Experiences in Long-Term Evaluation of Mercury Emission Monitoring Systems

Transcription

1 Energy & Fuels XXXX, xxx, A Experiences in Long-Term Evaluation of Mercury Emission Monitoring Systems Chin-Min Cheng,* Hung-Ta Lin, Qiang Wang, Chien-Wei Chen, Chia-Wei Wang, Ming-Chung Liu, Chi-Kuan Chen, and Wei-Ping Pan Institute for Combustion Science and EnVironmental Technology, Department of Chemistry, Western Kentucky UniVersity, 2413 NashVille Road, Bowling Green, Kentucky ReceiVed NoVember 10, ReVised Manuscript ReceiVed May 5, 2008 Six mercury continuous emission monitoring (CEM) systems provided by two leading mercury (Hg) CEM system manufacturers were tested at five coal combustion utilities. The linearity, response time, day-to-day stability, efficiency of the Hg speciation modules, and ease of use were evaluated by following procedures specified in the Code of Federal Regulation Title 40 Part 75 (40 CFR Part 75). Mercury monitoring results from Hg CEM systems were compared to an EPA-recognized reference method. A sorbent trap sampling system was also evaluated in this study to compare the relative accuracy to the reference method as well as to Hg CEM systems. A conceptual protocol proposed by U.S. EPA (Method 30A) for using an Hg CEM system as the reference method for the Hg relative accuracy (RA) test was also followed to evaluate the workability of the protocol. This paper discusses the operational experience obtained from these field studies and the remaining challenges to overcome while using Hg CEM systems and the sorbent trap method for continuous Hg emission monitoring. Introduction Mercury (Hg) is one of the 189 hazardous air pollutants listed in the 1990 Amendments to the Clean Air Act. Coal-fired power generation is the largest anthropogenic source of Hg in the U.S. and is responsible for the annual release of approximately 50 tons of Hg into the atmosphere. 1,2 The U.S. Environmental Protection Agency (EPA) announced the Clean Air Mercury Rule (CAMR) in 2005, which capped mercury emissions from coal-fired power utilities and established a mercury cap-andtrade program. Although the rule was vacated by the District of Columbia Circuit in February 2008, a more restrictive Hg emission regulation is expected to be implemented by the EPA using Maximum Achievable Control Technology (MACT) standards. To ensure the Hg emission reduction goals can be met, the implemented mercury emission regulation will also require affected electric utility units to continuously monitor mercury (Hg) mass emissions using available monitoring techniques [e.g., Hg continuous emission monitoring (CEM) system and sorbent trap method]. Any applied monitoring system will be subject to restrictive certification and quality assurance/quality control (QA/QC) procedures, which are currently described in the 40 CFR Part 75. The Institute for Combustion Science and Environmental Technology (ICSET) at Western Kentucky University (WKU), * To whom correspondence should be addressed. chin-min.cheng@ wku.edu. Current address: Department of Chemical Engineering, Ming-Chi University, Taipei, Taiwan. (1) U.S. Environmental Protection Agency. Mercury study report to Congress. Volume II: An inventory of anthropogenic mercury emissions in the United States. Technical Report, EPA-452/R b, Office of Air Quality Planning and Standards, Washington, D.C., (2) Brown, T. D.; Smith, D. N.; Hargis, R. A., Jr.; O Dowd, W. J. J. Air Waste Manag. Assoc. 1999, 49, in conjunction with Electric Power Research Institute (EPRI) and Illinois Clean Coal Institute (ICCI), conducted six field studies to evaluate using the Hg CEM systems and sorbent trap method for measuring Hg emissions. The study was carried out under a wide variety of coal-fired power generation operation conditions. Two types of pulverized coal (PC) boilers (i.e., wall-fired and cyclone) and a circulating fluidized bed (CFB) combustion system were studied. Bituminous, powder river basin (PRB) sub-bituminous, and lignite coals were burned by these tested facilities. Combinations of various air pollution control devices were also included in the study to provide comprehensive evaluation results. In these field studies, the initial certification and data quality assurance and quality control (QA/QC) procedures specified in the 40 CFR Part 75 were carried out to evaluate the capability of the tested Hg CEM systems to pass compliance requirements, as