Technology Transfer Network
Emission Measurement Center Spectral Database
Combine Cycle Gas-Fired Gas Turbine Emissions Test
|
FINAL REPORTHOUSTON LIGHTING AND POWER COMPANY T. H. WHARTON ELECTRIC GENERATING STATION HOUSTON, TEXAS EPA Contract No. 68D20163 Work Assignment No. I-34 Prepared by: Research Division Entropy, Inc. Post Office Box 12291 Research Triangle Park, North Carolina 27709 Prepared for: Lori Lay U. S. Environmental Protection Agency Emissions Measurement Branch Research Triangle Park, North Carolina 27711 May 27, 1994 DISCLAIMER This document was prepared by Entropy, Inc. under EPA Contract No. 68D20163, Work Assignment No. I-34. This document has been reviewed by the U.S. Environmental Protection Agency (EPA). The opinions, conclusions, and recommendations expressed herein are those of the authors, and do not necessarily represent those of EPA. Mention of specific trade names or products within this report does not constitute endorsement by EPA or Entropy, Inc. TABLE OF CONTENTS1.0 INTRODUCTION 1.1 BACKGROUND 1.2 DESCRIPTION OF THE PROJECT 1.3 PROJECT ORGANIZATION 2.0 PROCESS DESCRIPTION AND SAMPLE POINT LOCATIONS 2.1 PROCESS DESCRIPTION 2.2 CONTROL EQUIPMENT DESCRIPTION 2.3 SAMPLE POINT LOCATIONS 3.0 SUMMARY AND DISCUSSION OF RESULTS 3.1 OBJECTIVES AND TEST MATRIX 3.2 FIELD TEST CHANGES AND PROBLEMS 3.3 SUMMARY OF RESULTS 4.0 SAMPLING AND ANALYTICAL PROCEDURES 4.1 EXTRACTIVE SYSTEM FOR DIRECT GAS PHASE ANALYSIS 4.2 SAMPLE CONCENTRATION 4.3 CONTINUOUS EMISSIONS MONITORING 4.4 FLOW DETERMINATIONS 4.5 PROCESS OBSERVATIONS 4.6 ANALYTICAL PROCEDURES 5.0 QUALITY ASSURANCE/QUALITY CONTROL ACTIVITIES 5.1 QC PROCEDURES FOR MANUAL FLUE GAS TEST METHODS 5.2 QC PROCEDURES FOR INSTRUMENTAL METHODS 5.3 QA/QC CHECKS FOR DATA REDUCTION, VALIDATION, AND REPORTING 5.4 CORRECTIVE ACTIONS 6.0 CONCLUSIONS AND DISCUSSION 7.0 REFERENCES APPENDICES NOTE: Appendices A-D are not available 1.0 INTRODUCTIONThe U. S. Environmental Protection Agency (EPA) Office of Air Quality Planning and Standards (OAQPS), Industrial Studies Branch (ISB), and Emission Measurement Branch (EMB) directed Entropy, Inc. to conduct an emission test at Houston Lighting and Power Company's (HLPC) T. H. Wharton Electric Generating Station combined-cycle gas-fired gas turbine in Houston, Texas. The test was conducted on May 17 and 18, 1993. The purpose of this test was to identify which hazardous air pollutants (HAPs) listed in the Clean Air Act Amendments of 1990 are emitted from this source. The measurement method used Fourier transform infrared (FTIR) technology, which had been developed for detecting and quantifying many organic HAPs in a flue gas stream. Besides developing emission factors (for this source category), the data will be included in an EPA report to Congress. Before this test program, Entropy conducted screening tests using the FTIR method at facilities representing several source categories, including a coal-fired boiler. These screening tests were part of the FTIR Method Development project sponsored by EPA to evaluate the performance and suitability of FTIR spectrometry for HAP emission measurements. These tests helped determine sampling and analytical limitations, provided qualitative information on emission stream composition, and allowed estimation of the mass emission rates for a number of HAPs detected at many process locations. The evaluation demonstrated that gas phase analysis using FTIR can detect and quantify many HAPs at concentrations in the low part per million (ppm) range and higher, and a sample concentration technique was able to detect HAPs at sub-ppm levels. Following the screening tests, Entropy conducted a field validation study at a coal-fired steam generation facility to assess the effectiveness of the FTIR method for measuring HAPs, on a compound by compound basis. The flue gas stream was spiked with HAPs at known concentrations so that calculated concentrations, provided by the FTIR analysis, could be compared with actual concentrations in the gas stream. The analyte spiking procedures of EPA Method 301 were adapted for experiments with 47 HAPs. The analytical procedures of Method 301 were used to evaluate the accuracy and precision of the results. Separate procedures were performed to validate a direct gas phase analysis technique and a sample concentration technique of the FTIR method. A complete report, describing the results of the field validation test, has been submitted to EPA.[1] This report was prepared by Entropy, Inc. under EPA Contract No. 68D20163, Work Assignment No. I-34. Research Triangle Institute (RTI) provided the process information given in Sections 2.1 and 3.3.3. 1.2 DESCRIPTION OF THE PROJECT The FTIR-based method uses two different sampling techniques: (1) direct analysis of the extracted gas stream (hereafter referred to as the gas phase technique or gas phase analysis) and (2) sample concentration followed by thermal desorption. Gas phase analysis involves extracting gas from the sample point location and transporting the gas through sample lines to a mobile laboratory where sample conditioning and FTIR analyses are performed. The sample concentration system employs 10 g of Tenax® sorbent, which can remove organic compounds from a flue gas stream. Organic compounds adsorbed by Tenax® are then thermally desorbed into the smaller volume of the FTIR absorption cell; this technique allows detection of some compounds down to the ppb level in the original sample. For this test, approximately 850 dry liters of flue gas were sampled during each run using the sample concentration system. Section 4.0 describes the sampling systems. Entropy operated a mobile laboratory (FTIR truck) containing the instrumentation and sampling equipment. The truck was driven to the site at T. H. Wharton, and parked next to the sampling location. Three test runs were performed over a two-day period. Entropy tested the exhaust gases from one of the gas turbines operated by T.H. Wharton to generate electricity. The turbine burns natural gas. Hot gases from the combustion of the natural gas drive the turbine. Gases (about 1000 degF) exiting the turbine pass through an exhaust duct to a heat recovery steam generator (HRSG). Heat is recovered in the HRSG to produce steam, which in turn is used to drive a steam turbine. The cooled gases exit the HRSG to be exhausted through a short stack. The only control device is a water injection system used to minimize NOx emissions. Entropy installed sampling equipment in ports available on the gas turbine exhaust duct upstream of the heat recovery steam generator. Section 2.0 contains descriptions of the process and the sampling point locations. Direct gas phase analysis was used to measure carbon monoxide (CO), carbon dioxide (CO2), sulfur dioxide (SO2), nitrogen oxides (NOx), and ppm levels of other species. EPA instrumental test methods were used to provide concentrations of CO, CO2, O2, and hydrocarbons (HC). The sample concentration technique was used to measure HAPs at ppb levels. Entropy conducted three 4-hour sample concentration and gas phase runs at the turbine exhaust duct upstream of the HRSG. In addition, a single sample concen- tration run was also conducted at the HRSG stack simultaneous with Run 2 at the inlet of the HRSG. Combustion gas volumetric flows were calculated from fuel data provided by T. H. Wharton. Section 3.1 gives the test schedule. The test program was funded and administered by the Industrial Studies Branch (ISB) and the Emissions Measurement Branch (EMB) of the U.S. EPA. A representative from RTI collected process data. The following list presents the organizations and personnel involved in coordinating and performing this project.
