GLYCOL-REBOILER EMISSIONS-2 GLYCOL MASS-BALANCE METHOD SCORES HIGH FOR ESTIMATING BTEX, VOC EMISSIONS

May 31, 1993
Patrick L. Grizzle Oryx Energy Co. Dallas Evaluations by Oryx Energy Co., Dallas, of five reboiler-emissions estimating methods indicate that one employing rich/ lean-glycol mass balance yields excellent estimates of benzene, toluene, ethyl benzene, and xylene (BTEX) and acceptable estimates of total volatile organic compounds (VOCS). The company has confirmed the rich/lean-glycol estimating method from total-capture stack condensation sampling on small glycol units (less than 15 MMscfd) and
Patrick L. Grizzle
Oryx Energy Co.
Dallas

Evaluations by Oryx Energy Co., Dallas, of five reboiler-emissions estimating methods indicate that one employing rich/ lean-glycol mass balance yields excellent estimates of benzene, toluene, ethyl benzene, and xylene (BTEX) and acceptable estimates of total volatile organic compounds (VOCS).

The company has confirmed the rich/lean-glycol estimating method from total-capture stack condensation sampling on small glycol units (less than 15 MMscfd) and from partial-stack condensation sampling on large units.

BTEX emission data obtained from total-capture condensation and rich/lean mass balance are in excellent agreement. Agreement in total VOC emissions from the two methods depends upon inlet-gas composition.

Rich/lean-glycol BTEX emission estimates are independent of atmospheric and pressurized sample collection; higher estimates of VOC, on the other hand, are obtained from pressurized samples.

For industry, estimating air emissions and developing control technologies to mitigate their effects on the environment have become issues whose importance derives from the large number of units operated and variations in their sizes.

Although the actual number of glycol units operating in the U.S. is unknown, estimates have run as high as 40,000. In general, individual units have been designed to process from less than 1 MMscfd of natural gas up to several hundred. By far most of these units - approximately 90-95% - process less than 10 MMscfd.

ESTIMATING METHODS

In glycol dehydration, compressed wet inlet gas is passed counter-currently through a liquid-glycol contactor. The dry glycol, typically triethylene glycol (TEG), absorbs the water from the wet-gas stream, and the dry gas leaves the top of the contactor for further processing or transport.

The water in the rich-glycol stream is removed by vaporization in a reboiler, and the lean (dry) glycol is recycled into the contactor (Fig. 1).

In this closed loop system, flash tanks and heat exchangers are often incorporated to remove gaseous hydrocarbons and condensate from the rich glycol.

VOC and BTEX emissions from these glycol units vary depending upon the inlet feed composition (gas composition and water content) as well as the configuration, size, and operating conditions of the glycol unit (i.e., glycol type, pump type and circulation rate, gas and contactor temperatures, reboiler fire-cycles, and inlet scrubber and flash tank efficiencies).

Examination of Fig. 1 indicates four methods for estimating emissions:

  • Rich/lean-glycol mass balance

  • Inlet/outlet gas mass balance

  • Unconventional stack measurements:

    • Total-capture condensation

    • Partial stack condensation/flow measurement

  • Direct stack measurements:

    • Conventional stack measurements

    • Novel stack composition/flow measurements.

Both simplicity and cost indicate the indirect estimation methods as the methods of choice.

Conventional and unconventional stack measurements are labor intensive and require extensive analysis. In general, the cost of conducting these tests may exceed the cost of the technology, required to eliminate the emissions, especially for small units.

In addition, the small diameter of the steam vent and the low flows in the vent stream make conventional stack and flow measurement techniques unsuitable for estimating emissions from these units.

Although the development of novel direct-stack sampling procedures may have merit, the overall goal of Oryx's work has been to obtain field emissions data, evaluate potential emissions from these units, and develop and apply mitigation where needed.

In this approach, the work has concentrated on small field units. Larger gas plants, in general, will require control technology to reduce emissions under state and federal regulations.

Although located in isolated areas, the plants studied have electricity which broadens the range of available mitigation technologies. Considering the number of small production units in operation, our approach has been to concentrate on indirect methods and to use the stack measurement to test the accuracy of indirect mass-balance methods.

RICH/LEAN-GLYCOL BALANCE

Preliminary testing on units in Louisiana was conducted through the Louisiana Midcontinent Oil & Gas Association (Lamoga). This testing which used mass balance between the hydrocarbon content of rich and lean-glycol, samples indicated that some analytical development effort was needed.

