation program, contrary to the sequential modular based simulation programs PRO/Vision and ASPEN PLUS. However, the implementation of the recycle loops in HYSYS has been done by tearing (guessing) the process streams similar to the way it is done in a sequential modular simulation program. The suggested models should work well also in other simulation programs and with other vapour/liquid equilibrium models.


CONCLUSION Models for the most used glycol regeneration processes have been developed in the process simulation program HYSYS with the Peng Robinson equation of state. The models can also be implemented in other process simulation programs and with other vapour/liquid equilibrium models. Parameters in the models have been adjusted to simulate actual regeneration plant performance. The models have been used to evaluate alternative regeneration processes.

REFERENCES 1. Reid, L.S., "Coldfinger, an exhauster for removing trace quantities of water from glycol solutions used for gas dehydration," Ball-Reid Engineers, Inc., Oklahoma City 1975. 2. Smith, R., Skiff, T.B., "Drizo Gas Dehydration," Laurance Reid Gas Conditioning Conference, Norman Oklahoma, March 1990. 3. "Read Gas Dehydration Units," Read Process Engineering AS, Nesbru, Norway. 4. "GPSA Engineering Data Book," Section 20, Dehydration, 10th ed. Gas Processors Suppliers Association, Tulsa, Oklahoma 1987. 5. Kohl, A., Nielsen, R., "Gas Purification," 5th ed., Gulf Publications 1997. 6. Selstø, E.T., "Simulering av gasstørkeanlegg (Simulation of gas dehydration units)," M.Sc. thesis, Telemark University College, Porsgrunn 2001. 7. Hernandez-Valencia, V.N, Hlavinka, M.W., Bullin, J.A., "Design Glycol Units for Maximum Efficiency," Proc. 71st Ann. Convention of Gas Processors Association, p310, Tulsa 1992. 8. Øi, L.E., Bråthen, T., "Calculation of TEG Dehydration in HYSYS," Hyprotech 2000, Amsterdam November 6th-10th, 2000. 9. Billington, P., "Verification of the process control system for the Asgard B mono-ethylene-glycol (MEG) regeneration module using dynamic simulation software," Hyprotech 2000, Amsterdam November 6th-10th, 2000. 10. "Documentation for HYSYS 1.2," Hyprotech Ltd., Calgary 1987. 11. Øi, L.E., "Calculation of Dehydration Absorbers Based on Improved Phase Equilibrium Data," Proc. 78th Ann. Convention of Gas Processors Association, p32, Nashville 1999.