EPA-60*0/R-9 5-046 a
March 1995
Glycol Dehydrator BTEX and VOC Emissions Testing
Results at Two Units in Texas and Louisiana
Volume I. Technical Report
Curtis 0. Rueter
Dirk L. Reif
Duane B. Myers
Radian Corporation
Post Office Box 201088
Austin, TX 78720-1088
EPA Contract 68-01-0031
Work Assignment 067
EPA Project Officer: Charles 0. Mann
U. S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
Prepared for:
U. S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
ABSTRACT
Glycol dehydrators are used in the natural gas industry to remove water from natural gas,
and in the process, may also remove and emit significant quantities of benzene, toluene,
ethylbenzene, and xylenes (BTEX). The objective of this project was to collect emissions test data
at two triethylene glycol (TEG) units to provide data for comparison to GRI-GLYCalc™, a
computer program developed to estimate emissions from glycol dehydrators. Three analytical
techniques were used to determine emissions: Total capture condensation, pressurized glycol
cylinders, and atmospheric rich/lean glycol sampling.
Site T test results, using the various techniques, yielded BTEX emission estimates that
agreed reasonably well. Total volatile organic compound (VOC) emissions from the two glycol
methods did not match well with the total capture benchmark results; this is consistent with
previous results for systems without flash tanks. Site 2 atmospheric rich/lean glycol and
pressurized glycol emission results agreed closely with the total capture results for both BTEX and
total VOC. GRI-GLYCalc predictions using natural gas samples taken before the glycol absorber
agreed well with the total capture results for total BTEX emissions.
ii
-------
Section CONTENTS Page
Abstract ii
Figures v
Tables vi
Metric Equivalents viii
1. Executive Summary 1
Research and Development of New Methodologies 1
Process and Testing Descriptions 2
Results 3
2. Introduction 6
Background 6
Data Quality Objectives 9
Report Organization 10
3. Site Descriptions 11
Site 1 11
Site 2 13
4. Sampling and Analytical Matrix Design 17
5. Sampling Methods 23
Five Sampling Methods 23
Process Data Measurement and Collection 30
6. Analytical Methods 32
Pressurized Glycol Cylinder Handling 32
Analytical Techniques 34
Calibration Procedures 39
7. Data Reduction 44
Emissions Calculations 44
Data Review and Validation 54
Glycol Flow Rate 54
8 . Results 55
Site 1 55
Site 2 62
9. Data Quality Indicators for Critical Measurements 73
Definitions and Objectives 73
Calculation Methods 76
Precision Results 80
Bias Results 86
References 90
iii
-------
Section CONTENTS(continued) Pane
Appendix A. GPA Standards 2261-90 and 2286-86 A-l*
Appendix B. Daily Instrument Calibration Checks B-l
Appendix C. Site 1 Field Testing Results C-l
Appendix D. Site 2 Field Testing Results D-l
Appendix E. Data Quality Indicator Calculations E-l
'All appendices are in Volume II.
iv
-------
FIGURES
Number Paae
1. Typical glycol dehydration unit 7
2. Process flow diagram for Site 1 12
3'. Process flow diagram for Site 2 14
4. Sampling and analytical design for total capture liquid samples 18
5. Sampling and analytical design for total capture gas samples 19
6. Sampling and analytical design for liquid glycol samples 21
7. Sampling and analytical design for natural gas samples 22
8. Total capture condensation sampling apparatus 24
9. EPA Method TO-14 natural gas sampling apparatus 28
10. GPA Method 2166 natural gas sampling apparatus 29
11. Gas/glycol separation apparatus 33
12. Data reduction and validation approach for total capture,
pressurized glycol, and atmospheric glycol methods 45
13. Data reduction and validation approach for GRI-DEHY 46
V
-------
TABLES
Numbs it
1. Summary of Site 1 Emission Results 3A
2. Summary of Site 2 Emission Results 5
3. Dehydrator Field Testing Process Parameters 31
4. Analytical Conditions for the Analysis of Liquid Samples 35
5. Analytical Conditions for the Analysis of Hydrocarbon Gas Samples .... 37
6. Analytical Conditions for Analysis of Fixed Gases 37
7. Analytical Conditions for Subcontract Laboratories--GPA 2261 38
8. Analytical Conditions for Subcontract Laboratories--GPA 2286 38
9. Site 1 - Process Parameters 56
10. Site 1 - Total Capture Emission Results 57
11. Site 1 - Pressurized Glycol Emission Results 58
12. Site 1 - Atmospheric Glycol Emission Results 59
13. Site 1 - Natural Gas BTEX Composition 61
14. Site 1 - Natural Gas Audit Sample 62
15. Site 1 - Natural Gas Referee Sample 62
16. Site 1 - GRI-GLYCalc Results 64
17. Site 2 - Process Parameters 65
18. Site 2 - Total Capture Emission Results 66
vi
-------
TABLES (continued)
19. Site 2 - Pressurized Glycol Emission Results 66
20. Site 2 - Atmospheric Glycol Emission Results 67
21. Site 2 - Natural Gas BTEX Composition 70
22. Site 2 - Natural Gas Audit Sample 71
23. Site 2 - Natural Gas Referee Sample 71
24. Site 2 - GRI-GLYCalc Results 72
25. Summary of Data Quality Objectives for Critical (BTEX) Measurements .. 74
26. Pooled Precision for Total Capture Method - Emission Results 81
27. Pooled Precision for Glycol Methods - Emission Results 81
28. Pooled Precision (% RSD) for Natural Gas Results 83
29. Pooled Sampling and Analytical RSD {%} for Natural Gas Results 83
30. Process Data - Precision for Both Sites 84
31. Pooled Analytical Precision for Total Capture Samples Based on RPD
of Duplicate Analyses 85
32. Pooled Analytical Precision for Pressurized Glycol Samples Based on
RPD of Duplicate Analyses 85
33. Pooled Analytical Precision for Atmospheric Glycol Samples Based on
RPD of Duplicate Analyses 87
34. Pooled Analytical Precision for Canister Natural Gas Samples Based
on RPD of Duplicate Samples 87
vii
-------
tables (continued)
Number Page
35. Overall Relative Bias {%> for Site 1 Liquid Samples 88
36. Overall Relative Bias {%) for Site 1 Gas Samples (Based on Recovery
of cis-2-Butene) 89
37. Overall Relative Bias (%) for Site 2 Liquid Samples 89
38. Overall Relative Bias (%) for Site 2 Gas Samples (Based on Recovery
of cis-2-Butene) 89
METRIC EQUIVALENTS
EPA policy requires use of the metric system but, for the reader's
convenience, certain nonmetric units are used in this document. Readers more
familiar with the metric system may use the following factors to convert to
that system.
Komretnc Multiplied by Yields
kPa
Atm 9e#1
°F 5/9(°F-32)
ft 0.305
ft' 28.3
9^1. 3.79
in- 2.54
in- H9 3.38
lb 0.454
micron i
psi 6.89
°C
m
L
L
cm
kPa
*9
pm
kPa
viii
-------
SECTION 1
EXECUTIVE SUMMARY
RESEARCH AND DEVELOPMENT OF NEW METHODOLOGIES
The Emissions and Modeling Branch (EMB) of EPA's Air and Energy-
Engineering Research Laboratory (AEERL) was established to develop new and
improved emissions inventory methodologies for use by the states. New,
improved methods are reviewed by the Emission Factor and Inventory Group
(EFIG) of the Office of Air Quality Planning and Standards (OAQPS). New
methods approved by OAQPS are incorporated into EPA guidance documents for
state use in preparing emissions inventories required by the Clean Air Act.
Emissions estimation procedures for glycol dehydrators are not available
in current EPA emissions estimation guidance. EMB held discussions with the
Gas Research Institute (GRI) and the American Petroleum Institute (API) and
determined that they were actively involved in the development of process
models for estimation of emissions from glycol dehydrators. An industry
working group, chaired by GRI, had begun a program to develop field testing
methods and to collect emissions test data. The testing program and
associated emissions model development were of immediate interest to EMB as a
potential tool for estimating emissions from glycol dehydrators. EPA, GRI,
and API agreed that it would be appropriate for EMB to supplement the industry
program with an independent EPA testing program to assess the acceptability of
the process emissions model as an approved method for inclusion in EPA
emissions estimation guidance. This report describes two emissions tests
conducted by EMB to assess the current GRI glycol dehydrator emissions model
GRI-GLYCalc™. Additional tests, not discussed in this report, have been
performed by GRI and API at 8 other sites. These data will also be reviewed
prior to making a recommendation on including GRI-GLYCalc in EPA emissions
inventory guidance documents.
1
-------
PROCESS AND TESTING DESCRIPTIONS
Glycol dehydrators are used to remove water from natural gas, and in the
process of removing the water, may also remove and emit significant quantities
of benzene, toluene, ethylbenzene, and xylenes (BTEX). The most common glycol
dehydrator design employs an absorber, with triethylene glycol (TEG) used as
the absorbent, to remove water from natural gas. In the absorption step,
aromatic hydrocarbons such as BTEX are also absorbed into the glycol stream.
Following the absorption step, the glycol, rich with water and BTEX compounds,
is distilled to strip water from the glycol. Recovered lean (dry) glycol is
recycled for use in the absorber. Emissions of BTEX and other VOCs occur from
the glycol reboiler still vent. As a result of the 1990 Clean Air Act
Amendments (CAAA), hazardous air pollutant (HAP) emissions (primarily BTEX)
from the reboiler still vent stream of glycol dehydrators have become a
concern for the natural gas industry
Site 1 was located at a gas plant in west Texas and was processing 3.6
million standard cubic feet per day (MMSCFD) of gas without a flash tank and
using a gas-driven pump. Site 2 was located in southwest Louisiana and was
processing 4.9 MMSCFD of gas with a flash tank and using a gas-driven pump.
Testing was conducted over a two-day period at each site. Three emissions
measurement techniques were used at each site: total capture condensation
(the most accurate method), and two lower cost methods; pressurized glycol
cylinders, and atmospheric rich/lean glycol. The lower cost methods were
included in the test protocol to evaluate their applicability as emissions
screening tools for cases where use of the total capture method may not be
technically feasible or economically justifiable.
In total capture condensation the entire still vent stream was passed
through a 50-foot length of 1-inch diameter copper tubing coiled inside a 55-
gallon barrel and submerged in an ice/water mixture. Condensed hydrocarbons,
condensed water, and noncondensable gas were measured and sampled. Results
from total capture condensation were used as the benchmark against which other
methods were compared.
2
-------
The atmospheric rich/lean glycol method used samples of glycol from both
upstream (rich) and downstream (lean) of the reboiler collected at atmospheric
pressure in volatile organic analysis (VOA) vials. Emissions were calculated
using the difference in analyte concentrations in the rich and lean samples
and the glycol circulation rate. Based on past sampling experience in the GRI
project, the glycol methods may not produce a representative sample for total
VOC determination, particularly on systems without a flash tank.
The pressurized glycol cylinder method used samples of glycol (rich)
collected at line pressure upstream of the reboiler in stainless steel
cylinders. Emissions were calculated using the difference between the analyte
concentrations in the glycol cylinder and a lean glycol sample downstream of
the reboiler and the glycol circulation rate.
GRI-GLYCalc is a computer program developed by GRI as an alternative
screening tool to estimate emissions from glycol dehydrators using process
operating data and the composition of natural gas for the unit of interest.
To evaluate the use of GRI-GLYCalc and alternative natural gas sampling
methods, 5 types of natural gas samples were collected and analyzed.
Sub-atmospheric pressure canisters upstream of the absorber using a
sampling manifold;
High-pressure cylinders upstream of the absorber with and without a
sampling manifold; and
High-pressure cylinders downstream of the absorber with and without a
sampling manifold.
RESULTS
The results of the testing at Site 1, presented in tons per year plus or
minus one standard deviation are listed in Table 1. BTEX emission estimates
using the various techniques agreed reasonably well. Prediction by GRI-
GLYCalc of total BTEX emissions was close to the total capture results for
some of the gas sample types. Quality control data, however,.indicate that
the natural gas BTEXconcentrations for the cylinders may have been biased
3
-------
TABLE 1. SUMMARY OF SITE 1 EMISSION RESULTS*
Emissions (tpy)
Method
Bentene
Toluene
Ethylbenzene
Xylenes
Total BIEX
Total VOC
Total Capture Condensation
1.25 t 0.32
1.68 t 0.29
0.08 t 0.02
0.56 t 0.15
3.58 t 0.61
19.8 t *.0
Pressurized Glycol Cylin-
ders
1.22 t 0.16
1.81 i 0.25
0.08 t 0.01
0.61 i 0.09
3.71 i 0.51
10.7 * 1.9
Atmospheric Rich/Lean Gly-
col
1.2* t 0.20
1.85 i 0.28
0.08 t 0.01
0.62 t 0.10
3.79 » 0.59
11* * 1.8
GRI-GLYCale with Canister
Gas Samples
1.31
1.87
0.06
0.6*
3.88
21.8
GRI-GLYCale vlth Cylinder
Gas Samples Before Absorb-
er. with Manifold
2.50
3.68
0.2*
1.44
7.86
28.2
GRI-GLYCale with Cylinder
Cas Samples Before Absorb-
er, without Manifold
2.25
3.60
0.18
1.36
7.18
28.3
GRI-GLYCale with Cylinder
Gas Samples After Absorber,
with Manifold
1.60
2.29
0.06
1.68
5.71
25.5
CRI-GLYCale with Cylinder
Gas Samples After Absorber,
without Manifold
1.68
2.26
0.06
0.80
*.80
23.7
"Site 1 was a TEG dehydrator treating 3.6 MMSCFD of gas at 86°F and 659 psig; glycol circulation rate was
48.6 gallons per hour.
-------
high, which caused the high prediction by GRI-GLYCalc. Total volatile organic
compound (VOC) emissions from the two lower cost methods did not match well
with the total capture benchmark results; this is consistent with previous
results for systems without flash tanks (1).
Results of the sampling at Site 2 are listed in Table 2. Atmospheric
rich/lean glycol and pressurized glycol emissions results agreed closely with
the total capture results for both BTEX and total VOC. Removal of volatile
components in a flash tank upstream of the glycol sample point eliminates two-
phase gas/liquid flow in glycol lines, thus allowing a more representative
glycol sample. GRI-GLYCalc predictions using natural gas samples taken before
the glycol absorber agreed well with the total capture results for total BTEX
emissions.
