Patrick A. Thompson, Jeffrey A. Cunningham, Craig A. Berry
Radian Corp.
Austin
James M. Evans
Gas Research Institute
Chicago
An accurate method for estimating toxic emissions from glycol gas-dehydration units is available.
Estimates of emissions of benzene, toluene, ethyl benzene, and xylenes (BTEX) and of volatile organic compounds (VOCS) can be used to identify which units may be of regulatory concern under increasingly stringent state and federal environmental regulations.
In this conclusion to a three-part series covering such emissions from glycol reboilers, Radian Corp., Austin, and the Gas Research Institute, Chicago, report on development and capabilities of GRI-DEHY, a personal computer program that provides operators with a tool for estimating emissions of BTEX and other VOCs from triethylene glycol (TEG) dehydration units.
Part 1 (OGJ May 17, p. 28) covered the background to the problem and how some companies have responded. In Part 2 (OGJ, May 31, p. 61), Oryx Energy Co., Dallas, reported on its evaluation of five emissions-estimating methods.
GRI-DEHY is rapid and inexpensive in determining which dehydration units comply with applicable regulatory emission levels, which units may require minor control approaches, and which units may require retrofit controls.
Prerelease testing (April 1992) showed that GRI-DEHY provides good emissions estimates, especially for the BTEX compounds.
REBOILER'S ROLE
Glycol dehydration units are employed to remove water from produced natural-gas streams to prevent hydrate formation and corrosion in the pipelines. TEG is used in approximately 95% of glycol dehydration units and has gained wide acceptance in this application because of its high affinity for water, its chemical stability, and its low cost.
Estimates of glycol dehydration units operating in the U.S. run as high as 30,000; and as much as 100,000 worldwide.
Approximately 17-18 tcf/year of natural gas are dehydrated in North America, with the U.S. treating a large portion of that amount.1
Fig. 1 presents a flow diagram for a typical glycol dehydration unit.
The moist natural gas enters the bottom of the absorber where it is contacted countercurrently with the cool, lean glycol to remove the water. The dry gas exits the top of the absorber.
The rich glycol leaves the bottom of the absorber and, in larger units, often goes through a balance pump to a flash tank. Much of the natural gas that was captured at the high absorber pressures is separated from the glycol at the flash tank.
After a series of heat exchangers and filters, the rich glycol enters the still and reboiler where water is distilled or stripped from the glycol. Some units inject a stripping gas to produce higher purity glycol at normal reboiler temperatures.
Lean glycol from the surge tank is pumped back to the absorber.
During the absorption step, the glycol, which is a relatively good solvent for aromatic compounds, also removes benzene, toluene, ethyl benzene, and xylenes from the natural gas.
Because the atmospheric boiling points of BTEX range from 176 to 291 F., fewer of these compounds are flashed from the rich glycol in the flash tank which operates at lower temperatures and elevated pressures.
Most of the BTEX is separated from the glycol in the regenerator still. Other VOCs, which tend to be less soluble in glycol, are also emitted in lower quantities from the still vent.
Emissions of BTEX and other VOCs from dehydrator reboiler still vents have become a major concern for the natural-gas industry as a result of increasing regulatory pressure to reduce such emissions.
As noted in Part 1 of this series, Louisiana has already regulated still vents on large glycol units, and it has also recently promulgated an air toxics rule that may affect many smaller units. Texas, Oklahoma, Wyoming, and California are among the states which are also considering regulation of BTEX and/or VOC emissions from dehydration units.
The Clean Air Act Amendments of 1990, also as noted in Part 1, have provided additional impetus for regulating air toxics emissions.
Estimates of BTEX and VOC emissions from TEG dehydration units will be needed as more of these units are required to obtain permits. Emission estimates may also be required for compliance with other federal, state, or local regulations.
Furthermore, accurate estimates of the mass and composition of emissions will enable the natural-gas industry to select control technologies that will meet or exceed the regulatory requirements.
TECHNICAL APPROACH
GRI-DEHY is based on the simplified flow diagram shown in Fig. 2.
The glycol dehydrator System modeled consists of an absorber, a glycol pump, a flash tank (optional), and a regenerator. Viewing left to right across the figure yields the following observations:
- In addition to water, BTEX and other hydrocarbons are picked up in the absorber by the lean glycol.
- If a gas-driven (as opposed to electric) pump is used, a small amount of gas is mixed with the rich TEG that comes off the bottom of the absorber.
- If a flash tank is present, some of the water and hydrocarbons are flashed out of the rich TEG.
- Of the water and hydrocarbons that enter the regenerator, about 90-99% are emitted in the off gas, while only about 1-10% remain in the lean (regenerated) glycol.