The Lamoga initial tests were conducted with the standard U.S. Environmental Protection Agency (EPA) protocols for BTEX analysis (EPA 602 and EPA 8020).1 2

Although glycol samples can be analyzed by these methods, they were designed for low levels (ppb) of BTEX in water or soil. Rich-glycol streams contain much higher concentrations of BTEX.

For example, benzene concentrations range from 100-200 ppm up to 4,0005,000 ppm. Analyzing these samples with EPA methods requires the glycols to be significantly diluted in water.

These dilutions not only affect the accuracy of the analyses, but also can affect the analytical results for nonaromatic hydrocarbons because of their limited solubility in water. Consequently, total VOC estimates may be low with these methods.

Their use in collection of initial survey data conducted by Lamoga may have led to poor quality data and a bias against the rich/lean-glycol balance estimation method.

SIMPLIFIED CC METHOD

To avoid this problem, a simplified, direct-injection capillary gas chromatographic (GC) method for analyzing these samples was developed.

It is important to note that care must be taken in interpreting the data to obtain accurate estimates of VOCS. In TEG samples, several degradation products or contaminants (i.e., diethylene glycol and lube oil) are typically observed in the chromatograms. These must be removed from the VOC estimations to yield meaningful data.

The relative concentrations for various compound types obtained by this method for a typical sample are normal alkanes (6.5 wt branched alkanes (8 wt cyclic alkanes (12.5 wt and aromatic hydrocarbons (73.0 wt

The composition of these aromatics consisted of benzene (26.6 wt %), toluene (26.2 wt %), C2-benzenes (17.2 wt %), and C3-benzenes + (3.0 wt %).

As expected, the composition of the rich-glycol stream generally parallels the relative solubilities of the various compound types.

Although other compounds in addition to the aromatic hydrocarbons listed in the Clean Air Act of 1990 (CAA) are observed in the rich-glycol stream (i.e., hexane and ethylene glycol), the concentrations of these compounds are typically quite low compared to the BTEX compounds.

Although emissions of nonaromatic hydrocarbons could possibly be important from systems where the inlet gas contains a very low concentration of nonmethane, nonethane hydrocarbons (dry gas), Oryx did not observe this in its survey of more than 100 units in Louisiana, Texas, and Oklahoma.

SAMPLING

Because of the entrained and dissolved noncondensible gas in the rich-glycol stream, these samples when collected at atmospheric pressure, lose a significant amount of gas and potentially BTEX and VOCS. Consequently, the accuracy of the rich/lean mass balance method has been questioned by both industry representatives and regulatory agencies.

To address this issue Oryx conducted a set of studies as part of a task force for the Oklahoma-Kansas Midcontinent Oil & Gas Association (Okmoga). In this study, atmospheric and pressurized rich and lean samples were collected from two small production facilities in Oklahoma. Neither unit had a flash tank.

The atmospheric samples were collected in glass bottles. The rich samples were introduced into the bottle until the "foam" reached the top of the bottle. At this point, the bottle was tightly sealed and chilled.

To reduce the possible loss of hydrocarbons, no attempt was made completely to fill the bottle with the rich liquid glycol.

Lean samples were similarly collected and chilled. But because of the absence of degassing, the bottles were filled completely with the liquid glycol. Aliquots of the rich and lean-glycol samples were removed from the bottles and analyzed according to the direct-injection method.

Pressurized samples were collected in 1 l. piston cylinders. To each cylinder, 250 cc of rich or lean glycol were introduced into the cylinder and the cylinder returned to the lab for analyses.

At the lab, the liquid was expanded to 1 l., the cylinders were allowed to equilibrate overnight, and samples of the liquid phase and gas phase were analyzed. The liquids were analyzed by the direct-injection method.

Gas samples were introduced into a gas chromatograph with a 1-cc loop injector and analyzed on a capillary 25 m x 0.2 mm (ID) 100% methylsilicon column (0.33 ii film) with a flame-ionization detector (FID).

BTEX and other components were quantified from external calibrations using known gas standards.

In some cases, a small amount of condensate hydrocarbon was found on the glycol phase. Depending on the amount of the condensate, it was either analyzed separately or thoroughly mixed with the glycol prior to analyses of the glycol sample.

Combining the results of the individual analyses (gas, glycol, and condensate, when present) based on the relative volumes of each yielded the total amount of BTEX and VOCs in the pressurized sample. The volume of the gas phase in the cylinder was corrected to standard temperature and pressure based on the pressure in the cylinder following expansion.