HLPC Corporate Contact: Mr. Derek Furstenwerth (713) 897-8603
T. H. Wharton Plant Mr. Ron Jernigan
Contacts: Mr. Edward Wong (713) 897-2609
EMB Work Assignment Ms. Lori Lay (919) 541-4825
Managers: Mr. Dennis Holzschuh (919) 541-5239
Industrial Studies Branch Mr. Kenneth Durkee (919) 541-5425
(ISB) Contacts: Mr. William Maxwell (919) 541-5430
Entropy Project Manager: Dr. Thomas Geyer (919) 781-3551
Entropy Test Personnel: Mr. Scott Shanklin
Ms. Lisa Grosshandler
Dr. Laura Kinner
Mr. Greg Blanschan
Mr. Mike Worthy
RTI Representative: Mr. Jeffrey Cole (919) 990-8606
2.0 PROCESS DESCRIPTION AND SAMPLE POINT LOCATIONSThe process information was supplied by the T.H. Wharton Generating Station. Houston Lighting & Power Company's (HLPC) T.H. Wharton Unit Four is located in Houston, Texas. Unit Four began operation in 1974 and consists of four combined cycle gas turbines, numbered 41, 42, 43, and 44. Each combined cycle gas turbine consists of a General Electric MS-7000 simple cycle gas turbine and a heat recovery steam generator (HRSG). These components are described below. The four combined cycle gas turbines and their associated steam turbine are collectively referred to as a General Electric Stag 300 system (Figure 2-1). Gas turbine No. 41 (GT 41) was used for testing. One or more of these GT's normally operate during peak usage times, which vary depending on need. There is a 2-week planned outage for each GT once a year for an annual inspection. The primary fuel source for GT 41 is natural gas. GT 41 can also burn No. 2 fuel oil. Only natural gas was used during the test period. Each General Electric MS-7000 gas turbine is a 49 MW, single-shaft, three-bearing machine connected to its own generator. The hot gases exiting the combustion chambers drive the gas turbines, which in turn develop power to drive the axial compressor and to produce useful shaft output for driving the generator. The exhaust from all four GT's is used in the combined-cycle mode as the heat energy input to produce steam from HRSG feedwater. Each gas turbine has its own HRSG. The saturated steam in each HRSG drum is superheated by gas turbine exhaust. This superheated steam is collected from all four HRSGs and used to turn a 102.5 MW steam turbine generator. The gas turbine is equipped with a set of dampers which allow the turbine to operate in simple-cycle or combined-cycle mode. The bypass damper controls flow through the bypass or simple-cycle stack, and the isolation damper controls flow through the HRSG. During start-up operations the isolation damper is closed, preventing flue gas flow through the HRSG, and the bypass damper is open, allowing flue gas to exit through the bypass stack. This is referred to as simple-cycle operation. Once the turbine has completed start-up procedures the isolation damper is opened and the bypass damper is closed redirecting flue gas flow through the HRSG. The hot flue gas heats boiler feed water to produce steam, which, once it has reached sufficient quality, is used to drive a steam turbine to produce more electricity. This is referred to as combined-cycle operation. GT 41 can produce 48 to 62 MW depending on the time of year. In winter, the inlet air is colder and denser, allowing more fuel to flow to the turbines producing greater output. The opposite occurs during the summer when the inlet air is less dense. Unit Four is nominally rated at 300 MW. GT 41 can produce 210,000 lb/hr of steam from its HRSG for the steam turbine generator. All four GT's in Unit Four are capable of a combined 840,000 lb/hr of steam. 2.2 CONTROL EQUIPMENT DESCRIPTION 2.2.1 Nitrogen Oxides (NOx) Control A water injection system for NOx control is incorporated in the design of GT 41. The water injection system operates using demineralized water from the station feedwater treatment system. Water is vaporized when it is injected into the combustion air stream. The vaporization process removes some of the heat from the combustion chamber, thus lowering the peak flame temperature. The result of this temperature reduction is to reduce the formation of thermal NOx. 2.2.2 Sulfur Dioxide (SO2) Control Emissions of SO2 are considered negligible for natural gas firing. When using the alternate fuel (No. 2 distillate), SO2 emissions are controlled by the use of low sulfur content oil. The maximum sulfur content acceptable under current permit specifications is 0.5 percent sulfur by weight. 2.2.3 Particulate Control Particulate and visible emissions are limited by using natural gas as the primary fuel and No. 2 distillate oil as an alternate fuel. In addition to the low ash characteristics of No. 2 fuel oil, each gas turbine is equipped with a swirl plate to impart a swirl to the combustion air to ensure a thorough mixing of air and fuel so that complete combustion occurs. There were two options for a sampling location, at the turbine outlet (HRSG inlet) or the stack (outlet of the HRSG). The turbine outlet location was selected for testing because of concern that vents near the emission point on the stack would allow ambient air to dilute to sample stream. A second sampling train was available, and one sample concentration run of the system was performed at the stack concurrently with sample concentration Run 2 at the turbine outlet. 2.3.1 Turbine Outlet (HRSG Inlet) Figure 2-2 depicts the turbine outlet location. Six sample ports were available across the top of the 22.5-foot wide, 7.25-foot deep gas turbine exhaust duct. The ports were located 13 feet down-stream of a damper and 3.25 feet upstream of a widening in the duct (diffuser) that led to the HRSG. The separate sample probes were installed in the two middle ports (see Figure 2-2). A pitot probe was installed in the port adjacent to the sample concentration probe to provide indications of changes in the exhaust gas flow. These ports were the only ports used during the test. The flue gas conditions prohibited sample and pitot probe traverses to check for stratification in the gas stream, and determine flue gas volumetric flow. According to HLPC personnel, gas stratification was unlikely at this location on the process. A sample point traverse across the duct through a single port was performed using the gas phase probe to check for stratification and the results of the O2 traverse indicated the stream was not stratified (see Section 3.3.2). The sample concentration and the gas phase probe tips were both inserted to a depth of 3 ft. Flue gas conditions at the sample point location were 990 degF and about 20 inches of water positive pressure. Due to these extreme conditions, as a safety concern, the facility did not fire the turbine during the installation of the sampling probes, but fired the turbine once Entropy was ready to proceed with the test. 2.3.2 Stack (HRSG Outlet) Dimensions of the stack location are shown in Figure 2-3. The sample port was 1-foot below the top of the stack which was open. The gas flow presumably prohibited mixing between flue gas and air at the test point, but there were no CEM, Orsat or gas phase FTIR measurements performed to verify this. Horizontal vents around the stack just below the level of the sample port were closed during the test run, so there should have been no air leakage from this source. Although this location did not meet EPA Method 1 criteria, EPA believed the composition of the stack outlet to be of interest so a single sample concentration run was performed. The sample concentration probe extended through the port several feet into the opening at the top of the stack. The gas stream was between 290 and 300 degF during the test run. This allowed Entropy to set up the apparatus and insert the probe while the process was operating. 3.0 SUMMARY AND DISCUSSION OF RESULTS3.1 OBJECTIVES AND TEST MATRIX The purpose of the test program was to obtain information that will enable EPA to develop emission factors (for as many HAPs as possible) which will apply to electric utilities employing gas-fired gas turbines. EPA will use these results to prepare a report for Congress. The specific objectives were:
Measure HAP emissions (employing methods based on FTIR
spectrometry) in two concentration ranges, above 1 ppm using
gas phase analysis, and to sub-ppm levels using sample
concentration/thermal desorption.
Determine maximum possible concentrations for undetected HAPs
based on detection limits of instrumental configuration and
limitations imposed by composition of flue gas matrix.
Measure O2, CO2, CO, and hydrocarbons using gas analyzers.
Perform simultaneous testing at the inlet and outlet of the
HRSG and analyze data to assess effect (if any) of the HRSG on
HAP emissions.
Obtain process information from T. H. Wharton. This
information includes the rate of power production during the
testing periods.