For these two test sites, the GRI-GLYCalc model, using natural gas
sampled with evacuated canisters, agreed very well with measured emissions as
measured by the most accurate test method {total capture condensation} for
each site. As shown in Tables 1 and 2, the GRI-GLYCalc estimated emissions of
BTEX and total VOC are within 10 percent or less of the measured emissions.
4
-------
TABLE 2. SUMMARY OF SITE 2 EMISSION RESULTS"
Emissions (tpy)
Method
Benzene
Toluene
Ethylbenzene
Xy1enes
Total BTEX
Total VOC
Total Capture Condensation
6.02 « 104
9.87 t 1.50
0.8* i 0.16
6.1* ± 0.7*
22.9 * 3.2
36.9 i 3.1
Pressurized Glycol Cylin-
ders
6.71 t 0.98
11.1 » l.«
0.98 i 0.18
7.05 . 0.82
25.9 , 3.2
37.9 t 4.9
Atmospheric Rich/Lean Gly-
col.
5.62 t 0.76
9.25 * 0.93
0.80 t 0.12
5.74 t 0.40
21.4 * 2.0
30.8 t 3.4
GRI-GLYCalc vlth Canister
Cms Samples
5.22
8.63
0.89
7. 58
22.3
36.1
GRI-GLYCalc with Cylinder
Cas Samples Before Absorb-
er. vlth Manifold
5.55
8.9*
0.89
6 13
21.5
32.7
GRI-GLYCalc with Cylinder
Gas Samples Before Absorb-
er, without Manifold
5.62
8.51
0.82
5.33
20.3
31.1
GRI-GLYCalc with Cylinder
Gas Samples After Absorber,
vlth Manifold
J. 9J
5.69
0.*1
2.87
12.9
23.3
GRI-GLYCalc with Cylinder
Gas Samples After Absorber,
without Manifold
*.35
6.32
0.49
3.48
14.6
25.3
®Site 2 was a TEG dehydrator treating 4.9 MMSCFD of gas at lll'F and 1063 pslg; glycol circulation rate was
204 gallons per hour, and flash tank conditions were 205®F and 46 psig.
-------
SECTION 2
INTRODUCTION
BACKGROUND
Glycol dehydration units are commonly used in the natural gas industry to
remove water from natural gas streams to prevent corrosion and hydrate
formation in pipelines. Approximately 40,000 glycol dehydration units may be
in operation in the United States (2). A typical glycol dehydration unit, as
shown in Figure 1, consists of an absorber, a flash tank, heat exchanger(s),
filter(s), a glycol reboiler and still, and associated pumping and piping
equipment. Triethylene glycol (TEG) is most commonly used as the dehydrant
since it is more stable than other glycols (e.g., ethylene glycol) at high
temperatures. It can be dried more completely at the higher temperatures,
resulting in lower natural gas dewpoints.
In the absorption step of the dehydration process, aromatic hydrocarbons
such as (BTEX) are also absorbed. Following the absorption step, the glycol,
rich with water and BTEX compounds, is distilled to strip water from the
glycol. Recovered lean (dry) glycol is recycled for use in the absorber.
Emissions of BTEX and other VOCs occur from the glycol reboiler still vent.
Depending upon the volume of gas processed and the BTEX/VOC concentrations
present, emissions can be significant as defined by the 1990 Clean Air Act
Amendments (CAAA) discussed below.
3TEX compounds have been listed as hazardous air pollutants in the 1990
CAAA. According to the CAAA Section 112(a) 'definition, sources emitting over
10 tons per year (tpy) of any single hazardous air pollutant (HAP) or over 25
tpy of any combination of hazardous air pollutants are defined as major
sources subject to regulatory and EPA emission standard setting activities.
6
-------
Dry Natural Gas Still Vent Off-Gas
Burner
Exhaust
Stripping
Gas
Still
ABSORBER
Moist
Natural
Gas
Rcboiler
Natural
I
Gas/Glycol
Heat Exchanger
Gas
Flash Gas to Vent,
Fuel, or Stripping Gas
Surge Tank
Inlet
Separator
Lean Glycol
Rich
Glycol
FLASH
TANK
Glycol
Balance
Pump
To
Tanks
Rich Glycol
Filter
(Carbon
or Sock)
Figure 1. Typical glycol dehydration unit.
-------
Many glycol dehydration units may be classified as major sources based on this
definition. In addition, depending upon their emissions and location, some
glycol dehydration units may qualify as major VOC sources in ozone
nonattainment areas.
To provide a tool for the gas industry to use in estimating emissions
from TEG dehydration units, GRI developed a personal computer (PC) based
program called GRI-GLYCalc. The program is based on a simplified
representation of a typical glycol dehydration unit and is capable of
calculating BTEX and other VOC emission estimates. GRI-GLYCalc requires input
of specific values for various process variables, for example, natural gas
composition, natural gas tenperature, and glycol circulation rate. The
computer program employs vapor-liquid equilibrium models and mass balance
calculations to predict emissions.
Because no reference methods are available for sampling the still vent
stream of glycol dehydrators, GRI, in coordination with a working group of
representatives from several gas industry organizations, has been conducting a
research program to develop and evaluate sampling and analytical methods for
measuring emissions from glycol dehydration units. Data collected from
multiple sites in the sampling and analytical methods study will be used to
further validate GRI-GLYCalc.
The primary objectives of the GRI program have been to develop and
validate sampling and analytical methods for measuring emissions from glycol
dehydrator still vents. An industry working group, on which EPA was included,
was formed to oversee the program. Preliminary field experiments were
conducted to identify the most promising sampling methods. Subsequent field
experiments have centered on verification and validation of these methods
(atmospheric rich/lean glycol sampling, glycol sampling using pressurized
cylinders, and natural gas sampling with use of the GRI-GLYCalc model). GRI,
through the industry working group, funded initial method evaluation tests at
two sites, and tests at six additional sites for validation of the candidate
test methods.
8
-------
The objective of this work assignment is to collect additional emissions
test data at two sites to provide a broader base of field test data for 1}
comparison and validation of sampling and analytical methods; and 2)
comparison and validation of the GRI-GLYCalc computer program and its
subsequent revisions. The results from these two sites alone are not
sufficient to make generalizations about emissions from all glycol
dehydrators. However, when pooled with the information from eight other sites
in the GRI program, broader conclusions can be made and will be included in
the report to GRI.
DATA QUALITY OBJECTIVES
The two data sets generated from this work assignment will become part of
the larger effort presently funded by GRI. As stated previously, one of the
objectives of the GRI program is to determine the significance of the
difference between sample measurement results and estimates from GRI-GLYCalc.
Therefore, the measurement data collected during this program must be of
sufficient quality to determine whether GRI-GLYCalc can be used with
confidence to estimate emissions and to comply with environmental regulations.
To that end, the error in the measurement data collected must be controlled
and well-documented.
At the inception of the GRI program, candidate methods for sampling and
analyzing glycol dehydration unit streams were evaluated based on the general
approach and experimental design described in EPA's protocol for the field
validation of stationary sources (3) and in EPA Method 301, "Field Validation
of Pollutant Measurement Methods from Various Waste Media." (4) This design
featured six test runs with extensive analyte spiking and nested replicates.
After the most feasible sampling and analytical methods were chosen, the
subsequent test designs have followed the same general approach but with fewer
spikes and replicates.
9
-------
The glycol dehydrator sampling methods used in this study are pressurized
rich glycol cylinders and atmospheric rich/lean glycol. Emissions are
calculated by multiplying the difference between pressurized or atmospheric
rich glycol and lean glycol by the glycol circulation rate. These sampling
techniques are non-routine and are under development. One of the objectives
of the GRI program is to establish the bias and precision for the methods.
The sampling methods will be compared with total capture condensation (TCC),
chosen by the GRI industry working group as the benchmark for emission
sampling. TCC is considered the benchmark technique because the entire still
vent stream is passed through a condensation apparatus in which condensed
hydrocarbons, condensed water, and uncondensed gases are measured and sampled.
The data quality objective for this work assignment is to generate data
sets with a known data quality for inclusion in the overall GRI program. As a
result of this work assignment, data sets from ten test sites (with each data
set based on samples from six test runs) will be available for study. The
critical analytes are the BTEX compounds, although other HAPs are also
measured using these techniques.
This project has been defined as a Category II project whose results will
provide supporting data- for the design of rules, regulations, or policies
{e.g., strategic decision-making or standard setting).
REPORT ORGANIZATION
The remainder of this report is organized in the following sections:
Section 3 -- Site Descriptions;
Section 4 -- Sampling and Analytical Matrix Design;
Section 5 -- Sampling Methods;
Section 6 -- Analytical Methods;
Section 7 -- Data Reduction;
Section 8 — Results; and
Section 9 -- Data Quality Indicators for Critical Measurements.
10
-------
SECTION 3
SITE DESCRIPTIONS
This section consists of descriptions of the two sites sampled during the
field evaluation. A brief explanation of the upstream and downstream process-
es, unique process configurations, and sampling locations are included.
SITE 1
The first site was a triethylene glycol dehydrator at a gas plant in west
Texas. A diagram of the dehydrator with labeled sample points is shown in
Figure 2. Site 1 was sampled February 28 to March 1, 1994.
Process Flow
Sour gas (approximately 1000 ppm H2S) from numerous wells passed through
an amine contactor that removed the H2S. The sweet gas then flowed into the
glycol contactor where it countercurrently contacted triethylene glycol. The
dry gas was sent to a gas plant for additional processing. During the initial
site visit in early February, 1994, the dehydrator was processing about 6
million standard cubic feet per day (MMSCFD) of gas. The average gas rate
through the dehydrator during testing was about 3.6 MMSCFD. Natural gas flow
rate increased during the second day of sampling. This site processed gas
from a different set of wells during the first half the month than during the
second half, and the field test occurred as production switched to the new
wells. The average absorber temperature and pressure during sampling were
86°F and 659 psig.
The rich glycol proceeded from the contactor to a Kimray model 21015PV
gas-driven pump. After the rich glycol pressure was reduced through the pump,
it was sent through a sock filter. From the filter, the glycol flowed through
the glycol surge drum where it was heated by the hot lean glycol. The rich
11
-------
Vcnl
Burner
Exhaust
Dry Gas
to Plant
Total Capture
Outlet Natural Gas
Still
Vent
I'reheat
ABSORBER
REBOILER
^ Natural Cas
Swcti Cas
from Amine
Unit
Drain to
Sump
inlet Natiual Gas
Lean Glycol
tRKTOTKRT
SURGETANK
Kimray Rich Glycol Soek
Pump Filter
21015PV
® SAMPLE LOCATIONS
Figure 2. Process flow diagram for Site 1.
-------
glycol then passed through another preheater where it was warmed by vapors in
a closed stack above the reboiler before it proceeded into the still.
Lean glycol from the reboiler flowed to a surge drum located beneath the
reboiler. The lean glycol was then returned to the contactor at the inlet gas
pressure by the glycol pump.
Sampling Locations
The sampling locations for Site 1 are labeled on the process flow diagram
in Figure 2.
Prior to sampling, the still vent vapors flowed to a vertical pipe which
was routed to a sump. The vent was modified by site personnel by the
installation of a tee so that the stream could be diverted to the total
capture sampling apparatus or directed through the vertical pipe.
The rich glycol sampling port was located immediately after the glycol
pump discharge (i.e., on the low-pressure discharge side of the pump).
Atmospheric and pressurized rich glycol samples were taken at the same point
in the glycol line. The lean glycol sampling location was on the low-pressure
suction side of the glycol pump. There were existing valves and sample ports
at both locations, so no modifications were necessary.
Natural gas samples were taken on both the inlet and outlet streams of
the contactor. The natural gas sample ports were in place near ground level.
No modifications to the natural gas lines were necessary.
SITE 2
The second site was a triethylene glycol dehydrator in southwest
Louisiana. A process flow diagram showing sample points is presented in
Figure 3. Site 2 was sampled February 24-25, 1994.
13
-------
Rich Glycol
- ®—
To Fuel
PLASH
TANK
Dry Natural
Gas
ABSORBER
GLVCOL/GAS
HEAT
EXCHANGER
Gas from
Compressor
Outlet Naturj
Gas
Inlet Natural
Ixan Glycol
Rich Glycol
To Condenser
Total Capture
- ®—
STILI.
0~
Uurntr
Exhaust
KKBOILER
Kimray
Pump
45015PV
RICH/I.KAN
GLYCOL
HEAT
EXCHANGER
Lean Glycol
/
Fuel (Iroin
flash Isuik and
condenser)
(X> SAMPLE
LOCATIONS
Figure 3. Process flow diagram for Site 2
-------
Process Flow
The natural gas at Site 2 is associated gas resulting from oil produc-
tion. The gas was collected from numerous wells at low pressure (approximate-
ly 40 psig) and compressed before being sent to the glycol contactor. The dry
gas then passed through an outlet natural gas/lean glycol heat exchanger and
was sent to sales. During the Initial site visit the dehydrator was
processing 7-8 MMSCFD of gas. The average gas flow during testing was 4.9
MMSCFD, Problems with the gas line upstream of the compressor caused the
natural gas flow rate to fluctuate widely during the first day of sampling.
The average contactor temperature and pressure during sampling were 111°F and
1063 psig, respectively.
The rich glycol exited from the bottom of the contactor and passed
through a Kimray model 45015PV gas-driven balance pump in which the pressure
was reduced. From the pump the rich glycol proceeded to a section of piping
that contacted the reboiler still vapors. (This configuration is sometimes
called a "knockback" cooler because it condenses glycol vapors in the still
vent gas.) After the knockback, the rich glycol passed through a concentric
tube rich/lean glycol heat exchanger where it cooled the lean glycol stream
from the reboiler. The rich glycol then flowed to a flash tank where light
hydrocarbons were removed. The flash tank pressure was maintained at 45-50
psig; there was no temperature gauge on the flash tank, but measurement of the
rich glycol with a thermocouple at a sample point immediately after the flash
tank indicated a temperature of approximately 205°F. The flash gas was used
as fuel in the reboiler burner at this site. Before flowing to the reboiler,
the rich glycol passed through a pressure regulator. The piping configuration
between the pressure regulator and the reboiler allowed for a charcoal filter
to be installed; however, the filter was not in place during testing.
