Eliminating guess work

Ralph H.Weiland and John C. Dingman, Optimized Gas Treating, Inc, USA, discuss advances in process simulation, giving special consideration to the ProTreat mass transfer, rate based amine simulation model.

rocess simulation has become a commonplace tool in plant design and plant operations analysis. In most commercial hydrocarbon fractionation operations, it provides a fairly close approximation to what is observed in the field. One of the important factors contributing to its success is the engineers ability to estimate reasonably well the tray efficiencies and the HETPs of the column internals being used. This is because hydrocarbon systems are close to thermodynamically ideal, and efficiencies are well documented for such systems. Chemicals fractionation is less ideal, sometimes to the extent that efficiencies are hard to estimate without field experiences. The complexity and non ideality of amine based gas treating applications is not evident from the apparent simplicity of typical process configurations. However, sensible estimates for the CO2 and H2S mass transfer rates, tray efficiencies and HETPs very often cannot be made at all. In fact, these parameters

change with such variables as acid gas loading and temperature, so much so that efficiency and HETP usually show enormous variation across the height of a column. On the timescale of chemical engineering innovation, gas treating aimed at both the selective removal of H2S, with either maximum (or targeted) rejection of CO2, and the use of promoted solvents for bulk CO2 removal, are recent developments. While total acid gas removal processes are simpler to analyse, build, and operate, more often than not their economics are poorer. This frequently makes selective H2S and CO2 removal the preferred approach. The goal in selective H2S removal is to meet (usually) a ppmv specification on H2S and either produce a gas containing a maximum specified residual percentage of CO2 (normally 2 - 3%, which avoids the additional capital and operating costs for the removal of the extra CO2), or leave as much CO2 in the gas as possible, exemplified by tail gas treating. In either situation, column

Figure 1. Complex flowsheet.

HYDROCARBON ENGINEERING FEBRUARY 2001

Figure 2. Tail gas unit H2S profile.

Figure 3. High pressure absorber temperature profile.

over design is unacceptable, because firstly, a column with too many trays or too much packing will remove too much CO2, and secondly, because the additional CO2 absorption can result in failure to meet the H2S specification.

In a mass transfer rate model, there are intricate interrelated effects between the five key elements. They respond by a variety of chemical and physical, phase and component transport properties such as: Chemical kinetics. Diffusion coefficients, solvent viscosity, and other transport properties. Salting out effects on solubilities. The mass transfer characteristics of the actual hardware being used. When a ProTreat column simulation is run, the tower is being modelled in full detail as a piece of real equipment, not as an idealisation. From a separations standpoint, twopass and single-pass trays perform differently. Metal packing gives different results from plastics and ceramics. In addition, all solvent properties, including the changes in these properties caused by acid gas loading, affect mass transfer coefficients and thereby influence the actual separation that a given column under a given set of operating conditions will achieve. A mass transfer rate model is rich in detail and is a faithful mirror of the real world. In the language of process control, a true mass transfer rate simulation uses a distributed parameter model that consequently is mechanistic, detailed, and fully predictive. This is in contrast to the equilibrium stage approach which is a lumped parameter model (it assigns all the physical and chemical complexities to one or two parameters such as efficiency, or liquid residence time), and is accordingly non predictive.

Equilibrium stages vs. mass transfer rates

The rigorous, mass transfer rate approach used for all column calculations by ProTreat modelling entirely eliminates the need for empirical adjustments to simulate new applications correctly. The ProTreat mass transfer rate model is built on five key elements: Mass and energy balances around individual phases on a tray or in a packed segment. Conventional thermodynamic phase equilibrium. Equilibrium across interfaces. Chemical kinetics effects on mass transfer rates, particularly in the liquid phase. Mass and heat transfer rate models for transport across interfaces. In contrast, even the most sophisticated equilibrium stage model uses only two of these elements, namely mass and energy balances around an entire ideal stage, plus thermodynamic phase equilibrium. Often, an attempt is made to include reaction kinetics by empirical modelling via an adjustable parameter (H2S and CO2 tray efficiencies and/or liquid residence times), that forces the simulation to reproduce a conventionally operated columns treated gas composition. Some might call this a rate model but it is not a mass transfer rate model, it is an attempt to force a reaction parameter into a model that is inherently unrelated to rate processes in order to match some measure of plant performance. The fact is, regardless of any embellishments one might add, equilibrium stage models are non predictive. This makes them unable to deal with applications dependent on differences between mass transfer rates of CO2 and H2S, for which operating experience is lacking. Furthermore, all equilibrium stage models deal exclusively with ideal stages, with no accurate way to translate into the real world of actual tray counts and packed bed depths, let alone to deal with such questions as the effect of packing type, size and material, or tray type and configuration.

