Bryan Research and Engineering, Inc.

- Technical Papers

Sweetening LPG's with Amines

JOSEPH W. HOLMES, Bryan Research and Engineering, Inc., Bryan, Texas MICHAEL L. SPEARS, Champlin Petroleum Co., Bryan, Texas JERRY A. BULLIN, TexasA&M University, College Station, Texas

ABSTRACT As more LPG's (liquefied petroleum gas) are being produced, the demand for liquid hydrocarbon sweetening facilities has increased. The most common contaminants in LPG's are CO2, H2S, mercaptans, COS, CS2, and elemental sulfur. Each of these contaminants can cause problems in the finished products.

CEP May 1984

Bryan Research & Engineering, Inc.

Visit our Engineering Resources page for more articles.

Introduction As more LPG's (liquefied petroleum gas) are being produced, the demand for liquid hydrocarbon sweetening facilities has increased. The most common contaminants in LPG's are CO2, H2S, mercaptans, COS, CS2 and elemental sulfur. Each of these contaminants can cause problems in the finished products. H2S is corrosive and its presence can lead to deposition of free sulfur. Although mercaptans are used in small quantities for odorizing products, in larger quantities they impart a bad odor to the product as well as the combustion flue gases. According to Perry (1), COS does not react to copper strip tests. However, the COS can hydrolyze to H2S which can cause a product with no H2S to become sour after setting in a tank. Thus, if both COS and water are present in a product, removal of COS must be considered. The presence of C02 in LPG's can also cause problems. For example, if the product is treated with caustic, the presence of C02 can cause large caustic losses. Historically, several methods have been used to remove the contaminants. If the quantities of the contaminants are small, a simple caustic wash can be used. For larger quantities molecular sieves and amine treating systems can be used. While molecular sieves have the advantage of removing water and COS along with H2S and C02, their disadvantages include large capital and operating cost as well as catalyzing the formation of COS when both H2 S and C02 are present. The use of amines has become quite popular, especially in plants where amines are also being used to sweeten the gas stream before processing for the LPG's, Figure 1. In plants with multiple absorbers, a common stripper can be used to regenerate the amine. Although several LPG sweetening units using amines are currently in use,

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 1 of 8

Bryan Research and Engineering, Inc. - Technical Papers

many aspects of the design of these plants have been more of an art than a science. For the most part, the designs have involved approximations based on existing plants and guesswork.

This article discusses the design methods for LPG sweetening units using amines and recent design advances in process simulation. The results from the process simulation program for LPG sweetening are compared to data from operating plants. Design methodology Based on the operating experience for LPG sweetening plants using MEA (monoethanolamine), DEA (diethanolamine), and DGA, (1-5) the recommendations in Table 1 were made for design guidelines for the various amines. All of the amines are degraded to some extent by COS and C02. However, DEA is degraded less than MEA. In addition, less reboiler heat is required for DEA due to higher solution concentrations and loading as well as lower heats of absorption. Excessive corrosion could be encountered if the recommended loadings are exceeded. Table 1. Operating conditions for LPG sweetening with amines Amine MEA DEA DGA Wt. % Amine 5-20 25-35 50-70 Max Loading mol/ mol* 0.3 - 0.4 0.35 0.65 --

* Mol acid gas per mol amine. Current Methods. As pointed out by Honerkamp (4), two approaches have been generally used to determine necessary amine circulation rates and types of contact surface required. One is to ratio up or down from an existing plant not too different from the proposed new one. The other is to extrapolate available gas-amine data into the expected operating range. The scale-up or -down procedure is very straightforward and needs no further explanation. Obviously, any problems with the existing plant such as massive oversizing will be transmitted to the new unit. In the extrapolation method, a hypothetical vapor phase is assumed to be in equilibrium with both the LPG and

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 2 of 8

Bryan Research and Engineering, Inc. - Technical Papers

the amine solution, Figure 2. Henry's law and Raoult's law are used to calculate the partial pressure of the H2S and CO2 above the LPG. Using these laws for C02 as an example, the concentration of CO2in the LPG is the partial pressure of the C02 in equilibrium with the LPG divided by the vapor pressure of pure CO2 at the same temperature. The partial pressure of the C02 in the hypothetical vapor phase is assumed to be the equilibrium partial pressure of the C02 above the amine solution. A correction or safety factor of up to a factor of 12 is then applied to the equilibrium partial pressure. The correction factor is used to compensate for the nonideal solutions, unknown activity coefficients, and extrapolated partial pressure data for amines.

