July 2011
Environmental Technology
Verification Report
PlCARRO
MODEL G1103-c AMMONIA ANALYZER
Prepared by
Battelle
Battelie
Inc Business of Innovation
Under a cooperative agreement with
U.S. Environmental Protection Agency
-------
July 2011
Environmental Technology Verification
Report
ETV Advanced Monitoring Systems Center
PICARRO, INC. MODEL G1103-C AMMONIA
ANALYZER
by
Brad Goodwin, Ken Cowen, Thomas Kelly, Amy Dindal, Battelle
John McKernan, U.S. EPA
-------
Notice
The U.S. Environmental Protection Agency, through its Office of Research and Development,
funded and managed, or partially funded and collaborated in, the research described herein. It
has been subjected to the Agency's peer and administrative review. Any opinions expressed in
this report are those of the author(s) and do not necessarily reflect the views of the Agency,
therefore, no official endorsement should be inferred. Any mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
-------
FOREWORD
The EPA is charged by Congress with protecting the nation's air, water, and land resources.
Under a mandate of national environmental laws, the Agency strives to formulate and implement
actions leading to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, the EPA's Office of Research and
Development provides data and science support that can be used to solve environmental
problems and to build the scientific knowledge base needed to manage our ecological resources
wisely, to understand how pollutants affect our health, and to prevent or reduce environmental
risks.
The Environmental Technology Verification (ETV) Program has been established by the EPA to
verify the performance characteristics of innovative environmental technology across all media
and to report this objective information to permitters, buyers, and users of the technology, thus
substantially accelerating the entrance of new environmental technologies into the marketplace.
Verification organizations oversee and report verification activities based on testing and quality
assurance protocols developed with input from major stakeholders and customer groups
associated with the technology area. ETV consists of six environmental technology centers.
Information about each of these centers can be found on the Internet at http://www.epa. gov/etv/.
Effective verifications of monitoring technologies are needed to assess environmental quality
and to supply cost and performance data to select the most appropriate technology for that
assessment. Under a cooperative agreement, Battelle has received EPA funding to plan,
coordinate, and conduct such verification tests for "Advanced Monitoring Systems for Air,
Water, and Soil" and report the results to the community at large. Information concerning this
specific environmental technology area can be found on the Internet at
http://www.epa.gov/etv/centers/centerl.html.
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the
verification test, analyze the data, and prepare this report. We acknowledge the financial support
of the Electrical Power Research Institute (EPRI). Quality assurance oversight was provided by
Michelle Henderson, U.S. EPA, and Zachary Willenberg, Battelle. We would also like to thank
the Tennessee Valley Authority and the staff at the host facilities for their support and for
providing a venue for the performance of this verification test. Finally, the authors thank Dr.
Will Ollison of the American Petroleum Institute, Mr. Chuck Dene of EPRI, and Mr. Dennis
Mikel of U.S. EPA's Office of Air Quality Planning and Standards for their review of the
test/QA plan and this verification report.
in
-------
Contents
Acknowledgments iii
Table of Figures v
Table of Tables v
List of Abbreviations vi
Chapter 1 Background 1
Chapter 2 Technology Description 2
Chapters Test Design and Procedures 5
3.1 Introduction 5
3.2 Field Site 6
3.3 Test Procedures 6
3.3.1 Reference Method Sampling 7
3.3.2 Dynamic Spiking 8
Chapter 4 Quality Assurance/Quality Control 10
4.1 Reference Method QA/QC 10
4.1.1 Precision 10
4.1.2 Duplicate Analysis 11
4.1.3 Field Blanks 12
4.1.4 Spiked Trains 13
4.2 Audits 13
4.2.1 Performance Evaluation Audit 13
4.2.2 Technical Systems Audit 14
4.2.3 Data Quality Audit 14
4.3 QA/QC Reporting 14
Chapters Statistical Methods 15
5.1 Accuracy 16
5.2 Data Completeness 16
5.3 Operational Factors 16
Chapter 6 Test Results 18
6.1 Accuracy 21
6.2 Data Completeness 29
6.3 Operational Factors 29
6.3.1 Ease of use 29
6.3.2 Maintenance 30
6.3.3 Consumables Used/Waste Generated 30
Chapter 7 Performance Summary 31
Chapter 8 References 33
IV
-------
Figures
Figure 2-1. ThePicarro Model G1103-C ammonia analyzer 2
Figure 2-2. Block diagram of thePicarro Model G1103-C ammonia analyzer 3
Figure 3-1. Dilution probe installation and reference method sampling ports 7
Figure 6-1. Comparison of Model Gl 103-c readings to the boiler output and urea injection rate
between July 24 - September 10, 2009 18
Figure 6-2. Results of through-the-probe dynamic spiking 20
Figure 6-3. Comparison of CRDS readings to the boiler output and urea inj ection rate between
September 11 - October 29, 2009 21
Figure 6-4. Comparison of reference method results and Model Gl 103-c measurements from
October 27, 2009 22
Figure 6-5. Comparison of reference method results and Model Gl 103-c measurements from
October 28, 2009 22
Figure 6-6. Comparison of reference method results and Model Gl 103-c measurements from
October 29, 2009 23
Figure 6-7. Initial multi-point calibration of Picarro Model Gl 103-c 26
Figure 6-8. Final multi-point calibration of Picarro Model Gl 103-c 26
Tables
Table 4-1. Summary of Reference Method Results 11
Table 4-2. Summary of RPD in Duplicate Analyses for Reference Method Samples 12
Table 4-3. Results of Field Blank Analyses 13
Table4-4. Results of Spiked Train Analyses 13
Table 4-5. Performance Evaluation Audit Results of Ammonia Analysis by Ion
Chromatography 14
Table 6-1. Comparison of Field Test Reference Method Results and Average Model Gl 103-c
Measurements 24
Table 6-2. Comparison of Laboratory Calibration Check Results and Average Model Gl 103-c
Measurements 25
Table 6-3. Linearity Results for the Model Gl 103-c 26
Table 6-4. Results of Zero Checks of the Model G1103-C 27
Table 6-5. Results of Span Checks of the Model Gil 03-c 28
Table 6-6. Summary of Zero/Span Drift Checks of the Model Gl 103-c 28
Table 6-7. Summary of Rise/Fall Times of the Model Gl 103-c 29
Table 6-8. Summary of Maintenance Activities Performed During Verification Testing 30
Table 7-1. Summary of Verification Test Results for the Picarro Model Gl 103-c 31
-------
List of Abbreviations
AMS Advanced Monitoring Center
APH air preheater
°C Degrees Celsius (or Centigrade)
CRDS cavity ring-down spectroscopy
CTM Conditional Test Method
DFB distributed feedback laser
EPA U.S. Environmental Protection Agency
ESP electrostatic precipitator
ETV Environmental Technology Verification
1C ion chromatography
NIST National Institute of Standards and Technology
NOX nitrogen oxides
PE performance evaluation
QA quality assurance
QC quality control
QMP quality management plan
ppm parts per million
RPD relative percent difference
SCR selective catalytic reduction
SNCR selective non-catalytic reduction
TQAP test/QA plan
TSA technical systems audit
TVA Tennessee Valley Authority
|j,sec microseconds
VI
-------
Chapter 1
Background
The U.S. Environmental Protection Agency (EPA) supports the Environmental Technology
Verification (ETV) Program to facilitate the deployment of innovative environmental
technologies through performance verification and dissemination of information. The goal of the
ETV Program is to further environmental protection by accelerating the acceptance and use of
improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-
quality, peer-reviewed data on technology performance to those involved in the design,
distribution, financing, permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized testing organizations; with stakeholder groups
consisting of buyers, vendor organizations, and permitters; and with the full participation of
individual technology developers. The program evaluates the performance of innovative
technologies by developing test plans that are responsive to the needs of stakeholders,
conducting field or laboratory tests (as appropriate), collecting and analyzing data, and preparing
peer-reviewed reports. All evaluations are conducted in accordance with rigorous quality
assurance and quality control (QA/QC) protocols to ensure that data of known and adequate
quality are generated and that the results are defensible.