well as the data integrity and system stability during operation. Relative accuracy (RA) of the test Hg CEM systems was determined using the Ontario Hydro (OH) method. A sorbent trap method for monitoring the Hg emission was also tested following the procedures and data QA/QC criteria specified in the Appendix K of 40 CFR Part 75. In addition, a conceptual protocol (Method 30A) proposed by EPA for using Hg CEM systems as a reference test method for conducting RA testing was also carried out to evaluate the workability of the protocol. During the field study 1 (FS-1), stratification tests, using NO x, SO 2, and O 2 as substitutes for Hg, were also performed. This paper discusses the results obtained from instrument certification and data QA/QC verification tests, relative accuracy tests, and emission monitoring. This paper also discusses the /ef CCC: $40.75 XXXX American Chemical Society Published on Web 07/22/2008

2 B Energy & Fuels, Vol. xxx, No. xx, XXXX Cheng et al. field test number facility boiler fuel types Table 1. Configurations of Test Sites capacity (MWe) boiler types APCD configurations Hg in coal a (µg/g) Hg in stack gas (µg/dscm) CEM participated b FS-1 T unit 4 lignite 650 PC c ESP d + FGD e 0.20 ( ( 3 f,g A2 and B2 FS-2 S unit 123 bituminous 115 CFB h SNCR i + FF j ( ( 0.15 f B1 FS-3 C unit 31/32 bituminous 90 Cyclone SCR k + ESP + FGD ( ( 0.3 f A1 2.3 ( 1.6 g FS-4 S unit 4 bituminous 173 Cyclone SCR + ESP + FGD ( ( 0.16 g A1 FS-5 C unit 33 bituminous 205 PC SCR + ESP + FGD ( ( 0.18 g B1 FS-6 M unit 2 PRB l 140 PC ESP NA NA A3 and B3 a Dry based. b CEM A1 and B1 were provided by ICSET, and the others were provided by the facilities. c Pulverized coal. d Electrostatic precipitator. e Flue gas desulfurization. f Non-ozone season. g Ozone season. h Circulation fluidized bed. i Selective noncatalytic reduction. j Fabric filter. k Selective catalytic reduction. l Powder river basin. Table 2. Concentration of Hg Gases Generated by CEM A1 Verified by OH and Sorbent Trap Methods Hg concentration (µg/n m 3 ) Ontario Hydro sorbent trap C targeted ) 10.2 µg/n m ( ( 0.8 RD (%) Hg calibration source Table 3. Evaluation of Hg Calibration Gases response CEM targeted concentration (µg/dscm) response concentration (µg/dscm) RD (%) CEM A-1 CEM A ( CEM A-1 CEM B ( cylinder CEM A ( cylinder CEM B ( CEM B-3 CEM A ( CEM A-3 CEM B ( operational experience obtained from these field studies and the remaining challenges to be overcome. Experimental Section Testing Site. Evaluation studies were conducted at five coal combustion facilities. The configurations of each test unit are summarized in Table 1. As shown in the table, the five units tested in this study represented a wide range of operational configurations in terms of fuels, boiler types, and combinations of air pollution control devices (APCDs). The Hg concentrations in the coal used by the test boilers ranged from to 0.2 µg/g. The averages of Hg concentration in the flue gases of the test units during the evaluation period ranged from 0.19 to 14 µg/dscm. The results were determined by the OH method, which is one of the EPA reference methods for flue gas Hg measurement. Hg Continuous Emission Monitoring (CEM) Systems. Six Hg CEM systems were tested. Three of the systems were provided by manufacturer A and are referred to as A1, A2, and A3 in the Results and Discussion. The other three systems were provided by manufacturer B and are referred to as B1, B2, and B3 in the following section. Table 1 illustrates what Hg CEM system was used in each field evaluation. For a given Hg CEM system, the configuration includes an atomic fluorescence spectroscopy Hg analyzer, a calibration unit for providing elemental Hg gases, a sampling probe, a heated umbilical line for transporting flue gas samples and sampling parameters, and a gas conditioner for converting ionic Hg to elemental Hg in the flue gas and for removing acidic gas components. All Hg CEM systems used fast loop, inertial type sampling probes allowing for ash-free flue gas to be extracted from the stack. This type of probe extracts the flue gas by a stream of compressed air passing through an eductor installed at the exit end of the flue gas sampling loop. The movement of the compressed air creates a pressure differential across the eductor, which generates an axial flue gas flow in the sampling loop with high velocity. After being extracted, flue gas containing