Based on this method, the calculated equilibrium partial pressure of C02 above the amine is so far below the C02 partial pressure above the LPG that equilibrium considerations are not important. The only exceptions would be for very small C02 concentrations in the inlet or for very high amine loadings. In addition to the equilibrium considerations, the lean amine should be regenerated to less than 0.1 mol/mol. Heretofore, only correlations have been available to estimate the performance of the still and the lean amine concentrations. Computer Simulation. The design capability of LPG sweetening plants using amines has been greatly advanced with the expanded capabilities of a process simulation program called Tsweet. This program was originally written for gas sweetening plants using amines and has now been expanded to include LPG streams. Tsweet is an advanced process simulation program using the flexible flowsheet concept. Beginning with the process flowsheet and operating parameters, complete material and energy balances for the steady-state operating conditions can be performed. Rigorous tray-by-tray calculations using the Ishii-Otto (6) method are used to predict the design of the stripper and absorber columns. Some 24 chemical species are allowed, including MEA, DEA, DGA, MDEA, water, H2SC02 and an assortment of gases commonly present in natural and synthetic gases. Vapor pressures of H2S and C02 over the amine solutions are calculated by a modified Kent and Eisenburg (7) equilibrium-model. A kinetic model was developed to predict the effect of residence time, temperature, solution concentration, pressure and type of amine on the rate of C02 absorption. The program has been demonstrated to be very reliable in calculating the performance of amine gas sweetening plants. The program will accommodate multiple absorbers in LPG and/or gas service which feed a common stripper. The gas sweetening capabilities of Tsweet have been described by Polasek et al. (8) and Bullin and Polasek (9). To facilitate LPG sweetening in the program, the hypothetical vapor phase discussed was used. The hypothetical vapor is assumed to be in equilibrium with both the LPG and amine solution, Figure 3. The Soave-Redlich-Kwong (SRK) equation of state was used to predict the partial pressures above the LPG phase. The nonideal solution behavior is encompassed in the SRK equation. In addition, the theoretically-based Kent and Eisenburg model for the amine solutions represents a very large improvement over a simple extrapolation of the H2S and C02 partial pressure. Thus, no correction factor should be needed for the program results.

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 3 of 8

Bryan Research and Engineering, Inc. - Technical Papers

Figure 3.Amine LPG equilibria for computer simulation. Tsweet results vs. plant data The new simulation capability for LPG sweetening plants using amines is compared to data from several plants involving a wide variety of contactors. The types of contactors include packed columns, sieve tray columns, jet eductor mixers, and static mixers. According to Russell (5), about 6 to 8 ft (1.83 to 2.44 m) of packed bed with a distributor plate is required for one theoretical stage in liquid-liquid contactors. Honerkamp(4), Perry (1), and Russell (5) have noted that the separation efficiency for a packed column is significantly reduced at high turndown. Depending on the degree of turbulence in the contactor settling retention times from 10 to 30 minutes is required for good phase separation. Tsweet is compared to operating data from eight plants, Table 2. The operating data for plants 1 through 7 were reported by Honerkamp(4). The comparisons are discussed on an individual plant basis below. Plant 1. The absorber for this plant was a 30 in. (0.76 m) diameter column with three 12 ft (3.66 m) sections of 1 in. (2.54 cm) Intalox packing. Thus, three ideal stages were used for the simulation results, Table 2. The outlet C02 of 0.056 ppm calculated by Tsweet was close to the zero value reported in the data. The program was also run for one and two ideal stages. The calculated outlet C02 was 18.7 and 0.058 ppm, respectively. Thus, the absorber was severely oversized. The stripper calculations also matched the data well as shown by the lean solution loading. Table 2. Tsweet results vs, data for LPG sweetening with amines. 1 Case Hydrocarbon Rate, gpm T, F P, psig CO2 in, ppm H2S in, ppm CO2 out, ppm H2S out, ppm Amine Wt. % Flow, gal/min Ideal Stages ** Lean Loading * Rich Loading * Contactor Data TSweet Data 2 TSweet Data 3 TSweet

100 103 50 60 55 500 500 10,000 10,000 --0 0.056 --MEA MEA 20 20 10.8 10 -3 0.10 0.108 0.44 0.44 Three 12 ft. sections of packing

780 780 90 90 535 535 2,030 2,023 Trace 2.7 < 100 0.42 -0.01 MEA MEA 15 15 34 34 -2 0.10 0.13 0.28 0.32 Two 12 ft. sections of packing

208 210 77 -89 83 835 835 33,000 33,000 --Trace 0.4 --MEA MEA 20 20 54 50 -2 0.10 0.11 0.53 0.53 18 ft. of packing

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 4 of 8

Bryan Research and Engineering, Inc. - Technical Papers

Table 2. Tsweet results vs, data for LPG sweetening with amines (continued). 4 Case Hydrocarbon Rate, gpm T, F P, psig CO2 in, ppm H2S in, ppm CO2 out, ppm H2S out, ppm Amine Wt. % Flow, gal/min Ideal Stages ** Lean Loading * Rich Loading * Contactor Data TSweet Data 5 TSweet Data 6 TSweet