The EPA's National Risk Management Research Laboratory and its verification organization
partner, Battelle, operate the Advanced Monitoring Systems (AMS) Center under ETV. Real-
time ammonia monitoring systems for the measurement of excess ammonia (ammonia slip) were
identified as a priority technology category for verification through the AMS Center stakeholder
process. The AMS Center recently evaluated the performance of the Picarro, Inc., Model
Gl 103-c ammonia analyzer for analysis of ammonia in flue gas at a coal fired power plant.
-------
Chapter 2
Technology Description
This report provides results for the verification testing of the Picarro, Inc., Model Gl 103-c. The
following is a description of the Model Gl 103-c, based on information provided by the vendor.
The information provided below was not verified in this test.
The Model Gl 103-c, shown in Figure 2-1, is an analyzer designed to measure ammonia at the
parts-per-billion level in the presence of carbon dioxide, water vapor, and other gas species
present in flue gas streams. This analyzer is based on cavity ring-down spectroscopy (CRDS),
which is a technique in which a gas sample is introduced into a high-finesse optical cavity and
the optical absorbance of the sample is determined, thus providing concentration or isotopic ratio
measurements of a particular gas species of interest.1'2
Figure 2-1. The Picarro Model Gl 103-c ammonia analyzer.
-------
Figure 2-2 shows a block diagram of the Picarro CRDS analyzer. The components which make
up a basic CRDS instrument are a laser, a high-finesse optical cavity consisting of two or more
mirrors, and a photo-detector, where finesse is ratio of the free spectral range of the optical
cavity divided by the full width at half maximum of the transmission bands. Operationally, light
from a laser is injected into the cavity through one partially reflecting mirror. The light intensity
inside the cavity then builds up over time and is monitored through a second partially reflecting
mirror using a photo-detector located outside the cavity. The "ring-down" measurement is made
by rapidly turning off the laser and measuring the light intensity in the cavity as it decays
exponentially with a time constant, r, that depends on the losses due to the cavity mirrors and the
absorption and scattering of the sample being measured. Measurement of gas concentration is
based on the fact that the decay time constant is shorter when an absorbing gas is present in the
cavity than when no absorbing gas is present. After shutting off the laser, most of the light
remains trapped within the cavity for a relatively long period of time (i.e., microseconds [|j,sec]),
producing an effective path length of tens of kilometers through the sample.
Patented High-finesse Cavity
! >
Tunable
Diode Laser X
Patented
Wavelength
Monitor
(|
nr
it x.
T
i
Cell Volume = 35cc
x£: Sample Gas Inlet
Temperature Gauge
Pressure Gauge
Outlet Gas Flow
(to pump)
Photodetector
Laser Control
Electronics
Data Collection &
Analysis Electronics
Figure 2-2. Block diagram of the Picarro Model GllOS-c ammonia analyzer.
The Model Gl 103-c utilizes a telecom-grade distributed feedback (DFB) laser. Light from the
DFB laser is transported to a wavelength monitor via a polarization-maintaining optical fiber.
The analyzer is designed to simultaneously measure optical absorption using a proprietary
traveling wave cavity and the optical frequency at which the absorption occurs using a
proprietary wavelength monitor. The temperature and pressure of the ambient air sample
continuously flowing through the optical cavity are regulated at all times to 45 °C and 140 Torr,
respectively. A typical empty cavity decay constant, T, is 40 usec for this instrument. The
normalized reproducibility of the measured ring-down time constant (Ar/r) is better than 0.02%.
With a ring-down acquisition rate of 100 Hz, the typical sensitivity (1 sigma) of the instrument is
1.6>
-------
about 0.025 ppb with five of minutes of averaging. The Model Gl 103-c has an operational range
of 0-10 ppm NHs, with an optional extended range up to 50 ppm.
The analyzer continuously scans the laser over individual ammonia spectral features and records
the absorption loss and wavelength at each spectral point. Each spectrum is comprised of
absorption loss as a function of optical frequency. The concentration is proportional to the area
under each measured spectral feature. Concentration measurements are provided approximately
every second, corresponding to a total of 100 ring-down and wavelength monitor measurements.
The wavelength monitor used in the analyzer is solid-state in design and has no moving parts. It
is designed to provide wavelength measurements over a frequency range corresponding to
greater than 100 nm. The wavelength precision (defined as the repeatability of the wavelength
measurement at a single spectral point) is approximately 1MHz (la) or approximately 3 x 10"5
cm"1. The relative accuracy, defined as the repeatability of the difference of the wavelength
measurement between two spectral points separated by approximately 1 GHz or approximately
0.03 cm"1 (the width of a typical absorption line at a typical operating pressure of 140 Torr)
during a spectral scan is approximately 0.3MHz (approximately 1 x 10"5 cm"1). The size and
shape of the ammonia spectral line at 6,548 cm"1 is a sensitive function of the temperature and
pressure of the sample. Therefore, the analyzer is designed to control the sample gas temperature
to a precision of a few hundredths of a degree (Icr) over ambient temperatures ranging from 10 to
35°C and the sample pressure to a precision of 0.05 Torr (Icr). In the analyzer, a combination of
proportional valves (for flow control) is used to maintain the cavity at a known constant pressure.
Ammonia is a toxic, reactive, and corrosive compound that is soluble in water. These
characteristics are known to prolong the transport time, thereby slowing gas monitor response
times in a closed-path system that may contain water. Care has been taken to use Teflon for all
wetted materials to keep this effect to a minimum.
The Model Gl 103-c has dimensions of 43 x 25 x 59 cm (17" x 9.75" x 23") including the base,
and can be rack mounted or operated on a bench top. The approximate purchase price of the
Model Gl 103-c is $55,000.
-------
Chapter 3
Test Design and Procedures
3.1 Introduction
Selective catalytic reduction (SCR) and selective non-catalytic reduction (SNCR) technologies
are commonly used on coal-fired power plants to reduce nitrogen oxides (NOX) emissions
through chemical reaction with ammonia (NHs). These technologies require the introduction of
large quantities of NHa to reduce the NOX emissions in a stoichiometric fashion. However,
frequently a small amount of excess NHa is introduced that subsequently is emitted in the flue
gas. This excess ammonia, called slip, is frequently difficult to measure because of its relatively
low concentration. CRDS systems have been proposed as a potentially viable means of
monitoring ammonia slip because of the high sensitivity of the technique.
The purpose of this verification test was to generate performance data on CRDS monitoring
technologies with a particular focus on monitoring of ammonia under normal operating
conditions in a full-scale coal-fired power plant utilizing SCR or SNCR NOx control technology.
The test was conducted over a period of approximately 90 days and involved the continuous
operation of the Picarro Model Gl 103-c at an operational coal-fired power plant. During testing,
the Model Gl 103-c continuously monitored ammonia slip concentration in the flue gas
downstream of NOX control technology.
This verification test was conducted according to procedures specified in Test/QA Plan for
Verification of Cavity Ring-down Spectroscopy Systems For Ammonia Monitoring in Stack Gas 4
(TQAP) and adhered to the quality system defined in the ETV AMS Center Quality Management
Plan (QMP)5. The testing conducted satisfied EPA QA Category III requirements. The TQAP
and/or this verification report were reviewed by:
• Charles Dene, Electric Power Research Institute (TQAP only)
• William Ollison, American Petroleum Institute
Dennis Mikel, U.S. EPA
Kristen Benedict, U.S. EPA (report only)
The Picarro Model Gl 103-c was evaluated in the field test on the following performance
parameters:
> Accuracy
> Data completeness
> Operational factors including ease of use, maintenance requirements, and
consumables used/waste generated.