Table 3-1 presents the testing schedule that was followed at T. H. Wharton. 3.2 FIELD TEST CHANGES AND PROBLEMS On the initial test day Entropy experienced difficulties aligning the FTIR cell. It was important to achieve satisfactory alignment of the cell because this determined the intensity of the signal reaching the detector, which in turn influenced the signal to noise ratio (S/N) of the data. The sensitivity of the instrument depends, in part, on S/N. The problem was corrected but not before set-up of the sample concentration apparatus was complete and the plant was operating under conditions suitable for testing. It was decided to begin sample concentration Run 1 and begin direct gas phase testing as soon as possible during Run 1. As a result there are direct gas phase and CEM data covering only a portion of Run 1. This was deemed acceptable because there was ample opportunity to obtain gas phase data on the second test day. The second change was introduced to permit the completion of two test runs in a single day. Initially, the plan called for the completion of two 4-hour sample concentration runs with concurrent gas phase runs performed over the entirety of the two 4-hour periods. Instead, sample concentration Run 1 commenced as soon as the system was ready and the turbine was operating at full capacity. Direct gas phase testing commenced about 40 minutes after the beginning of Run 2. Gas phase analysis continued through the end of Run 2 and into Run 3 but was stopped before Run 3 was completed. This plan was the best way to accomplish the test objectives and complete the test runs within the originally scheduled time. Also, it was not necessary for the gas phase analysis to run for the entire 4 hours of each Run to collect enough data to characterize the flue gas stream. 3.3.1 FTIR Results Gas phase and sample concentration data were analyzed for the presence of HAPs and other species. All spectra were visually inspected and absorbance bands were identified. Then spectra were analyzed, using analysis procedures developed by Entropy, to determine concentrations of any species detected. These results are presented in Table 3-2. Maximum possible concentrations were determined for undetected HAPs. These results are presented in Tables 3-3, 3-4, and 3-5. 3.3.1.1 Gas Phase Results Each gas phase FTIR spectrum was separately analyzed for the presence of HAPs or other species. The spectra revealed that the gas phase samples were composed primarily of;
water vapor
CO2
Smaller amounts of NO (an average of about 15 ppm) and CO were also
detected.
NO2 was detected but not quantified because quantitative reference
spectra are not currently available.
Calculated concentrations of NO for each spectrum will be included in a table with the final report. No SO2 or HCl was detected in the gas phase spectra. The spectra were analyzed for the presence of HAPs that are currently represented in the quantitative reference spectra library. None were detected. Previously, Entropy developed analysis programs to analyze for HAPs in FTIR spectra of samples extracted from a coal-fired boiler stack. Statistical analyses showed that the programs were successful in measuring some HAPs in hot/wet and condenser samples.[1] The major interferant species detected at the coal-fired boiler are very similar to those that have been identified at the gas-fired gas turbine (with the exception that SO2 was not detected in the gas-fired exhaust). Therefore, the same programs were used to analyze the data obtained in this test. The results of the analyses are presented in Appendix C. A set of subtracted spectra was generated so that maximum possible (minimum detectible) concentrations could be calculated for HAPs that were not identified in the sample stream. Reference spectra of water vapor and CO2 were scaled and subtracted from each of the field spectra. The remaining base lines were then analyzed for every compound represented in the quantitative spectral library to determine the maximum possible concentrations of HAPs that were undetected. The calculations were performed according to the procedures described in Section 4.6.3. Results for hot/wet and dry (treated with the condenser or PermaPure® dryers) spectra are presented in Tables 3-3 and 3-4 respectively. The results are averages of the calculated values for all of the spectra over the 3 sample runs. The hot/wet gas phase spectra are the most difficult to analyze because there is strong interference from water vapor. Even so, in results from the hot/wet gas phase data, 92 compounds gave minimum detectible concentrations below 10 ppm, and of these, 77 are below 5 ppm, and 26 are 1 ppm or lower. The results represent upper limits for in-stack concentrations of the HAPs listed. This means that, for a HAP to be present in the gas stream, its concentration must have been below the calculated maximum possible concen- tration. 3.3.1.2 Sample Concentration Spectra The sample concentration spectra represent integrated samples taken over each 4-hour run.
Ammonia (NH3) was detected in samples from all three runs at the
turbine outlet and in the sample taken from the HRSG outlet. It
was also present in the ambient samples collected at each location.
HCl was detected in the sample from Run 1 and in both ambient
samples at the turbine outlet.
Evidence of hexane was observed in samples from both locations and
the ambient samples. Absorbances similar to hexane are often
observed in spectra of desorbed samples. These features may be due
to a mixture of alkane hydrocarbons, including hexane, the sum of
whose spectra gives absorbances which appear similar to hexane.
A siloxane compound was detected that Entropy first identified in
spectra of samples taken at the coal-fired boiler validation test.[1]
This compound was shown to be a product of a reaction between HCl
or water vapor in the gas stream and materials in the filter
housing of the Method 5 box. Entropy took steps to eliminate this
problem and the siloxane, if it is a contaminant, is present at
very low levels relative to validation data.
Nitrous acid (HNO2) was detected in spectra from Runs 1 and 2 and
from the HRSG outlet. This compound was probably formed by
reaction of the NOx in the gas stream with the water condensed in
the collection tube.
Table 3-2 shows calculated concentrations of HCl, NH3, and hexane in samples where these species were detected. In-stack concentrations are also given for the same species. In-stack concentrations were determined by dividing the in-cell concentration by the concentration factor (see Section 4.6.4). The in-stack concentrations are based on the volume of gas sampled and do not account for effects of the sampling system or the adsorption/ desorption efficiencies of HCl and NH3. Therefore, the values in Table 3-2 represent lower limits on the concentrations for these species. Upper limits are provided by the values given in the gas phase data ( Tables 3-3 and 3-4). Table 3-5 gives minimum detectible concentrations for species not detected using Tenax® and the maximum in-stack concentrations are based on the amount of gas sampled (See Section 4.6.5) Other absorbance bands were also observed which still remain unidentified. None of these bands were attributed to HAPs for which Entropy currently has reference spectra. When these bands are identified, it should become clear whether they are due to emissions from the process or were formed by conditions unrelated to the process (i.e. by contamination). These bands do not consistently appear in every sample so it is possible that concentrations of some species varied during the test Runs. Spectral analysis programs were also developed for sample concentration spectra. The analysis programs were used to evaluate the sample concentration data for HAPs. The results, presented in Appendix C, give calculated concentrations only for those HAPs that Entropy has proven in a field validation study can be measured using Tenax®. 3.3.2 Instrumental and Manual Test Results In accordance with standard turbine emission test requirements (i.e., Subpart GG and EPA Method 20), a preliminary O2 traverse was conducted immediately prior to initiating Run 1 to determine an appropriate measurement point location. Other probes were installed and could not be removed while the turbine was in operation, therefore, only three of the six sample ports were traversed during this check. The traverse point locations and corresponding O2 measurements are presented in Figure 3-1. These results indicated no change in the O2 levels across the duct; therefore, the probes were positioned at a depth of 3 ft. within the duct for the testing. Table 3-6 summarizes the results of the EPA Methods 3A and 10 tests as described in Section 4.3. All CEM results in the table were determined from the average gas concentration measured during the run and adjusted for the pre- and post-test run calibration check results (Equation 6C-1 presented in EPA Method 6C, Section 8). Although not required by Method 10, the same data reduction procedures as that in Method 3A were used for the CO determinations to improve the quality of the data. All measurement system calibration bias and calibration drift checks for each test run met the applicable specifications contained in the test methods. No HC data were available from the test because the analyzer malfunctioned during the first test run. Each test run CO emission rate was computed using the averaged concentration measurement for the test run, the flue gas volumetric flow rate, and the appropriate conversion factors. The turbine exhaust gas flow rates used to compute mass emissions in units of lb/hr were determined using EPA Method 19 procedures and the measured flue gas O2. An on-line process gas chromatograph analyzes a natural gas sample every hour at the Wharton facility. The fuel analysis data supplied by the source are included in Appendix B. The analysis data collected for each test period were averaged. This information and the amount of fuel fired by the turbine were used to compute the heat consumption and Fd-factor needed to compute the dry exhaust gas volumetric flow rate (in units of dry standard cubic feet per minute, dscfm) for each test run (see Table 3-7). Wet basis flow rates (wscfm) were computed based on 13% H2O in the flue gas. As a quality assurance check of the O2 and CO2 data, Fo factors were calculated for each test run. The calculated Fo results presented in Table 3-8 are within the range of acceptable values. 