Lean glycol exited the bottom of the reboiler and passed through the
rich/lean glycol heat exchanger where it was cooled by the rich glycol. The
lean glycol was pumped to the inlet gas pressure by the Kimray pump, then
passed through an outlet natural gas/lean glycol heat exchanger before it
entered the top of the contactor.
15
-------
The still vent vapors were sent to a forced-draft aerial condenser. The
noneondensable gases were burned as reboiler fuel, while the liquids were
recovered and processed.
Sampling. Locations
The sampling locations for Site 2 are labeled on the process flow diagram
in Figure 3.
Modifications to the still vent were necessary to allow bypass of the
existing condenser. A tee was placed just upstream of the condenser so that
the still vent flow could be routed to the total capture apparatus during
sampling.
The rich glycol samples (atmospheric and pressurized) were taken after
the flash tank from piping that was in place for a charcoal filter. No
modification was necessary to sample the rich glycol.
The low pressure lean glycol sample port chosen before the start of the
test was located immediately downstream of the reboiler before the rich/lean
glycol heat exchanger. As a result of the temperature at this point (300+oF)
and restricted space around the port, the lean glycol sample location was
moved to the high-pressure line. A loop of 0.25 inch sample line immersed in
ice water was used as a cooling coil. Moving the sample location should not
have affected the results because the lean glycol composition should not
change between the low and high pressure points.
Natural gas samples were taken on both the inlet and the outlet gas lines
to the absorber. The sample ports were available near ground level, and no
modifications were necessary.
16
-------
SECTION 4
SAMPLING AND ANALYTICAL MATRIX DESIGN
This section describes the sampling and analytical approach for the
sampling techniques utilized during the field tests. These included:
• Total capture condensation (TCC);
• Pressurized glycol cylinders;
• Atmospheric rich/lean glycol; and
• Natural gas sampling.
Included in this section is a discussion of the number of runs at each site,
duplicate samples, duplicate analyses, and spiked samples.
Six runs were conducted at each site, nominally three runs per day for
two days (Because of time constraints due to process upsets at Site 2, two
runs were performed the first day and four runs were performed the second
day). Samples for a given run for all methods were collected over as short a
time as possible to ensure that the samples reflected the same process
conditions and could be compared on a side-by-side basis. Total capture
required an extended sampling period (typically one hour), and samples for all
the other methods and sampling points were collected while the total capture
sample was taken.
The sampling and analytical approaches for total capture condensation
samples are presented in Figure 4 (hydrocarbon liquids) and in Figure 5
(noncondensable gases). Total capture condensation also produced a liquid
water product, and the wpter from two of the runs at each site was analyzed
for hydrocarbons. Problems with the total capture condensation apparatus
during Run 4 at Site 1 caused the hydrocarbon liquid sample to be lost, so a
seventh run was added.
17
-------
Run
Sample
Analysis
x ^ j J 4
t A A A A
cs d e f g h is
I
s=spiked
Figure 4. Sampling and analytical design for total capture liquid samples.
-------
Run
Sample
Analysis
I 1
3
A
c d
>
es
s=spiked
* Total capture gas canister spiked prior to use at Site 1,
after sampling at Site 2.
Figure 5. Sampling and analytical design for total capture gas samples.
-------
Duplicate analyses were conducted using split samples on Run 3
hydrocarbon liquid fraction. Liquid hydrocarbon samples from Runs 2 and 4
were split and spiked in the laboratory before analysis. (However, since Run
4 was lost at Site 1, these splits were performed on the samples from Runs 2
and 5). A spiked duplicate sample was collected for the Run 3 gases. For
Site 1, the spike was introduced into the container before sample collection,
and at Site 2 the spike was added to the gas after sample collection but prior
to analysis.
Figure 6 presents the sampling and analytical approach for the
pressurized glycol and atmospheric rich/lean glycol testing. This design was
similar to the total capture approach except that field duplicate samples were
taken for Run 3. These duplicate samples were analyzed as nested duplicates.
Samples from Runs 2 and 5 at Site 1 (to match the total capture condensation
spikes) and from Runs 2 and 4 at Site 2 were split in the laboratory and one
aliquot of the split was spiked. The gas obtained from the pressurized glycol
cylinders was handled according to the same sampling and analytical approach
as the total capture noneondensable gas (Figure 5).
The natural gas sampling and analytical design is depicted in Figure 7.
Simultaneous field duplicate samples were collected during Run 3 for all
methods with one of the Run 3 samples analyzed in duplicate. The field
duplicates for the Run 3 high-pressure natural gas samples were both analyzed
by the subcontract laboratories; however, there were no analytical duplicate
analyses or spikes performed on the high-pressure samples. The sampling and
analytical procedures are despribed in more detail in Sections 5 and 6.
20
-------
Run
Sample
Analysis
^ j j ¦ 4
A A A A
b cs d e f g h is
s=spiked
Figure 6. Sampling and analytical design for liquid glycol samples.
-------
Run
N3
N3
Sample
Analysis
1
I t
c111 d111
3 ' <2>
t
5
es
1
h
s=spiked
(1) Subcontract laboratories did not analyze in duplicate.
(2) Natural gas canister spiked prior to use in field; cylinders were not spiked.
Figure 7. Sampling and analytical design for natural gas samples.
-------
SECTION 5
SAMPLING METHODS
This section describes the sampling methods used and process data
collected during the field testing.
FIVE SAMPLING METHODS
The sampling methods explained below were used to collect one sample for
each run except when field duplicates were collected as specified in Section
4.
Total Capture Condensation
The total capture condensation (TCC) apparatus was set up as shown in
Figure 8. The TCC apparatus consists of a 50-foot length of one-inch inside
diameter copper tubing coiled inside a 55-gallon barrel. The inlet to the
coil was connected to the still vent outlet piping using a short piece of 2-
inch Tygon tubing. The barrel was filled with an ice/water mixture during
sampling to condense steam and organics in the still vent stream.
Condensed liquids were separated from the vapors in a gravity separator
(liquid knockout). The liquids were collected in one-gallon glass bottles and
allowed to partition into organic and aqueous phases. The total volume of
each phase was measured with a two-liter graduated cylinder. Two 40 ml
Volatile Organic Analysis (VOA) vials were filled to overflowing with organic
liquid and capped for shipment to the laboratory. A separatory funnel was
used to separate the aqueous phase from the remaining hydrocarbon liquid and
to obtain two 40 ml VOA vials of aqueous phase sample. The vials were placed
in a cooler with ice for transport to the laboratory.
23
-------
Ice/lee Water
Knock-
out
Dry
Gss
Meter
T
Water/
Hydrocarbon
Sample
Gas
Sample
Figure 8, Total capture condensation sampling apparatus.
-------
The volume of liquid hydrocarbon collected during total capture sampling
at Site 1 was less than the volumes observed at most of the other sites in the
GRI program. During the second day of sampling, it was decided a more
accurate means of measuring the hydrocarbon liquid was to determine the mass
with a balance. This procedure was initiated because of the error associated
with measuring a small volume of liquid in the two-liter graduated cylinder
normally used in the field. The volume of hydrocarbon collected was
calculated after the density was determined gravimetrically in the laboratory.
The total noncondensable gas (NCG) volume was measured with a calibrated
dry gas meter. A portion of the NCG was collected in evacuated 2.8 L SUMMA™
polished canisters for analysis in the laboratory. The NCG samples were
collected over approximately a fifteen minute period during each run. Flow
was regulated to the canisters using a needle valve to achieve a final
pressure of -7" Hg gauge.
Total capture samples for each run nominally were collected over a one-
hour period. The cooling coil and knockout were not cleaned between runs but
were purged for at least thirty-five minutes before beginning another run.
Pressurized Glycol Cylinders
Pressurized glycol samples were collected in conjunction with the
corresponding TCC run. Sample port locations for each site were shown
previously in Figures 2 and 3. The sample valve and the line were allowed to
purge for at least one minute before sample collection. The valve was closed
and a pre-weighed, evacuated stainless steel cylinder was connected to the
port. The sample valve and then the cylinder inlet valve were opened to fill
the cylinder. The cylinder was allowed to equilibrate at line pressure for at
least seven minutes before the cylinder inlet valve was closed. The cylinder
was removed from the sample port and capped for transport to the laboratory.
Filled cylinders were stored at ambient conditions.
The gas and glycol fractions in the cylinder were separated in- the
laboratory before analysis. This procedure is explained in Section 6.
25
-------
Atmospheric Rich/Lean Glycol
Rich Glycol Samples--
Atmospheric rich glycol samples were collected at the same port as the
pressurized glycol samples. To minimize the loss of volatiles from the
atmospheric samples, the glycol was cooled by passing it through a coil
immersed in an ice-water bath. Rich glycol typically foams when reduced to
atmospheric pressure due to the release of volatile components). The sample
valve was opened, and the line was purged for at least one minute before
sampling. Without adjusting the flow, duplicate samples were collected in 40
mL VOA vials. The vials were filled to overflowing with liquid and foam then
were immediately tightly capped. The foam layer was not allowed to break, and
the vial was not refilled with liquid. Therefore, a headspace developed above
the liquid in the vial as the foam dissipated. Foaming was minimal at Site 2,
so there was no observable headspace. Vials were stored in a cooler with ice
for transport.
The cooling coil was not used for Run 1 and Runs 4 through 7 at Site 1
due to the high viscosity of the glycol. It was decided that the decreased
collection time (i.e., higher flow) achieved by not cooling the glycol was
more crucial to retaining volatile components than the cooling effect. The
rich glycol temperature increased significantly during Runs 2 and 3 (>85°F),
so the cooling coil was used.
Lean Glycol Samples--
Atmospheric lean glycol samples were collected at the points shown in
Figures 2 and 3. Sampling on the low pressure side of the glycol pump was not
feasible at Site 2, so samples were collected on the high-pressure side. The
sample valve was opened, and the line was purged for at least one minute
before sampling. Without adjusting the flow, duplicate samples were collected
in 40 mL VOA vials. The vials were filled to overflowing then tightly capped
and stored in a cooler with ice for transport.
26
-------
Natural Gas Samples
Five natural gas samples were collected for each run. The approaches
used for each sample are summarized as follows:
• Modified EPA Compendium Method TO-14 (5) before the glycol absorber
with Gas Processors Association (GPA) Method 2166 (6) sampling
manifold;
• GPA Method 2166 without probe before and after the absorber; and
• GPA Method 2166 without sampling manifold and probe before and after
the absorber.
The modified EPA TO-14 apparatus is shown in Figure 9. EPA Compendium
Method TO-14 is typically used for time-integrated ambient air sampling in
SUMMA polished stainless steel canisters. The sampling time for Method TO-14
has been reduced from twenty-four hours to fifteen minutes for use in natural
gas sampling. The GPA Method 2166 sampling manifold was used to remove any
liquid droplets that may have been present in the line. The bulk of the gas
stream was vented continuously to the atmosphere. After the manifold was
purged for at least two minutes, a slipstream of the flowing gas was diverted
to an evacuated 2.8 L SUMMA polished canister through a 7 micron filter and a
Veriflow flow regulator. The regulator was set so that the canister would be
filled to a pressure of approximately -7" Hg gauge pressure over a 15 minute
period. Initial and final canister pressures were recorded in the field.
Actual fill times ranged from 14 to 17 minutes, and final gauge pressures
ranged from -5.5" Hg to -7" Hg.
The equipment used in GPA Method 2166 is depicted in Figure 10. GPA
Method 2166 specifies the use of a sampling probe inserted into the gas line
to collect the sample in the center of the pipe and away from the walls where
liquids may accumulate. The probe could not be used at either site due to the
configuration of the existing sample ports. Samples were collected using the
manifold shown in Figure 10 both upstream and downstream of the glycol
absorber. The sampling manifold was purged, the vent was closed, and the
cylinder was rapidly filled to line pressure, and the final pressure in the
27
-------
Vacuum
Gauge
Flow
Regulator
1J U
7jim Filter
Vent
Evacuated
Canister
Sample
Port
Liquid Drain
Natural Gas
Line
Figure 9. EPA Method TO-14 natural gas sampling apparatus
-------
High Pressure
Sample Cylinder
Vent
fO
VD
Sample
Port
Liquid Drain
Natural Gas Line
Figure 10. GPA Method 2166 natural gas sampling apparatus.
-------
cylinder was recorded. Samples collected without the manifold were connected
directly to the purged sample port and filled to line pressure.
Evacuated cylinders were used at Site 1. Cylinders initially filled to
approximately 5 psig with helium were utilized at Site 2. After the sample
was collected at high pressure, the cylinder valves and fittings were immersed
in water to check for leaks. If a leak could not be repaired by tightening
the fittings, the cylinder sample was voided, and the sample was retaken.
Natural gas samples were stored and transported at ambient conditions.
The high pressure cylinders were shipped to the laboratories following IATA
shipping regulations for compressed natural gas.
PROCESS DATA MEASUREMENT AND COLLECTION
The process parameters listed in Table 3 were monitored and recorded
during the field testing. Natural gas (i.e., absorber) temperature and
pressure, flash tank temperature and pressure, and reboiler temperature were
taken from gauges on the vessels. Natural gas flow rate at Site 1 was
recorded at the adjacent gas plant. A printout listing the flow rate during
each hour was obtained at the end of testing. At Site 2, the dry gas flow
rate was acquired from a computer screen located in a building on site.
The glycol circulation rate was determined by counting pump strokes for a
set period of time and converting the stroke rate to a flow rate using the
manufacturer's specifications. This flow rate determined from the pump stroke
rate was used in all calculations requiring glycol circulation rate. A
measurement was taken to check the flow rate by shutting off the rich glycol
line to the reboiler and collecting the glycol in a covered container. The
liquid collected for a specific period of time was weighed and converted to a
volumetric flow rate. One such flow confirmation was conducted at Site 1, and
two flow measurements were performed at Site 2. Rich and lean glycol water
content was determined for each run by Karl Fischer analysis.
30
-------
TABLE 3. DEHYDRATOR FIELD TESTING PROCESS PARAMETERS
Process Variable Units
Absorber Temperature "F
Absorber Pressure Ps*-g
Natural Gas Flow Rate MMSCFD
Glycol Circulation Rate Gallons pe'r hour
Flash Tank Temperature (if "F
applicable)
Flash Tank Pressure (if applicable) psig
Dry Gas Water Content (if available) lb H20/MMSCF gas
Lean Glycol Water Content wt % Hz0
The dry gas water content was not measured at Site 1. At Site 2, the dry
gas water content was monitored by an in-line analyzer. Because the analyzer
was battery operated and required 15-30 minutes for equilibration, readings
were taken periodically.