fly ash particles continue to travel in a straight direction. A sample stream is withdrawn from the main flue gas flow by a vacuum created by a second eductor at a very low filter face velocity, which separates the sample stream from fly ash. High velocity axial gas flow and low radial velocity prevent particles entrained in the flue gas from depositing on or penetrating into the porous filter wall. The sample stream was diluted with zero air before it was transported to the analyzer. The dilution ratios of the participating Hg CEM systems ranged from 28-41, which were adjusted on the basis of calibration results. For Hg CEM systems from manufacturer A, the diluted gas sample was transported through a heated umbilical line (180 C) to a conditioner located more than 100 m away from the sampling probe. In the conditioner, the gas sample was separated into two streams: i.e., Hg(0) and Hg(T). The Hg(0) stream delivered a gas sample containing only elemental Hg and other insoluble components to the analyzer after passing through a scrubber, in which ionic Hg and acidic components in the gas were removed. The Hg(T) stream passed a catalyst tube where ionic Hg was reduced to elemental Hg at a temperature of 700 C. After removing acidic components in another scrubber, the Hg(T) stream was sent to the analyzer. A chiller (5 C) was used to remove excess moisture in the conditioned sample gas. For Hg CEM systems from manufacturer B, the sample conditioning unit was located in the probe assembly. The flue gas sample was immediately treated after being diluted. Instead of using deionized water, the system used sorbent material to remove acidic components from the sample stream. Both A and B Hg CEM systems used atomic fluorescence spectroscopy as the Hg detector. The difference between the two types of analyzers was the application of gold traps. Two parallel gold traps were used by the analyzers in the Hg CEM systems from manufacturer A to selectively capture the elemental mercury in the sample gas prior to the detector. Elemental mercury in the sample gas formed an amalgam with the gold surface. Once loaded, the trap was heated and flushed to the detector with ultra pure argon. One complete adsorption and desorption cycle took approximately 150 s. In the case of Hg CEM systems from manufacturer B, flue gas was continually delivered into the detector without passing through gold traps and, therefore, generated continuous Hg readings. The B Hg CEM systems were set to provide a 5 min average reading every 1 min. Only compressed air was required to operate the Hg CEM systems from manufacturer B. In the case of the A Hg CEM systems, in addition to compressed air, supplies of ultra high purity argon and deionized water were also required. Sorbent Trap Mercury Sampling System. A sorbent trap sampling system provided by Apex Instruments, Inc. (Fuquay- Varina, NC) was used in this study. A known volume of dry flue gas was extracted from the stack through paired sorbent traps with a constant flow rate of L/min. Each trap consisted of three sections of activated carbon. The first section was designed to capture Hg in the flue gas. The second and third sections were used for sampling and analytical QA/QC purposes. After sampling, each section of the trap was analyzed for Hg using an RP-M324 mercury analyzer (Ohio Lumex, OH). The analyzer decomposes the sorbent with a known mass at a temperature between 600 and 800 C. The

3 EValuation of Mercury Emission Monitoring Systems Energy & Fuels, Vol. xxx, No. xx, XXXX C Table 4. Summaries of Hg CEM Setup, Certification, and Data Validation Results total study duration setup/ adjustment duration a (day) system down time b (day) 7 day calibration error check linearity check system integrity check cycle time check (min) Observed from Field Studies 1 and 2 CEM A-1 (FS-1) c /90 NA NA NA NA CEM B-1 (FS-2) /51 2/24 2/2 1/1 13 CEM A-2 (FS-1) /74 0/2 4/4 2/ CEM B-2 (FS-1) /71 1/2 5/6 2/3 8 Observed from Field Studies 3, 4, and 5 after Modification CEM A-1 (FS-3) /51 2/2 1/1 1/1 10 CEM B-1 (FS-4) /65 2/2 1/1 1/1 8 CEM A-1 (FS-5) /66 2/2 1/1 1/1 10 a The duration is the days required since the system was powered up to the day when the first test of the 7 day calibration error check was started. It excludes the time required for umbilical line installation and infrastructure construction. b System down time reports the days that the system was brought