The first step in the program is to use the input data for the gas and TEG to prepare a mass balance for water around the absorber. This is done in one of two ways.
- If the user inputs the dry-gas water specification, then it is assumed that the system operates so that this specification is exactly met. This information is used to calculate the number of theoretical stages (N) in the absorber.
- If the user inputs the number of theoretical stages, then this number is used to calculate the dry-gas water content.
In either case, it is assumed that the wet gas entering the absorber is saturated with water. Once the water balance is completed and the number of theoretical stages is known, the program performs vapor-liquid equilibrium calculations to estimate the K-value (volatility) of each component in the system.
The K-values are used with the vapor and liquid flow rates to calculate absorption factors for each component. The absorption factors are used with the Kremser-Brown approximation to calculate the moles of organic compounds absorbed in the dehydrator.2
MASS CALCULATIONS
After the absorber calculations are completed, the program performs some simple mass-balance calculations for either electric or gas-driven pumps.
If a gas-driven pump is specified by the user in the flowsheet, GRI-DEHY adds a small amount of wet gas to the rich TEG upstream of the flash tank. The composition and flow rate of the pump discharge are then used as inputs into the flash tank (if one is present in the system).
A standard flash calculation is performed at the user-specified temperature and pressure to determine the amount and composition of vapor flashed from the rich TEG.
The computed composition of the glycol solution from the flash tank is used as an input to estimate the performance of the regenerator. In order to avoid the complex heat and material balances that would be required if the regenerator were rigorously modeled, GRI-DEHY uses a simple empirical model.
Preliminary results from GRI research on sampling and analysis of glycol dehydrator emissions have shown that essentially all the light hydrocarbons (methane to hexane) and nearly all (90-99%) the BTEX are stripped from the rich TEG in the regenerator.
The regenerator emissions are estimated by a simple mass balance. The composition of VOCs (including BTEX) in the lean TEG to the absorber is then known. The entire series of calculations is repeated until the computed composition of the lean TEG converges.
The present version of GRI-DEHY uses a simplified component data base that consists of water, TEG, BTEX compounds, and normal aliphatic hydrocarbons CH4-C6H14.
Noncondensable gases such as H2S, N2, and CO2 are modeled as methane; isobutane and isopentane are modeled as n-butane and n-pentane, respectively. All aliphatics heavier than n-C5 are modeled as n-hexane.
The GRI-DEHY output has two parts:
- The "Emissions Report" gives estimates for the emissions of BTEX and other hydrocarbons from the still vent.
- The "Stream Report" gives estimated flow rates and compositions for all major streams in the dehydration unit.
Table 1 shows the inputs required of the user and also shows the outputs returned by GRI-DEHY.
In order for GRI-DEHY to be as user friendly as possible, the program was designed to be operated through use of pull-down menus. Except for specific information which must be input by the user, the program is completely menu-driven.
THERMODYNAMIC CALCULATIONS
Key to the process calculations made by GRI-DEHY are its thermodynamic calculations.
Accurate modeling of both the absorber and the flash tank requires accurate calculation of the distribution coefficients (Ki) which can be described by Equations 1a and 1b found in the accompanying equations box.
Equation 2 is derived from vapor-liquid equilibrium equations.
For GRI-DEHY to evaluate values of Ki, it must have good models for each of the thermodynamic variables in these equations. Following are the models used in the calculation of key variables:
- Pisat is calculated with the extended Antoine equation.
- fi and fisat are calculated from the Peng-Robinson equation of state.
- hi is estimated by assuming that the parties molar volume equals the pure component molar volume.
- gi is calculated with the Universal Quasi Chemical Activity Coefficient (Uniquac) model.
Since the methods for evaluating saturation pressures and Poynting factors are well known and the Peng-Robinson equation of state has been used widely in the oil and gas industry,3 the remainder of this discussion focuses on how GRI-DEHY calculates activity coefficients.
As noted, GRI-DEHY uses the Uniquac model to calculate the liquid-phase activity coefficients.4 This model can estimate g by expressing the excess Gibbs energy of the liquid phase as a function of temperature and composition.
The result is Equation 3.
In Equations 3 and 4, z is a coordination number representing the number of molecules surrounding any other molecule in the liquid solution. GRI-DEHY follows the conventions of Prausnitz and Abrams4 and assumes z = 10.
Also, the parameters qi and ri are molecular parameters for surface area and volume, respectively.
The variables tij in the Uniquac equation account for interactions between different molecules in solution. In GRI-DEHY, tij is assumed to have the form shown in Equation 5, where R is the ideal gas constant, T is the system's temperature, and Aij and Bij are adjustable parameters.
It is important to note that the matrices A and B are not symmetrical, so that Aij = Aji and Bij = Bji.