3.3.3 Process Operation During Testing 3.3.3.1 Process Results Table 3-9 and Figures 3-2, 3-3, and 3-4 present the process results and can be found immediately following this section. 3.3.3.2 Problems and/or Variations during Testing During Run 1 (2:45 p.m. to 6:45 p.m., 5/17/93), there were no process operations that would interfere with testing. During Run 2 (11:00 a.m. to 3:23 p.m., 5/18/93), a piece of test equipment overheated and was replaced with a reserve unit. The Run was stopped during the down time, then restarted. The Run was extended to achieve a total Run time of 4-hours. During Run 3 (3:55 p.m. to 7:55 p.m., 5/18/93), the turbine's megawatt output increased. This increase was due to a thunderstorm that passed over the plant. GT 41 obtains its combustion air from ambient air outside the unit. As the temperature dropped, the air density increased and the mass flow through the turbine also increased, although fuel flow stayed essentially the same. This increased mass flow provided more power to the turbine and, thus, greater megawatt generation. 4.0 SAMPLING AND ANALYTICAL PROCEDURESThe FTIR analysis is done using two different experimental techniques. The first, referred to as direct gas phase analysis, involves transporting the gas stream to the sample manifold so it can be sent directly to the infrared cell. This technique provides a sample similar in composition to the flue gas stream at the sample point location. Some compounds may be affected because of contact with the sampling system components or reactions with other species in the gas. A second technique, referred to as sample concentration, involves concentrating the sample by passing a measured volume through an absorbing material (Tenax®) packed into a U-shaped stainless steel collection tube. After sampling, the tube is heated to desorb any collected compounds into the FTIR cell. The desorbed sample is then diluted with nitrogen to one atmosphere total pressure. Concentrations of any species detected in the absorption cell are related to flue gas concentrations by comparing the volume of gas collected to the volume of the FTIR cell. Desorption into the smaller FTIR cell volume provides a volumetric concentration related to the volume sampled. This, in turn, provides a corresponding increase in sensitivity for the detection of species that can measured using Tenax®. Sample concentration makes it possible to achieve lower detection limits for some HAPs. Infrared absorbance spectra of gas phase and concentrated samples were recorded and analyzed. In conjunction with the FTIR sample analyses, measurements of (HC), (CO), (O2), and (CO2) were obtained using gas analyzers. Components of the emission test systems used by Entropy for this testing program are described below. 4.1 EXTRACTIVE SYSTEM FOR DIRECT GAS PHASE ANALYSIS An extractive system, depicted in Figure 4-1, was used to transport the gas stream from the turbine exhaust duct directly to the infrared cell. 4.1.1 Sampling System Flue gas was extracted through a stainless steel probe. In order to protect the Teflon® sampling system components, a thermocouple was installed at the outlet of the probe to verify that the sample gas temperatures had been lowered to approximately 350 degF before entering the heated line. A Balston® particulate filter rated at 1 micron was installed at the outlet of the sample probe. A 100-foot length of heated 3/8-inch O.D. Teflon® sample line connected the probe to the heated sample pump (KNF Neuberger, Inc. model number N010 ST.111) located inside the mobile laboratory. The temperature of the sampling system components was maintained at about 300 degF. Digital temperature controllers were used to control and monitor the temperature of the transport lines. All connections were wrapped with electric heat tape and insulated to ensure that there were no "cold spots" in the sampling system where condensation might occur. All components of the sample system were constructed of Type 316 stainless steel or Teflon® . The heated sample flow manifold, located in the FTIR truck, included a secondary particulate filter and valves that allowed the operator to send sample gas directly to the absorption cell or through a gas conditioning system. The extractive system can deliver three types of samples to the absorption cell. Sample sent directly to the FTIR cell is considered unconditioned, or "hot/wet." This sample is thought to be most representative of the actual effluent composition. The removal of water vapor from the gas stream before analysis was sometimes desirable; therefore, a second type of sample was provided by directing gas through a condenser system. The condenser employed a standard Peltier dryer to cool the gas stream to approximately 38 degF. The resulting condensate was collected in two traps and removed from the conditioning system with peristaltic pumps. This technique is known to leave the concentrations of inorganic and highly volatile compounds very near to the (dry-basis) stack concentrations. A third type of sample was obtained by passing the gas stream through a series of PermaPure® dryers. This system utilized a network of semi-permeable membranes. Water vapor was drawn through the membrane walls by a concentration gradient, which was established by a counter flow of dry air along the outside of the membrane walls. In addition to protecting the absorption cell, water removal relieved spectral interferences, which could limit the effectiveness of the FTIR analysis for particular compounds. 4.1.2 Analytical System The FTIR equipment used in this test consists of a medium-resolution interferometer, heated infrared absorption cell, liquid nitrogen cooled mercury cadmium telluride (MCT) broad band infrared detector, and computer (see Figure 4-2). The interferometer, detector, and computer were purchased from KVB/Analect, Inc., and comprise their base Model RFX-40 system. The nominal spectral resolution of the system is one wavenumber (1 cm-1). Samples were contained in a model 5-22H infrared absorption cell manufactured by Infrared Analysis, Inc. The inside walls and mirror housing of the cell were Teflon® coated. Cell temperature was maintained at 240 degF using heated jackets and temperature controllers. The absorption path length of the cell was set at 22 meters. 4.1.3 Sample Collection Procedure During operation of the gas turbine, the flue gas temperatures of 990 degF and a positive pressure of about 20 inches of water at the sampling location presented a safety concern. Therefore, according to agreement with the plant, the turbine was not operated until installation of the sampling probes was completed. Once installation was completed, the plant fired the turbine and the test proceeded. During all three test runs, direct gas phase analysis was performed at the stack concurrent with the sample concentration testing. Over each 4-hour test run, flue gas continuously flowed through the heated system to the sample manifold in the FTIR truck. A portion of the gas stream was diverted to a secondary manifold located near the inlet of the FTIR absorption cell. The cell was filled with sample to ambient pressure and the FTIR spectrum recorded. After analysis, the cell was evacuated so that a subsequent sample could be introduced. The process of collecting and analyzing a sample and then evacuating the cell to prepare for the sample required less than 10 minutes. During each run, about 12 gas phase samples were analyzed. Sample concentration was performed using the adsorbent material Tenax®, followed by thermal desorption into the FTIR cell. The sample collection system employed equipment similar to that of the Modified Method 5 sample train. 4.2.1 Sampling System Figure 4-3 depicts the apparatus used in this test program. Components of the sampling train included a heated stainless steel probe, heated filter and glass casing, stainless steel air-cooled condenser, stainless steel adsorbent trap in an ice bath, followed by two water-filled impingers, one knockout impinger, an impinger filled with silica gel, a sample pump, and a dry gas meter. All heated components were kept at a temperature above 120 degC to ensure no condensation of water vapor within the system. The stainless steel condenser coil was used to pre-cool the sample gas before it entered the adsorbent trap. The trap was a specially designed stainless steel U- shaped collection tube filled with 10 grams of Tenax® and plugged at both ends with glass wool. Stainless steel was used for the construction of the adsorbent tubes because it gives a more uniform and more efficient heat transfer than glass. Each sampling run was 4-hours at approximately 0.12 to 0.13 lpm for a total sampled volume of about 30 to 40 dcf. The sampling rate depended on the sampling train used and was close to the maximum that could be achieved. Collection times provided a volumetric concentration that was proportional to the total volume sampled. The resulting increase in sensitivity should allow detection to concentrations below 1 ppm for some HAPs. 4.2.2 Analytical System Before analysis condensed water vapor was removed from the collection tubes using a dry nitrogen purge for about 15 minutes. Sample analyses were performed using thermal desorption-FTIR. The sample tubes were wrapped with heat tape and placed in an insulated chamber. One end of the tube was connected to the inlet of the evacuated FTIR absorption cell. The same end of the tube that served as the inlet during the sample concentration run served as the outlet for the thermal desorption. Gas samples were desorbed by heating the Tenax® to 250 degC. A preheated stream of UPC grade nitrogen was passed through the adsorbent and into the FTIR absorption cell. About 7 liters of nitrogen (at 240 degF) carried the desorbed gases to the cell and brought the total pressure of the FTIR sample to ambient pressure. The infrared absorption spectrum was then recorded. The purging process was repeated until no evidence of additional sample desorption was noted in the infrared spectrum. 