offline for maintenance. c FS ) field study. Hg concentration was determined by measuring the mercury vapor released during the decomposition using a Zeeman atomic adsorption spectroscopy. Sampling Location. For a given test, the sampling location for the Hg CEM system was above the nearest flow restriction by a distance of greater than 10 duct diameters and away from the flue gas outlet by at least 22 duct diameters. The gas sample (treated or untreated depending upon the CEM systems) was transported to the analyzer through a heated umbilical line ( m in length). While carrying out OH method sampling, the stack gas was collected at either 90 or 180 from the Hg CEM sampling probe with the exception of field test 5 (FS-5). In the FS-5 test, the flue gases were extracted from the two flue gas desulfurization (FGD) unit outlet ducts for OH method sampling. Certification Tests. A series of inertial certification tests described in the Appendix A of 40 CFR Part 75 were carried out (with the exception of FS-6), which included a 7 day daily calibration error check, linearity check, system integrity check, and cycle time test. The relative accuracy of a given evaluated Hg CEM system (except for FS-6) with respect to the OH method results was also determined. A total of 12 runs of paired OH method sampling were carried out for the relative accuracy test. The data QA/QC procedures provided in Appendix B of Part 75 were also followed to evaluate the stability of the CEM systems. Inertial Certification Test. Detailed procedures for each test can be seen in Appendix A of 40 CFR Part 75. In summary, the 7 day calibration error test was to evaluate the accuracy and stability of the calibration of the Hg gas monitor over an extended period of operation. For the linearity check test, three concentration levels of elemental Hg gases, which were chosen based on the fuel burned by the coal combustion utility, were measured by an Hg CEM system to determine whether the response of a gas monitor is linear across its range. Instead of using elemental Hg, three levels of ionic Hg gases were used in the system integrity check to verify the effectiveness of the sample gas conditioning unit. The cycle time test was to measure whether a gas monitoring system is capable of completing at least one cycle of sampling, analyzing, and data recording every 15 min. RelatiVe Accuracy Test. The relative accuracy (RA) test was performed using the OH reference method. Detailed RA test procedures are described in Appendix A of 40 CFR Part 75. In short, while an RA test was conducted in this study, 12 runs of OH method samplings were carried out within a period of 96 consecutive unit operating hours. The evaluated Hg continuous monitoring system remained operating during the OH method sampling. No system adjustment was performed during the RA test. After the RA test, at least 9 sets of the OH method data were chosen based on the criteria specified in the procedures and those data were used as the reference values. A total of 9-12 sets of Hg monitoring results from the test CEM system, which were the mean values of the Hg readings collected from perspective OH method sampling periods, were compared to the reference values. If the instrument passed the RA test, the bias test was performed to determine whether the monitoring system was biased low with respect to the OH reference method. Data QA/QC Procedures. Three types of tests (i.e., daily calibration error check, weekly one point system integrity check, and linearity check) were performed on the test CEM after the first RA test was completed. Testing procedures were very similar to the tests described in the initial certification tests. Instrumental Reference Method (EPA Method 30A). The procedures described in the conceptual method using the Hg CEM system as a reference method for conducting relative accuracy testing were evaluated in field study 1 (FS-1). Both CEM A2 and B2 were used as the IRM systems. Results and Discussion Calibration Gas Evaluation. The National Institute of Standards and Technology (NIST)-traceable gas standards and protocols for mercury gas generators and cylinders were not available when the studies were conducted. To evaluate the Hg gas provided by the calibration units of the CEM A and B systems, a series of comparison tests were performed. Two concentration levels of Hg gas cylinders provided by Spectrum Gases, Inc. (West