RESULTS AT TWO UNITS

Concentration and emission data for BTEX and VOCs from the two units are shown in Table 1.

Unit 1 was processing approximately 2.2 MMscfd of gas; Unit 2 approximately 1.8 MMscfd. As noted in the table, the concentrations of benzene in the rich-glycol samples from these units are relatively low when compared with Gulf Coast production units.

On the other hand, these samples contain relatively high concentrations of toluene and xylenes. This presumably reflects the chemical composition of the inlet gas. This table also indicates the relative fluctuations that are inherent to glycol dehydration units.

Although the reproducibility of the analytical method is approximately 3 to 5%, the variation in the data is on the order of 10%. This is not unusual in estimating emissions from glycol units.

Because of fluctuations in inlet gas and contactor temperatures, reboiler fire cycles, and other operating conditions, the uncertainty in the emission measurements primarily results from system variations, not analytical methodology.

As noted in Table 1, Unit 1 was sampled three times during the day; Unit 2 four times. The larger variation in the concentration data for Unit 1 and compared with Unit 2 suggests that the former unit was more variable during the day than the latter.

Comparison of the pressurized and atmospheric data clearly indicates that sample collection is an insignificant variable for the estimation of BTEX emissions from these units. BTEX emissions are virtually identical for pressurized vs. atmospheric sampling.

This is not true, however, for VOC estimates. Although the emissions are similar for Unit 2 by both sampling methods, VOC estimates for the pressurized samples are approximately 20% higher for Unit 1. The same conclusion is seen from the total-capture condensation data (discussed presently).

In the total-capture condensation method, the entire steam vent is condensed and the resulting hydrocarbon condensate, water, and noncondensible gas phases are analyzed and combined to yield total BTEX and VOC emission estimates.

Comparison of the stack condensation with the glycol mass-balance data indicates excellent agreement in the BTEX emission data between the two methods, independent of sample-collection methods.

Although the VOC estimate from stack condensation appears to agree well with the atmospheric glycol sampling, the uncertainty in the value-approximately 20% -is in the range of potential underestimation from the atmospheric sampling.

It is important to note that the data presented in Table 1 are from glycol units with a relatively dry (low condensate content) inlet gas. Based on total-capture condensation studies, it appears that the loss of VOCs from atmospheric glycol samples is less for dry inlet-gas units (1020%) and higher for wet inlet-gas units (50%).

This difference probably reflects a mass-transport effect. When the concentration of heavy hydrocarbons is low in the rich glycol, degassing of the rich glycol does not remove much of the heavy components. Alternatively, when the concentration of heavy hydrocarbons is high, degassing removes more.

Based upon these studies, rich/lean-glycol mass balance is a highly convenient, cost-effective method for estimating air emissions from glycol dehydration units.

The method yields excellent estimates of BTEX from the units and fair to good estimates of VOCS, depending on the inlet-gas composition. Good estimates of BTEX emissions can be obtained with atmospheric sampling.

The amount of VOC emissions is underestimated with atmospheric sampling and can be improved with pressurized samples. Use of pressurized samples complicates the analytical method, however, and requires analyses of multiple phases (glycol, condensate, and noncondensible gases).

Oryx's experience indicates that one of the major advantages of the glycol mass-balance method is that it is a "time averaged method." As a result of the volume of glycol in the system, short-tern variations in operating conditions of the unit are minimized because of this averaging effect.

Consequently, the results are more representative of the actual emissions from the unit.

INLET-OUTLET BALANCE

The use of inlet (wet) and outlet (dry) gas mass balance has been proposed as one method for estimating air emissions from glycol units without flash tanks. On units with flash tanks, the method can still be used but the flash tank's effluent must be analyzed and proper adjustments made to emission estimates. Unfortunately, Oryx's experience with this method indicates that, generally; emissions are either greatly overestimated or grossly underestimated. In fact, in some cases, BTEX or VOCs appear to be 'created" in the contactor.

Unlike the rich/lean-glycol mass balance method, the problem with the gas mass balance method results because estimates are being made from small differences in relatively large numbers.

For example, the 8-hr average (six determinations) benzene contents for the wet and dry gas streams for a small Louisiana unit (2.2 MMscfd) were 219 15 ppm and 213 14 ppm, respectively. The difference in the benzene contents is obviously quite small (4 ppm) compared with the uncertainty (about 15 ppm).

Whether the uncertainty in the data reflects analytical uncertainty or short-term variations in the benzene content is unclear from the data set. Whatever the source, however, the relative magnitude of the uncertainty as compared with the difference in the inlet and outlet gas content can significantly affect emission estimates.