4.2.3 Sample Collection Procedure During each 4-hour run, sample concentration testing was conducted at the turbine outlet. During Run 2 a sample was also collected simultaneously at the stack. The sample concentration test apparatus was set up at the location after Entropy performed leak checks of the system. Sample flow, temperature of the heated box, and the tube outlet temperature were monitored continuously and recorded at 10-minute intervals. At the end of each run, sampling was interrupted and the collection tube was removed. The open ends were tightly capped and the tube was stored on ice until it was analyzed. In most cases, the tubes were analyzed within several hours after the sample run. 4.3 CONTINUOUS EMISSIONS MONITORING Entropy's extractive measurement system and the sampling and analytical procedures used for the determinations of HC, CO, O2, and CO2 conform with the requirements of EPA Test Methods 25A, 10, and 3A, respectively, of 40 CFR 60, Appendix B. A heated extractive sampling system and a set of gas analyzers were used to analyze flue gas samples extracted at the turbine outlet sample point location. The analyzers received gas samples delivered from the same sampling system that supplied the FTIR cell with condenser sample. These gas analyzers require that the flue gas be conditioned to eliminate any possible interference (i.e., particulate matter and/or water vapor) before being transported and analyzed. All components of the sampling system that contact the gas sample were Type 316 stainless steel and Teflon® . A gas flow distribution manifold downstream of the heated sample pump was used to control the flow of sample gas to each analyzer. A refrigerated condenser removed water vapor from the sample gas analyzed by all the analyzers except for the HC analyzer (Method 25A requires a wet basis analysis). The condenser was operated at approximately 38 degF. The condensate was continuously removed from the traps within the condenser to minimize contact between the gas sample and the condensate. The sampling system included a calibration gas injection point immediately upstream of the analyzers for the calibration error checks and also at the outlet of the probe for the sampling system bias and calibration drift checks. The mid- and high-range calibration gases were certified by the vendor according to EPA Protocol 1 specifications. Methane in air was used to calibrate the HC analyzer. A computer-based data acquisition system was used to provide an instantaneous display of the analyzer responses, as well as compile the measurement data collected each second, calculate data averages over selected time periods, calculate emission rates, and document the measurement system calibrations. Table 4-1 presents a list of the analyzers that Entropy used during the test program. Figure 4-1 presents a simplified schematic of Entropy's reference measurement system. The test run values were determined from the average concentration measurements displayed by the gas analyzers during the run and are adjusted based on the zero and upscale sampling system bias check results using the equation presented in Section 8 of Method 6C. The CEM data are presented in Appendix A. Because of the high flue gas temperature and pressure conditions, it was not possible to perform velocity traverses at point locations according to EPA Method 1 specifications. In lieu of pitot measurements, flue gas volumetric flow was determined using mass balance calculations based on the natural gas fuel usage rate, fuel composition, exhaust gas diluent concentrations, and an F-factor as outlined in EPA Method 19 (40 CFR 60). The natural gas feed rate to the turbine was a process parameter recorded by the RTI representative during the test program. The rates were recorded at 15-minute intervals and then averaged for each test run period. The Wharton facility operates an on-line gas chromatograph that analyzes a natural gas sample every hour. This analysis data was supplied to EPA so that the gross calorific value (GCV, in units of Btu/ft3) and Fd-factor (in units of dry standard cubic feet of combustion gas generated per million Btu of heat input, dscf/MM-Btu) could be determined for the computation of the flue gas volumetric flow rates. A pitot tube was positioned adjacent to the point where the sample concentration probe was inserted. Single point P values were recorded at 10 minute intervals to verify that flow characteristics, at the sampling point, were not changing significantly during the test. Heat consumption of the turbine was calculated from the fuel data:
where:
HC = Heat Consumption (mmBtu/hr)
GCV = Gross Calorific Value of Fuel (Btu/ft3) from fuel analysis
data provided by Wharton.
FQT = Fuel Flow Rate (mmft3/day) provided by Wharton.
The dry exhaust gas flow rate was calculated using EPA Method 19 procedures:
where:
Fd = Dry basis F-factor (dscf/mmBtu) determined from fuel analysis.
%O2d = dry basis concentration measurement from EPA Method 3A
During the testing, an RTI representative monitored the process operations so that emissions test data could be correlated with process conditions. The process observations are presented in Section 3.3.3. 4.6.1 Description of K-Matrix Analyses K-type calibration matrices were used to relate absorbance to concentration. Several descriptions of this analytical technique can be found in the literature[2]. The discussion presented here follows that of Haaland, Easterling, and Vopicka[3]. For a set of m absorbance reference spectra of q different compounds over n data points (corresponding to the discrete infrared wavenumber positions chosen as the analytical region) at a fixed absorption pathlength b, Beer's law can be written in matrix form as
The quantity which is sought in the design of this analysis is the matrix K, since if an approximation to this matrix, denoted by K, can be found, the concentrations in a sample spectrum can also be estimated. Using the vector A* to represent the n measured absorbance values of a sample spectrum over the wavenumber region(s) of interest, and the vector C to represent the j estimated concentrations of the compounds comprising the sample, C can be calculated from A* and K from the relation
Here the superscript t represents the transpose of the indicated matrix, and the superscript -1 represents the matrix inverse.
The standard method for obtaining the best estimate K is to minimize the square of the error terms represented by the matrix E. The equation represents the estimate K which minimizes the analysis error. Reference spectra for the K-matrix concentration determinations were de- resolved to 1.0 cm-1 resolution from existing 0.25 cm-1 resolution reference spectra. This was accomplished by truncating and re-apodizing[4] the interfer- ograms of single beam reference spectra and their associated background interferograms. The processed single beam spectra were recombined and converted to absorbance (see Section 4.3). 4.6.2 Preparation of Analysis Programs To provide accurate quantitative results, K-matrix input must include absorbance values from a set of reference spectra which, added together, qualitatively resemble the appearance of the sample spectra. For this reason, all of the Multicomponent analysis files included spectra representing interferant species and criteria pollutants present in the flue gas. Several factors affect the detection and analysis of an analyte in the stack gas matrix. One is the composition of the stack gas. The major spectral interferant in the gas-fired boiler effluent are water and CO2. At CO2 concentrations of about 10 percent and higher, weak absorbance bands that are normally not visible begin to emerge. Some portions of the FTIR spectrum were not available for analysis because of extreme absorbance from water and CO2, but most compounds exhibit at least one absorbance band that is suitable for analysis. Significant amounts of NO, and NO2 were also present in the samples and these species needed to be accounted for in any analytical program. A second factor affecting analyses is the number of analytes that are to be detected because the program becomes more limited in distinguishing overlapping bands as the number of species in the sample increases. A third factor depends on how well the sample spectra can be modeled. The best analysis can be made when reference spectra are available to account for all of the species detected in the sample. When reference spectra are not available for a compound which has been identified, then it becomes more difficult to quantify other species. A set of Multicomp program files had been previously prepared for analysis of data collected at a coal-fired utility for the purpose of performing statistical validation testing of the FTIR methods. Separate programs were prepared to measure 47 different compounds. Four baseline subtraction points were specified in each analytical region, identifying an upper and a lower baseline averaging range. The absorbance data in each range were averaged, a straight baseline was calculated through the range midpoint using the average absorbance values, and the baseline was subtracted from the data prior to K-matrix analysis. Before K-matrix analysis was applied to data all of the spectra were inspected to determine what species had been detected. Program files were constructed that included reference spectra representing the detected species and were then used to calculate concentrations of the detected species. Sample concentration spectra were also analyzed using program files that were shown by the validation testing to be suitable for measuring some HAPs. 4.6.3 Error Analysis of data The principal constituents of the gas phase samples were water, CO2, NO, and NO2. A program file was prepared to quantify each of these compounds. Other than these species and N2O no major absorbance features were observed in the spectra. After concentrations of the main constituents were determined, the appropriate standard was scaled and subtracted from the spectrum of the sample mixture. This helped verify the calculated values. New spectra were generated from the original absorbance spectra by successively subtracting scaled standard spectra of water, CO2, NO, and NO2. The resulting "subtracted" spectra were analyzed for detectible absorbencies of any HAPs and, for undetected species, the maximum possible concentrations that could have been present in the samples. Maximum possible (minimum detectible) concentrations were determined in several steps. The noise level in the appropriate analytical region was quantified by calculating the root mean square deviation (RMSD) of the baseline in the subtracted spectrum. The RMSD was multiplied by the width (in cm-1) of the analytical region to give an equivalent "noise area" in the subtracted spectrum. This value was compared to the integrated area of the same analytical region in a standard spectrum of the pure compound. The noise was calculated from the equation:
where:
RMSD = Root mean square deviation in the absorbance values within a
region.