Branchburg, NJ) were also used. First, the Hg gas from the CEM A2 calibration unit was evaluated by three runs of the OH method test. The concentration was also evaluated by the sorbent trap method. The results are summarized in Table 2, where the relative deviation (RD, %) is calculated by the following equation: RD (%) ) C measured 100% (1) C targeted The results shown in the table have been corrected to 20 C. As can be seen in the table, the target values of the evaluated Hg gas was 9.7 and 0.9% higher than the averages of the three OH method and sorbent trap method measurements, respectively. The relative standard deviations of the three OH method and sorbent trap method results were 2.2 and 7.9%, respectively. The test illustrated that the Hg gas generated from the calibration unit of the A1 Hg CEM system was somehow close to the reference value. The second test of the series consisted of measuring the Hg gas from the A1 calibration unit using the A2 and B2 systems. While doing measurements, the A1 Hg gas was delivered through the calibration gas line, inertial filter, sample conditioner, heated sample delivery line, and the analyzer of the two measuring Hg CEM systems. In addition, a Hg cylinder gas (9.5 µg/n m 3 ) provide by Spectra Gases (West Branchburg, NJ) was also used. Results obtained from the study are listed in Table 3.

4 D Energy & Fuels, Vol. xxx, No. xx, XXXX Cheng et al. Table 5. Summary of Operation Difficulties of Mercury CEM Systems in Field Tests period problem observed solutions A1 in FS-1 7/30-8/8 1. incorrect umbilical line temperature replaced umbilical line 2. no sample flow incorrect orifice differential pressure 8/12-9/25 1. unstable proportion valve operation 2. unstable loop flow rate 3. incorrect absolute venture pressure 1. cleaned sampling loop 2. sampling loop and calibration line were found reversely connected to the stingers. A2 in FS-1 7/19-7/28 1. probe was not heated 2. low Hg 2+ readings 1. replaced heating elements 2. replace catalyst 8/20-8/22 incorrect probe pressure response brushed sampling loop frequently 9/13-9/25 1. low Hg 2+ conversion rate 2. low sampling flow rate 3. software crash 1. replaced catalyst 2. replaced peristaltic pump tubings 3. upgraded instrument software B2 in FS-1 7/18-7/29 low calibration recovery 1. replaced probe stinger 2. replaced heating element 3. replaced calibration unit 4. replaced lamp 5. added a humidifier in the calibration gas path 6. cleaned umbilical line 7. replaced catalyst 8/23-9/9 low calibration recovery 1. replaced inertial filter 2. adjusted calibration unit 9/14-9/17 1. drifting issue of the analyzer 1. controlled reaction cell temperature B1 in FS-2 11/17-12/9 1. incorrect probe temperature high zero air background 2. low sample flow 3. unstable calibration gas flow rate 1. replaced probe 2. cleaned umbilical line 3. removed blockage found in the S2 valve in the analyzer 12/20-12/22 1. unstable calibration results and Hg readings 1. replaced the lamp with a heating element 2. upgraded instrument software 12/25-1/14 1. unstable and negative Hg readings 1. installed nitrogen generator 2. used nitrogen as carrier gas for the analyzer 3. replaced the calibration unit It was found that the response of the A2 system to the A1 Hg gas was within 2% of standard deviation. The response of the B2 system to the A1 Hg gas was 34% higher than the target value. Using the Hg gas cylinder, the response of the B2 system was closer than the response from the A2 system. According to the test results, in general, the response from Hg CEM systems provided by manufacturer B were systematically higher than the results from the A systems. The observed discrepancy was due to the concentration bias between the Hg calibration gases generated from the two units. At a specified concentration, the calibration gas generated from the A calibration units was higher than the Hg gas provided by the B units. The conclusion is illustrated by field study 6 (FS-6), in which Hg gases from the calibration units of A3 and B3 were measured against each other. As can be seen in Table 3, an approximate 20% of disagreement was observed between the tag value and the response of the measuring Hg CEM. Mercury CEM System Setup, Certification, and Operation. The observations during the system setup and initial certification period are discussed. Table 4 provides an overview of the evaluation results from four Hg CEM