For comparison, the average rich and lean-glycol benzene data for the same day (10 determinations) were 1,425 102 ppm and 21 t 5 ppm, respectively. Although the uncertainties are of the same relative magnitude, the glycol mass-balance method clearly should be less affected by the uncertainty in the da ta because of the large difference in concentrations between the rich and lean samples.

It is important to note that the emission estimates from gas mass balance can worsen for larger units which process more gas. Studies indicated that the gas mass-balance results have been 'm error by more than an order of magnitude for large gas plant units.

UNCONVENTIONAL MEASUREMENTS

Although the rich/lean-glycol mass balance provides good emission estimates from glycol dehydration units, Oryx conducted studies of unconventional stack measurements for confirmation and to provide a testing method for units with control devices.

Two approaches were developed depending on the size of the dehydration unit.

TOTAL CAPTURE

For small production units (

In this method, the entire steam vent effluent passes through a condenser (50 ft of coiled copper tubing; 1-1.5 in. ID) which is cooled with ice and water to approximately 45-50- F.

The condensate (water and hydrocarbon) collects in an accumulator, and the noncondensible gas collects in a stainless steel sample-bomb for analysis.

The noncondensible gas flow is measured with a dry test meter. To average variations in the unit, each total-capture measurement is run for at least 30 min, and typically three measurements are made on each unit.

Analyses of the various phases and associated volumes collected indicated emission estimates. The water-phase samples are analyzed by the standard EPA 8020 method and the gas-phase samples by capillary gas chromatography (described previously) with a 1-cc loop injector.

The hydrocarbon condensate samples are analyzed with the same procedure as used for the rich/lean glycol with the exception that the sample weight is reduced by a factor of 10 (0.2 g).

Typical emission data collected with this method appear in Table 2. For comparison, the rich/lean-glycol mass balance (atmospheric sampling) data also appear.

Unit 1 is located in Oklahoma and processes relatively dry gas. Data for this unit appear in Table 1.

Unit 2, located in Louisiana, was studied in detail in conjunction with the Louisiana Department of Environmental Quality. Nine determinations were made on this latter unit over 8 hr to learn the short-term variations which could be expected in emission estimates.

As noted previously, there is excellent agreement between the rich/lean-glycol mass balance and total-capture condensation methods for BTEX. VOC data agree less well.

For the unit in Oklahoma processing relatively dry gas, VOC data from the two methods agree well, although the uncertainty in the condensation data suggests that the VOC estimates by the glycol method may be as much as 20% low.

The Louisiana unit is processing gas with more heavy hydrocarbons. In this case, the VOC data obtained by the glycol method is about 50% of that by total-capture condensation.

Reasons for the larger variance for hydrocarbon-rich inlet gas streams vs. relatively dry gas streams have been discussed previously.

Fig. 3 shows the variation in benzene emission over about 8 hr for the Louisiana unit (Unit 2). The benzene-emission estimates varied from 0.63 to 0.92 tons/year (tpy) with the glycol mass-balance method and from 0.57 to 1.21 tpy with the total-capture method.

Both variations are well outside the analytical uncertainty associated with the two methods. Consequently, they reflect typical fluctuations which correspond to variations in the unit's operating conditions.

Both sets of data indicate that the benzene emissions dropped slightly from early morning to late afternoon. This slight decrease probably reflects variations in the inlet-gas temperature (water content) through the day and, thus, solubility differences of the benzene in the TEG in the contactor.

It is interesting to note the significant increase in benzene emissions as estimated by the total-capture method during mid-day (Runs 3-6). This same trend is not seen in the glycol data.

This larger variation in the data suggests that the total-capture method is more sensitive to short-term operational variations in the unit than the glycol method.

In the total-capture method, attempts are made to time-average the data by collecting the effluent for a minimum of 30 min. These data would suggest this time averaging approach may still be insufficient to reduce the short-term variations (inlet temperatures, low-high fire cycles, etc.).

Because of the volume of glycol in the system, the glycol mass-balance method is less sensitive to these operational fluctuations. It should be noted that the glycol data appear to increase late in the day when the total-capture data are decreasing.

This increase in the glycol method data later in the day could reflect a time-delayed response to the high total-capture data collected earlier in the day (Runs 3-6), although this explanation remains speculative because data have not been collected long enough to confirm the trend.