Understanding what affects selectivity

In processes where H2S and CO2 are present together and selectivity towards H2S removal is an issue, the extent to which each gas is removed in the column is related directly to the rate at which that species is absorbed by the solvent. So, what affects mass transfer rates? If we rely on an equilibrium stage model for the answer, we will be disappointed because rate is not part of the equilibrium vocabulary, no matter how much we embellish the model. In an equilibrium stage model, phase equilibrium and the stageto-stage L/V ratio are logically the only factors affecting the separation. However, in reality, these parameters are no HYDROCARBON ENGINEERING FEBRUARY 2001

Figure 4. High pressure absorber gas-phase H2S profile.

more important than reaction kinetics, diffusion, reaction equilibrium, and especially the mass transfer characteristics of the tower internals. Why are these other factors so important in reactive absorption? The key to understanding the selectivity of a particular process is appreciating the fact that all alkaline solvents are thermodynamically selective towards CO2 but kinetically selective towards H2S. If process performance depended solely on phase equilibrium, CO2 would always be absorbed preferentially over H2S, and selectivity wouldn't be possible at all. However, CO2 and H2S react quite differently in solution, and their absorption rates are controlled by resistances in entirely different phases. CO2 forms reaction products with the amine in the solvent by binding chemically to the amine at finite rates of reaction. Consequently, after a long enough time, very little CO2 remains free (unreacted) in solution and the CO2 back pressure is very low. (It should be noted that the CO2 back pressure is determined by the concentration of free, molecular, unreacted CO2 in solution, not by the total CO2 content of the solution. Furthermore, the reacting species are free physically dissolved CO2 and unreacted molecular amine and free CO2 is always at a very low concentration because it is a sparingly soluble gas.) The slowness of the CO2 reaction can be used to advantage in achieving selectivity because when short contact times are used, only part of the CO2 has a chance to react, the balance passes through the contactor in the treated gas. CO2 absorption is itself controlled by the reaction-diffusion process in the liquid phase; therefore, changing the reaction kinetics profoundly affects the CO2 absorption rate. When H2S absorbs into an amine, it converts immediately to sulfide and bisulfide ions via simple instantaneous protonation reactions without directly involving the amine at all. The important factor for H2S absorption is solvent alkalinity, not reaction kinetics. Alkalinity is affected by the amount of CO2 already absorbed and the amount of molecular (unreacted) amine still available in the solution. Indeed, if CO2 continues to be absorbed, it will cause the alkalinity to decrease with the result that already absorbed H2S will be expelled from the solution. In addition, H2S absorption is controlled by the diffusional resistance in the

Figure 5. Regenerator (column 5) temperature profile.

gas phase. This gives us another variable to manipulate in achieving a desired degree of selectivity because different types of column internal components have different vapour, and liquid, side mass transfer characteristics. For example, flows over packing typically show much lower liquid-side mass transfer coefficients than trays, which makes them better at rejecting CO2 because CO2 absorption is liquidphase controlled. The kinetics of CO2-amine reactions and the fact that CO2 absorption is liquid-phase controlled are crucial to achieving and controlling CO2 slip. The careful selection of the amine or amine blend formulation (for kinetic selectivity), as well as the column internals (for equipment selectivity) are important in any optimisation for a desired level of treating. The secret lies in choosing an amine with the right alkalinity and the right reactivity towards CO2, allowing gas-liquid contact for the right length of time, and using equipment with the right internals. Equilibrium stage models simply cannot capture these effects.