-------
An attempt was made to evaluate precision, calibration/zero drift, linearity, and response time in
the field test through dynamic spiking with compressed standard gases. However, because of
complications associated with delivering the compressed gases to the analyzer in the field (e.g.,
long equilibration times), these parameters were evaluated in laboratory testing subsequent to the
field study. Duplicate reference method samples collected during the last week of field testing
were used to assess the comparability of the CRDS measurements with the standard reference
method results. Data completeness was determined from a review of the valid data collected
during the verification testing period. Operational performance parameters such as ease of use,
maintenance requirements, and consumables used/waste generated were determined from
observations by the Battelle field testing staff and from on-site staff. This test was not intended
to simulate long-term performance of the Model Gl 103-c at a monitoring site.
3.2 Field Site
Testing was performed at a full-scale coal fired power plant owned and operated by the
Tennessee Valley Authority (TVA). Initially, the Model Gl 103-c was installed at a unit of the
Kingston Fossil Plant in Kingston, TN on June 25, 2009, at a location downstream of the SCR
and upstream of the air preheater (APH). However, because of low power demand, the unit on
which the Model Gl 103-c was installed was not in use for several consecutive weeks. This
prompted a move of the Model Gl 103-c to a different unit at the plant on July 20, 2009. After
that installation, reference method sampling was performed on July 20-21, 2009, using a
modified version of EPA CTM-0273 to assess the ammonia concentration in the flue gas. The
results of those reference measurements indicated that there was no detectable ammonia slip in
the sample stream. Consequently, the decision was made to relocate the testing to a facility
where appreciable ammonia slip levels were likely to be present.
An alternative test facility was identified and the Model Gl 103-c was moved and installed at the
alternate power plant on July 24, 2009. Testing was conducted through the end of October. The
power plant where testing was performed included multiple 200 megawatt boilers, each
configured with SNCR NOX reduction capabilities that involved the injection of an aqueous urea
solution into each boiler. During testing, the Picarro Model Gl 103-c system was installed in an
environmentally controlled instrument shelter to maintain temperature stability of the analyzer.
The Model Gl 103-c was installed along with other continuous emission monitors (CEMs) and
supplied with flue gas that was sampled from the superheated section of one boiler. The flue gas
was drawn from the duct using a dilution probe (100:1 dilution) incorporating a particulate filter,
and was delivered to the Model Gl 103-c through approximately 150-200 feet of heated (100 °C)
Teflon tubing. Dilution of the flue gas was deemed necessary to condition the sample and
minimize potential condensation processes in the sample line.
3.3 Test Procedures
The flue gas delivered to the Model 1103-c was drawn from a section of duct work upstream of
the APH and electrostatic precipitator (ESP). The ports for the reference method sampling were
located directly in-line with the dilution probe delivering the flue gas to the Model Gl 103-c.
Figure 3-1 shows the locations of the dilution probe and reference method sampling trains at the
test facility. The dilution probe extended inward from the blue box mounted on the wall of the
duct as shown in this figure. The diluted flue gas was delivered to the instrument trailer through
-------
the heated umbilical cord protruding from the box. The port used for the collection of the
reference method samples was directly below the dilution sampling point, and is shown in Figure
3-1 with a sampling probe installed.
3.3.1 Reference Method Sampling
Reference sampling was performed according to CTM-0273, with the following modifications:
Sampling was not conducted isokinetically since only gaseous ammonia was measured
by the Model Gil03-c.
• Sampling was conducted with a nozzle since isokinetic sampling was not necessary.
To instrument trailer
Duct wall
To impingers
Reference method probe
Figure 3-1. Dilution probe installation and reference method sampling ports.
Each reference method test run involved the simultaneous collection of samples from two
collocated trains. Thus each reference method test run provided two reference ammonia samples
-------
for comparison to the Model Gl 103-c data. The sampling duration for each run was typically
between 20 and 30 minutes. Each collected sample was analyzed on-site by ion chromatography
(1C) for ammonia as ammonium ion (NH4+). For each test run the Model Gl 103-c results and
the analytical results of the individual reference method measurements were normalized to 7%
O2. The reference method sampling schedule was compressed to three days to allow for set-up
and tear-down activities for the test crew.
In addition to the reference method samples, field blank samples and field spike samples were
recovered from separate sampling trains on each day that reference method samples were
collected. Each field blank and spike train was transported to the sampling location and then
recovered without sampling the flue gas. The samples were analyzed by the same procedures as
normal samples.
3.3.2 Dynamic Spiking
The Picarro Model Gl 103-c was challenged in the field with ammonia compressed gas
standards. Initially, the ammonia gas standard was supplied directly to the Model Gl 103-c (i.e.,
disconnected from the dilution probe) after off-line dynamic dilution with zero air using an
Environics Model 6100 mass flow dilution system. However, this method was found impractical
for routine testing activities because of the time required for equilibration of the dilution system
and an apparent effect from residual ammonia gas in the sample delivery system.
Additionally, on-line dynamic spiking was attempted in the field by introducing the ammonia gas
standard into the probe tip upstream of the particulate filter such that the ammonia spike passed
through the dilution probe where it was mixed with diluent gas at a ratio of 1:100. However,
because of the length of tubing required to deliver the compressed gas standard from the gas
cylinder to the probe tip and then back to the analyzer, and the "sticky" nature of the ammonia
gas, the time required for the ammonia concentration to equilibrate also made this method of
dynamic spiking impractical. Furthermore, since several process-control continuous emission
monitors (CEMs) sampled from the same gas stream, through the probe spiking required that
these CEMs remain off-line during the dynamic spiking procedure. Extended periods of off-line
operation was not acceptable to the plant operator. Consequently dynamic spiking was not
performed in the field during this verification test. Instead, laboratory testing was conducted as
described in Section 3.3.3 to assess precision, linearity, zero/calibration drift, and response time.
3.3.3 Laboratory Testing
After completion of the field testing, the Picarro Model Gl 103-c was challenged under
laboratory conditions by supplying zero air and an ammonia reference standard diluted over a
range of target concentrations. During this laboratory testing, ammonia was delivered to the
Picarro Model Gl 103-c using a permeation oven (Vici Metronics, Dynacalibrator Model 340)
operated at 30 ± 1 °C with a certified ammonia permeation tube (Vici Metronics, Part Number
181-055-0140-F56-C30, certified rate 1,351 ng/min ± 0.15%). Dilution air was supplied using a
zero air generator (Aadco, Model 737). Separate lines were used to deliver the ammonia and
zero air to the Model Gl 103-c, to avoid delays due to equilibration of the delivery lines.
Ammonia from the permeation oven was continuously flowed through the ammonia delivery
line, even when not connected to the Model Gl 103-c, to prevent potential variations in ammonia
concentrations caused by adsorption and desorption in the line. At each concentration level, the
-------
air flow rate through the permeation oven was measured using a calibrated flow meter (Bios,
DryCal DC-2).
Reference ammonia concentrations [or C, in ng/L (or equivalently ug/m3)] were calculated
according to:
c = K-^ m
F ' '
where K is equal to 1.439 and is determined from the ideal gas constant and the molecular weight
of ammonia, P is the permeation rate of the ammonia permeation tube (ng/min), and F is the
dilution gas flow rate (L/min).
-------
Chapter 4
Quality Assurance/Quality Control
QA/QC procedures and all verification testing were performed in accordance with the test/QA
plan for this verification test4 and the quality management plan (QMP) for the AMS Center.5
QA/QC procedures and results are described below.
4.1 Reference Method QA/QC
The following sections describe the QA/QC procedures employed in the collection and analysis
of reference samples.
4.1.1 Precision
The precision of duplicate reference method results was calculated for each test run as
RPD= Ri~Ri xlOO ,,i
(R,+R2)/2 W
where Rj and R2 are the reference method results for the duplicate trains. Table 4-1 presents the
results of the reference method analyses. Note that the two separate train results presented are
each the average of duplicate analyses for each sample (see Section 4.1.2).
Because the boiler load varied throughout the testing period as the demand changed, the
ammonia levels in the facility also varied. Consequently, during the reference testing the
ammonia levels are likely to have varied considerably more than the anticipated ± 35%
prescribed in the test/QA plan.4 As a result, the reference data were not screened to identify
outliers relative to the mean of the reference method data, but were screened to identify those
paired samples that exceeded 35% RPD.