systems (A1, A2, B1, and B2). It was found that during the FS-1 and FS-2 studies, except for CEM A1, the systems required at least 2 weeks for setup and performance adjustment before the first day of the 7 day calibration error check began. The setup and adjustment duration did not include the time spent on umbilical line installation and infrastructure construction (e.g., compressed and/ or argon gas lines, deionized water supply, power supply, and/ or instrument shelter). Although the A1 system started the inertial certification test after 5 days of setup and adjustment, a combination of multiple operational difficulties, including heated umbilical line malfunction and probe blockage, prevented the A1 system from participating in the test after approximately 2 weeks of operation. The probe blockage was later found because of a reversed connection of the sampling loop and calibration gas loop at the heated probe tip. A summary of the operational difficulties found during the testing period can be seen in Table 5. As demonstrated in the table, Hg CEM systems A2, B1, and B2 all encountered multiple system malfunctions during the first two field studies, i.e., FS-1 and FS-2. In the case of the A2 system, the catalyst in the sample conditioner was replaced twice during the testing period because low conversion efficiencies were observed. However, no catalyst was replaced for the A1 system in the FS-3 and FS-5 studies, in which the system was evaluated for a longer operational time. The observed different lifetime of the catalyst might be due to the presence of certain element(s) or/and compound(s) in the flue gases. For example, the lignite coal used in the FS-1 study contained a much higher Se concentration level (8.1 ( 0.9 µg/g) compared to the Se contents (ranging from 0.08 to 0.2 µg/g) in the coals used in the FS-3 and FS-5 studies. However, the effect of flue gas compositions on the lifetime of the catalyst needs further investigation. In general, sampling loop blockage and contamination, electronic parts failure, and malfunctions of the calibration unit were the most commonly found problems causing unsatisfactory calibration results. When the system encountered operational difficulties, various corrective actions (e.g., parts replacement, probe cleaning and maintenance, operational software update, changing unit, to maintain the Hg CEM) were taken. Similar operational difficulties were also found by the other testing

5 EValuation of Mercury Emission Monitoring Systems Energy & Fuels, Vol. xxx, No. xx, XXXX E Figure 1. Effect of lamp temperature on the stability of the B1 Hg CEM system observed in the FS-2 study. group. In a 10 month long-term field evaluation conducted by EPA, 3 similar operational problems were also reported. In addition, significant drifting was observed on the B1 system tested in the FS-2 study (Figure 1), where the average stack Hg concentration was 0.35 µg/dscm. As shown in Figure 1, the stack Hg readings fluctuate between 1.8 and -1.2 µg/dscm within a 10 h period. The drifting and high background noise problems were found to be correlated to the variation of lamp temperature and inappropriate carrier gas for the Hg analyzer, which both affected the intensity stability of the Hg lamp. The problems were improved by using N 2 as the carrier gas and controlling the lamp temperature at 35 C. The monitoring results after the modification remained relatively constant at a level of 0.4 µg/dscm (Figure 1), which was very close to the OH method measurement results. Before the modification, the system used purified compressed air as the carrier gas, which might reduce the sensibility of the analyzer because of a quenching effect. Using an atomic fluorescence spectroscopy detector for Hg analysis is based on the measurement of fluorescence emanating from excited mercury atoms contained in the gas sample. However, fluorescence can be quenched quickly through the collision of excited mercury atoms with other components of the sample gas, especially oxygen, which affected the monitoring readings. After modifications and maintenances in the FS-1 and FS-2 studies, the CEM A1 and B1 systems later participated in three other field studies (FS-3, FS-4, and FS-5). Different from the operational experiences of the FS-1 and FS-2 studies, both systems required much less time (less than 3 days) to setup and perform adjustments before precertification tests