Anatomy of a mass transfer rate model

In principle, operating parameters like temperature and tray hydraulics, chemical kinetics, reaction equilibria, and mass transfer equipment characteristics can be accounted for via their effect on stage efficiencies in the case of trays, or via HETP and HTU values for packing. In actual practice, however, the relationships are so extraordinarily complex in amine-based gas treating that they are beyond our ability to generalise because of the very limited amount of basic data available. In other words, there is no realistic possibility of translating amine-treating experience into valid correlations for efficiency or HETPs. The equilibrium stage approach will therefore remain mired in numbers of ideal stages. For example, tray efficiency depends markedly on the species being transferred (CO2 vs. H2S), on the operating conditions (temperature, pressure, acid gas loading of the solvent and phase flow rates) and on the physical characteristics of the tray (geometric factors such as tray type, weir heights, approach to jet flood, not to mention trays versus packing, and so on). The more direct, albeit more difficult, basic approach used by mass transfer rate modelling is not subject to the limitations associated with equilibrium stages. HYDROCARBON ENGINEERING FEBRUARY 2001

the large number of chemical species versus only one temperature variable. Species transfer Heat transfer Thus, mass transfer involves an order b - y *) b - Tw ) Phase V/inside tube Rv = kga(y i i Q = hiAi(T i i of magnitude more equations than heat * b w b Phase L/outside tube RL = Eikla(x i - x i) Q = hoAo(T o - T o) transfer, and to make matters worse, b * * b b w w b Equality of fluxes R = kga(y i - y i) = Eikla(x i - x i) Q = hiAi(T i - T i) = hoAo(T o - T o) the majority of the equations (e.g., Interfacial equilibrium y*i = Hix*i Twl = Two reaction equilibrium) are non linear. But Notes: the mass transfer rate equations can kl Liquid side mass transfer a Vapour-liquid interfacial area i Inside of tube (heat transfer); be solved for complete columns and (mass transfer) film coefficient component identity (mass entire flowsheets containing multiple A Surface area of tubes (heat Q Heat transfer rate in heat transfer) columns quite routinely and in computtransfer) exchanger L Liquid side of the interface E Enhancement factor, accountR Mass transfer rate on a tray o Outside of tube (heat ing times that have shrunk to the point ing for effect of reaction or segment of packing transfer) where time is no longer an issue. kinetics on mass transfer rates T Temperature V Vapour side of the interface h Tube side or shell side heat x Mole fraction in liquid w At the tube wall (heat While modelling of absorber opertransfer film coefficient y Mole fraction in vapour transfer) ations has traditionally received the kg Vapour side mass transfer b In the bulk phase (far from * In physical equilibrium at the most attention, it is equally important film coefficient the interface) interface (mass transfer) to be able to model solvent regeneration accurately. There are two major The most direct way to develop a feel for the basics of concerns involved in the regeneration of the rich amine solmass transfer rate modelling of columns is by drawing a vents. The first is the need for an accurate prediction of the comparison with heat exchanger design and analysis. ShellH2S and CO2 loadings of the lean solvent produced in the and-tube heat exchangers, for example, are designed using regeneration step. This is critical to an absorber's ability to computer software (or sometimes still by hand) that contains: meet treated gas specifications. The second concern is the ability to predict accurately the amount of reboiler heat duty A database of correlations for heat transfer film coeffi(energy requirement) needed to obtain the desired level of cients for the tube and shell sides of the exchanger as solution regeneration. Simulating regenerators is inherently functions of hydraulics (as measured for example by more difficult for all models due to pinch conditions at the Reynolds numbers), tube bundle layout, tube and shell bottom. Equilibrium stage models dont work very well here passes, and physical arrangements of shell-side baffles. either because the chemical reactions affect mass transfer in A database of relevant fluid properties such as heat just as important ways as in absorption. None of the trays in capacity, viscosity, thermal conductivity and density. a stripper comes even close to an equilibrium stage. The The heat transfer properties of the tubes. desorption rate of each acid gas and its concentration in the In the simplest case of analysing an existing exchanger vapour and liquid phases influences the stripping of the without phase change, the general approach is to determine the other. From a technical standpoint, the ProTreat stripper overall heat transfer rate, or perhaps rates at different positions model treats amine regenerators every bit as rigorously as along the flow paths if properties vary with temperature in some absorbers. It produces the best possible predictions of stripunusual way. What is important to recognise is that the calculaper performance, without the need for empirical adjustment tion hinges on determining the heat transfer rate. Mass transfer parameters. When the absorber and stripper models are tied is strongly analogous, but for a number of reasons it is enortogether in a recycle flowsheet, one obtains the best possimously more complex. This is particularly important when ble prediction of total treating plant performance, again withchemical reactions play a key role, as they do in amine treating. out the use of empirical corrections of any kind whatever. It scarcely needs to be pointed out that never is there talk of This complete freedom from gross empiricism permits the equilibrium heat exchangers, either with or without efficiency. All reliable design of new facilities for which absolutely no operheat exchanger calculations are strictly rate based, and have ating data or prior experience exists. been for nearly a century. With the advent of heavy duty comThe ProTreat package does the mass transfer equivalent puting, mass transfer equipment has become as amenable to of heat exchanger design calculations in the setting of acid completely rational design as is heat exchange equipment. gas treating with amines. In this model, the mass transfer The equations that describe the transfer of a component coefficients for the two phases are calculated from film coefi between phases are compared with their heat transfer ficient correlations for specific trays and packings. The correcounterparts in Table 1. The similarities in the equations are lations depend on tray type (bubble, sieve, valve), tray geomquite striking, but of course, there are also differences. One etry (weir height and length, tray active area), on packing of the important differences is that the bulk phase compositype (Intalox saddles, Pall rings, Cascade Mini-Rings, and so tion in the liquid is determined by a complex set of reaction on), packing size and material (metals, plastic, ceramic), and equilibrium relations. Another is the appearance of the factor on the flow rates and physical properties of the phases. Even Ei, called the enhancement factor. This is a dimensionless the effect of acid gas loadings on physical and transport number that accounts for the effect of reaction on diffusional properties is included. The ProTreat mass transfer rate mass transfer, found by solving the differential equations that model is detailed and mechanistic and does its calculations describe the particulars of that sub process. There are literin the context of a fully flexible flow sheeting environment. ally dozens of different sets of kinetics for which solutions are Case studies available, with each in the context of several theories of Three sets of examples are presented. These are: mass transfer such as the two film theory, Danckwerts surface renewal theory, and so on. A third difference is that A somewhat complex, integrated plant containing four although temperatures in the two phases at the tube surface contactors, two regenerators and a flash unit. are the same (in the absence of significant tube wall resisA packed column treating a high pressure gas with a tance and scale), the concentrations of an acid gas on the hypothetical three amine mixture. two sides of an interface are related by physical solubility. Yet A commercial absorber removing CO2 from high presanother difference, and this is the one responsible for the sure natural gas. almost one century lag between heat and mass transfer, is
Table 1. Comparison between equations for mass and heat transfer rates

