Warren R. True
Pipeline/Gas Processing Editor
A research branch of the U.S. Environmental Protection Agency (EPA) has conducted field tests to evaluate a computer program's effectiveness in estimating emissions of hazardous air pollutants (HAPs) and volatile organic compounds (VOCs) at glycol dehydrators.
Test results of the program, GRI-GLYCalc, have led the laboratory to recommend, in an intra-agency memorandum issued earlier this year, that the program be included in EPA's guidance for state and local agencies that are developing emissions inventories to meet Clean Air Act requirements.
GRI-GLYCalc is a personal computer-based emissions model developed by the Gas Research Institute (GRI), Chicago, as a tool the gas industry can use to estimate emissions from glycol dehydrators. It is successor to another model, GRI-DEHY, also developed by GRI (OGJ, June 14, 1993, p. 36).
The model runs on an IBM-compatible personal computer, with minimal system requirements. It is available free to academic institutions and government agencies from the Radian Corp., Austin, Tex. (contractor to GRI). The cost is $25 to all others.
Memo routing
The memorandum was written by Larry G. Jones, chief of the emissions and modeling branch at EPA's air and energy research laboratory, Research Triangle, N.C. It was addressed to J. David Mobley, group leader of the emission factor and inventory group in EPA's office of air quality planning and standards.
Copies were released by Radian to attendees of an industry forum in Houston in August.
The office of air quality planning and standards, Jones explained to Oil & Gas Journal, issues federal standards for air pollution from stationary sources and approves state standards.
The office has since accepted the recommendation of Jones' group and approved GRI-CLYCalc for use by state agencies.
Jones' branch conducted field tests at two units in Texas and Louisiana and in conjunction with additional glycol dehydrator emissions tests sponsored by GRI and API, Washington. The primary objective of the field tests was to learn the effectiveness of GRI-GLYCalc for estimating HAP and VOC emissions.
Jones' memorandum accompanied a report "Glycol Dehydrator BTEX and VOC Emissions Testing Results at Two Units in Texas and Louisiana."
The emission standards division of the office of air quality planning and standards is developing maximum achievable control technology (MACT) emissions standards for glycol dehydrators.
Dehydrator population
The memo explains that glycol dehydrators are used in natural-gas processing to remove water from natural gas and reports that an estimated 40,000 may be in use.
Currently, the most common glycol dehydrator design uses an absorber, with triethylene glycol (TEG) as the absorbent, to remove the water from natural gas.
In absorption, such aromatic hydrocarbons as benzene, toluene, ethylbenzene, and xylenes (BTEX) present in the processed natural gas are also absorbed into the glycol.
The BTEX compounds and other VOC species may be emitted to the atmosphere when, in subsequent processing, the glycol stream is distilled to recover the glycol for reuse. Emissions of BTEX and other VOCs occur from the reboiler still vent.
For many dehydrators, emissions of BTEX and other HAPs may exceed the 'major source' HAP emissions thresholds cited in Section 112 (a) of the Clean Air Act Amendments (1990).
GRI-GLYCalc operation
Jones describes GRI-GLYCalc as employing fundamental engineering thermodynamics and experimental data to estimate emissions.
Inputs to the model include natural-gas composition, gas flow rate, and such dehydrator design parameters as glycol-circulation rate, dehydrator operating temperature and pressure, glycol pump type, and presence or absence of a flash tank.
According to the memo, a user-friendly interface allows data input. Except for composition data, input data should normally be available from company operators. Default values may be assumed, if necessary, for some process input values.
GRI-GLYCalc can print reports showing tons per year or pounds per hour emissions of BTEX, HAP, and VOC. It can estimate emissions for either TEG or ethylene glycol-based units. TEG units represent about 95% of glycol units in use, says Jones.
The model can also estimate effects of condensation and incineration used as emission controls and adjust emissions to account for use of stripping gas in the glycol regenerator.
Field tests
The memo reports that for the two EPA tests sites, use of GRI-GLYCalc with measured gas-composition data produced emissions estimates that agreed closely with test results for the most accurate test method.
Emissions estimated by GRI-GLYCalc for both BTEX and VOC were within 10% of the measured emissions at both sites (Table 1(29951 bytes).
The tests found the most accurate test method to be "total capture" in which the entire still vent gas stream is captured, condensed, and analyzed. This method is used as a benchmark against which the results from other emissions testing and estimation methods may be compared.
Other simpler test methods were also evaluated as part of this project, says the memo. Table 1(29951 bytes) shows test results for the atmospheric rich/lean method and pressurized glycol cylinder method.
These methods both involve collection of glycol samples upstream and downstream of the reboiler. Emissions are calculated by material balance, and the methods use glycol flow rate and glycol composition data based on laboratory analysis of glycol samples.
For the emissions and modeling branch's test sites, says the memo, emissions measured with these simpler methods were generally consistent with each other and agreed well with the total capture results. This was true except for the VOC emissions estimated for Site 1, where the simpler test methods underestimated the VOC emissions.
More glycol dehydrator emissions tests have been sponsored by GRI and the API at eight other test sites; results are summarized in Table 2(57409 bytes). The emissions and modeling branch's test sites appear as Sites 9 and 10 in these tables.
For many of the GRI/API sites, states the memo, GRI-GLYCalc tended to overestimate emissions, compared to the total capture benchmark, by as much as 50-100% for BTEX emissions.
Because use of default values for model inputs will often be necessary, some overestimation of emissions is unavoidable. This may be acceptable for emissions-inventory purposes. Determining the applicability of regulatory standards or compliance determinations may require source emissions tests to quantify emissions more accurately.