Of the 24 reference method sampling runs, 10 exhibited RPDs that exceeded 35% (Table 4-1).
Of the test runs that exceeded 35% RPD, the RPD values ranged from 38 to 83% and had an
average of 53%. No assignable cause was identified for the observed differences between the
duplicate trains. These results are in contrast to previous reference method sampling at this facility
during July 24 -25 that showed no significant differences between duplicate trains. Battelle's Quality
Manager was on-site during the collection of several of the reference method samples to conduct
a technical systems audit and observed nothing in the sample collection or analytical procedures
to account for the observed differences. Each reference method sample was analyzed on-site by
10
-------
duplicate 1C instruments (see Section 4.1.2) and showed good precision between the duplicate
analyses. Previous spot checks of the ammonia concentration from different ports suggested a
gradient across the width of the duct. However, since the duplicate probes were sampling from
the same port, the proximity of the probes to one another during the sample collection
(approximately 2-3 inches) was likely not sufficient to account for the observed differences. In
three test runs, it appeared that one of the duplicate trains experienced a leak which resulted in
measured ammonia concentrations near zero. Therefore, these test runs were flagged as outliers
in Table 4-1 and were not used in data comparisons to the Model Gl 103-c. Thus, 21 sets of
duplicate reference method results were compared to the results generated by the Model Gl 103-
c.
Table 4-1. Summary of Reference Method Results
Date
10/27/2009
10/28/2009
10/29/2009
Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Ammonia Slip (ppm @ 7% O2)
Train 1
11.7
17.8
22.1
19.3
15.8
1.5*
2.2*
16.4
19.0
21.2
17.1
18.6
16.4
14.1
13.4
13.6
9.0
9.5
13.9
15.9
14.1
13.1
10.0
9.1
Train 2
28.4
36.0
27.1
20.5
13.5
19.1
16.6
17.7
15.3
11.7
10.0
19.2
12.9
18.6
9.9
13.9
0.2*
15.6
21.6
24.7
20.6
15.0
16.5
14.5
RPD
83.1%
67.7%
20.2%
6.1%
16.0%
*
*
7.6%
21 .6%
57.9%
52.6%
3.0%
24.0%
27.7%
30.6%
2.7%
*
48.8%
43.4%
43.4%
37.6%
13.9%
49.6%
46.1%
* Leak suspected in sampling train during sample collection. No RPD calculated.
4.1.2 Duplicate Analysis
Each reference method sample collected was analyzed on-site on duplicate, collocated 1C
instruments. To assess the analytical precision of the reference method, the RPD of the duplicate
11
-------
analyses for each sample were calculated and are presented in Table 4-2. On average, the results
of the duplicate analyses from the collocated 1C instruments were within 5%. In all but three
instances among the 48 RPD values in Table 4-2, the calculated RPD was within 10%.
4.1.3 Field Blanks
Field blank samples were collected each day during reference method testing and analyzed to
assess the level of potential contamination in the preparation, handling, and analysis of the
reference method trains. For these field blanks, reference method trains were prepared and
transported to the duct for sampling but were not used for sampling. The trains were recovered
and analyzed on-site as normal samples using duplicate 1C instruments. The results of the
analysis of these samples are presented in Table 4-3 in terms of the equivalent flue gas ammonia
concentration based on typical reference method sampling conditions. In all cases, the measured
ammonia in the blank samples was below 1% of the calibration range and far below the expected
concentration in the reference samples (typically 10 to 20 ppm).
Table 4-2. Summary of RPD in Duplicate Analyses for Reference Method Samples
Date
10/27/2009
10/28/2009
10/29/2009
Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Average
Standard Deviation
Maximum
RPD
Train 1
3.0%
1 .2%
3.9%
3.4%
6.7%
8.5%
5.0%
0.6%
1.8%
5.1%
4.1%
2.2%
5.4%
5.5%
10.0%
7.1%
7.9%
5.8%
4.7%
4.3%
3.8%
4.7%
5.2%
3.5%
4.7%
2.3%
10.0%
Train 2
5.0%
2.6%
0.3%
1.0%
1 .6%
6.6%
1.5%
13.9%
2.4%
1 .2%
2.3%
2.3%
1.1%
8.8%
0.1%
4.0%
31 .9%
0.2%
2.2%
4.3%
0.8%
3.1%
1.7%
5.4%
4.3%
6.6%
31 .9%
12
-------
Table 4-3. Results of Field Blank Analyses
Date
10/27/2009
10/28/2009
10/29/2009
Ammonia (ppm)
IC1
0.007
0.011
0.028
1C 2
0
0
0.025
4.1.4 Spiked Trains
On each day of reference method sampling, a spiked reference method train was prepared using a
National Institute of Standards and Technology (NIST)-traceable ammonia standard. These
trains were prepared and transported to the sampling duct but were not used for sampling. The
trains were recovered and analyzed on-site as normal samples using duplicate 1C instruments.
The results of the analysis of these samples are presented in Table 4-4 in terms of equivalent flue
gas ammonia concentration. All spike recoveries were within 10% of the target NHa spikes.
Table 4-4. Results of Spiked Train Analyses
Date
10/27/2009
10/28/2009
10/29/2009
Spike Target,
(ppm)
1.66
1.67
1.45
NH3 Analysis, (ppm)
IC1
1.50
1.65
1.38
1C 2
1.53
1.55
1.43
Percent Difference
IC1
-9.6%
-1 .5%
-5.1%
1C 2
-8.1%
-7.7%
-1.5%
4.2 Audits
4.2.1 Performance Evaluation Audit
Performance evaluation (PE) audits were made to ensure the quality of the critical
measurements. The thermocouples used for the stack temperature measurements were audited
by making collocated measurements using an independent thermocouple with a NIST-traceable
calibration. The results of these audits showed agreement within 2% in absolute temperature
between the test and audit thermocouples.
Two balances were used during the preparation and analysis of the reference method samples,
one for measurements of large masses (e.g., > 100 g) and the other for more precise
measurements of smaller weights (e.g., < 50 g). Both balances were audited using a set of NIST-
traceable weights. The results of the PE audit showed differences between the measured and
actual weights of less than 1%.
PE audit samples were prepared from an independent ammonium standard solution and were
analyzed by the duplicate 1C instruments used for the analysis of the reference method samples.
Table 4-5 summarizes the results of those analyses. In all cases, the results of the PE audit
showed differences between the measured and nominal concentrations of less than 10%.
13
-------
Table 4-5. Performance Evaluation Audit Results of Ammonia Analysis by Ion
Chromatography
Date
10/26/2009
10/27/2009
Nominal
Concentration
2.86 ppm
2.86 ppm
Measured Concentration
1C 1 (% Diff)
2.88 (0.5%)
2.89 (0.8%)
1C 2 (% Diff)
3.13(9.4%)
2.95(3.1%)
4.2.2 Technical Systems Audit
A technical systems audit (TSA) was performed on October 28, 2009 by the Battelle Quality
Manager, Mr. Zachary Willenberg. Because of difficulties in implementation of the field testing
portion of this verification test and scheduling of the stack testing crew, the TSA was conducted
near the end of the verification test rather than near the beginning of the test as called for in the
TQAP. A checklist was prepared by Battelle QA Manager, and approved by the EPA QA
Manager and used for performance of the audit. Minor discrepancies were noted between the
actual procedures and those specified in CTM-027.3 In particular, isokinetic sampling was not
performed, the recovery of the sample train occurred at the collection point not at a different
laboratory, and a single sample bottle was used for the collection of all sections of the sample
train, as opposed to the use of separate bottles for the different sections. Also, a duct transverse
was not conducted prior to collection of the reference samples. Rather, the sampling probes
were inserted to the same depth as the dilution probe to allow for sampling under similar
conditions. None of the items noted had any apparent bearing on the sampling or analytical
results. An audit report was prepared and provided to the Verification Test Coordinator for
review of any findings.