HYDROCARBON ENGINEERING FEBRUARY 2001

Case study 1: Complex flowsheet

For economy of presentation, a number of cases of absorption and regeneration have been combined into the single flowsheet shown in Figure 1 and several of these columns are examined here in detail. The purposes of this example are to show the importance of flexibility in a flowsheeting package and to indicate how a number of different operations can be very effectively integrated. With the exceptions of Control and Recycle, the unit operations blocks in the Figure have names that make block functionality self explanatory. Control is a block (usually one per flowsheet) that allows the solvent composition and flow rate to be fixed, monitored, and controlled somewhere in the plant. Solvent makeup requirements are also reported here. The ProTreat simulator solves recycled flowsheets iteratively and sequentially. The blocks labelled Recycle permit input of an estimate for stream composition to initiate the solution process and act as stations where progress to a solution is monitored. The sequence in which blocks are to be solved is internally generated. In this example, the entire processing scheme uses a nominal 45 wt% generic MDEA solution for treating gas streams ranging from tail gas (Column 6) to 900 psig natural gas (Column 2) at H2S to CO2 ratios from 2.5 (35 mol% total acid gas) to 0.25 (10 mol% total acid gas). Three of the absorbers (Columns 1, 2 and 3) operate in split flow mode and the rich amine from two of them (1 and 6) is used as mid column semi lean feed for two others (2 and 3, respectively). One of the regenerators is used in split flow mode with the semi lean stream (about 40% of the rich feed to the regenerator) being used as intermediate feed to a low pressure absorber (Column 1); the fully stripped solvent from this regenerator is fed to the top of the tail gas unit and the low and intermediate pressure contactors. The tail gas unit is packed with 30 ft of FLEXIMAX 400 packing and is modelled as 10 segments, each containing 3 ft of packing. The column is automatically sized for 80% flood. It operates at a bottom pressure of 3 psig and treats 2 MMSCFD of Claus-plant tail gas at 100 F containing 8% CO2 and 2%

Figure 6. Effect of solvent flow on packed bed absorber temperature profiles.