n = Number of absorbance values in the region.
Ai = Absorbance value of the ith data point in the analytical
region.
AM = Mean of all the absorbance values in the region.
If a species is detected, then the error in the calculated concentration is given by:
where:
Eppm = Noise related error in the calculated concentration, in ppm.
x2 = Upper limit, in cm-1, of the analytical region.
x1 = Lower limit, in cm-1, of the analytical region.
AreaR = Total band area (corrected for path length, temperature, and
pressure) in analytical region of reference spectrum of
compound of interest.
CONR = Known concentration of compound in the same reference
spectrum.
This ratio provided a concentration equivalent to measured area in the subtracted spectrum. For instances when a compound was not detected, the value Eppm was equivalent to the minimum detectible concentration of that (undetected) species in the sample. Some concentrations given in Tables 3-3 3-4, and 3-5 are relatively high (greater than 10 ppm) and there are several possible reasons for this.
The reference spectrum of the compound may show low absorbance at
relatively high concentrations so that its real limit of detection
is high. An example of this may be acetonitrile.
The region of the spectrum used for the analysis may have residual
bands or negative features resulting from the spectral subtraction.
In these cases the absorbance of the reference band may be large at
low concentrations, but the RMSD is also large (see Equation 7).
An example of this is methyl chloride. If the maximum possible
concentrations for the hot/wet samples (14.42 ppm) and the
condenser samples (6.52 ppm) are compared for methyl chloride, the
drier spectra give a significant improvement because it is easier
to perform good spectral subtraction on spectra where absorbance
from water bands is weaker.
The chosen analytical region may be too large, unnecessarily
including regions of noise where there is no absorbance from the
compound of interest. An example of this may be ethyl benzene
where the chosen analytical region is more than 250 cm-1.
In the second and third cases the stated maximum possible concentration may be lowered by choosing a different analytical region, generating better subtracted spectra, or by narrowing the limits of the analytical region. Entropy has already taken these steps with a number of compounds. If more improvements can be made, they will be included in the final report. 4.6.4 Concentration Correction Factors Calculated concentrations in sample spectra were corrected for differences in absorption pathlength between the reference and sample spectra according to the following relation:
Corrections for variation in sample pressure were considered, and found to affect the indicated HAP concentrations by no more that one to two percent. Since this is a small effect in comparison to other sources of analytical error, no sample pressure corrections were made. 4.6.5 Analysis of Sample Concentration Spectra Sample concentration spectra were analyzed in the same manner as spectra of the gas phase samples. To derive flue gas concentrations it was necessary to divide the calculated concentrations by the concentration factor (CF). As an illustration, suppose that 10 ft3 (about 283 liters) of gas were sampled and then desorbed into the FTIR cell volume of approximately 8.5 liters to give concentration factor of about 33. If some compound was detected at a concentration of 50 ppm in the cell, then its corresponding flue gas concentration was about 1.5 ppm. When determining the concentration factor it was also important to consider that the dry gas meter was cool relative to the FTIR cell. Also, the total sampled volume was measured after most of the water was removed. The total volume of gas sampled was determined from the following relation:
The concentration factor, CF, was then determined using Vflue and the volume of the FTIR cell (Vcell) which was measured at an absolute temperature (Tcell) of about 300 K: Finally, the in-stack concentration was determined using CF and the calculated concentration of the sample contained in the FTIR cell, Ccell. 5.0 QUALITY ASSURANCE/QUALITY CONTROL ACTIVITIESQuality assurance (QA) is defined as a system of activities that provides a mechanism of assessing the effectiveness of the quality control procedures. It is a total integrated program for assuring the reliability of monitoring and measurement data. Quality control (QC) is defined as the overall system of activities designed to ensure a quality product or service. This includes routine procedures for obtaining prescribed standards of performance in the monitoring and measurement process. The specific internal QA/QC procedures that were used during this test program to facilitate the production of useful and valid data are described in this section. Each procedure was an integral part of the test program activities. Section 5.1 covers method-specific QC procedures for the manual flue gas sampling. Section 5.2 covers the QC procedures used for the instrumental methods. QC checks of data reduction, validation and reporting procedures are covered in Section 5.3, and corrective actions are discussed in Section 5.4. 5.1 QC PROCEDURES FOR MANUAL FLUE GAS TEST METHODS This section details the QC procedures that were followed during the manual testing activities. 5.1.1 Pitot Tube QC Procedures The QC procedures for pitot tube P measurements during the test runs followed guidelines set forth by EPA Method 2. The following QC steps were followed during these tests:
The S-type pitot tube was visually inspected before sampling.
Both legs of the pitot tube were leak checked before and after
sampling.
Proper orientation of the S-type pitot tube were maintained while
making measurements. The roll and pitch axis of the S-type pitot
tube was maintained at 90 deg to the flow.
The magnehelic set was leveled and zeroed before each run.
The pitot tube/manometer umbilical lines were inspected before and
after sampling for leaks and moisture condensate (lines were cleared
if found).
Reported duct dimensions and cross-sectional duct area were verified
by on-site measurements.
The stack gas temperature measuring system was checked by observing
ambient temperatures prior to placement in the stack.
The QC procedures that were followed in regards to accurate sample gas volume determination are:
The dry gas meter is fully calibrated immediately before the field
test using an EPA approved intermediate standard.
Pre-test and post-test leak checks were completed and were less than
0.02 cfm or 4 percent of the average sample rate.
The gas meter was read to a thousandth of a cubic foot for the
initial and final readings.
Readings of the dry gas meter and meter temperatures were taken every
10 minutes during sample collection.
Accurate barometric pressures were recorded at least once per day.
Post-test dry gas meter checks were completed to verify the accuracy
of the meter full calibration constant (Y).
5.1.2 Sample Concentration Sampling QC Procedures QC procedures that allowed representative collection of organics by the sample concentration sampling system were:
Only properly cleaned glassware and prepared adsorbent tubes that had
been kept closed with stainless steel caps were used for any sampling
train.
The filter, Teflon® transfer line, and adsorbent tube were maintained
at +-10 degF of the specified temperatures.
An ambient sample was analyzed for background contamination.