31
-------
SECTION 6
ANALYTICAL METHODS
This section describes the analytical methods used for the samples
collected during the field testing. Sample preparation and equipment set-up
are explained in detail.
PRESSURIZED GLYCOL CYLINDER HANDLING
The total volume of the gas and the liquid inside the pressurized
cylinders was measured, equilibrated, and collected using a water displacement
apparatus shown in Figure 11.
Initially, the vent to the gas sample vessel was closed and the valve on
the sample cylinder was opened, which allowed the gas to escape from the
cylinder into the one-liter burette. The separatory funnel was lowered below
the height of the liquid In the burette in order to pull a slight vacuum on
the system. The volume of the gas released from the cylinder is equal to the
measured volume of water displaced in the burette plus any gas remaining in
the cylinder. The volume of water displaced was measured after the height of
the liquid level in both the burette and the separatory funnel had been
adjusted to the same level. After a one minute period, the vent to the gas
sample container was opened, and the gas from the burette was displaced into
the container. The procedure was repeated until no gas was released from the
cylinder, and the volume of gas remaining in the cylinder was calculated based
on the glycol volume in the cylinder and the cylinder size. A sample of the
glycol solution was collected in a VOA vial from the bottom valve on the
sample cylinder.
Typically the gas obtained from the glycol pressurized cylinders is
collected in Tedlar bags for analysis. Due to the unusually low pressure in
the cylinders, all but one of the gas samples from Site 1 were collected in
-------
Vent to Tedlar for
Gas Capture
Initially Filled with Water
3-Way Valve
T3
T3
•a
2L Separatory
Funnel
Glycol Drain
Figure 11. Gas/glycol separation apparatus.
33
-------
evacuated canisters similar to those used for the total capture noncondensable
gas. (The vacuum in the evacuated canisters allows for the collection of gas
samples at or below atmospheric pressure, while Tedlar bag collection requires
a source above atmospheric pressure.) There was no measurable gas phase in
the cylinders from Site 2; consequently, gas samples were not collected. This
was consistent with the absence of foam in the atmospheric rich glycol samples
collected at Site 2.
ANALYTICAL TECHNIQUES
Liquid Hydrocarbon and Aqueous Samples
Gas chromatographic systems with flame ionization detectors were used to
quantitate the BTEX target analytes as well as n-hexane, 2,2,4-trimethyl-
pentane, C5 to C10 normal alkanes, cyclohexane, and total hydrocarbons. A
Hewlett-Packard 5880 gas chromatograph equipped with a split/splitless
injector and a capillary column was used to analyze the liquid samples.
Quantitation of the target and non-target analytes was based on individual
response factors (see calibration procedures at the end of this section). The
n-hexane response factor was used to quantify the unidentified hydrocarbons.
Hydrocarbon samples collected from the total capture method were diluted
1 part in 100 in methylene chloride prior to analysis. The samples from Run 3
at both sites were analyzed in duplicate. Duplicate analysis results were
obtained by doing two independent dilutions of the same sample.
Samples 2 and 5 from Site 1 and samples 2 and 4 from Site 2 were used to
generate spiked samples. Sample spiking was done during the sample dilution
process. Spiking levels were determined by first analyzing the samples to
determine the concentrations of the target analytes. A subset of the analytes
(hexane, octane, benzene, and toluene) were spiked into the samples at levels
that approximately doubled their concentrations. All of the spiked samples
were analyzed using the conditions found in Table 4.
34
-------
TABLE 4. ANALYTICAL CONDITIONS FOR THE ANALYSIS OF LIQUID SAMPLES
Instrument
Detector
Injection System
Column
Hewlett-Packard 5880
FID
Capillary; Split 50 to 1
60 m x 0.32 mm id
3.0 pm film Restec RTx-1
Helium @ 20 psig
-3 mL/min
Helium @ 30 mL/min
40°C for 2 min; programmed at 8°C/min to 225°C,
final hold 10 min
275°C
300°C
L/iL
Carrier Gas
Carrier Gas Flow Rate
Make-up Gas
Temperature Program
Injector Temperature
Detector Temperature
Injection Volume
The density of the hydrocarbon samples was determined gravimetrically
after the samples had been analyzed. The sample was weighed in either a 5 or
10 mL volumetric flask and then the density was reported in units of grams per
milliliter.
Two aqueous samples from each site were analyzed. Analytical duplicates
and spikes were not performed on these samples. Eighteen milliliters of
aqueous samples was extracted with 3 mL of dichloromethane. The extracted
organic phase was analyzed using the conditions in Table 4.
Both hydrocarbon and aqueous samples collected in VOA vials were analyzed
within 14 days of sampling.
Glycol Samples
The glycol samples from the stainless steel cylinders were transferred to
VOA vials prior to analysis. Analysis of these glycol samples and the
atmospheric glycol samples was accomplished by diluting the glycol in a 1:1
ratio with methylene chloride and using the same analytical conditions as for
the liquid hydrocarbon samples (Table 4). The identified analytes are the
same as for the liquid hydrocarbons and aqueous samples. Analytical
35
-------
duplicates and spikes were done as stated for the hydrocarbon samples. All of
the pressurized glycol samples were analyzed within 14 days of the laboratory
separation, while all of the atmospheric samples were analyzed within 14 days
of collection.
The atmospheric rich and lean glycol samples were analyzed for water
content by an subcontracting laboratory. The procedure for determining water
content was ASTM Method D 1744-83, Karl Fischer analysis.
Gas-Phase Samples
The canister samples (from natural gas, pressurized glycol cylinders, and
total capture noncondensable gases) were analyzed using the conditions found
in Tables 5 and 6.
Table 5 lists the analytical conditions for hydrocarbons analysis. The
target analytes were BTEX, n-hexane, C2-C6 alkanes, n-heptane, and 2,2,4-
trimethylpentane. Concentrations of ethane, propane, n-butane, n-pentane, n-
hexane, benzene, toluene, and ethylbenzene were calculated using compound-
specific response factors. Xylene constituent concentrations were calculated
with the o-xylene response factor. Concentrations of all other compounds,
speciated and unidentified, were based on the n-hexane response factor. Table
6 shows the conditions for fixed gas analysis, which includes methane, carbon
dioxide, nitrogen, and oxygen. The gas samples were analyzed within 21 days
of sampling.
Designated canisters were spiked prior to being sent to the field at Site
1. At Site 2 spiked canisters were not available in the field so these
canisters were spiked in the laboratory after sampling. The gas samples were
spiked with a known volume of cis-2-butene. The exact concentration of cis-2-
butene in the sample varied with the sample volume collected. The target
concentration of cis-2-butene was between 100-300 ppmv.
36
-------
TABLE 5. ANALYTICAL CONDITIONS FOR THE ANALYSIS OF HYDROCARBON GAS SAMPLES
Instrument
Data System
Detectors
Injection System
Column
Carrier Gas
Carrier Flow Rate
Make-up Gas
Temperature Program
Cryogenic Trap Temperature
Trap Desorption Temperature
Hewlett-Packard 5880 or Varian 3700
Nelson 2600 Series with 900 Series
A/D Converter
FID and PID
Cryogenic Focusing/Thermal Desorption
60 m x 0.32 mm id 1 /im film J&W
DB-1 fused silica capillary
Helium @ 16 psig
-2 mL/min
Nitrogen @ 30 mL/min
-50°C for 2 min; programmed at 6®C/min to 150°C,
then 150°C to 250°C at 20°C/min; no final hold
-186°C
180°C
TABLE 6. ANALYTICAL CONDITIONS FOR ANALYSIS OF FIXED GASES
Instrument
Data System
Detector
Injection System
Columns
Carrier Gas
Carrier Gas Flow Rate
Temperature Program
Hewlett-Packard 5710A
HP 3392A Integrator
Thermal Conductivity (TCD)
Fixed loop 1.0 mL
1. 5 ft x 0.25 in o.d. glass packed with
Molecular Sieve 5A 80/100 mesh
2. 7 ft x 0.25 in o.d. glass packed with
Porapak Q 80/100 mesh
Helium
40 mL/min (both columns)
50°C Isothermal
The high-pressure natural gas cylinders were sent to subcontract
laboratories for analysis. Analytical conditions for the subcontract
laboratories are given in Tables 7 and 8 (7). GPA Method 2261 was used to
quantify nitrogen, oxygen, carbon dioxide, and Cx through C7+ hydrocarbons.
GPA Method 2286 was used for C5 through C12+ hydrocarbons and BTEX. The
compounds detected by both methods (C5 and C6) were used to check the
consistency of the two methods.
37
-------
TABLE 7. ANALYTICAL CONDITIONS FOR SUBCONTRACT LABORATORIES--GFA 2261
Laboratory 1 Laboratory 2
Reference Method
Sample Container
Sample Preparation
Separation Column
Detection Method
Reporting Limit
GPA 2261
Positive-pressure
cylinders
Heat to 140®F
Molesieve 13x/
Silicon oil
DC200-500
GC-TCD
0.01%
GPA 2261
Positive-pressure
cylinders
Heat to 140®F or to
50°F above sampling
Temperature
Chromosorb 200-500/
30% Silicon oil
GC-TCD
0.02%
TABLE 8. ANALYTICAL CONDITIONS FOR SUBCONTRACT LABORATORIES--GPA 2286
Laboratory 1 Laboratory 2
Reference Method
Sample Container
Sample Preparation
Separation Column
Detection Method
Reporting Limit
GPA 2286
Positive-pressure
cylinders
Heat to 1409F
Methyl siloxane
50 m X 0.32 mm ID
2 ym film thickness
GC-FID
0.0001% (1 ppmv)
GPA 2286
Positive-pressure
cylinders
Heat to 140°F or to
50°F above sampling
temperature
Methyl siloxane
100 m X 0.21 mm ID
5 pm film thickness
GC-TCD
0.001% (10 ppmv)
38
-------
CALIBRATION PROCEDURES
Liquid Samples
The target analytes were benzene, toluene, ethylbenzene, and m-,p-, and
o-xylene. In addition, n-pentane through n-decane, cyclohexane, and 2,2,4-
trimethylpentane were calibrated and speciated but were not treated as target
analytes. The non-target analytes were not subject to the same quality
control checks. All other hydrocarbons were reported as unidentified VOC.
Procedures - -
For each target analyte, calibration standards were prepared at a minimum
of five concentrations by adding volumes of stock standards to a volumetric
flask and diluting to volume with methylene chloride. The calibration
standards bracketed the expected range of concentrations found in real
samples. A calibration factor (CF), defined as the ratio of the response to
the amount injected, was calculated for each identified analyte at each
standard concentration. The linearity of the calibration curve was considered
acceptable if the correlation coefficient for the least square regression was
greater than or equal to 0.995. The validity of the calibration standards was
verified by injection of an independent (second-source) mid-level standard
once following each multi-point calibration using a subset of the target
analytes.
Analytical Reference Standards--
The primary standards containing the target analytes for gas
chromatographic quantitation were prepared from a stock solution and then
checked against a secondary standard purchased from Supelco. Dichloromethane
was the diluent and a blank chromatogram was obtained to determine impurities
inherent to the solvent. Each individual analyte was purchased from Aldrich
with at least 99.0% purity.
Preparation of Stock Solution--
The stock solution was prepared as follows: Dichloromethane
(approximately 20 mL) was added to a volumetric flask (50 mL). Next, each
39
-------
analyte (lOQpL each of BTEX and the other organics) was added to the
volumetric flask with a Hamilton gas tight syringe (25QjjL). The analytes were
added in the order of descending molecular weight, with the tip of the syringe
below the surface of the solvent. After addition of the analytes the solution
was diluted to volume with dichloromethane, and the flask was capped and
shaken. The quality of the stock solution was verified by comparison of the
standards made from the stock solution with a second reference standard (see
below).
Preparation and Analysis of Standards--
Five standards, each with a different concentration, were prepared by
removing (with a gas-tight syringe) an aliquot from the stock solution and
injecting the aliquot below the surface of the dichloromethane In a volumetric
flask. The flask contents were diluted to volume, capped, and shaken to
ensure mixing. Each primary standard was injected into a GC-FID, and a
calibration factor was calculated for each analyte based on total peak area
per mass injected. A second source standard purchased from Supelco containing
BTEX was injected into the GC-FID, and a calibration factor was determined for
the secondary standard.
Quality Control Checks--
The working calibration factor was verified by the injection of a mid-
level calibration standard at the beginning and end of each working day or
analysis sequence. If the response for any target analyte varied from the
predicted response by more than ±7%, a second analysis was conducted of the
calibration standard. If two calibration checks in a row varied from the
predicted response by more than 7%, a new calibration factor (or curve) was
prepared. A complete listing of the daily calibration checks is contained in
Appendix B.
Gas Samples
Different calibration procedures and quality control checks are used for
the canister samples based on modified EPA TO-14 and the high pressure natural
gas cylinders. Each is discussed separately below.
40
-------
Modified EPA TO-14 Canister: Non-Methane Hydrocarbon--
The BTEX compounds and n-hexane were the target analytes that were
subject to the quality control requirements, but other compounds, including
C2-C6 hydrocarbons, were also identified. A GC with a flame ionization
detector (FID) was used to analyze non-methane hydrocarbons; a photoionization
detector (PID) was used to validate the FID results.
Calibration standards containing BTEX and n-hexane were prepared and
analyzed at three different levels. Calibration standards were made from
concentrated stock gas cylinders. An aliquot of stock gas was placed in an
evacuated canister and diluted with ultra high-purity helium. The size of the
stock gas aliquot and final dilution pressure were varied to produce the
desired final concentration; the calibration standards bracketed the range of
responses expected during the sample analysis. The calibration standard
canisters were handled in the same manner as the sample canisters (i.e., same
analysis routine and mode of sample introduction). Each standard was compared
to the previous standard to ensure proper preparation.
Two types of slope response factors were derived from the multipoint
calibration -- individual response factors for n-hexane and BTEX (p-xylene
response factor was used to calculate concentrations for each xylene
identified) and a global n-hexane-based carbon response factor for all
remaining non-target compounds. An external standard quantitation method was
used. The acceptance criteria for the slope calibration factors was a
correlation coefficient of greater than 0.995.