Figure 7. Effect of solvent flow rate on H2S content of treated gas.

Figure 8. Effect of solvent flow rate on CO2 content of treated gas.

Figure 9. ProTreat modelled temperature profile vs. plant performance measurements.

H2S in nitrogen. Solvent is the fully stripped amine from the reboiler of Column 5 at 80 USgpm, 120 F, containing about 49 wt% MDEA loaded to 0.00024 and 0.0019 mol/mol H2S and CO2, respectively. The column picks up very little acid gas (total loading increase of only 0.0275) but, as shown in the plot of Figure 2, it manages to produce a 35 ppmv H2S treated gas. It also slips fully 97% of the CO2. The rich solvent leaving this contactor is actually still quite lean and is perfectly suitable as a solvent for injection into the middle of Column 3. The vapour phase H2S profile is typical of H2S removal using amines, the H2S concentration falls off almost logarithmically with position in the column. Note that the concentration shown for any segment is the concentration of the gas leaving that segment. Thus, the gas leaving segment 7 is about 4000 ppmv which is the concentration at a position in the column 18 ft from the top of the packing (the bottom of segment 7 is 21 ft below the top of the bed). Column 3 presents an interesting case of the effect of using a semi lean amine to a mid column tray to do bulk removal in the bottom of a column and a fully regenerated lean amine stream to the top of the absorber to do the final cleanup. This column uses 20 generic valve trays to process 5 MMSCFD of 900 psig natural gas containing 10% CO2 and 3% H2S. Treated gas is 0.26 ppmv H2S and 0.125% CO2. The tray-to-tray temperature and gas-phase H2S profiles are shown in Figures 3 and 4, respectively. The semi lean stream to tray 12 is somewhat hotter than the liquid from tray 11 and this causes the temperature on tray 12 to rise several degrees more than it otherwise would. The temperature and H2S profiles track each other because, of course, the temperature rise of the solvent is directly related to the amount of acid gas absorbed. Note that these profiles show conditions on the actual trays in the column. The mass transfer rate model even distinguishes between the temperatures of the vapour and liquid on the same tray. Although in this particular instance, the differences are not great, in some cases they are.

HYDROCARBON ENGINEERING FEBRUARY 2001

Table 2. High pressure sour gas stream and lean solvent Sour gas Temperature (F) Pressure (psig) Flow (MMSCFD) H2S (mole %) CO2 (mole %) CH4 (mole %) C2H6 (mole %) C3H8 (mole %) C4H10 (mole %) Water (% saturation) 100 1000 100 0.25 10.00 80.00 5.00 3.00 1.75 100 Lean solvent Temperature (F) Pressure (psig) Flow (USgpm) H2S (mol/mol loading) CO2 (mol/mol loading) MDEA (wt%) DEA (wt%) Piperazine (wt%) Hydrocarbons 120 1050 Various 0.0005 0.05 17 15 8 Nil

Table 3. Stream conditions for commercial absorber Sour gas Temperature (F) Pressure (psig) Flow (MMSCFD or USgpm) Carbon dioxide (mol% or loading) Methane (mol%) Ethane (mol%) Propane (mol%) n-Butane (mol%) i-Butane (mol%) n-Pentane (mol%) Nitrogen (mol%) DEA (wt%) MDEA (wt%) 80 910 48.76 1.99 85.44 6.66 2.36 0.77 0.89 1.70 0.19 0 0 Lean solvent 118 910 175 0.037 0 0 0 0 0 0 0 15 28