5.1.3 Manual Sampling Equipment Calibration Procedures 5.1.3.1 Type-S Pitot Tube Calibration -- EPA has specified guidelines concerning the construction and geometry of an acceptable Type-S pitot tube. If the specified design and construction guidelines are met, a pitot tube coefficient of 0.84 is used. Information pertaining to the design and construction of the Type-S pitot tube is presented in detail in Section 3.1.1 of EPA document 600/4-77-027b. Only Type-S pitot tubes meeting the required EPA specifications were used. The pitot tubes were inspected and documented as meeting EPA specifications prior to field sampling. 5.1.3.2 Temperature Measuring Device Calibration -- Accurate temperature measurements are required during source sampling. The bimetallic stem thermometers and thermocouple temperature sensors used during the test program were calibrated using the procedure described in Section 3.4.2 of EPA document 600/4-77-027b. Each temperature sensor is calibrated at a minimum of three points over the anticipated range of use against a NIST-traceable mercury-in-glass thermometer. All sensors were calibrated prior to field sampling. 5.1.3.3 Dry Gas Meter Calibration -- Dry gas meters (DGMs) were used in the sample trains to monitor the sampling rate and to measure the sample volume. All DGMs were fully calibrated to determine the volume correction factor prior to their use in the field. Post-test calibration checks were performed as soon as possible after the equipment was returned as a QA check on the calibration coefficients. Pre- and post-test calibrations should agree within 5 percent. The calibration procedure is documented in Section 3.3.2 of EPA document 600/4-77-237b. 5.2 QC PROCEDURES FOR INSTRUMENTAL METHODS The flue gas was analyzed for CO, O2, CO2, and HC. Prior to sampling each day, a pre-test leak check of the sampling system from the probe tip to the heated manifold was performed and was less than 4 percent of the average sample rate. Internal QA/QC checks for the instrumental test method measurement systems are presented below. Method 3A requires that the tester : (1) select appropriate apparatus meeting the applicable equipment specifications of the method, (2) conduct an interference response test prior to the testing program, and (3) conduct calibration error (linearity), calibration drift, and sampling system bias determinations during the testing program to demonstrate conformance with the measurement system performance specifications. The performance specifications are identified in Table 5-1. A three-point (i.e., zero, mid-, and high-range) analyzer calibration error check is conducted before initiating the testing by injecting the calibration gases directly into the gas analyzers and recording the responses. Zero and upscale calibration checks are conducted both before and after each test run in order to quantify measurement system calibration drift and sampling system bias. Upscale is either the mid- or high-range gas, whichever most closely approximates the flue gas level. During these checks, the calibration gases are introduced into the sampling system at the probe outlet so that the calibration gases are analyzed in the same manner as the flue gas samples. Drift is the difference between the pre- and post-test run calibration check responses. Sampling system bias is the difference between the test run calibration check responses (system calibration) and the initial calibration error responses (direct analyzer calibration) to the zero and upscale calibration gases. If an acceptable post-test bias check result is obtained but the zero or upscale drift result exceeds the drift limit, the test run result is valid; however, the analyzer calibration error and bias check procedures must be repeated before conducting the next test run. A run is considered invalid and must be repeated if the post-test zero or upscale calibration check result exceeds the bias specification. The calibration error and bias checks must be repeated and acceptable results obtained before testing can resume. Although not required by Methods 10 and 25A, the same calibration and data reduction procedures required by Method 3A were used for the CO and HC determinations to improve the quality of the reference data. 5.3 QA/QC CHECKS FOR DATA REDUCTION, VALIDATION, AND REPORTING Data quality audits were conducted using data quality indicators which require the detailed review of: (1) the recording and transfer of raw data; (2) data calculations; (3) the documentation of procedures; and (4) the selection of appropriate data quality indicators. All data and/or calculations for flow rates, moisture content, and sampling rates were spot checked for accuracy and completeness. In general, all measurement data have been validated based on the following criteria:
Acceptable sample collection procedures.
Adherence to prescribed QC procedures.
Upon completion of testing, the field coordinator was responsible for preparation of a data summary including calculation results and raw data sheets. 5.3.1 Sample Concentration The sample concentration custody procedures for this test program are based on EPA recommended procedures. Because collected samples were analyzed on-site, the custody procedures emphasize careful documentation of sample collection and field analytical data. Use of chain-of-custody documentation was not necessary, instead careful attention was paid to the sample identification coding. These procedures are discussed in more detail below. Each spectrum of a sample concentration sample has been assigned a unique alphanumeric identification code. For example, Tinl125A designates a Tenax® spectrum of a sample taken at the turbine outlet (HRSG inlet), from sample tube number 25. The A means this is the spectrum of the first desorption from this tube. Every adsorbent tube has been inscribed with a tube identification number. The project manager was responsible for ensuring that proper custody and documentation procedures were followed for the field sampling, sample recovery, and for reviewing the sample inventory after each run to ensure complete and up-to-date entries. A sample inventory was maintained to provide an overview of all sample collection activities. Two ambient samples were prepared at the turbine outlet. One was obtained before the test began and a second after the test was completed. One ambient sample was collected at the HRSG outlet. The ambient samples were run through the identical trains used in the runs. This provided a check for contamination of the sampling train. The charged ambient tube was stored and analyzed in the same manner as those obtained during testing. Ambient runs were 1-hour. The volume of air drawn for the blanks was sufficient to verify that the sampling train was clean and performing properly. Because relatively minor contamination was identified from the ambient samples, it was accounted for in the subsequent analyses of the sample spectra by using spectral subtraction. Major contamination was not observed in the samples. Sample flow at the dry gas meter was recorded at 10 minute intervals. Results from the analyzers and the spectra of the gas phase samples provided a check on the consistency of the effluent composition during the sampling period. 5.3.2 Gas Phase Analysis During each test run a total of 12 gas phase samples were collected and analyzed. Each spectrum was assigned a unique file name and a separate data sheet identifying sample location and sampling conditions. A comparison of all spectra in this data set provided information on the consistency of effluent composition and a real-time check on the performance of the sampling system. Effluent was directed through all sampling lines for at least 5 minutes and the CEMs provided consistent readings over the same period before sampling was attempted. This requirement was satisfied any time there was a switch to a different conditioning system. When the cell was being evacuated, the FTIR was continuously scanning to provide a spectral profile of the empty cell. A new sample was not introduced until there was no residual absorbance remaining from the previous one. The FTIR was also continuously scanning during sample collection to provide a real-time check on possible contamination in the system. 5.3.3 FTIR Spectra For a detailed description of QA/QC procedures relating to data collection and analysis, refer to the "Protocol For Applying FTIR Spectrometry in Emission Testing." A spectrum of the calibration transfer standard (CTS) was collected at the beginning and end of each data collection session. The CTS gas was 100 ppm ethylene in nitrogen. The CTS spectrum provided a check on the operating conditions of the FTIR instrumentation, e.g. spectral resolution and cell path length. Ambient pressure was recorded whenever a CTS spectrum was collected. Two copies of all interferograms and processed spectra of backgrounds, samples, and the CTS were stored on separate computer disks. Additional copies of sample and CTS absorbance spectra were also stored for use in the data analysis. Sample spectra can be regenerated from the raw interferograms, if necessary. FTIR spectra are available for inspection or re-analysis at any future date. Pure, dry ("zero") air was periodically introduced through the sampling system in order to check for contamination. Contamination was not observed, but on one occasion water condensed in the cell manifold. The lines and cell were purged with dry N2, until the contamination was eliminated. As successive spectra were collected the position and slope of the spectral base line were monitored. If the base line within a data set for a particular sample run began to deviate by more than 5 percent from 100 percent transmittance, a new background was collected. During the course of the test program, it was the responsibility of the field coordinator and the sampling team members to see that all measurement data procedures were followed as specified and that measurement data met the prescribed acceptance criteria. 6.0 CONCLUSIONS AND DISCUSSIONEntropy conducted an emissions test at T.H. Wharton Electric Generating Station in Houston, Texas. Direct gas phase analysis, and sample concentration testing were both performed over two days. At the same time gas analyzers were used to measure CO, O2, CO2, and hydrocarbons in the gas streams. Three 4-hour sample concentrations runs were conducted at the gas turbine outlet. Direct gas phase analyses and CEM measurements were performed concurrently with the sample concentration runs. Additionally, one 4-hour sample concentration run was conducted at the stack during Run 2 at the turbine outlet. Gas phase analysis revealed the presence of water vapor, CO, CO2, NO2 and NO. HCl, ammonia and nitrous acid (HNO2) were detected in sample concen- tration spectra. Also, some unidentified absorption bands were observed in the sample concentration spectra. A primary goal of this project was to use FTIR instrumention in a major test program to measure as many HAPs as possible or to place upper limits on their concentrations. Four other electric utilities were tested along with the T.H. Wharton facility. Utilities present a most difficult testing challenge for two reasons:
1) They are combustion sources so the flue gas contains high levels of
moisture and CO2 (both are spectral interferants).