4.2.3 Data Quality Audit
Records generated in the verification test received a one-to-one review before these records were
used to calculate, evaluate, or report verification results. Data were reviewed by a Battelle
technical staff member involved in the verification test. The person performing the review added
his/her initials and the date to a hard copy of the record being reviewed. At least 10% of the data
generated were reviewed along with the laboratory record book. Data were reviewed from initial
acquisition through final reporting. An audit report was prepared and provided to the
Verification Test Coordinator for review of any findings.
4.3 QA/QC Reporting
Each audit was documented in accordance with Sections 3.3.4 and 3.3.5 of the QMP for the ETV
AMS Center.5 The results of the TSA and ADQ were submitted to the EPA.
14
-------
Chapter 5
Statistical Methods
The statistical methods used to evaluate the quantitative performance factors listed in Section 3.1
are presented in this chapter. Qualitative observations were also used to evaluate verification test
data.
5.1 Accuracy
The accuracy of Model Gl 103-c readings was evaluated in two ways. Firstly, the RPD of the
Model Gl 103-c readings relative to the field reference results and the laboratory calibration
check results was calculated to assess accuracy. The RPD was calculated by adapting Equation 2
in Section 4.1.1 (i.e., by using the field reference method results or the laboratory calibration
check results and Model Gl 103-c results, instead of the duplicate reference results in the RPD
calculation.)
Additionally, the relative accuracy (RA) of the Picarro Model Gl 103-c readings was also
assessed by comparison to the reference method results from the field test and the laboratory
calibration check results based on Equation 3:
RA =
d +ta ^-
a + l"-i r- (3)
x
where d refers to the difference between the calculated ammonia concentration from the
reference method or calibration check and the average of the CRDS measurements recorded
during the respective measurement periods, and x corresponds to the mean of the measured
reference method results or calibration check concentration. Sd denotes the sample standard
deviation of the differences, while tan-i is the t value for the 100(1 -a)th percentile of the
distribution with n-1 degrees of freedom. The relative accuracy was determined for a a value of
0.025 (i.e., 97.5 percent confidence level, one-tailed). The RA calculated in this way can be
PI
interpreted as an upper confidence bound for the relative bias of the analyzer, i.e., L=^-, where the
x
superscript bar indicates the mean value of the differences, or of the reference values.
15
-------
5.2 Linearity
Linearity was assessed from a linear regression analysis of the multipoint calibration data
collected during the laboratory testing. The analysis used the theoretical ammonia
concentrations delivered to the Picarro Model Gl 103-c as the independent variable and the
measured results from the Model Gl 103-c as the dependent variable. The theoretical ammonia
concentrations were based on delivery of a certified ammonia mass emission rate in a calibrated
dilution flow rate. The results of the multipoint calibration were plotted and linearity was
expressed in terms of slope, intercept, and coefficient of determination (r2).
5.3 Precision
Since dynamic spiking was not performed in the field, precision was calculated in a different
fashion than described in the TQAP. For this verification test, precision was calculated in terms
of the percent relative standard deviation (RSD) of the CRDS measurements made during the
zero/span checks conducted during the laboratory testing. During each zero/span check, the
mean and standard deviation of the readings recorded for each check were calculated during the
last five minutes of the check. The RSD was calculated as the standard deviation of the mean
multiplied by 100 and divided by the mean, for each check. This measure of precision differs
from the originally planned measure of precision in that the Model Gl 103-c was not measuring
flue gas during the addition of the gas standard. Thus, precision measured here is a more direct
measure of the instrument precision as it does not include variability in ammonia concentrations
from the flue gas.
5.4 Zero/Calibration Drift
Calibration and zero drift were reported in terms of the mean, RSD, and range (maximum and
minimum) of the readings obtained from the Model Gl 103-c daily measurement of the same
ammonia standard gas, and of zero gas. These results, along with the range of the data, indicate
the daily variation in zero and standard readings over the two week period of laboratory
zero/calibration drift measurements.
5.5 Response Time
Response time was assessed in terms of both the rise and fall times of the Model Gl 103-c when
sampling the ammonia gas standard delivered during the zero/calibration drift checks. Rise time
(i.e., 0% - 95% response time) was determined from the response when the gas delivered to the
Model Gl 103-c was switched from zero gas to the ammonia standard. After a stable reading had
been achieved, the fall time (i.e., the 100% to 5% response time) was determined from the
response when the gas delivered to the Model Gl 103-c was switched from the ammonia standard
back to zero gas.
5.6 Data Completeness
Data completeness was assessed based on the overall data return achieved by the Model Gl 103-c
during the testing period. This calculation determined the total number of apparently valid data
points reported by the monitoring system divided by the total number of data points potentially
available in the entire field period. The causes of any incompleteness of data return were
investigated based on operator observations or vendor records, and were noted in the discussion
of data completeness results.
16
-------
5.7 Operational Factors
Operational factors regarding ease of use, data output, maintenance needs, consumables used,
etc., were evaluated based on observations recorded by Battelle and on-site support staff, and
were explained by the vendor as needed. Battelle or testing staff recorded all activities
performed on the monitoring systems in a laboratory record book maintained at the test site,
including observations on the performance factors given above. Examples of information
recorded in the record books include the use or replacement of any consumables; vendor effort
(e.g., time on site) for repair or maintenance; the duration and causes of any down time or data
acquisition failure; and observations about ease of use of the Model Gl 103-c. These
observations are summarized in this report to aid in describing the performance of the Model
Gl 103-c.
17
-------
Chapter 6
Test Results
Figure 6-1 shows the ammonia concentrations measured by the Model Gl 103-c along with the
relative boiler load and urea injection rates at the host facility during the period from July 24-
September 10, 2009. During this period the Model Gl 103-c sampled from the reheat side of
Unit 2 at the host facility. The ammonia readings shown represent approximately 10-minute
averages of the data from the Model Gl 103-c and are corrected for the 100-fold dilution
introduced by the sampling probe. The boiler output and urea injection rate values are presented
in arbitrary units. The periodic spikes in the Model Gl 103-c readings coincide with periods
when the daily zero/span calibration checks were performed. During the morning of August 27,
operation of the boiler was stopped because of low demand and the unit remained out of
operation until September 11, when the Model Gl 103-c was moved to a different boiler.
0.4
CRDS Readings
BoilerOutput (Arb.)
Urea Inj. Rate (Arb.
07/24/09 07/31/09 08/07/09 08/14/09 08/21/09 08/28/09 09/04/09 09/11/09
Figure 6-1. Comparison of Model GllOS-c readings to the boiler output and urea injection
rate between July 24 and September 11, 2009.
In general, the ammonia concentrations measured by the Model Gl 103-c during the July 24-
September 11 period were substantially below the concentrations expected based on 1C reference
18
-------
method sampling/analysis performed on Unit 2 to spot-check ammonia concentrations during
instrument installation and shakedown on July 24-25. During that period, results from the
reference method sampling indicated ammonia concentrations of approximately 1 to 5 ppm,
whereas the Model Gl 103-c measurements of the flue gas during this period showed ammonia
concentrations of approximately 0.020 to 0.060 ppm. The cause for the discrepancy could not be
determined. To assess whether this was the result of delivering sample through the relatively
long sampling line, the Model Gl 103-c was moved from the instrument trailer to the duct and
several attempts were made to deliver conditioned flue gas to the Model Gl 103-c without the use
of long sampling lines. In one instance, the flue gas was sampled through the reference method
probe and conditioned using a Baldwin Environmental Model 10410 electronic water condenser
prior to delivery to the Model Gl 103-c. Also, flue gas was collected directly into a Tedlar
sampling bag and delivered to the Model Gl 103-c, both with and without dilution with nitrogen.