For GRI-DEHY, many of the Uniquac adjustable parameters were determined from experimental data using the data-regression capability of the process simulator ASPEN/SP.5 Interaction parameters for binary pairs of other component; were estimated with the Uniquac Functional-group Activity Coefficient (Unifac) method.6
For those components, such as methane and ethane, that are present at higher than their critical temperatures, GRI-DEHY uses a Henry's law expression to calculate the liquid-phase fugacity, as shown in Equation 6.
The Henry's law constant (Hi) is a function of both composition and temperature. To model the composition dependency, GRI-DEHY calculates an overall Hi for each supercritical component based on the composition of the liquid.
The temperature dependency of Hi is assumed to be of the form of Equation 7 in which T is the system temperature and a, b, and c are constants.
CASE STUDIES
During prerelease testing of GRI-DEHY, comparisons were made between results from GRI-DEHY estimations and the actual measured emissions of the dehydration units being studied.
Table 2 shows the operating conditions of some of these units. The units considered in Table 2 cover a wide range of operating conditions, such as gas production rate and BTEX composition in the wet gas.
Much of this table's data were obtained from API.
For GRI-DEHY to yield accurate emissions estimates, the input data to the program must be measured or estimated as accurately as possible.
Based on a sensitivity analysis of GRI-DEHY (discussed presently), it is especially important to have accurate data for wet-gas temperature, glycol circulation rate, and wet-gas composition. Errors in measuring or estimating these values will strongly affect the reliability of the emissions estimates.
Furthermore, in comparisons of GRI-DEHY results with measured emissions, it is important to remember that measured emissions are, in many cases, time-averaged values. Emissions estimates from GRI-DEHY, however, represent emissions from a unit that is operating at exactly the conditions input by the user.
Therefore, achieving good agreement between field measurements and results from GRI-DEHY requires that the emissions measurements be made at the same operating conditions as those which are input to the program.
This is especially important if the unit is subject to significant operational variability, such as naturally occurring fluctuations in the concentrations of trace constituents in the inlet natural gas.
In some cases the concentration of benzene or other trace components have been found to vary as much as 50-100%. This type of variation directly affects the uncertainty of the estimated emissions.
Fig. 3 shows comparisons for measured emissions with predictions made by GRI-DEHY for the units described in Table 2. Separate graphs are presented for benzene, toluene, ethyl benzene, xylenes, total BTEX, and nonaromatic hydrocarbons (i.e., CH4-n-C6H14).
Each graph has a solid line drawn at 45 to the horizontal axis. These lines represent perfect agreement between measurements and model predictions.
Points above the line represent model predictions that were higher than the measured emissions; points below the line represent model predictions that were lower than measured emissions. The closer a point is to the 45 line, the better the agreement between measured and estimated emissions.
As the graphs show, the difference between experimental and estimated BTEX emissions is greatest for ethyl benzene. This is expected because the ethyl benzene concentration of natural gas is usually near the analytical detection limit. As a result, the uncertainties in experimentally measured ethyl benzene concentrations are usually very high.
Fig. 3 demonstrates that results from GRI-DEHY compare reasonably to experimental data even when a wide range of gas-production rates and BTEX compositions are considered.
GRI-DEHY is sufficiently accurate to meet its objective of providing a screening tool to allow gas producers to identify which TEG dehydration units may be of concern due to BTEX emissions.
In assessing the uncertainty associated with the emissions estimates made by GRI-DEHY, a distinction should be made between different classes of compounds. For the BTEX compounds, GRI-DEHY's thermodynamic vapor-liquid equilibrium models have been calibrated with experimental data on the solubility of aromatic hydrocarbons in TEG.4
As a result, the emissions estimates for the BTEX compounds are likely to compare to experimental data better than the estimates for the C1-C6 alkanes compare to experimental data (Fig. 3f).
Future releases of GRI-DEHY will incorporate the results of ongoing GRI projects to obtain more experimental data on the solubility of aliphatic hydrocarbons in TEG.
SENSITIVITY ANALYSIS
Programs like GRI-DEHY can be used to evaluate the effects of process fluctuations on dehydrator emissions, an application borne out by a sensitivity analysis.
An actual dehydrator in South Texas provided the base case reference point for evaluating the effects of changing individual variables. Extensive sampling and analytical work performed around this unit verified that GRI-DEHY performs reasonably well in modeling dehydrator emissions.
Table 3 lists the operating data for the base-case unit.
Each of the key elements listed in Table 3 was varied above and below the base value, while the other variables were held constant. Based on this analysis, the variables that most affect emissions are gas concentration, gas temperature, and glycol-circulation rate.