PARTIAL STACK

On larger units (greater than 15 MMscfd), a total-capture condensation method for estimating emissions of the stack is impossible. Although Oryx's studies have concentrated on the smaller production units due to the much higher number of these units in operation, the company developed a partial-stack condensation method for testing these larger units.

In these studies, a 0.25-in. probe was centered in the steam vent stack and a vacuum was applied to the system to pull the vent effluent through the sampling train (Fig. 4). After the system was thermally equilibrated, the condensible hydrocarbons and water were removed by passing the effluent through a condensation coil which was chilled in ice and water.

The condensate and water collected in a cooled flask, the noncondensibile gas which passed through the vacuum pump collected for analyses, and the flow of noncondensible gas was measured with a wet test meter.

After the sampling train had collected material for approximately 30 min, the inlet probe was removed from the steam vent and the total gas flow measured with a turbine meter. As in the case of the total-capture method, the composition of each collected phase (hydrocarbon, water, and gas) and the relative volumes are used with the total flow data to estimate the BTEX and VOC emissions.

Although this procedure yields data which compare reasonably well with the glycol mass-balance data, the validity of the total flow measurement from the vent is a major concern.

In some cases, Oryx believes that the turbine flow meters give relatively accurate data. On some smaller units with lower flows, however, condensation on the turbine blades causes a problem. Accurate measurement of these vents' streams is clearly, where additional research is needed.

Oryx is aware that hot-wire anemometers and orifice plates have also been used successfully to determine the flows from these larger units, but the company believes that all of these methods need more thorough evaluation.

Some analytical results obtained collected by the sampling train procedure are given in Table 3. As noted there, compared with the total-capture condensation data, the partial-condensation results agree less well with those of the rich/lean-glycol method.

And as mentioned, accurate measurement of the flow from the steam vent is of concern. In the case of Unit 1, there was an additional problem. Here was the only case in which Oryx found the glycol mass-balance results to be significantly higher than the stack-condensation field test data.

Although the uncertainty in the flow could be part of the problem, the primary problem in this unit was the homogeneity of the glycol stream. In this large gas plant unit, a stratified phase of hydrocarbon condensate was found in the glycol stream.

This "contamination" of the rich-glycol stream, consequently, gave rise to these high emission estimates by this method. The data from the much smaller unit is more typical: There is reasonable agreement between the data on BTEX emissions and relatively poor agreement in the VOC emission estimates.

It should be noted that the uncertainty in the partial-stack condensation data is significantly higher than either the total-capture condensation or rich/lean mass-balance methods. This larger uncertainty is probably related to many factors but certainly to the positioning of the probe in the steam vent, the reproducibility of the side-stream flow through the vacuum system, and the uncertainty in the total flow measurements.

CONVENTIONAL MEASUREMENT

Although Oryx's experience with direct sampling and analyses of the steam vent and flow measurement has been limited, a contractor has conducted two studies to determine the potential of the method for emission estimates.

In these studies, the concentrations of benzene, toluene, and VOCs in the steam vent were directly determined by gas chromatography. Although some problems with condensation arose, the condensate that collected in the instrumentation was analyzed and the compositional analysis data combined with the gas analyses data.

From moisture content determinations and stack velocity and flow data obtained with orifice plates, estimates of the emissions were obtained with conventional stack measurement.

Typical data obtained by the method are given in Table 4.

For comparison, the table includes data obtained from glycol mass balance. As compared to the rich/lean-glycol mass balance results, the agreement in total VOCs by the two methods is relatively good (+20%). Agreement in the BTEX emissions, however, is very poor.

Emission estimates of benzene, toluene, and ethylbenzene by the two methods differ generally by factors of 2 or 3. In addition, xylene has not been found in either study by use of conventional stack measurement, although significant xylene emissions were estimated from the glycol mass balance.

In general, additional research is needed and appropriate modifications required for conventional stack measurements to be acceptable for estimating emissions from these units.

ACKNOWLEDGMENTS

The author wishes to thank Ken Konvicka and Donna Sablotny for assistance in data collection; Les Brigham, Bessco Inc., for development and assistance in evaluation of control devices; and Oryx Energy Co., Dallas, for permission to publish this article.

REFERENCES

  1. "Methods for Chemical Analysis of Water and Wastes," U.S. Environmental Protection Agency, EPA No. 600/4-79-020 (Rev.), March 1983.

  2. "Test Methods for Evaluating Solid Waste, Physical/Chemical Methods," U.S. Environmental Protection Agency, SW-846 (Third Ed.), September 1986.

Copyright 1993 Oil & Gas Journal. All Rights Reserved.