However, in all instances upon cooling the vapor phase ammonia appeared to co-condense with
the water vapor in the flue gas. Analysis of the condensate indicated ammonia concentrations
consistent with the reference method measurements, indicating the presence of ammonia in the
flue gas. No attempt was made to collect an impinger sample of the diluted gas (similar to the
reference method collection) since the 100-fold dilution of the flue gas would have resulted in a
sampling time of approximately 2,000 minutes. Thus, during the time when the Model G-l 103c
was installed on Unit 2, there was no clear evidence that ammonia concentrations measured by
the Model G-l 103c in the dilution probe were representative of the actual concentrations in the
duct.
On September 11, the sampling point was switched such that the Model Gl 103-c sampled from
the superheat side of Unit 1 rather than the reheat side of Unit 2 at the host facility. Prior to
routine sampling from Unit 1, several dynamic spiking tests were performed using compressed
gas standards to confirm proper operation of the Model Gl 103-c. Off-line dynamic spiking was
performed by disconnecting the Model Gl 103-c from the dilution probe and delivering known
concentrations of ammonia to the Model Gl 103-c using an off-line dynamic dilution system
(Environics, Model 6100) with zero air as the ammonia diluent. The results of those tests
indicated that although the Model Gl 103-c responded to the ammonia gas standards that were
supplied at the expected concentrations, there were several drawbacks to the sample delivery that
precluded routine use of this testing method for challenging the Model Gl 103-c. In particular, a
substantial equilibration time was associated with delivery of either ammonia or zero air, and
successive delivery of a target ammonia concentration separated by delivery of zero air
suggested a potential "memory effect". Subsequent laboratory testing used separate sample lines
for the NHa standard and the zero air with a continuous flow through the NHa line to prevent the
need for equilibration within the calibration system.
In addition to the off-line dynamic spiking tests, on-line dynamic spiking was performed using
through-the-probe delivery of the compressed ammonia gas standard. The calibration
compressed gas (100 ppm NHa in N2) was introduced to the calibration line at the instrument
shelter and sent through the calibration line to the probe where it was subsequently diluted
(100:1) and travelled in dilute form back through the sampling line to the analyzer. Both the
calibration line and sampling lines were heated Teflon tubing approximately 150 feet in length.
Figure 6-2 shows the response of the Model Gl 103-c to the ammonia standard during the
through-the-probe spiking.
19
-------
1000
750
500
250
Introduction of NH3 to calibration line
forthrough-the-probe introduction to
the Model G1103-C analyzer
(100 ppm std with 100:1 dilution in probe
should result in 1.000 ppb at analyzer)
13:05
13:15
13:25
Time
13:35
13:45
Figure 6-2. Results of through-the-probe dynamic spiking.
Figure 6-2 illustrates a gradual increase in the measured ammonia beginning several minutes
after initial introduction of the standard gas. The relatively long period (approximately 30
minutes) required for the ammonia signal to equilibrate is likely the result of equilibration of the
calibration and sampling lines. Although the ammonia signal did not reach equilibrium during
the spiking, it does not appear that the signal would have reached the expected 1,000 ppb level.
It is not clear whether the apparent low response is attributable to the Model Gl 103-c or to
sample loss in the calibration or sampling lines. However, subsequent laboratory testing using
short sampling lines indicated response times on the order of 2-5 minutes rather than the much
longer response time observed here. Continued through-the-probe dynamic spiking was not
practical because of the large volume of gas consumed during this procedure, due to the required
over-pressurization of the probe to prevent entrainment of flue gas. Furthermore, through-the-
probe spiking required that the several process control CEMs remained off-line during the
dynamic spiking procedure, which was not acceptable to the plant operator.
Figure 6-3 shows the ammonia concentrations measured by the Model Gl 103-c along with the
relative boiler load and urea injection rates during the period from September 11- October 29,
2009. During this period the Model Gl 103-c was sampling from the superheat side of Unit 1 at
the host facility. The ammonia readings shown in Figure 6-3 represent approximately 10-minute
averages of the data from the Model Gl 103-c and are corrected for the 100-fold dilution
introduced by the sampling probe. The boiler output and urea injection rate values are presented
in arbitrary units. Although less obvious than in Figure 6-1, there are periodic spikes in the
Model Gl 103-c readings which coincide with periods when the daily zero/span calibration
checks were performed. Operation of the boiler was stopped because of low demand on the
evening of October 22 and remained off until the evening of October 25. During this period, the
readings from the Model Gl 103-c showed a gradual decay to a concentration corresponding to
approximately 0.9 ppm. Since there was no urea injection during this period, these readings may
20
-------
indicate the release of adsorbed ammonia from the walls of the sampling lines, or from the
internal surfaces of the facility duct work.
40
Q.
Q.
O
u
CRDS Readings
Boiler Load (Aib.)
Urea Inj. Rate (Arb.)
10
9/11/09
9/21/09
10/1/09
10/11/09
10/21/09
Figure 6-3. Comparison of CRDS readings to the boiler output and urea injection rate
between September 11 - October 29, 2009.
6.1 Accuracy
Reference method sampling was performed from October 27 to October 29, 2009 and included a
total of 24 sampling runs. Note that although some reference method samples were collected on
July 24-25*, because various manipulations to the Model Gl 103-c analyzer were performed
during the collection of those samples, they are not included in the evaluation of accuracy.
These manipulations included attempts to bypass the dilution probe and sampling line and
deliver conditioned flue gas directly to the Model 1103-c from the reference method probe.
During the majority of the reference sampling periods the Model 1103-c was not installed and
sampling from the dilution probe system.
Figures 6-4 to 6-6 show the results from the duplicate reference method trains, along with the
Model Gl 103-c measurements for each day of sampling presented as 2-minute rolling averages.
From these figures, the significant disparity between the duplicate reference method sampling
trains for some of the test runs is evident; those differences were noted in Section 4.1.1 and
Table 4-1. For example the first two test runs on both October 27 and 28 (Figures 6-6 and 6-7)
show substantial differences between the duplicate trains.
21
-------
40
-CRDS
—Train 1
—Train 2
6:00
9:00 12:00
Time
15:00
Figure 6-4. Comparison of reference method results and Model GllOS-c measurements
from October 27, 2009.
CRDS
— Train 1
—Train 2
6:00
9:00
12:00
Time
15:00 18:00
Figure 6-5. Comparison of reference method results and Model GllOS-c measurements
from October 28, 2009.
22
-------
30
20
£
Q.
Q.
C
o
E
E
-------
Table 6-1. Comparison of Field Test Reference Method Results and Average Model
GllOS-c Measurements
Date
10/27/2009
10/28/2009
10/29/2009
Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Ave.
Max.
Min.
Ammonia Concentration (ppm)
Ave.
CRDS
18.8
19.2
19.7
11.7
5.40
12.9
10.5
11.5
11.5
13.0
12.8
13.6
9.20
10.1
10.3
11.0
8.90
10.2
13.6
15.3
14.5
11.2
10.6
9.50
12.3
19.7
5.40
Ave. Ref.
(Train 1, Train 2)
20.0(11.7,28.4)
26.9(17.8,36.0)
24.6(22.1,27.1)
19.9(19.3,20.5)
14.7(15.8, 13.5)
19.1 (*, 19.1)
16.6 (*, 16.6)
17.1 (16.4, 17.7)
17.2(19.0, 15.3)
16.4(21.2, 11.7)
13.6(17.1, 10.0)
18.9(18.6, 19.2)
14.6(16.4. 12.9)
16.3(14.1, 18.6)
11.7(13.4,9.90)
13.8(13.6, 13.9)
9.00 (9.00, *)
12.5(9.50, 15.6)
17.8(13.9,21.6)
20.3(15.9,24.7)
17.3(14.1,20.6)
14.1 (13.1, 15.0)
13.3(10.0, 16.5)
11.8(9.10, 14.5)
16.8
26.9
11.7
RA
RPD
-6.3%
-28.6%
-19.9%
-41 .4%
-63.0%
-
-
-32.5%
-33.2%
-20.9%
-5.6%
-28.1%
-37.5%
-38.3%
-11.3%
-19.8%
-
-18.9%
-23.7%
-24.6%
-16.6%
-20.1%
-19.8%
-19.5%
-25.2%
-63.0%
-5.6%
31 .9%
* Outlier. Leak in sampling train suspected but not confirmed.