The instrument calibration was checked daily prior to sample analysis
using the midpoint calibration standard. If the daily calibration response
did not agree with the multipoint slope within 15%, corrective action was
taken to verify the instrument operation, and new calibrations were performed.
A method blank sample containing helium was analyzed immediately after each
calibration check standard to verify system cleanliness. Total non-methane
hydrocarbons for the blank were limited to carryover of less than 5% from the
standard. Results above this level required system maintenance before
analyses resumed.
41
-------
Modified EPA TO-14 Canister: Fixed Gas Analysis by TCD--
A TCD calibration standard was prepared and analyzed at three levels.
These standards contained percent levels of nitrogen, carbon dioxide, oxygen,
and methane at varying concentrations in a helium matrix. A multipoint slope
response factor was calculated for each of these analytes using linear
regression through a zero point. These response factors were used for
quantitation, and the correlation coefficient for each regression was greater
than 0.995. Corrective action for any analyte failing to meet this criteria
was 1) re-analysis if an assignable cause was easily determined; or 2)
instrument maintenance and recalibration as warranted.
The instrument calibration was checked daily using a calibration standard
prior to analysis. If the daily calibration response did not agree within 15%
for the target analytes, corrective action was taken to verify the instrument
operation. New calibration curves were performed as necessary.
A method blank canister containing helium was analyzed after the
calibration check standard to verify system cleanliness. Observance of target
analytes at levels > 10% of the lowest calibration point indicated maintenance
was necessary. The lowest calibration point for oxygen was 0.99%; for
nitrogen, 8.97%; for carbon dioxide, 1.98%; and for methane, 1.98 percent.
Procedures for High Pressure Samples
High pressure natural gas samples were sent to an outside laboratory for
analysis according to GPA Methods 2261 and 2286 (see Appendix A). The
subcontract laboratories have performed similar analyses during previous
sampling in the GRI program. An extensive quality assurance effort was
conducted to validate the data produced from these previous sites. The QA
effort included a round-robin sample analysis and the analysis of performance
evaluation samples. Comparison of the the performance evaluation samples
indicated that the various analytical methods produce equivalent results.
In this work assignment, quality assurance procedures for the subcontract
laboratories included field duplicate samples, an audit (quality assurance)
42
-------
sample prepared by Radian to be analyzed as a field sample, and shipment of
one of the field samples analyzed by each subcontractor to Radian for a second
analysis and comparison of results. A protocol specification detailing the
analytical procedures to be used, target analytes, and data quality
specifications was provided to each subcontractor prior to sample analysis.
43
-------
SECTION 7
DATA REDUCTION
The data generated in this project were used to calculate emissions of
BTEX and VOC from the glycol dehydration units. Figure 12 summarizes the data
reduction and validation for the total capture, pressurized glycol, and
atmospheric glycol methods, while Figure 13 summarizes the data reduction and
validation for the GRI-GLYCalc model. The reduction, validation, and
reporting of the data are discussed separately below.
EMISSIONS CALCULATIONS
Emission calculations for total capture condensation, pressurized glycol,
and atmospheric rich/lean glycol are presented below, followed by the
calculations for the data summaries. The data summary calculations also apply
to the process and analytical data used in GRI-GLYCalc.
Total Capture Condensation
Field data required: Volume hydrocarbons (HC), liters (L)
Volume water, L
Volume noncondensable gas, standard cubic
feet (SCF)
Sample collection period, min
Analytical data required: Analyte concentrations in HC, mg/L
Analyte concentrations in water, mg/L
Analyte concentrations in noncondensable
gas, ppmv
Hydrocarbon density, mg/L (determined
gravimetrically in the laboratory)
44
-------
Field
Data
Analytical
Results
\
Summary of
Results
Review for
Outliers and
Data Quality
Final
Emission
Estimates
Emission
Calculations
for Each Run
Figure 12. Data reduction and validation approach for total capture,
pressurized glycol, and atmospheric glycol methods.
45
I
-------
Summary of
Field
Data
Review for
Outliers and
Data Quality
Review for
Outliers and
Data Quality
GRI-DEHY
Model
Summary of
Analytical
Data
Final
Emission
Estimates
Analytical
Results
Field
Data
Figure 13. Data reduction and validation approach for GRI-GLYCalc.
46
-------
Compounds of interest were BTEX and VOC.
Duplicates:
Results from duplicate analyses were
averaged to determine a single value for
the sample concentration. Duplicate sample
results were averaged to report a single
value for each run.
Emissions Calculations:
Step 1. Calculate flow (in lb/hr) for analytes in hydrocarbon and
noncondensable gas phases as below,
a. For specific compounds:
Compound in HC, lb/hr
analyte concentration,
r L
(volume HC, L) x
sample collection period, min
X
'
60 min
x
1 lb
hr
453,600 mg
compound in gas phase, lb/hr
analyte concentration, ppmv
106
(volume gas, SCF) x
sample collection period, Qin
1 lb mole compound
x
^ lb compound
x
' >
60 min
379 SCF
lb mole compound
hr
47
-------
where: 379 SCF - volume of one lb-mole of gas at 60°F and 1 atmosphere
pressure.
MW - molecular weight.
b. For total VOC:
For HC phase, the HC density in mg/L was substituted for the
analyte concentration.
In the gas phase, total VOC was determined by summing emissions
of all C3+ compounds and unidentified VOC.
Unidentified VOC, ppmv -
TNMHC, ppmv - sum of all identified C2+ compounds, ppmv
Unidentified VOC was assumed to have a molecular weight of 86
(i.e., hexane).
Step 2. Aqueous phase emissions were calculated using the same equation
as for the hydrocarbon phase in Step 1.
Aqueous phase hydrocarbons were not used in emission
calculations if they accounted for less than 3% of the combined
hydrocarbon phase and noncondensable gas phase hydrocarbons
(see Section 8).
Step 3. Total emissions in tons per year (tpy) were calculated for BTEX
and VOC.
46
-------
Compound-specific or total VOC emissions, tpy =
(HC phase, lb/hr + noncondensable gas phase, lb/hr) x
1 ton
x
» ^
8,760 hrs
2,000 lbs
1 yr
Pressurized Glycol
Field data required: Average glycol flow rate, gallons/hour
(gph)
Analytical data required: Analyte concentrations in pressurized rich
glycol, mg/L
Analyte concentrations in atmospheric lean
glycol, mg/L
Analyte concentrations in gas, ppmv
Volume of glycol in sample cylinder, mL
Volume of gas in sample cylinder at
laboratory conditions, mL
Laboratory ambient temperature, K
Duplicates: Results from duplicate analyses were
averaged to determine a single value for
the sample concentration. Duplicate sample
results were averaged to report a single
value for each run.
Emissions Calculations:
Step 1. Mass (in mg) of analytes present in glycol phase was
determined.
49
-------
Mass of compound, mg In glycol phase = (Analyte concentration mg/L) x
(volume of glycol in cylinder, mL) x
1 L
1,000 mL
Step 2. Mass (in mg) of analyte in gas phase was calculated.
Mass of compound, mg in gas phase «
Analyte concentration ppmv
10"
(volume of gas collected at laboratory conditions mL) x
273 K
laboratory temperature K
1 mmole gas
22.4 mL gas
mg
mmole
Step 3. Masses of analyte from glycol and gas phases were combined to
determine emissions in tons per year (tpy).
50
-------
Compound emissions, tpy
mass compound in glycol, mg + mass compound in gas phase, mg
volume of glycol, mL
(avg. glycol flow rate, gph) x
3,785.6 mL
1 gallon
1 lb
453,600 mg
1 ton
2,000 lbs
8,760 hr
1 yr
Atmospheric Rich/Lean Glycol
Field data required:
Average glycol flow rate, gph
Analytical data required:
Analyte concentrations in rich glycol, mg/L
Analyte concentrations In lean glycol, mg/L
Duplicates:
Results from duplicate analyses were
averaged to determine a single value for
the sample concentration. Duplicate sample
results were averaged to report a single
value for each run.
Emissions Calculations:
Compound emissions were determined directly from the concentrations
in rich and lean glycol and the glycol flow rate.
51
-------
Compound emissions, tpy -
(Analyte concentration in rich glycol, mg/L -
analyte concentration in lean glycol, mg/L) x
(avg. glycol flow gph) x
r
3.7856 L
x
*
1 lb
1 gallon
453,600 mg
1 ton
2,000 lbs
8 ,760 hrs
1 yr
Natural Gas
The natural gas results generated from all five sets of data were handled
according to a four step process:
• Data (reported in ppmv or mole percent) were normalized to remove
any air contamination from the samples;
• Results were reviewed for outliers and data quality (i.e., QA/QC
data review);
• Mean results and RSDs were calculated for natural gas composition
and process parameters that were input to GRI-GLYCalc using the
equations in the data summary section below; and
• GRI-GLYCalc was run for each of the five natural gas data sets for a
particular site. Result were in tons/year.
Data Summaries
The following procedure was used for both emissions calculations (all
sampling methods) and data input to the GRI-GLYCalc model (including natural
gas composition and process operation),
52
-------
Step 1. The average over all runs was calculated.
N
i-l
N-
Duplicate analyses and duplicate samples were handled as
described previously.
Step 2. The sample standard deviation was calculated.
SD =
N
E (xi -x)2
i =1
N - 1
Step 3. The relative standard deviation was determined.
RSD = _ x 100%
Step 4. Data were reviewed and validated (see validation section
below).
Step 5. Steps 1-3 were repeated, omitting any data that did not pass
the review and validation step.
53
-------
DATA REVIEW AND VALIDATION
Suspected outliers in a data set were tested using the Dixon criteria for
test of extreme observation (8). Once identified, these outliers were
evaluated to determine the cause or condition, if possible. As appropriate,
outliers were eliminated during the data review to preclude incorrect
conclusions from being drawn from the data set.
GLYCOL FLOW RATE
Glycol circulation rate calculated from the glycol pump stroke rate was
checked by collecting glycol for a specified period of time and determining
the mass. The measured glycol circulation rate was calculated using the
following equation:
Glycol Circulation Rate, gph «
Glycol mass, g mL
Collection period, min
1.12 g
f *
1 gallon
' *
60 min
3785.6 mL
1 hr
where: 1.12 g/mL
TEG Density
54
-------
SECTION 8
RESULTS
This section presents the results of the field testing at Site 1 and Site
2, respectively. Average results over all runs are reported with the
corresponding standard deviations.
SITE 1
Individual run and analyte data determined for Site 1 are given in
Appendix C.
Process Data
Table 9 lists the process data readings taken during Site 1 sampling.
The reported glycol circulation rate was determined by recording the pump
stroke rate and converting to gallons/hour using the manufacturer's pump
curve. Much of the process variability associated with the natural gas
properties (flow rate, temperature, and pressure) appeared during the second
day of sampling when new gas wells were being brought on-line. The glycol
circulation rate also fluctuated significantly on the second day. The glycol
pump was set below the minimum stroke rate recommended by the manufacturer; at
sites tested previously in the GRI program, low pump rate settings cause the
pump operation to be unstable.
Independent measurement of the glycol circulation rate was conducted by
diverting the glycol into a container for a specified period of time and
measuring the mass of the collected liquid. The measured value was
approximately 25% lower than the circulation rate determined from the pump
curve. However, because the glycol passed through a 0.25-inch sample line to
atmospheric pressure during the measurement instead of the approximately one-
inch process line pumping against some higher head pressure, the measurement
55
-------
TABLE 9. SITE 1 - PROCESS PARAMETERS
Relative
Number
Parameter
Average
Standard
Deviation
(%)
Maximum
Minimum
o£ Data
Points
Gas Flow Rate, MMSCFD
3 .64
22 . 8
6.68
2.87
33
Gas Temperature, °F
86
11.4
99
68
15
Gas Pressure, psig
559
2.4
680
640
15
Glycol Circulation
Rate, gph
48.6
26.8
76.6
25.3
15
Dry Gas H20 Content"'
lb/MMSCF
7.0
N/A
N/A
N/A
N/A
Lean Glycol Water
Content, wt %
1.06
24.3
1.44
0.71
5
Gas Pump Volume
Ratiob, acfm/gpm
0.0288
N/A
N/A
K/A
N/A
Reboiler Temperature,
°F
359
1.1
365
353
14
a Not recorded during the field test but required as input to GRI-GLYCalc; value is typical gas
specification used as default in GRI-GLYCalc.
b Taken from puirp ir.anufacturer' s specifications.
was not fully representative of the process conditions. It was decided that
the measured circulation rate was similar enough to the punp curve to use the
pump curve rate in emission calculations and in GRI-GLYCalc.
Sampling
Total Capture Condensation--
The average emissions calculated for the six total capture runs are
listed in Table 10. Seven runs had been conducted, but one sample was lost
due to problems with the sampling apparatus. Average total BTEX emissions
were 3.58 ± 0.61 tons per year Ctpy), and total VOC emissions were 19.8 ± 4.0
tpy. Other non-target HAP emissions were 0.49 ± 0.11 tpy n-hexane and 0.15 ±
0.03 tpy 2,2,4-trinethylpentane. There were no outliers in the total capture
56
-------
TABLE 10. SITE 1 - TOTAL CAPTURE EMISSION RESULTS
Compound
Emission (tpy)
Standard
Deviation
Benzene
1.25
0.
32
Toluene
1.68
0.
29
Ethylbenzene
o.oe
0.
02
Xylenes8
0.56
0.
15
Total BTEX
3.58
0.
61
n-Hexane
0.49
0.
11
2,2,4-Trimethylpentane
0.15
0.
03
Total VOC
19.8
4
.0
•Contributions to xylene emissions were 0.45 tpy m,p-xylenes and
0.11 tpy o-xylene.
data set. The majority of total capture emissions were measured in the
noncondensable gas fraction of the still vent stream {61% for BTEX, 89% for
total VOC).
Two aqueous samples were analyzed, and the hydrocarbon contained in the
aqueous phase contributed approximately 0.7% (0.025 tpy) of the hydrocarbon
phase total BTEX emissions. As specified in Section 7, since the aqueous
contribution was less than 3%, no additional analyses of water samples were
conducted.
Pressurized Glycol--
The average emissions determined from the seven pressurized glycol runs
are listed in Table 11. Emissions were 3.71 ± 0.51 tpy BTEX and 10.7 ± 1.9
tpy total VOC. The pressurized glycol BTEX emission result was consistent
with the total capture result, but the total VOC emissions were significantly
lower. The difference could be attributed to the collection of a non-
representative sample due to the presence of two-phase flow in the glycol line
(see Section 9). Emissions of n-hexane (0.39 ± 0.06 tpy) and 2,2,4-
trimethylpentane (0.11 ± 0.02 tpy) were lower than but within one standard
deviation of the total capture results.