Fig. 4a shows the effect of varying the BTEX concentration from 500 to 1,500 ppm. As expected, an increase in the BTEX content increases emissions significantly.
While this result is expected, it is important because the BTEX content of the gas entering the dehydrator may vary significantly over time.
For example, the BTEX concentration of the inlet natural gas to the South Texas dehydrator was measured as a function of time with an on-line Fourier Transform Infrared (FTIR) spectrometer and concentration fluctuations of 50-75% were routinely observed. These concentration variations will in turn cause significant fluctuations in the BTEX emissions.
Dehydration temperature is another key variable which affects BTEX emissions. Fig. 4b shows BTEX emissions as a function of temperature. As the graph shows, an increase in temperature from 75 to 95 F. decreases emissions by more than 30%. This occurs because the amount of BTEX absorbed in the dehydrator decreases as the temperature increases.
The glycol circulation rate also has a significant influence on dehydrator emissions. Fig. 4c shows the effect of varying the circulation rate from 0.6 to 2.0 gpm.
As that graph shows, increasing the circulation rate by a factor of four also increases emissions by approximately a factor of four. Again, this result is not surprising because BTEX absorption is a function of the liquid-to-gas (L/G) ratio in the dehydrator.
In lieu of installing air-pollution controls on the regenerator, the dehydrator operator may try to reduce emissions from his units by setting the glycol-circulation rate at the minimum amount required to produce gas within pipeline-water specification.
Flash tank temperature and pressure are other process variables that a glycol unit operator can use to reduce regenerator emissions.
Increasing the flash temperature or decreasing the flash pressure will cause a higher fraction of BTEX and other VOCs to be flashed from the glycol before it goes to the regenerator. This is particularly true if a gas-driven pump is used in the dehydrator system because the pump-driven gas will help strip the hydrocarbons from the glycol. This will help reduce emissions, however, only if the flash gas is used as fuel or burned in a flare. As Fig. 4d shows, a high flash temperature and low flash pressure in a system with a gas-driven pump can volatilize as much as 15-20% of the BTEX.
REFERENCES
- Rueter, Curtis O., and Murff, M.C., Results of the Glycol Waste Survey, Draft Report, GRI Contract No. 5088-254-1709, October 1991.
- Wankat, Phillip C., Equilibrium Stages Separations, New York, Elsevier Science Publishing, 1987.
- Peng, Ding-Yu, and, Robinson, D.B., "Two and Three Phase Equilibrium Calculations for Systems Containing Water," Canadian Journal of Chemical Engineering, Vol. 54, December 1976.
- Abrams, Denis S., and, Prausnitz, J.M.,"Statistical Thermodynamics of Liquid Mixtures: A New Expression for the Excess Gibbs Energy of Partly of Completely Miscible Systems," AlChE Journal, Vol. 21, No. 6, November 1975.
- Ng, H.J., and, Chen, C.J., The Solubility of Selected Aromatic Hydrocarbons in Triethylene Glycol, Gas Processors Association Research Report 131, Project 895, December 1991.
- Fredenslund, A., Jones, R.L., and Prausnitz, J.M., AlChE Journal, Vol. 21, No. 6, November 1975.
Editor's note: Readers may receive a complimentary 3.5 in. and 5.25 in. diskette, each containing the program, and a user's manual by writing to GRI-DEHY PROGRAM, c/o OIL & GAS JOURNAL, 3050 Post Oak Blvd., Suite 200, Houston, TX 77056. Requests may also be sent by telefax to 713/963-6285. In either form, please provide a full and accurate shipping address (not a postal box number) and telephone number.
No packages will be sent out after July 1, 1994.
Hardware requirements for GRI-DEHY Version 1.0 are minimal: an IBM PC computer or other fully IBM-compatible model of personal computer, DOS version 2.0 or higher, one 3.5 in. or 5.25-in. diskette drive, and 0.512 mb of memory (RAM) with 0.400 mb free.
Radian says that program execution time for GRI-DEHY is significantly less when run on a computer equipped with a math coprocessor chip. On a 20 mhz PC with an Intel 80386 main processor and an Intel 387 mobile math coprocessor, execution time for GRI-DEHY is typically 10-15 sec.
Future versions of GRI-DEHY, says Radian, will expand the component data base to include H2S, CO2, N2, and aliphatic hydrocarbons heavier than C6H14.
Also, capabilities for modeling ethylene glycol units and regenerator emissions control devices such as condensers will be included.
Radian Corp. and the GRI have stipulated that GRI-DEHY is published and sponsored by the Gas Research Institute, Chicago, but not sponsored by, approved by, or affiliated with any other party that may use DEHY in promoting goods and services.
Copyright 1993 Oil & Gas Journal. All Rights Reserved.