265 -305 2,460 270 Trace nil DEA 20 20

273 80 305 2,450 270 1.2 5.9 DEA 20 20

42 80 - 90 -5,000 -100 -MEA 8 5.3

44 85 200 5,020 -47 -MEA 8 5.3

21 -265 nil 1,080 -nil MEA 5

<p align=center>5

22 85 265 -1,080 -9.2 MEA 5 5

-2 0.07 0.05 0.30 0.27 Two 8 ft. sections of packing

-1 0.10 0.10 0.44 0.43 10 perforated trays

-1 0.07 0.07 0.13 0.13 Jet eductor-mixer

Table 2. Tsweet results vs, data for LPG sweetening with amines (continued). 7 Normal TSweet 7 Maximum Data TSweet 8 Data TSweet

Case Hydrocarbon Rate, gpm T, F P, psig CO2 in, ppm H2S in, ppm CO2 out, ppm H2S out, ppm Amine Wt. % Flow, gal/min Ideal Stages ** Lean Loading * Rich Loading * Contactor

Data

250 250 55 55 435 435 2,260 2,200 --nil 0.90 --MEA MEA 15 15 100 110 -1 0.20 0.20 0.22 0.22 Jet eductor-mixer

200 205 55 55 435 435 6,675 6,630 --138 74 --MEA MEA 15 15 100 93 -1 0.38 0.38 0.46 0.45 Jet eductor-mixer

105 105 100 100 385 385 7,260 7,260 --660 158 --DEA DEA 30 27 14 14 -1 0.30 0.30 0.25 0.32 Static mixer

* Loadings are moi of acid gas per mol of amine. A lean solution loading of 0.1 mol/moi was assumed if no information was available. ** Equilibrium stages used in the absorber.

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 5 of 8

Bryan Research and Engineering, Inc. - Technical Papers

SI Conversion: kPa psi X 6.89; L - gal X 3.79; m - ft X 0.3048; C - (F - 32)/1.8. Plant 2. A vertical column absorber with two 12 ft (3.66 m) sections of packing was used. Two ideal stages were used for the Tsweet results, Table 2. Since a trace of H2S in the inlet was reported, an H2S feed rate of 0.01 Ib mol/h (4.54 gmol/h) which corresponds to 2.7 ppm was assumed for Tsweet. For this inlet rate, 0.01 ppm H2S was calculated for the sweet LPG. The calculated C02 was 0.42 ppm compared to the reported value of less than 100 ppm. Plant 2 was also simulated using one ideal stage. The outlet concentrations were 4.2 ppm C02 and 0.24 ppm H2S. Thus, one 12-ft (3.66-m) section of packing would have been adequate to meet the plant requirements. Plant 3. A 48 in. (1.22 m) diameter absorber with 18 ft (5.49 m) of packing was used in Plant 3. The inlet C02 was 3.3 mol % and the outlet was reported as a trace. For one and two ideal stages, the program results were 848 and 0,4ppmCO2out, respectively. The 18 ft (5.49 m) of packing with only one distributor probably represented between one and two ideal stages. Thus, one ideal stage would not be adequate for this plant. Plant 4. In this case, a 66 in. (1.68 m) diameter absorber with two 8 ft (2.44 m) sections of ceramic packing was used. The design hydrocarbon rate was 525 gal/min (1,987 L/min) compared to the actual rate of 265 gal/min (1,003 L/min). DEA was used as the amine. For two ideal stages, the outlet H2S was about 5 ppm compared to the trace reported in the data. Since no inlet LPG temperature was reported in the data, a value of 60F (15.6C) was also used in Tsweet in an effort to match the data. The outlet H2S dropped to only about 4 ppm. In another effort to match the data, the lean amine concentration was lowered to 0.04 mol/mol. However, the outlet H2S fell only to 2.9 ppm. Since a value of "nil" was reported for H2S, a copper strip corrosion test was probably used. According to Perry (7), a 1A copper strip is generally thought to represent H2S concentrations below about 1 to 2 ppm. Baumer(3) suggested that a 1A copper strip can be obtained at H2S levels as high as 3 or 4 ppm. The only other potential explanation for the plant's performing better than the program calculation for two ideal stages was that the two 8 ft (2.44 m) packed sections performed better than two ideal stages. However, three ideal stages along with the other conditions shown in Table 2 resulted in an outlet H2S of 3.2 ppm. Thus, the plant appears to be performing better than equilibrium. Plant 5. A 30 in. (0.76 m) diameter absorber with ten perforated trays was used. The design LPG rate was 85 gal/min (322 L/min) compared to the operating rate of 42 gal/min (159 L/min). The calculated value for the 42 gal/min (159 L/min) case using one ideal stage was 47 ppm compared to the 100 ppm data value. The fact that liquid-liquid contactors lose efficiency rapidly with turndown is well known. When the contactor was operated at the design rate of 85 gal/min (322 L/min), the outlet C02 was 50 ppm. For one and two ideal stages for the 85 gal/min (322 L/min) case, the program values were 1,130 and 540 ppm for the outlet C02 and 0.601 and 0.677 mol/mol for the rich amine loading, respectively. In both cases the rich amine was loaded to 100% of equilibrium. Thus, the amine flow must have been increased when the 85 gal/min (322 L/min) data were taken. Plant 6. The contactor for this plant was a jet eductor-mixer. The eductor was followed by a settling tank with a residence time of about 30 min. Very little C02 was present in the feed while the H2S was about 1,080 ppm. The calculated values for the outlet H2S were 9.2 and 2.9 ppm for one and two ideal stages, respectively. The reported concentration was "nil." Unless two or more jet eductors are used, more than one ideal stage is not possible. In fact, a single jet eductor seldom performs as one full ideal stage. No additional information was available on this