57
-------
TABLE 11. SITE 1 - PRESSURIZED GLYCOL EMISSION RESULTS
Compound
Emission (tpy)
Standard Deviation
Benzene
1.22
0.16
Toluene
1.81
0.25
Ethylbenzene
0.08
0.01
Xylenes®
0.61
0 .09
Total BTEX
3.71
0.51
n-Hexane
0.39
0.06
2,2,4-Trimethylpentane
0.11
0.02
Total VOC
10 .7
1.9
"Contributions to xylene emissions were 0.49 tpy m,p-xylenes and 0.12 tpy o-xylene.
The cylinder off-gas for Run 7 was not included in the average because of
problems analyzing the small sample volume. However, the off-gas contributed
only a small fraction of the emission results. Off-gas volumes from all runs
were smaller than is typically observed because the pressure in the cylinders
was less than about 5 psig as compared with 40 psig at some other sites.
Atmospheric Glycol--
Average emissions calculated from the atmospheric glycol method are given
in Table 12. BTEX emissions of 3.79 ± 0.59 tpy compared well to both the
total capture and pressurized glycol method results. Total VOC emissions were
11.4 ± 1.8 tpy which compared well with the pressurized glycol results but
were lower than the total capture results. Loss of VOC is expected due to the
nature of the sample collection in open vials. Emissions of n-hexane at 0.39
± 0.05 tpy and of 2, 2,4-trimethylpentane at 0.12 ± 0.02 tpy emissions were
similar to the pressurized glycol results, but slightly lower than the
emissions measured by tQtal capture.
58
-------
TABLE 12. SITE 1 - ATMOSPHERIC GLYCOL EMISSION RESULTS
Compound
Emission (tpy)
Standard Deviation
Benzene
1.24
0.20
Toluene
1.85
0.28
Ethylbenzene
0.08
0 . 01
Xylenes"
0 .62
0.10
Total BTEX
3 .79
0.59
n-Hexane
0.39
0 . 05
2,2,4-Trimethylpentane
0.12
0.02
Total VOC
11.4
1.8
"Contributions to xylene emissions
were 0.50 tpy n,p-xyler.es and 0.
.12 tpy o-xylene.
Natural Gas and GRI-GLYCalc Results
Natural Gas Analyses--
Table 13 lists the natural gas BTEX concentrations determined using the
five different sampling approaches. The evacuated canister method (modified
EPA TO-14) yielded results significantly lower than the GPA Method 2166
techniques. It is believed that the analytical procedures used by laboratory
#1 for the cylinders may produce BTEX concentrations that are biased high.
Comparison of the cylinder samples taken with and without the sampling
manifold indicates that there was no observable effect of the manifold on the
gas analysis results. There did, however, appear to be removal of BTEX by
glycol in the absorber as evidenced by the generally lower BTEX concentrations
in the post-absorber gas samples.
Results of the audit natural gas sample analyses conducted by Radian and
the subcontract laboratory are given in Table 14. Results of the audit
analysis did not show any significant differences between the laboratories.
The in-house laboratory and the subcontract laboratory results, respectively,
were approximately 11% higher and 3% lower than the certified benzene
concentration. As a referee check, the in-house laboratory also analyzed one
of the field cylinder samples initially analyzed by the subcontract
59
-------
laboratory. The results from both laboratories are listed in Table 15. BTEX
concentrations determined by the subcontract laboratory were approximately 65%
higher than the Radian results.
GRI-GLYCalc Results--
GRI-GLYCalc input process parameters for Site 1 were listed in Table 9.
The reported lean glycol water content is an average of the Karl Fischer
analyses from each run. Dry gas water content was not measured at Site 1, so
the default value of 7 lb H20/MMSCF gas was chosen. The pump gas-to-glycol
volume ratio was determined from data supplied by the manufacturer.
The GRI-GLYCalc program was run with each of the five average gas
analyses input as the wet gas composition, even though samples taken after the
absorber were actually "dry." The GRI-GLYCalc emission estimates for the BTEX
compounds are compared with the total capture sampling results in Table 16.
The GRI-GLYCalc prediction of 3.88 tpy for total BTEX using the canister gas
sample was within 7% (less than one standard deviation) of the total capture
result.
GRI-GLYCalc estimates for total BTEX with the cylinder gas samples were
all significantly (30% or more) higher than the total capture result. The
high emission calculation was a consequence of the higher BTEX concentrations
determined for the natural gas in the cylinder samples. Results of the
referee sample discussed previously and the good agreement between GRI-GLYCalc
using the canister gas analysis and the total capture results indicate that
the subcontract laboratory for Site 1 may have biased BTEX concentrations high
in the actual natural gas samples. A possible explanation for the bias in
actual samples but not in the audit sample may be a matrix interference in the
natural gas. However, there was no significant difference between the in-
house sample and the referee sample for Site 2 (Table 23). This result tends
to discredit the theory of matrix interference. In fact, there are some
differences between the analysis protocols used by the two subcontractor
laboratories. It is probable that the gas cylinder sample analytical
60
-------
TABLE 13. SITE 1 - NATURAL GAS BTEX COMPOSITION
Component
Concentration (ppnrv)
Canister*
Cylinder
Before
Absorber,
with
Manifold6
Cylinder
Before
Absorber,
without
Manifold*5
Cylinder
After
Absorber,
with Mani-
fold6
Cylinder
After
Absorber,
without
Manifold"
Benzene
57.9
± 4.1
109.6
±
18.4
98.6
+
19.3
73.5
+
22.2
74.1
+
33.4
Toluene
45.6
± 9.0
89.7
+
16.2
82.7
±
36.6
56.1
±
20.5
54.9
+
31.7
Ethylbenzene
1.2
± 0.8
3.7
+
4.1
2.5
±
3.8
1.4
±
2.5
0.9
+
1.0
Xylenes
7.9
± 2.4
18.3
±
10.7
16.8
"fr*
9.6
21.3
±
31.9
9.9
+
7.8
*ALkene composition Input to GRI-GLYCale: 1.90X nitrogen, 83.9X methane, 7.90X ethane, 3.86X propane, 1.70X butane*. 0.
-------
TABLE 14. SITE i - NATURAL GAS AUDIT SAMPLE
Laboratory Benzene Concentration
(ppmv)
In-House 336
Subcontract
Laboratory #1
Certified
"jr>2
Concentration
TABLE 15. SITE 1
- NATURAL GAS
REFEREE SAMPLE
Component
Concentration (ppmv)
In-House
Subcontract
Laboratory #1
Benzene
57
96
Toluene
42
73
Ethylbenzene
2
ND
Xylenes
6
10
procedures used by laboratory #i did produce 3TEX concentration measurements
that are biased high.
SITE 2
Individual run and analyte results determined for Site 2 are given in
Appendix D.
Process Data
Table 17 lists the process data readings taken during Site 2 sampling.
All the process conditions except natural gas flow rate were relatively
constant throughout the test. The gas flow rate variation was caused by
62
-------
factors riot related to the dehydrator operation. Failure of a valve upstream
of the compressor caused gas flow to be shut off to the dehydrator temporarily
before sampling had begun on the first cay. During the night between the
first and second days of sampling, one of the wells stopped producing, causing
a drop in the gas flow rate.
Independent measurement of the glycol circulation rate was conducted by
diverting the glycol into a container for a specified period of time and
measuring the mass of the collected liquid. The measured value was
approximately 20% higher than the circulation rate determined from the pump
curve. However, because the glycol was flowing to atmospheric pressure during
the measurement instead against several feet of glycol head (entering the
reboiler still), the measurement was not fully representative of the process
conditions. It was decided that the measured circulation rate was close
enough to the pump curve rate to justify use of the pump curve in emission
calculations and in GF.I-GLYCalc.
Sampling
Total Capture Condensation--
The average emissions calculated for the six total capture runs are
listed in Table 18. Average total BTEX emissions were 22.9 ± 3.2 tpy, and
total VOC emissions were 3 6.9 ± 3.1 tpy. Because the flash tank removed
virtually all of the noncondensable gases, the large majority of total capture
emissions were measured in the hydrocarbon liquid fraction (99% of BTEX and
97% of total VOC). There were no outliers in the emission results.
Two aqueous samples were analyzed, and the hydrocarbon contained in the
aqueous phase contributed approximately 0.4% (0.10 tpy) of the hydrocarbon
phase total BTEX emissions. As specified in Section 7, since the aqueous
contribution to emissions was less than 3%, no additional analyses of water
samples were conducted.
63
-------
TABLE 16. SITE 1 - GRI-GLYCalc RESULTS
Emissions (tpy)
Natural Gas
Sampling Method
Benzene
Toluene Ethylbenzene Xylenes Total BTEX Total VOC
Canister
Cylinder Before
Absorber, with
Manifold
Cylinder Before
Absorber, without
Manifold
Cylinder After
Absorber, with
Manifold
Cylinder After
Absorber, without
Manifold
Total Capture
Benchmark
1.31
2.50
2.25
1.68
1.68
1.25
1.87
3.68
3.40
2.29
2.26
1.68
0.06
0.24
0.18
0.06
0.06
0.08
0.64
1.44
1.36
1.68
0.80
0.56
3.88
7.86
7.18
5.71
4.80
3.58
21.7
28.2
28.3
25.5
23.7
19.9
-------
TABLE 17. SITE 2 - PROCESS PARAMETERS
Parameter
Relative Number
Average Standard Maximum Minimum of
- Deviation Points
(%)
Gas Flow Rate, MMSCFD
4.89
18.5
6.24
3.38
12
Gas Temperature, °F
111
5.8
118
98
12
Gas Pressure, psig
1063
0.4
1070
1060
12
Glycol Circulation
Rate, gph
204
0
204
204
12
Dry Gas H20 Content,
lb/MMSCF
12
7
12.9
11.3
3
Lean Glycol Water
Content, wt %
1.31
24
1.51
0.68
6
Gas Pump Volume
Ratio", acfm/gpm
0.033
N/A
N/A
N/A
N/A
Flash Tank
Temperature, °F
205
6.5
214
190
3
Flash Tank Pressure,
psig
45.5
1.2
46
45
6
Reboiler Temperature,
377
0.5
380
375
6
°F
Not recorded during the field test but required as input to GRI-GLYCalc; taken from
manufacturer's specifications.
Pressurized Glycol--
The average emissions determined from the six pressurized glycol runs are
listed in Table 19. Emissions were 25.9 ± 3.2 tpy BTEX and 37.9 ± 4.9 tpy
total VOC. The pressurized glycol BTEX emission results were higher than the
total capture results for both BTEX and total VOC. No assignable reason could
be determined to explain the higher results of the pressurized glycol
emissions. Calibration checks before and after the sample analyses were
within acceptable limits, and there was no observable matrix interference in
the spike analysis (see Section 9).
65
-------
TABLE 18. SITE 2 - TOTAL CAPTURE EMISSION RESULTS
Compound Emission (tpy) Standard Deviation
Benzene
6.02
1.04
Toluene
9.87
1.50
Ethylbenzene
0. 84
0.16
Xylenes®
6.14
0.74
Total BTEX
22.9
3.2
n-Hexane
0.22
0.05
2,2,4-Trimethylpentane
0.12
0 .02
Total VOC
36.9
3.1
"Contributions to xylene emissions were
4.61 tpy m,p-xylenes and 1.53 tpy
o-xylene.
TABLE 19. SITE 2 -
PRESSURIZED GLYCOL EMISSION
RESULTS
Compound
Emission (tpy) Standard Deviation
Benzene
6.71
0.98
Toluene
11.1
1.6
Ethylbenzene
0.98
0.18
Xylenes8
7.C5
0.82
Total BTEX
25.9
3.2
n-Hexane
0.23
0.04
2,2,4-Trimethylpentane
0.07
0.01
Total VOC
37.9
4.9
'contributions to xylene emissions were 5.26 tpy m,p-xylenes and 1.79 tpy o-xylene.
There was no recoverable off-gas from the Site 2 glycol cylinders, so the
emissions were determined from liquid glycol only. This is consistent with
66
-------
the absence of foaming in the atmospheric glycol samples; that is, most of the
volatile gases had been removed in the flash tank upstream of the glycol
sample location.
Atmospheric Glycol--
Average emissions calculated from the atmospheric glycol method are given
in Table 20. At 21.4 ± 2.0 tpy and 30.8 ± 3.4 tpy, the BTEX and total VOC
emissions compared well with the total capture results. The unit at Site 2
had a flash tank operating at relatively high temperature, which removed
upstream of the glycol sample point volatile components typically lost from
atmospheric samples.
TABLE 20. SITE 2 -
- ATMOSPHERIC GLYCOL
EMISSION RESULTS
Compound
Emission (tpy)
Standard Deviation
Benzene
5.62
0.76
Toluene
9.25
0.93
Ethylbenzene
0.80
0.12
Xylenes"
5.74
0.40
Total BTEX
21.4
2.0
n-Hexane
0.21
0.04
2,2,4-Trimethylpentane
0.12
0.02
Total VOC
30.8
3.4
'Contributions to xylene enissions were 4.29 tpy m,p-xylenes and 1.45 tpy o-xylene.
Natural Gas and GRI-GLYCalc Results
Natural Gas Analyses--
Table 21 lists the natural gas BTEX concentrations determined using the
five different sampling approaches. The evacuated canister method (modified
EPA TO-14) yielded results similar to the GPA Method 2166 result upstream of
67
-------
the absorber both with and without manifold. The use of a sampling manifold
did not appear to influence the natural gas BTEX composition result.
Consistent with the Site 1 results, comparison of gas samples taken
before and after the absorber showed significant removal of BTEX by the glycol
in the absorber. It was probable that significant removal of BTEX was
observed because of the high glycol-to-gas ratio in the absorber. The gas
industry standard practice for glycol circulation to provide adequate gas
drying is 2 to 3 gallons of glycol per pound of water removed from the gas
(9,10). Glycol circulation at Site 2 was approximately 15 gallons of glycol
per pound of water removed.