ranged from 100 ppmv (at 1600 USgpm) to 1.0% (at 1100 USgpm). As the solvent flow is reduced, the temperature bulge shifts up the column and becomes increasingly wider, so that at 1000 USgpm, most of the column is experiencing high temperatures. The large broad temperature profile is a result of both the low L/V ratio and the high acid gas loadings through most of the column. The higher solvent flows carry an increasingly higher proportion of the heat of absorption down the column and this moves the position of the temperature bulge downward; however, the CO2 absorbed at low solvent flow rates causes higher loadings, and therefore it produces a hotter solvent so the temperature at the bulge itself is higher. As shown in Figures 7 and 8, solvent flow rate has a profound impact on both the H2S and CO2 contents of the treated gas. Variability in solvent flow rate has little or no effect on the treated gas when the solvent flow is high. However, as it is reduced, a value is reached (approximately 1300 USgpm), below which the H2S in the treated gas rises exponentially with further reductions in solvent rate. At about this same solvent flow, the residual CO2 in the gas also begins to rise exponentially. Of course, normal column operation is usually at the lowest possible solvent flow rate, in this example at about 1100 USgpm, where the H2S can be kept below 4 ppmv. This is exactly where column performance is most sensitive to the amine flow. Again, it should be emphasised that these temperature profiles and performance predictions are based on the way 40 ft of FLEXIMAX 400 packing would be expected to behave in this particular service. Theoretical stages and HETPs have not been used.

Case study 3: Commercial absorber

Because gas plant performance is markedly affected by the quality of the lean amine being fed to the top of the absorbers, it is just as important to be able to simulate regenerators accurately, as it is absorbers. Column 5 is a 24 tray regenerator with rich amine to tray 3 and a 40% semi lean amine draw from tray 14. Its temperature profile is shown in Figure 5, where it can be seen that the top 2 trays are used to heat reflux, but it takes the top 10 trays for the solvent to reach anything approaching a constant temperature. Using the semi-lean amine drawn from Column 5 for bulk acid gas removal from the very sour low pressure gas being fed to Column 1, then feeding the top of Column 1 with a portion of the very well stripped lean amine from the same regenerator, permits extremely good H2S removal. Split flow solvent regeneration produces an unusually high effective stripping ratio in the bottom half of the regenerator. The final example compares a ProTreat mass transfer rate based simulation with actual field performance data for an operating column. The owners asked that the facility not be identified. Having already seen how sensitive absorber temperature profiles can be to operating conditions, a good test of any treating model is its ability to predict both observed temperature profiles and treated gas quality. The gas and lean solvent are as shown in Table 3. The solvent is a blend of 15 wt% DEA and 28 wt% MDEA. The 5 ft diameter contactor contains 22 single pass sieve trays having 8% open area and treats the gas to about 1600 ppmv CO2. The simulated temperature profile is shown in Figure 9 where it is compared with the temperatures measured in several downcomers. The agreement between the mass transfer rate model and the measured data is excellent. This is especially the case in light of the sensitivity of temperature profiles to operating conditions, as discussed in the preceding example.

Case study 2: Mixed solvent contactor

The second example is a hypothetical case using a three component solvent (17 wt% MDEA, 15 wt% DEA plus 8 wt% piperazine), to remove H2S and CO2 from a high pressure natural gas described in Table 2. The column was simulated for solvent flows ranging from 1000 - 1600 USgpm in 100 USgpm increments. It contained 40 ft of FLEXIMAX 400 steel packing and the column was sized for 80% flood. The results of this set of simulations reveal several interesting aspects of treating with a mixed amine solvent. First, the temperature profiles shown in Figure 6 at four solvent rates show that the temperature profile is extremely sensitive to solvent flow. At 1000 USgpm, the column produced a treated gas containing 8 ppmv H2S and a residual CO2 content of just under 2%; at solvent rates above 1100 USgpm, the H2S content was less than 4 ppmv (usually less than 1 ppmv); the residual CO2 in the treated gas

Conclusion
Mass transfer rate based modelling of amine treating operations produces the most accurate facsimile possible of the actual physical behaviour of the columns and the effect of the details of their internals on performance. In the fully flexible flowsheeting environment of the ProTreat package, detailed performance predictions can be easily obtained in the familiar Windows environment.
ProTreat was developed with the partial financial support of the Gas Research Institute under Contract Number 5092-222-2409, Dennis Leppin, GRI Project Manager.

HYDROCARBON ENGINEERING FEBRUARY 2001