began. During approximately 2 months of operation, no downtime was required for system maintenance. The operational stability of Hg CEM systems A1 and B1 after modification are demonstrated by the daily calibration error check results from the FS-3, FS-4, and FS-5 studies (Figure 2). As shown in the figure, both A1 and B1 passed every daily calibration error check performed in the FS-3 and FS-4 studies. (3) U.S. Environmental Protection Agency. Long-term field evaluation of mercury continuous emission monitoring systems: Coal-fired power plant burning eastern bituminous coal and equipped with selective catalytic reduction (SCR), electrostatic precipitator (ESP), and wet scrubber: Field activities from November 2004 to September Final Report of EPA Contract GS-10F-0127J, Office of Air Quality Planning and Standards Air Quality Assessment Division, Research Triangle Park, NC, In the FS-5 study, two problems (i.e., probe blockage and DI water overflow) occurred during the 68 day operation, which resulted in eight unsatisfactory daily calibration error checks. The operational experiences obtained from these studies show that the Hg systems provided by the two manufacturers were still in the development stage during the testing period. Many instrument parts had not yet been standardized or qualitycontrolled. The systems required a long time to install and perform adjustments because both manufacturers needed to search for suitable setups. The stability study demonstrates that the Hg CEM systems were able to run without major maintenance for a period of time after the necessary modifications were made. Relative Accuracy and Bias Tests. During each field test, two series of relative accuracy (RA) tests were performed on every evaluated CEM system with the exception of the FS-5 and FS-6 studies. Only one series of RA tests was carried out at the middle of the FS-5 study, and no test was performed on the FS-6 study. For the other four field studies, the first series of RA tests was conducted after the initial certification tests were completed. The second series was carried out during the last week of each field study. Results obtained from the RA tests conducted in the four field tests are summarized in Table 6. The bias test results were also calculated and listed in the table. As shown in Table 6, the RA results of A2 and B2 were less than 20%. In the other field studies, although the RA values were greater than 20%, the evaluated Hg CEM systems also passed the RA test because the differences between the mean values of the CEM measurements and the OH method mean values did not exceed 1.0 µg/dscm. According to the criteria specified in the procedure, the results were acceptable when the average of the Hg concentration measured by the OH method during the RA test was less than 5.0 µg/dscm. By comparing the CEM readings to the OH method results, it was found that the two CEM systems provided by manufacturer A (i.e., A1 and A2) responded to lower Hg values, except for the readings from the first series of the RA tests in the FS-3 study. In the five RA tests carried out at the CEM A1 system, three sets of the results show the system was bias low. In the case of the Hg CEM systems provided by manufacturer B, the readings were consistently higher than the results from the OH method. The lower responses observed in the A Hg CEM systems and higher responses observed in the B Hg CEM systems were likely due to the bias caused by the Hg calibration gases provided by the two systems. As discussed in the previous section, the A Hg CEM systems were calibrated with a higher concentration of Hg gas compared to the Hg gas provided by the B systems with the same tag value. As a result, the A Hg CEM systems responded with lower stack Hg monitoring results. EPA Method 30A (IRM) Evaluation. An Hg CEM system is required to meet a series of certification and data quality assurance tests (specified in the EPA Method 30A) when it is used as an instrument reference method for an RA test. Detailed procedures for performing the tests are described in a draft version of EPA Method 30A. Two Hg CEM systems, A2 and B2, were used to evaluate how the two Hg CEM systems respond to the tests specified in the method. The time required to perform the IRM protocol was evaluated. Table 7 summarizes the results obtained from the Hg 0 and Hg 2+ calibration error check tests. As shown in the table, the two evaluated CEM systems passed the two precertification tests. The CEM B2 passed the Hg 2+ system calibration error check