Comparison of the audit and referee samples for Site 2 indicate that
subcontract laboratory #2 may have had a slight high bias for the BTEX
concentrations in natural gas. Results of the audit natural gas sample
analyses are given in Table 22. The in-house laboratory benzene result was
within 2% of the certified concentration. The subcontract laboratory result
was approximately 22% greater than the certified benzene concentration. The
referee cylinder sample results given in Table 23, however, showed close
agreement for benzene concentration between the two laboratories. The
subcontract laboratory did report slightly higher levels of toluene and
xylenes.
The mean results of the three sets of samples collected before the
absorber (one canister and two cylinders), however, show excellent agreement
for BTEX concentration. The slight bias shown in the referee and audit
analyses may not have been representative of all the results.
GRI-GLYCalc Results--
GRI-GLYCalc input process parameters for Site 2 were listed previously in
Table 17. The reported lean glycol water content is an average of the
individual Karl Fischer analyses for each run. Dry gas water content was
measured by the operating company at Site 2, so the measured value was used in
place of the default. The pump gas-to-glycol volume ratio was determined from
data supplied by the manufacturer.
68
-------
The GRI-GLYCalc program, was run with each of the five gas analyses input
as the wet gas composition, even though samples taken after the contactor were
actually "dry." The GRI-GLYCalc emission estimates for the BTEX compounds
compared with the total capture sampling results are shown in Table 24. The
GRI-GLYCalc prediction for total BTEX using both the canister and cylinder
samples collected before the absorber using the manifold were within 7% of the
total capture result. The GRI-GLYCalc estimate for total BTEX with the
cylinder gas sample taken before the absorber without using the manifold was
approximately 11% lower than the total capture result. Program predictions
with the gas sair.ples taken after the absorber are all significantly lower than
the total capture result. The low emission calculation was a consequence of
the lower BTEX concentrations in the outlet natural gas analyses, probably
caused by absorption of BTEX into the glycol. Total VOC emission predictions
by GRI-GLYCalc were lower than the total capture benchmark for all natural gas
analyses but within 16% for samples taken before the absorber.
69
-------
TABLE 21. SITE 2 - NATURAL GAS BTEX COMPOSITION
Concentration (pprav)
Component
Benzene
Toluene
Ethylbenzene
Xylenes
Cylinder Cylinder Before Cylinder Cylinder
Before Absorber, Absorber, with- After Absorber, After Absorber,
Canister' with Manifold6 out Manifold" with Manifold5 without Manifoldb
143 ± 17
142 ± 13
11.2 ± 2.5
73.5 ± 10.6
152 ± 24
147 ± 43
10.9 + 5.0
60.4 ± 33.4
150 ± 13
136 ± 12
10.0 ± 0.0
51.9 ± 7.7
116 ± 10
101 ± 11
5.8 ± 2.1
33.6 ± 12.6
105 ± 10
91.4 + 6.3
5.0 ± 0.0
27.5 ± 13.0
* Alkane composltion Input to CHI-GLYCalci 0.472 carbon dioxide, 1.301 nitrogen, 91.932 methane, 3.802 ethane,
1,231 propone, 0,702 butmea, 0.232 pentanes, 0.29X he Kane 5 plus,
b AvATAge alkms composition Input to Gftl-GL.YCa.ic: 94.OX methane, 3.65% ethane, 1.242 propane, 0,712 butanes,
0.222 pentanes, 0.222 hexanea plus.
-------
TABLE 22.
SITE 2
- NATURAL GAS AUDIT SAMPLE
Laboratory
Subcontract
Laboratory #2
Certified
Concentration
Benzene Concentration
(ppmv)
In-House 307
370
302
TABLE 23. SITE 2 - NATURAL GAS REFEREE SAMPLE
Component
Concentration (ppmv)
In-House
Subcontract
Laboratory #2
Benzene
Toluene
Ethylbenzene
Xylenes
165
107
4
24
170
130
ND
30
-------
TABLE 24. SITE 2 - GRI-GLYCale RESULTS
Emissions (tpy)
Natural Gas
Sampling Method
Benzene
Toluene
Ethylbenzene
Xylenes
Total BTEX
Total VOC
Canister
5.22
8.64
0.89
7.58
22.3
36.1
Cylinder before Absorber
with Manifold
5.55
8.94
0.89
6.13
21.5
32.7
Cylinder before Absorber
without Manifold
5,62
8.51
0.82
5.33
20.3
31.1
Cylinder after Absorber,
with Manifold
3.93
5.69
0.41
2.87
12.9
23.3
Cylinder after Absorber
without Manifold
4.35
6.32
0.49
3.48
14.6
25.3
Total Capture Benchmark
6.02
9.8?
0.84
6.14
22.9
36,9
-------
SECTION 9
DATA QUALITY INDICATORS FOR CRITICAL MEASUREMENTS
DEFINITIONS AND OBJECTIVES
Variables of interest in determining data quality were bias and
precision. This section describes the calculation methods for and the
objectives for bias and precision in the sampling and analytical portions of
the projects, and discusses the results of the data quality calculations.
Table 25 presents the precision and bias objectives for the critical
measurement parameters at each site. The bias and precision objectives for
the results pooled over both sites were narrower ranges than for the
individual sites; objectives for the pooled results are discussed in the
sections that follow. The BTEX compounds were considered critical chemical
parameters. Glycol circulation rate and natural gas flow rate were critical
process parameters.
The bias and precision objectives were specified for the BTEX emissions
(as measured in tons/year) and for the natural gas BTEX concentrations (as
measured in parts per million volume). Previous results from the GRI project
had suggested that the precision objectives were not feasible for BTEX
emissions less than 1 ton per year and natural gas BTEX compositions below 20
pprr.v. At these levels, the analytes were at or near the reporting limits and
uncertainty in the measurements greatly increased.
Overall bias in sampling could not be directly measured since the "true"
value in the process streams was not known, so a definitive determination of
total measurement bias could not be obtained during this work assignment.
Therefore, spiked sample results were used as a measure of bias. An
insufficient nuiri>er of spiked samples were collected to assess whether spike
recoveries indicated statistically significant bias independent of random
73
-------
TABLE
25. SUMMARY OF DATA QUALITY OBJECTIVES
FOR CRITICAL (BTEX) MEASUREMENTS
Matted or
Snpla or
Analyeia
Overall Data Quality Objactiw
Analytical Data Quality ObjKtln
Par me tar
Far«Ur
T«faoiqui
OA Measure
Precision
Bias
OA Haaaora
Precision
Bias
Total Capture
Hydrocarbon and
aqueous liquid
Roncondensable
gas
GC/FID
GC/FID,PID
Preclaion-RSD of
alx runa; Bias-
Maan spike re-
covary
<501 RSD
±101
Praciaion-RPD for
duplicate analyses;
Bias-Mean of all
Dally Lab Standard
Analysaa
<101 RPD
±?r
±5*
Pressurized
Glycol
Glycol liquid
Offgat
GC/FID
GC/FID.PID
Precision-RSD of
six runs; Bias-
Ms an spike re-
covary
<50X RSD
*102
Preclfllon-RFD for
duplicate analyses;
Bias-Hean of all
Daily Lab Standard
Analysis
<101 RPD
±7:
±51
Atmospheric
Rich/Lean
Glycol
Glycol liquid
GC/FID
Precision-RSD of
six runs; Bits-
Mean spike re-
covery
<501 RSD
±101
Precision-RPD for
duplicate analyses;
Bias-Mean of all
Daily Lab Standard
Analysis
<10X RPO
±71
Hatural Gat
Canlstar
GPA Mathod
GC/FID.PID,TCD
GC/FID
Praciaion-RSD;
Bias-Mean spike
racovary
Pracision-RSD;
Bias-Raferea
Analysis
<501 RSD
±101
Praoiaion-RPD for
duplicate analyses;
Bias-Mean of all
Dally Lab Standard
Analyaaa
<101 RPD
±5X
Glycol Unit
Process Data
Natural gas flow
rata
Glycol circula-
tion rat*
Pracislon-RSD of
six runs; Bias-
based on plant
calibrations.
<501 RSD
(<107 RSD
defines
steady-
stata oper-
ation)
Manufacturer
specs.
HA
NA
NA
-------
error; however, previous experience in the GRI project had indicated that no
significant bias due to matrix effects would occur. Analytical bias was
quantified through calibration and continuing checks of analytical system
performance.
Process conditions during testing were considered to have a high degree
of precision when the system was operating under steady-state conditions,
which was determined by collecting process operating data and using
engineering judgment. Generally, a steady-state process will have less than
10% relative standard deviation in its operating data. Although at least some
of the process data collected at each site did not meet the steady-state
criteria, the process fluctuations were gradual and did not demonstrate
extreme instantaneous deviations from steady-state. The process fluctuations
were all within the 50% relative standard deviation (RSD) objective given in
Table 25. A more complete discussion of the process variability is included
later in this section.
Overall precision was determined by calculating the RSD for the emission
results. This incorporated process, sampling, and analytical variability.
Duplicate (collocated) samples and duplicate analyses (repeatability) were
used to separate the sampling and analytical components of the variability, as
necessary. Previous experience had indicated that process fluctuations in
glycol dehydrators were typically the greatest sources of variability. The
project team had essentially no control over these fluctuations. Analytical
precision was determined by calculating the relative percent difference (RPD)
of duplicate analyses.
Another data quality indicator for the field testing was completeness.
The completeness objective was to collect valid data for at least five of the
six planned samples of each type for each site. The objective was met for all
sample types at both sites.
A final data quality indicator was sample representativeness. Sample
representativeness is the extent to which the sample collected represents the
75
-------
true value of emissions from the units. Past experience had indicated that
pressurized glycol samples were not representative due to the existence of
two-phase flow in the rich glycol line. The atmospheric rich and lean glycol
samples may not have been representative due to the loss of VOC (and BTEX to a
lesser extent) during the sampling. By contrast, total capture condensation
measured the entire still vent stream. Sample representativeness was not an
issue for this technique. Thus, GRI and its Industry Working Group used the
total capture results as the benchmark to which the glycol methods and the
GRI-GLYCalc results were compared. Emission results reported in Section 8
indicated that representativeness was a concern for both pressurized and
atmospheric glycol samples at Site 1 but not at Site 2. The absence of
volatile hydrocarbons in the rich glycol at Site 2 produced representative
glycol samples.
As the target analytes, BTEX compounds were considered to be of primary
importance and all sample dilutions and injection volumes were tailored to
guarantee the quantification of these compounds; however, concentrations of
ethylbenzene and the xylenes were sometimes at or below the reporting limit,
which affected the precision of the measurements.
Reporting limits for the non-BTEX hydrocarbons were initially a concern.
A number of the hydrocarbons were present at levels that required large
dilutions to avoid overloading the system or to attempt to measure values
outside the calibration range of the analytical instrument. Therefore, some
of the lower concentration analytes (e.g., 2,2,4-trimethylpentane) were
difficult to detect at these dilutions. Every effort was made to ensure that
the reporting limits were sufficiently low so that those compounds not
detected did not significantly contribute to the total emissions from that
site.
CALCULATION METHODS
As discussed previously, the key data quality indicators for this project
were bias and precision. Calculations for these indicators are discussed
below.
76
-------
Bias
The formula used to calculate absolute bias of a test value compared to a
reference value was
» • K " *r)
where:
B — absolute bias,
X,. - test value, and
X,. - reference value.
Relative bias (RB) was defined as the absolute bias expressed as a fraction or
percentage of the reference value:
RB « JL
Spike recovery was a special case of relative bias, calculated as follows:
s - V
%Rec. 2. x 100
CS
where:
Sm - spiked sample result,
Uu «= unspiked sample result, and
CS — calculated value of the spike amount.
For this project, bias was based on spike recovery data. Recovery bias in a
spiked sample was expressed in terms of percent error, given by:
Recovery Bias (% Error) ¦= % Recovery - 100.
77
-------
For the combined data sets, pooled bias was calculated for each analysis type
in a particular method as follows:
^®pooL«d
£ RBX
where N is the number of measurements. For spike recovery, % error (i.e.,
recovery bias) was used as the relative bias term.
Precision
The precision for a particular method was defined as the relative
standard deviation (RSD) of the emission results over the numerous runs^ The
RSD was calculated as follows:
SD
N
E
i -1
N - 1
xf
where:
SD - standard deviation;
XL - result for a particular run,
X - mean result; and
N - number of runs.
Then:
RSD - ££
78
-------
The data quality indicator goals were that the RSD be < 50 percent for a
particular data set from a specific site and < 20 percent for the pooled
results of the entire program (pooled precision was calculated similarly to
pooled relative bias above). If the precision exceeded these goals, a further
analysis of the components of the RSD was performed. This analysis was based
on the following equation:
RSD2 = RSDp + RSDi
where:
RSDp = relative standard deviation due to the process
variability, and
RSDsa = relative standard deviation due to sampling and analytical
variability.
RSDga was determined by using the duplicate sample results to calculate
the standard and relative standard deviations as shown above. RSDP was then
calculated by difference. RSDEa was broken into its two parts by a similar
approach:
RSDs2a = RSD? + RSD?
where:
RSDs = relative standard deviation due to sampling variability, and
RSDa = relative standard deviation due to analytical variability.
RSDa was calculated from the duplicate (replicate) analyses.
79
-------
Analytical Precision
Analytical precision was determined by calculating the relative percent
difference (RPD) for duplicate analyses. The RPD was calculated using the
following equation:
RPD i = ( Initial Result - Duplicate Result | x 10Q%
Mean of Initial and Duplicate Results
Analytical Bias
Analytical bias was determined by calculating the RPD from the original
calibration curve of a mid-level calibration standard analyzed before and
after a group of samples. The calibration standard was compared with the
original calibration curve as follows;
RPD % - Ga^*-krat*-on Check, ppm - Original Standard, ppra ^
Original Standard, ppm
PRECISION RESULTS
Overall
Total Capture--
The pooled precision for BTEX emissions from the total capture sampling
method is given in Table 26, Since only the toluene and total BTEX precisions
were less than the objective limit of 20%, the RSD was broken down into
process, sampling, and analytical parts. As expected based on previous
experience, process variability was the greatest determinant to the overall
precision.