RA = Relative accuracy (Section 5.1)
24
-------
Table 6-2. Comparison of Laboratory Calibration Check Results and
Average Model Gl 103-c Measurements
Theoretical
Date Concentration
(ppb)
Measured Span
Concentration
(ppb)
% Difference
from Theoretical
8/6/1 0
8/9/1 0
8/10/10
8/11/10
8/12/10
8/13/10
8/16/10
8/17/10
8/18/10
Mean
Maximum
Minimum
1,530
1,520
1,502
1,521
1,518
1,521
1,515
1,518
1,523
1,521
1,519
1,521
1,519
1,530
1,502
1,598
1,589
1,585
1,586
1,581
1,586
1,587
1,585
1,582
1,579
1,580
1,579
1,585
1,598
1,579
4.4%
4.5%
5.5%
4.3%
4.1%
4.3%
4.8%
4.4%
3.9%
3.8%
4.0%
3.8%
4.3%
5.5%
3.8%
RA
4.6%
6.2 Linearity
Figures 6-7 and 6-8 show the linear regression results of the multipoint calibration checks of the
Model Gl 103-c conducted at the beginning and end of the laboratory testing, respectively. In
these figures, the measured concentrations are plotted as a function of the theoretical ammonia
concentrations delivered to the Model Gl 103-c based on the measured flow rates from the
dilution system. The results of these linear regressions are presented in Table 6-3.
For both the initial and final multipoint calibration of the Model Gl 103-c, these results show
slopes within 4% of unity, with intercepts that are statistically indistinguishable from zero, and r2
values of greater than 0.995. It should be noted that the accuracy of the output flow rate of the
permeation oven used to deliver the ammonia is lowest at the highest ammonia concentration
(i.e., lowest dilution flow) and has an expected uncertainty of ~2 to 5% at the highest
concentrations.
25
-------
5,000
y = 0.998x+ 19.5
R2 = 0.997
£ 3,000
0)
o
c
O 2,000
1,000 2,000 3,000 4,000
Theoretical Concentration (ppb)
5,000
Figure 6-7. Initial multi-point calibration of Picarro Model GllOS-c.
4,000
y=1.033x-12.3
R2 = 1.000
1,000 2,000 3,000
Theoretical Concentration (ppb)
4,000
Figure 6-8. Final multi-point calibration of Picarro Model GllOS-c.
Table 6-3. Linearity Results for the Model GllOS-c
Initial
Final
Slope
0.998 (0.021)
1.033(0.009)
Intercept (ppb)
19.5(41.0)
-12.3(13.3)
r2
0.997
1.000
26
-------
6.3 Precision
Tables 6-4 and 6-5, respectively, shows the results of the zero and span checks of the Model
Gl 103-c conducted during the laboratory testing. The results presented in these tables are
calculated from the 2-minute average output values of the Model Gl 103-c. Included in Table 6-
5 are the known ammonia concentrations delivered during the span checks and the calculated
difference between the measured and known ammonia concentrations. Each table also presents
the mean, standard deviation, and range of the respective zero/span checks.
As shown in Table 6-4 over the course of the two week laboratory testing, the average
concentrations reported by the Model Gl 103-c during the daily zero checks were between 0.21
and 2.19 ppb, with an average value of 1.03 ppb. The standard deviations of the measurements
during these zero checks were between 0.04 and 0.11 ppb, with an average value of 0.07 ppb.
Although presented, the relative standard deviations have little meaning when assessing
precision for zero checks, since as the mean concentration approaches zero the relative standard
deviation dramatically increases. The results in Table 6-5 show the average measured
concentrations during the span checks were between 1,579 and 1,598 ppb, with an average value
of 1,585 ppb. The measured concentrations of the span checks exceeded the theoretical
concentrations in all cases. The percent differences between the measured and theoretical values
were between 3.8% and 5.5%, with an average value of 4.3%. The standard deviations of the
measurements during these zero checks were between 1.1 and 7.5 ppb, with an average value of
2.0 ppb. The calculated relative standard deviations ranged from 0.07% to 0.47%, with an
average value of 0.13%.
Table 6-4. Results of Zero Checks of the Model Gl 103-c
Measured Zero
Concentration (ppb)
8/6/1 0
8/9/1 0
8/10/10
8/11/10
8/12/10
8/13/10
8/16/10
8/17/10
8/18/10
Mean
Maximum
Minimum
Standard Deviation
0.770
2.12
2.08
0.750
2.19
0.600
0.550
1.40
0.390
1.81
0.250
0.210
0.280
1.03
2.19
0.210
0.770
Standard Deviation Dcn /0/.
(ppb) RSD(/o)
0.06
0.06
0.06
0.04
0.11
0.08
0.07
0.06
0.05
0.07
0.04
0.05
0.09
0.07
0.11
0.04
0.02
8.0%
3.0%
2.8%
5.7%
4.9%
13.1%
12.2%
4.2%
12.6%
4.1%
16.6%
24.9%
30.9%
1 1 .3%
30.9%
2.8%
9.1%
27
-------
Table 6-5. Results of Span Checks of the Model Gl 103-c
Theoretical
Date Concentration
(ppb)
Measured Span
Concentration
(ppb)
% Difference
from Theoretical
Standard
Deviation (ppb)
RSD
8/6/1 0
8/9/1 0
8/1 0/1 0
8/11/10
8/12/10
8/1 3/1 0
8/16/10
8/17/10
8/18/10
Mean
Maximum
Minimum
St. Dev.
1,530
1,520
1,502
1,521
1,518
1,521
1,515
1,518
1,523
1,521
1,519
1,521
1,519
1,530
1,502
6.5
1,598
1,589
1,585
1,586
1,581
1,586
1,587
1,585
1,582
1,579
1,580
1,579
1,585
1,598
1,579
5.3
4.4%
4.5%
5.5%
4.3%
4.1%
4.3%
4.8%
4.4%
3.9%
3.8%
4.0%
3.8%
4.3%
5.5%
3.8%
0.5%
1.9
1.3
2.1
2.4
1.2
1.1
1.4
1.2
1.1
1.1
7.5
1.8
2.0
7.5
1.1
1.8
0.12%
0.08%
0.13%
0.15%
0.08%
0.07%
0.09%
0.07%
0.07%
0.07%
0.47%
0.11%
0.13%
0.47%
0.07%
0.12%
6.4 Zero/span drift
The results of the zero/span checks were used to assess zero/span drift of the Model Gl 103-c.
Table 6-6 presents the results of the zero/span check measurements along with the calculated
differences between successive zero/span checks. These results show no clear trends in drift of
either the zero or span readings of the Model Gl 103-c over the course of the laboratory testing.
Table 6-6. Summary of Zero/Span Drift Checks of the Model Gl 103-c
Date
8/6/1 0
8/9/1 0
8/10/10
8/11/10
8/12/10
8/13/10
8/16/10
8/17/10
8/18/10
Measured Zero
Concentration
(ppb)
0.770
2.08
0.750
2.19
0.600
0.550
1.40
0.390
1.81
0.250
0.210
0.280
Change from
Previous
(ppb)
1.31
-1.33
1.44
-1.59
-0.05
0.85
-1.01
1.42
-1.56
-0.04
0.07
Measured Span
Concentration
(ppb)
1,598
1,589
1,585
1,586
1,581
1,586
1,587
1,585
1,582
1,579
1,580
1,579
Change from
Previous
(ppb)
-9
-4
1
-5
5
1
-2
-3
-3
1
-1
28
-------
6.5 Response Time
Table 6-7 presents the calculated rise and fall times for the Model Gl 103-c, for the no-average,
30-second average, and 2-minute average output results from the daily zero/span checks.