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 6 of 8

Bryan Research and Engineering, Inc. - Technical Papers

plant Thus, the differences between the data and the program could not be explored further. Plant 7. A jet eductor-mixer and settling tank combination was also used for Plant 7. Most of the rich amine stream was recycled directly back to the eductor. A slip stream was regenerated and returned to the eductor. This plant was operated at two levels of C02 in the feed corresponding to the "normal" of 2,260 and the "maximum" of 6,675 ppm. For the normal C02 content, the program calculated an outlet C02 of 0.9 ppm compared to the data value of "nil." For the maximum CO2 content, the program predicted 74 ppm compared to 138 ppm. The amine circulation rates were kept constant at 15 gal/min (56.8 L/min) for both cases. However, the LPG rate was decreased from 250 gal/min (946 L/min) for the normal case to 200 gal/min (757 L/min) for the maximum case. As expected, the jet eductor is probably not acting as a full ideal stage. Plant 8. A static mixer consisting of five mixing elements was used to contact the LPG and DEA solution. The static mixer was 3 in. (7.62 cm) in diameter with an overall length of 15 ft (4.57 m). The mixing elements were grouped 3 at the inlet and 2 near the middle of the mixer. Each element was 3 in. (7.62 cm) long and reversed the direction of the swirl. The contactor was followed by a separator and a final coalescer. The CO2 content of the sweet LPG was 660 ppm. Using one ideal stage, the program predicted 158 ppm CO2, The performance of the static mixer system was excellent to approach an ideal stage this closely. In summary The previous design methods for LPG sweetening plants using amines were based on either a simple scale-up or -down from an existing plant or an extrapolation of gas-amine information. The extrapolation method requires a correction factor of up to a value of 12. Another method based on computer simulation has become available. The capabilities of the amine-gas sweetening program, Tsweet, were expanded to handle LPG streams. The program uses the SRK equation of state for the LPG and modified Kent and Eisenburg model for the amine solutions. The entire plant including the stripper is solved by Tsweet. Results from the program were compared to data from eight plants with a wide variety of contactors including packed columns, perforated plate columns, jet eductor-mixer, and static mixers. The predicted values from Tsweet matched the data quite well except for the one plant with a jet eductor. Tsweet was also used to examine the extent of contact needed in the various plants. Based on the program and the data from eight plants, one ideal stage is sufficient to remove the C02 up to about 0.5% for most applications using MEA. REFERENCES 1. Perry, C. R., Oil and Gas J., p. 76 (May, 1977). 2. Bacon, K. H., "Liquid Treating," Warren Petroleum Co., Tulsa, Ok. 3. Baumer, J. A., "Large Scale Liquid-Liquid DEA Treating for Fractionation Plants," Regional GPA Meeting, Houston (Nov., 1981). 4. Honerkamp, J. D., **Treating Hydrocarbon Liquids with Amine Solutions," Proc. Gas Conditioning Conf. (1975). 5. Russell, R. M., Oil and Can /., p. 1.15 (Dec., 1980). 6. Ishii, Y., and F. D. Otto, Can J. of Chem. /#.. 51, p. 601 (Oct., 1973). 7. Kent, R. L., and B. Eisenburg. Hydro. Proc., p. 87 (Feb., 1976).

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 7 of 8

Bryan Research and Engineering, Inc. - Technical Papers

8. Polasek, J. C., S. T. Donnelly, and J. A. Bullin, Chem. Eng. Prog., p. 63 (Mar., 1983). 9.- Bullin J. A., and J. C. Polasek, "Selective Absorption with Amines," Proc. 6lst Annual Gas Processor's Com;. (1982). copyright 2001 Bryan Research & Engineering, Inc.

Copyright 2006 - All Rights Reserved Bryan Research and Engineering, Inc.

Page 8 of 8