Jones told Oil & Gas Journal that the overestimation, although large, is acceptable because no other tool for estimating emissions is currently available.
For the GRI/API test sites, pressurized glycol and atmospheric rich/lean glycol tests compared reasonably well to the total capture results for BTEX but tended to underestimate VOC emissions at some sites.
Sampling techniques
A second objective of the emissions and modeling branch's project was to assess different sampling techniques for the collection of natural-gas samples for laboratory analysis.
The GRI-GLYCalc method, as the memo notes, requires gas composition data as an input. Because the model is sensitive to BTEX concentrations in wet-gas streams, accurate composition data are essential for accurate prediction of emissions by the model.
Five different gas-collection methods were assessed. For the emissions and modeling branch's test sites, the modified EPA Compendium Method TO-14 with GPA sampling manifold, with the gas sample collected before the absorber, gave the best results when used with GRI-GLYCalc.
The other collection methods tested were GPA Standard 2166 without sampling manifold before and after the absorber. Standard 2166 is the industry-accepted method for natural gas sampling.
These methods differ primarily in the apparatus used for sample collection.
Jones' memo says that existing EPA stack-testing methods are unsuitable for use with glycol dehydrators because of the typically low, fluctuating flow rate of the still gas vent stream and high water-vapor levels in this stream. Accurate measurement of organic compound concentrations in a gas stream that may be more than 90% water is problematic.
By capturing the entire still vent gas stream, the total-capture technique overcomes the flow problem. A total-capture test run lasted for 60 min, adequate time to account for flow variations and reboiler on/off firing cycles. All condensable hydrocarbons and water from the still vent stream were collected during a test run.
The volume of noncondensable gas was measured with a dry gas meter. Grab samples of the noncondensable gas were collected from laboratory analysis.
The memo states that, although the total-capture technique provides the most accurate estimate of emissions, it is more hazardous, expensive, and difficult to perform, and is unsuitable for use with dehydration units that process more than 10 MMscfd of natural gas.
Alternative test methods, pressurized glycol and atmospheric rich/lean glycol, are simpler and cheaper to perform. Test results from both were usually similar.
While both methods compared favorably to the total-capture benchmark for BTEX (Table 2(57409 bytes), both alternative methods tended to underestimate VOC emissions. This underestimation most likely results from the difficulty in obtaining a representative sample of the glycol stream for units with a high noncondensable gas component in the glycol stream.
The flash tank removes entrained or absorbed gas from the glycol stream. For units without flash tanks, a high noncondensable gas flow, possibly as two-phase gas/liquid flow, is likely.
A glycol stream sample that accurately includes the noncondensable gas component may not, therefore, have been collected.
In Table 2(57409 bytes), Sites 2, 3, 8, and 10 had flash tanks in operation. For Sites 2, 3, and 10, atmospheric/rich lean results and total capture results for VOC emissions agree better than for most of the other sites which lacked flash tanks. Site 8 had added stripping gas which, the memo says, may explain why VOC emissions for this site were underestimated.
Need for accurate data
For the best application of GRI-GLYCalc, the most significant concern is for the collection of accurate input data: accurate estimates of emissions by any model can be expected only if input data to the model are accurate.
For GRI-GLYCalc, says the memo, the most critical data are natural-gas composition and glycol circulation rate.
An estimate of the dry-gas water content, or a specified number of equilibrium stages in the absorber, is needed to run GRI-GLYCalc. At most glycol dehydrator sites, however, the dry gas water content is not routinely measured. Dry gas water content or the number of equilibrium stages in the absorber may be specified from design values.
For Site 10 (Table 2(57409 bytes), continuous measurements of dry-gas water content were available. For Site 10, GRI-GLYCalc-predicted BTEX and VOC emissions are within 3% of the total-capture benchmark, a much better result than for the other sites.
Thus, the memo concludes, the uncertainty of emissions estimated by GRI-GLYCalc appears to be significantly reduced if actual dry-gas water contents are input to the model.
For BTEX-emissions estimation, GRI-GLYCalc is sensitive to BTEX concentrations in the inlet gas stream. These concentrations must be determined by collection and analysis of a natural-gas sample.
As discussed previously, the project evaluated different approaches for sample collection.
Although the modified EPA Compendium Method TO-14 measurement approach gave the best results for the emissions and modeling branch's sites, GPA Standard 2166's approaches, with samples collected before the absorber, produced results similar to those produced by GRI-GLYCalc for Site 10.
Other GRI/API locations, where different gas sample collection methods were tested, had similar experiences. The different sampling approaches sometimes agreed closely, sometimes differed.
GPA 2166 may be more susceptible to bias caused by collection of condensed liquids in the gas cylinders. Use of a manifold may remove entrained aerosols and liquids from the collected samples, but the manifold must be maintained hotter than the natural gas to prevent condensation of BTEX compounds on the manifold walls.
The memo states that GRI-GLYCalc is relatively sensitive to glycol circulation rate. Ideally, the rate would be a measured value, but such measurements may be unavailable or impractical for some sites.
If the glycol flow rate is unknown, it may be estimated by counting glycol pump strokes per minute and using pump manufacturer's specifications for volume of glycol per pump stroke.
Alternatively, a design rule-of-thumb for the ratio between volume of glycol and weight of water to be removed (recirculation ratio) may be used.
The reference manual for GRI-GLYCalc suggests a typical value for the recirculation ratio, as well as default values for selected other process variables where specific measurements may not be available.
Copyright 1995 Oil & Gas Journal. All Rights Reserved.