Glycol Methods--
Pooled precision results of the BTEX emissions determined by the two
glycol methods are listed in Table 2?. Since all the results (except for
80
-------
TABLE 26. POOLED PRECISION FOR TOTAL CAPTURE METHOD - EMISSION RESULTS
RSD (%)
Analyte
Overall
Process Sampling
Analytical
Benzene
21.5
21.4 1.7
0.4
Toluene
16.2
16.2 1.3
0 . 6
Ethylbenzene
23 .1
23.0 1.6
0.6
m,p-Xylenes
23 . 6
23.5 1.6
0.6
o-Xylene
25.1
25.1 1.7
0.5
Total BTEX
15 . 6
15.6 1.5
0.5
Objective
20. 0
-
-
TABLE 27.
POOLED PRECISION FOR GLYCOL METHODS - EMISSION RESULTS
Overall RSD (%)
Analyte
Pressurized Glycol Atmospheric Glycol
Benzene
16.9
15.6
Toluene
16.9
11.9
Ethylbenzene
22.9
18.5
m, p-Xylenes
18.0
10.9
o-Xylene
21.9
13.0
Total BTEX
14.9
11.1
Objective
20.0
20.0
ethylbenzene and o-xylene) were below the objective of 20%, no additional
calculation was performed. Ethylbenzene and o-xylene concentrations in the
glycol samples were much lower than other target analyte concentrations. As
discussed previously, the precision objectives were not feasible for such low
concentrations.
81
-------
Natural Gas--
Pooled precision values for the five natural gas sampling methods are
listed in Tables 28 and 29. The precision objective of 20% RSD was met for
total BTEX but for only a few other analyses. Because the concentrations of
BTEX in the natural gas were in general near the reporting limit, inherent
uncertainty in the result was greatly increased. The sampling and analytical
contribution to the precision was calculated and is listed in Table 29.
Cylinder samples analyzed by subcontract laboratories were not analyzed in
duplicate, so the sampling and analytical precision was calculated from
sampling variation only. The laboratory that analyzed cylinder samples from
Site 2 reported BTEX concentration to only the nearest 10 ppmv, which tended
to mask actual variability at low concentrations. The cylinder samples from
Site 1 showed the greatest variability, with all of the compounds except
benzene above the 50% RSD objective (single-site) for at least some of the
samples (see Appendix E).
Process Data--
The precision of glycol unit process data was not pooled. Relative
standard deviation of process parameters for both sites are listed in Table
30.
At Site 1, the RSD for contactor pressure was below the 10% level
specified for steady-state operation, but those for all other process
variables exceeded the precision objective. It should be noted that there was
a distinct difference in operation between the first and second days of
sampling.
The natural gas flow rate, glycol circulation rate, and contactor
temperature were relatively constant for the first day. However, from the
first to the second day, the natural gas flow rate increased and the average
contactor temperature dropped from approximately 95°F to 78°F from the first
to the second day. These changes were probably due to the processing of gas
82
-------
TABLE 28. POOLED PRECISION (% RSD) FOR NATURAL GAS RESULTS
Sample Type
Analyte
Canister
Cylinder
Pre-
Cylinder
Pre-
Cylinder
Post-
Cylinder
Post-
absorber
absorber
absorber
absorber
with
without
with
without
manifold
manifold
manifold
manifold
Benzene
9.3
16.2
13.5
20.1
28.5
Toluene
14.3
26.5
25.8
22.8
32 .5
Ethylbenzene
45.3
62.1
44.9
66.0
25.0
Xylenes
22.4
60.7
25.6
57.7
56.1
Total BTEX
7.8
15.2
12.4
15.4
19.8
Objective
20.0
20.0
20.0
20.0
20.0
TABLE 29. POOLED
SAMPLING AND ANALTYICAL RSD(%) FOR
NATURAL GAS
RESULTS"
Sample Type
Analyte
Canister
Cylinder
Pre-
Cylinder
Pre-
Cylinder
Post-
Cylinder
Post-
absorber
absorber
absorber
absorber
with
without
with
without
manifold
manifold
manifold
manifold
Benzene
2.0
4.7
26.1
25.7
0.6
Toluene
3.5
12.8
27.8
24.8
9.1
Ethylbenzene
34.7
61.3
0
0
59.8
Xylenes
14.3
12.9
26.1
64.0
18.9
Total BTEX 3.0 9.6 26.6 26.7 5.6
•Cylinder cample results are sampling RSD only.
from different wells for the first half of each month, as explained in Section
8. Additional wells were initially brought on-line in the midst of the field
testing. The glycol circulation rate also became somewhat erratic on the
second day of testing. The glycol pump was set to operate at a rate below the
minimum suggested by the manufacturer, which has been shown at another site in
the GRI program to cause instability in pump operation.
83
-------
TABLE 30. PROCESS DATA - PRECISION* FOR BOTH SITES
RSD (%)
Parameter
Site 1
Site 2
Natural Gas Flow Rate
22.8
18.5
Absorber Temperature
11.4
5.8
Absorber Pressure
2.4
0.4
Glycol Circulation Rate
26.8
0.0
Flash Temperature
N/A
6.5
Flash Pressure
N/A
1.2
Objective
50
50
Process operation at Site 2 was much more consistent than at Site 1.
Natural gas flow rate showed greater variability the first day as wells were
brought on-line after a valve failure upstream of the compressor. Before
sampling was begun on the second day, one of the wells stopped producing gas,
which caused a drop in the flow rate. All the other process variables were
within the objective limits.
Analytical
The analytical precision was defined as the relative percent difference
(RPD) of duplicate analyses.
Total Capture--
Analytical precision results for both the total capture gas and liquid
analyses are listed in Table 31. The RPDs for all analytes were less than 2%,
below the objective of 10 percent.
Pressurized Glycol--
Analytical precision values for both the pressurized glycol liquid and
gas analyses are given in Table 32. A value of one-half the reporting limit
was used in calculations for analytes not detected. The RPDs for all the
84
-------
TABLE 31. POOLED ANALYTICAL PRECISION FOR TOTAL CAPTURE SAMPLES BASED ON RPD
OF DUPLICATE ANALYSES
RPD (%)
Analyte
Liquid Samples
Gas Samples
Benzene
0.78
0.71
Toluene
0.72
1.12
Ethylbenzene
0.70
1.43
m,p-Xylenes
0.68
1.40
o-Xylene
0.55
1.15
Total BTEX
0.66
0.92
Objective
10.0
10.0
TABLE 32.
POOLED ANALYTICAL PRECISION FOR PRESSURIZED GLYCOL
SAMPLES BASED ON RPD OF DUPLICATE ANALYSES
RPD (%)
Analyte
Liquid Samples
Gas Samples
Benzene
0.58
1.41
Toluene
0.45
2.05
Ethylbenzene
0.86
163*
m,p-Xylenes
0.93
85.3"
o-Xylene
0.74
N/D
Total BTEX
0.48
1.79
Objective
10.0
10.0
*Valu«« were et or Dear the reporting Limit.
analytes in the liquid phase were below 1%, well within the specified
objective of 10 percent.. The precision values for ethylbenzene and xylenes in
the gas phase were outside the objective limits. No corrective action was
considered necessary because the sample concentrations were at or near the
reporting limit and had no effect on the emissions to the reported number of
significant figures.
85
-------
Atmospheric Glycol--
Atmospheric rich and lean glycol analytical precision values are given in
Table 33. The objective for atmospheric glycol RPDs was also 10%; the RPDs
for all analytes were less than 6 percent.
Natural Gas-
Analytical precision values listed in Table 34 for natural gas canister
samples were below the objective of 10% for benzene and toluene. The low
concentrations of ethylbenzene and xylenes in the natural gas create high
uncertainties in the results. Most of the variability in the pooled results
was due to the ethylbenzene and xylene results at Site 1; Site 2 results were
all within the objective (see Appendix E). A value of one-half the reporting
limit was used in calculations for analytes not detected. Duplicate
laboratory analyses were not performed for the cylinder gas samples analyzed
by subcontract laboratories.
BIAS RESULTS
As calculated in this study, bias could be positive or negative.
Positive bias indicated that the analysis detected more of an analyte than was
actually present, and the calculated emissions were subsequently biased high.
Negative bias indicated that the analysis detected less of an analyte than was
actually present, and the calculated emissions were biased low.
Overall
Overall relative bias was calculated separately for the two sites due to
the unique sample matrix at each location.
Site 1--
Overall bias values based on spike recovery in each of the liquid sample
types from Site 1 are given in Table 35. The objective limit for the absolute
value of the bias was 10 percent. The results from all the sample types and
86
-------
TABLE 33. POOLED ANALYTICAL PRECISION FOR ATMOSPHERIC GLYCOL SAMPLES BASED ON
RPD OF DUPLICATE ANALYSES
RPD (%)
Analyte
Rich Glycol
Lean Glycol
Benzene
0.85
1.39
Toluene
0.87
0.99
Ethylbenzene
2.55
5.20
m,p-Xylenes
1.84
0.70
o-Xylene
1.29
CO
Total BTEX
1.10
1.12
Objective
10.0
10.0
TABLE 34. POOLED
ON RPD
ANALYTICAL PRECISION FOR CANISTER
OF DUPLICATE SAMPLES
NATURAL GAS SAMPLES BASED
Analyte RPD (%)
Benzene 2.25
Toluene 4.54
Ethylbenzene 40.00"
m,p-Xylenes 14.0
o-Xylene 6.22e
Total BTEX 4.16
Objective 10.0
a
•Analyt* tt oz below reporting limit.
analytes met the objective with the exception of benzene and toluene in lean
glycol. Analyte concentrations in the lean glycol were near the reporting
limit, so there was greater inherent uncertainty in spike preparation and
analysis. Because the bias results were only slightly above the objective and
analyte concentration in the lean glycol is not a significant factor in
emissions calculations, no corrective action was deemed necessary.
87
-------
TABLE 35. OVERALL RELATIVE BIAS (%) FOR SITE 1 LIQUID SAMPLES
Analyte
Sample Type
Benzene
Toluene
Hexane
Octane
Total Capture
-4.6
-2.3
-3.6
-1.7
Pressurized Rich Glycol
-4.6
-3.6
- B. 1
-5.1
Atmospheric Rich Glycol
-0.2
0.2
-4.3
5.4
Atmospheric Lean Glycol
11.0
12.7
-4.8
-1.4
Overall bias results for the gas samples from Site 1 are given in Table
36. Analytical spike recoveries were outside the objective limit for both
natural gas and total capture gas canister samples. The natural gas spike
recovery was only slightly below the limit, so no corrective action was taken.
The total capture gas spike recovery was significantly outside the objective
limit. No explanation for the low recovery was apparent; however, the close
agreement between total capture target analyte emissions and the glycol
methods emissions indicate that recovery of the target analytes was not a
problem.
Site 2--
Overall bias results for the Site 2 liquid samples are listed in Table
37. Total capture and atmospheric rich and lean glycol biases were within the
objective of +10 percent. Bias for the pressurized glycol analytes was
slightly outside the specified objective. Because the atmospheric -glycol
samples were well within the bias objectives, it is not probable that there
was a matrix interference in the Site 2 glycol, and no corrective action was
taken.
Overall bias values for the Site 2 gas samples are listed in Table 38.
There appeared to be minimal bias with these samples.
88
-------
TABLE 36. OVERALL RELATIVE BIAS (%) FOR SITE I GAS SAMPLES
(BASED ON RECOVERY OF cis-2-Butene)
Sample Type
Relative Bias
Natural Gas Canister
-13.7
Total Capture Gas
44.5
Pressurized Glycol Gas
-1.8
TABLE 37. OVERALL
RELATIVE BIAS (%) FOR SITE 2 LIQUID SAMPLES
Analyte
Sample Type
Benzene
Toluene Hexane Octane
Total Capture
1.1
1.4 0.8 -3.2
Pressurized Rich Glycol
-13.4
-11.3 -17.6 -13.4
Atmospheric Rich Glycol
0.1
-1.0 3.1 1.4
Atmospheric Lean Glycol
4.6
3.8 -5.8 -4.0
TABLE 38. OVERALL RELATIVE BIAS
(BASED ON RECOVERY OF
(%) FOR SITE 2 GAS SAMPLES
cis-2-Butene)
Sample Type
Relative Bias
Natural Gas Canister
-8.9
Total Capture Gas
-6.1
89
-------
REFERENCES
1. Reif, D.L., C.O. Rueter, and J.M. Evans. Emission Measurement Results
from GRI's Glycol Dehydrator Sampling and Analytical Methods Development
Program. In: Proceedings of the 1994 GRI Glycol Dehydrator/Gas
Processing Air Toxics Conference, Gas Research Institute, Chicago,
Illinois, 1994. pp. 59-82.
2. Graham, J.F., et al. Natural Gas Dehydration: Status and Trends. GRI-
94/0099, Gas Research Institute, Chicago, Illinois, 1994.
3. Protocol for the Field Validation of Emission Concentrations from
Stationary Sources. EPA-450/4-90-015 (NTIS PB91-216184). U.S.
Environmental Protection Agency, Office of Air Quality Planning and
Standards, Research Triangle Park, North Carolina, April 1991.
4. Federal Register, 40CFR Part 63. Vol. 57, No. 250, pp. 61998-62002.
December 29, 1992.
5. Compendium Method TO-14, The Determination of Volatile Organic Compounds
(VOCs) in Ambient Air Using Summa™ Passivated Canister Sampling and
Chromatographic Analysis. In: Second Supplement to Compendium of Methods
for the Determination of Toxic Organic Compounds in Ambient Air, EPA-
600/4-89-018 (NTIS PB90-127382), U.S. Environmental Protection Agency,
Research Triangle Park, North Carolina, June 1988.
6. Obtaining Natural Gas Samples for Analysis by Chromatography. GPA
2166-86. Gas Processors Association, Tulsa, Oklahoma, 1986. 13 pp.
7. Carlson, R.L., et al. Preliminary Comparison of Natural Gas Sampling and
Analytical Methods. In: Proceedings of the 1994 GRI Glycol
90
-------
Dehydrator/Gas Processing Air Toxics Conference, Gas Research Institute,
Chicago, Illinois, 1994. pp. 209-242.
8. Dixon, W.J. Processing Data for Outliers. Biometrics, 9(1);89. 1953.
9. Pearce, R.L. and C. Richard Sivalls. Fundamentals of Gas Dehydration,
Design, and Operation with Glycol Solutions. In: Laurence Reid Gas
Conditioning Conference Fundamentals Manual, third printing. The
University of Oklahoma, Norman, Oklahoma, 1990. Section I.
10. Ballard, D. "Operator Talk...Glycol Dehydration." Hydrocarbon
Processing, April 1977, pp. 111-118.
91
------- |