Table 6-7. Summary of Rise/Fall Times of the Model Gl 103-c
Date
8/6/10
8/9/10
8/10/10
8/11/10
8/12/10
8/13/10
8/16/10
8/17/10
8/18/10
Mean
Maximum
Minimum
Std. Deviation
No
average
03:54
01:52
02:36
01:57
02:37
02:31
02:11
02:16
02:06
02:15
02:35
02:21
02:26
03:54
01:52
00:32
Rise Time
30 -sec
average
03:59
02:06
02:46
02:12
02:57
02:41
02:21
02:36
02:16
02:35
02:55
02:41
02:40
03:59
02:06
00:30
2-min
Average
05:33
04:18
04:43
04:19
04:48
04:42
04:27
04:38
04:22
04:36
04:50
04:46
04:40
05:33
04:18
00:20
Fall Time
No
average
00:39
01:17
00:34
00:39
00:44
00:39
00:35
00:39
00:39
00:39
00:39
00:38
00:42
01:17
00:34
00:11
30 -sec
average
01:04
01:42
01:03
01:08
01:08
01:03
01:04
01:04
01:03
01:04
01:03
01:03
01:07
01:42
01:03
00:11
2-min
average
03:47
04:18
03:44
03:49
03:47
03:43
03:40
03:44
03:44
03:44
03:43
03:44
02:26
03:54
01:52
00:32
Table 6-6 shows that the fall time of the Model Gl 103-c was consistently shorter than the rise
time, with the mean no-average rise and fall times being 2 min 26 sec and 42 sec, respectively.
6.6 Data Completeness
Throughout the verification test periods including the field and laboratory testing, the Model
Gl 103-c recorded data approximately every three seconds with no gaps in the data. Except for
periods when installation, relocation, or maintenance activities were performed, the Model
Gl 103-c exhibited 100% data completeness.
6.7 Operational Factors
6.7.1 Ease of use
Subsequent to the initial installation, the instrument was uninstalled, repackaged, and reinstalled
twice during the verification test. These subsequent installations were performed by Battelle and
on-site support staff, after receiving training from Picarro representatives during the initial
installation. Operation of the Model Gl 103-c was automated upon instrument startup and
required no external intervention. Individual space delimited text data files were automatically
generated for each day of testing and saved on an internal hard drive in separate data folders
identified with the corresponding date.
29
-------
6.7.2 Maintenance
Table 6-8 presents a summary of the activities performed on the Model Gl 103-c system during
the verification test. The installation of the instrument was initially performed by two
representatives of Picarro, who completed the installation in approximately one hour, after which
the Picarro representatives performed a variety of diagnostic tests to ensure proper operation and
optimize system performance. Those diagnostic tests were performed over the course of
approximately two days, although the durations of the individual activities were not recorded.
Table 6-8 shows that the Model Gl 103-c had minimal down time, even with multiple
installations in different sampling locations.
Table 6-8. Summary of Maintenance Activities Performed During Verification Testing
Date
6/25/09
6/25/09 - 6/26/09 1
7/20/09
7/22/09
7/24/09
7/24/09-9/11/09
9/11/092
9/11/09
9/11/09
9/11/09-10/29/09
10/26/09-
10/29/09
10/30/09
Duration
55 minutes
Various
~45 minutes
~45 minutes
~45 minutes
49 days
~5 minutes
~2.5 hours
~45 minutes
48 days
Various
~60 minutes
Activity
Instrument unpacking, installation
Instrument diagnostic checks
Instrument relocation to operational boiler
Tear-down and instrument repacking
Instrument installation at new facility
Routine operation
Switch sampling line from Unit 2 to Unit 1
Dynamic spiking tests
Through-the-probe calibration tests
Routine operation
Reference method sampling
Instrument shutdown, removal, and
packaging
Down Time
NA
NA
~45 minutes
~45 minutes
~45 minutes
NA
~5 minutes
NA
NA
NA
NA
NA
Although the Model Gl 103-c was installed and operating, vendor representatives elected to perform a variety
of diagnostic checks on June 25 and 26 to ensure proper operation. The durations of the individual diagnostic
activities were not recorded separately.
2 Unit 2 was off-line for two weeks because of low demand. The sampling line was switched to sample from
the Superheat side of Unit 1.
6.7.3 Consumables Used/Waste Generated
The Model Gl 103-c uses no compressed gases, reagents, or supplies for normal operation. Thus
during routine monitoring activities, no consumables were used and no waste was generated.
Compressed gas standards and dilution gas are needed for delivery of calibration standards to
the analyzer.
30
-------
Chapter 7
Performance Summary
Table 7-1 presents a summary of the results of the verification of the Picarro Model Gl 103-c
during this verification test.
Table 7-1. Summary of Selected Verification Test Results for the Picarro Model Gl 103-c
Performance
Parameter
Linearity
Accuracy
Precision
Zero/Span Drift
Response Time
Data
Completeness
Method of
Evaluation
Linear regression
of multipoint
calibration results
Comparison to
reference method
results
Evaluation of
daily zero/span
check results
Evaluation of
daily zero/span
check results
Calculated from
daily zero/span
check results
Ratio of number
of data points
collected to
number of
potential data
points that could
have been
collected
Results
Initial
Final
RSD
RA
Mean
St. Dev.
RSD
% Diff. from Theory
Slope Intercept
0.998 (±0.021) 19.5 (±41
1.03 (±0.009) -12.3 (±13
Laboratory Testing
4.3%
4.6%
Zero Check
1.03 ppb
0.77 ppb
11.3%
N/A
• No apparent trend in changes between zero and
Ave Time 0 sec
Mean 02:26
Std. Dev. 00:32
• Completeness =
Rise Time
30 sec 2 min 0 sec
02:40 04:40 00:42
00:30 00:20 00:11
100%
r2
0) 0.997
.3) 1.000
Field Testing3
-25.2%
31 .9%
Span Check
1590 ppb
5.3 ppb
0.13%
4.3%
span checks
Fall Time
30 sec 2 min
01:07 02:26
00:11 00:32
31
-------
Ease of use
Maintenance
Consumables/
waste
generated
Operator
observations
Operator
observations
Operator
observations
• Initial installation was completed in ~45 minutes by vendor
representatives
• Subsequent installations were performed by Battelle and on-site
support staff in ~45 minutes
• Operation is automated upon powering and requires no external
intervention
• Operated unattended for duration of testing period
• Daily space delimited data files are generated automatically and
stored in separate data files on an internal hard drive
• No routine maintenance activities were performed during testing
• Non-routine maintenance included diagnostic tests performed after
initial installation but prior to routine monitoring periods
• No consumables were used and no waste was generated during
routine monitoring activities
• Compressed gas standards were used for dynamic spiking tests;
the waste gas stream from the dynamic spiking was combined with
the excess flue gas
These results are based on measurements from the entire field sampling system, including the dilution probe, the transfer line
from the duct, and the Model Gl 103-c, compared with reference method measurements taken directly from the duct.
32
-------
Chapter 8
References
1. Busch KW, Busch MA, Cavity Ring-down Spectroscopy: An Ultratrace Absorption
Measurement Technique. ACS Symposium Series 720, Oxford (1997).
2. Atkinson, D. B., "Solving chemical problems of environmental importance using cavity ring-
down spectroscopy," The Analyst 128, 117-125 (2003).
3. U.S. EPA, Conditional Test Method (CTM-027) Procedure for Collection and Analysis of
Ammonia in Stationary Sources, August 1997. Available at:
http ://www. epa.gov/ttn/emc/ctm. html
4. Battelle, Test/QA Plan for Verification of Cavity Ring-down Spectroscopy Systems For
Ammonia Monitoring in Stack Gas, prepared by Battelle, Columbus, Ohio, June 2009.
5. Battelle, Quality Management Plan for the ETV Advanced Monitoring Systems Center,
Version 7.0, U.S. EPA Environmental Technology Verification Program, prepared by
Battelle, Columbus, Ohio, November 2008.
33
------- |