2) The large volumetric flow rates typical of these facilities can lead
to mass emissions above regulated limits even for HAPs at very low
concentrations. This places great demand on the measurement method
to achieve low detection limits. Furthermore, with natural gas as
the combustion fuel, concentrations of any HAPs formed in the process
would be expected to be very low.
This represents the first attempt to use FTIR spectroscopy in such an ambitious test program. The program accomplished very significant achievements and demonstrated important and fundamental advantages of FTIR spectroscopy as an emissions test method:
Using a single method quantitative data were provided for over 100
compounds.
Software was written to analyze a large data set and provide
concentration and detection limit results quickly. The same or
similar software can be used for subsequent tests with very little
investment of time for minor modifications or improvements.
The original data are permanently stored so the results can be
rechecked for verification at any time.
A single method was used to obtain both time-resolved (direct gas)
and integrated (sample concentration) measurements of gas streams
from two locations simultaneously.
The two techniques of the FTIR method cover different concentration
ranges.
Preliminary data (qualitative and quantitative) are provided on-site
in real time.
With little effort at optimization (see below), detection limits in
the ppb range were calculated for 26 HAPs and between 1 and 5 ppm for
77 other HAPs using direct gas phase measurements of hot/wet samples,
which present the most difficult analytical challenge. Sample
concentration provided even lower detection limits for some HAPs.
A compound detect is unambiguous.
It is appropriate to include some discussion about the "maximum possible concentrations" presented in Tables 3-3 to 3-5. These numbers were specifically not labeled as detection limits because use of that term could be misinterpreted, but they will be referred to as "detection limits" in the discussion below. In FTIR analysis detection limits are calculated directly from the spectra (see Section 4.6.3 and the "FTIR Protocol"). These calculated numbers do not represent fundamental measurement limits, but they depend on a number of factors. For example:
Some instrumental factors
Spectral resolution.
Source intensity.
Detector response and sensitivity.
Path length that the infrared beam travels through the sample.
Scan time.
Efficiency of infrared transmission (through-put).
Signal gain.
Some sampling factors
Physical and chemical properties of a compound.
Flue gas composition.
Flue gas temperature.
Flue gas moisture content.
Length of sample line (distance from location).
Temperature of sampling components.
Sample flow.
Instrumental factors are adjustable. For this program instrument settings were chosen to duplicate conditions that were successfully used in previous screening tests and the validation test. These conditions provide speed of analysis, durability of instrumentation, and the best chance to obtain measurements of the maximum number of compounds with acceptable sensitivity. Sampling factors present the same challenges to any test method. An additional consideration is that the numbers presented in Tables 3-3 to 3-5 are all higher than the true detection limits that can be calculated from the 1 cm-1 data collected at T.H. Wharton. This results from the method of analysis: the noise calculations were made only after all spectral subtractions were completed. Each spectral subtraction adds noise to the resulting subtracted spectrum. For most compounds it is necessary to perform only some (or none) of the spectral subtractions before its detection limit can be calculated. With more sophisticated software it will be possible to automate the process of performing selective spectral subtractions and optimize the detection limit calculation for each compound. (Such an undertaking was beyond the scope of the current project.) Furthermore, the detection limits represent averages compiled from the results of all the spectra collected at the sampling location. A more realistic detection limit is provided by the single spectrum whose analysis gives the lowest calculated value. It would be more accurate to think of "maximum possible concentrations" as placing upper boundaries on the HAP detection limits provided by these data. Another important sampling consideration is the sample composition. In Table 3-3 benzene's detection limit is quoted as 4.83 ppm. This was determined in the analytical region between 3020 and 3125 cm-1. Benzene exhibits a much stronger infrared band at 673 cm-1 but this band was not used in the analysis because absorbance from CO2 strongly interfered in this analytical region. At a lower CO2 emission source an identical FTIR measurement system would provide a benzene detection limit below 1 ppm for direct gas analysis (even ignoring the consideration discussed in the previous paragraph). Any difficulties associated with measuring particular compounds are related to the sampling conditions and not the FTIR analysis. It was necessary to cool the flue gas from about 1000 degF down to about 300 degF. This introduced the possibility of condensing relatively non-volatile species in the sampling line. The moisture content of the flue gas was estimated to be about 15 percent and this should have caused no problem with condensation in the sampling line. However, water soluble species are more difficult to measure at higher moisture levels. FTIR techniques offer a good way to measure unstable or reactive species because FTIR spectrometry can be readily used to monitor the sampling system integrity. That was not done in this test because the primary goal was the general one of measuring as many compounds as possible, not optimizing the measurement system for any particular compound or set of compounds. 7.0 REFERENCES
1) "FTIR Method Validation at a Coal-Fired Boiler," EPA Contract No.
68D20163, Work Assignment 2, July, 1993.
2) "Computer-Assisted Quantitative Infrared Spectroscopy," Gregory L.
McClure (ed.), ASTM Special Publication 934 (ASTM), 1987.
3) "Multivariate Least-Squares Methods Applied to the Quantitative Spectral
Analysis of Multicomponent Mixtures," Applied Spectroscopy, 39(10), 73-
84, 1985.
4) "Fourier Transform Infrared Spectrometry," Peter R. Griffiths and James
de Haseth, Chemical Analysis, 83, 16-25,(1986), P. J. Elving, J. D.
Winefordner and I. M. Kolthoff (ed.), John Wiley and Sons,.
|
where:
A = The n x m matrix representing the absorbance values of the m
reference spectra over the n wavenumber positions, containing
contributions from all or some of the q components;
K = The n by q matrix representing the relationship between absorbance
and concentration for the compounds in the wavenumber region(s) of
interest, as represented in the reference spectra. The matrix
element Knq = banq, where anq is the absorptivity of the qth
compound at the nth wavenumber position;
C = The q x m matrix containing the concentrations of the q compounds
in the m reference spectra;
E = The n x m matrix representing the random "errors" in Beer's law for
the analysis; these errors are not actually due to a failure of
Beer's law, but actually arise from factors such as
misrepresentation (instrumental distortion) of the absorbance
values of the reference spectra, or inaccuracies in the reference
spectrum concentrations.
where:
Ccorr = The pathlength corrected concentration.
Ccalc = The initial calculated concentration (output of the Multicomp
program designed for the compound)
Lr = The pathlength associated with the reference spectra.
Ls = The pathlength (22m) associated with the sample spectra.
Ts = The absolute temperature of the sample gas (388 K).
Tr = The absolute gas temperature at which reference spectra
were recorded (300 to 373 K).
where:
Vflue = Total volume of flue gas sampled.
Vcol = Volume of gas sampled as measured at the dry gas meter after
it passed through the collection tube.
Tflue = Absolute temperature of the flue gas at the sampling location.
Tcol = Absolute temperature of the sample gas at the dry gas meter.
W = Fraction (by volume) of flue gas stream that was water vapor.