6 F Energy & Fuels, Vol. xxx, No. xx, XXXX Cheng et al. Figure 2. Daily calibration results of Hg CEM systems during operation: (a) B-1 at field study 2, (b) A-1 at field study 3, (c) A-1 at field study 4, and (d) B-1 at field study 5. Table 6. Results of RA and Bias Tests field study unit A2 B2 B1 A1 B1 A1 series Hg CEM System OH method available runs OH a average CEM a average NA RA b (%) NA djc NA cc d NA conclusion pass pass pass pass NA pass pass pass pass pass pass bias test e fail fail pass pass NA pass pass pass pass pass fail Sorbent Trap Method ST f average NA RA (%) NA dj NA cc NA conclusion fail pass failed pass NA pass pass pass pass pass pass bias test pass pass NA fail pass pass pass pass pass d g CEMS NA a Units of µg/dscm. b Relative accuracy ) ([difference arithmetic mean] + [confidence coefficient])/rm arithmetic mean 100. c Difference arithmetic mean. d Confidence coefficient. e Difference arithmetic mean should be less than the confidence coefficient to pass the bias test. f Average of Hg monitoring results from the sorbent trap method. g CEMS average - ST average. on the second attempt after the system was recalibrated using ionic Hg gases. The system response time was determined by measuring the time required for an evaluated system to respond 95% of the stable value after the calibration gas was changed from a low level to a high level. The results can be seen in Figure 2. The response time for the CEM A2 and B2 was determined to be 5 and 4 min, respectively. Also shown in the figure is the time required for each Hg CEM system to give a stable response at each testing concentration level. It was found that the A2 system responded to the elemental Hg gas approximately 2 times faster than to the Hg 2+ vapor. In the case of B2, no significant difference was found. The dynamic spiking (DS) test was performed on the tested CEM systems after the systems passed both system calibration error checks. The spiking Hg gas was delivered into the sampling loop at a flow rate approximately 1 / 10 of the loop flow rate. Two spiking levels were required. At the high-level spike, the Hg concentration in the spiking ionic Hg vapor was calculated so that the Hg concentrations in the spiked flue gases

7 EValuation of Mercury Emission Monitoring Systems Energy & Fuels, Vol. xxx, No. xx, XXXX G Table 7. Summaries of the Results Obtained from 3-pt Hg 0 and Hg 2+ System Calibration Error Checks certification concentrationa system responseb a y has to be within % to meet spike recovery criteria. were times higher than the native concentrations. During the low-level spike, the Hg concentration in the flue gas was elevated to a level that was times higher than the native concentration. Figure 3 shows the temporal trends of the responses from the A2 and B2 systems during the dynamic spiking tests. As absolute difference a - b calibration error [ a - b ]/CS 100 conclusion T complete (min) calibration gas level 3-pt Hg 0 System Calibration Error Check low pass 5 A2 mid pass 5 high pass 5 low pass 3 B2 mid pass 3 high pass 3 A2 B2 target level high low high low 3-pt Hg 2+ System Calibration Error Check low pass 10 mid pass 10 high pass 10 low pass 4 mid pass 3 high pass 3 Q probe (lpm) Table 8. Results Obtained from Dynamic Spiking Tests Performed on A2 and B2 Q spike (lpm) actual C spike value (µg/dscm) Figure 3. Temporal responses of Hg CEM systems on the 3-pt Hg 0 and Hg 2+ error checks. expected C ss (µg/dscm) actual C ss C native (mg/m 3 ) (µg/dscm) pre post average percent spike recoveryy a CEM A CEM B shown in the figure, the time required for a stable response by each spike was after about 10 min for A2 and 5-7 min in the case of B2. Results obtained from the test are summarized in Table 8. In the table, the C ss and C native values were the averages of the three readings after the responses of the systems were stable. As shown in the table, the spike recoveries of the A2 system ranged from 69 to 76%. In the case of the B2 system, it was approximately 120% at the high span and 104% at the low span. Neither test system passed the dynamic spiking test based on the resulting recoveries, which all exceeded the 5% criterion with the exception of the low span at the B2 system. It is suspected that the observed low recoveries for the A2 system were due to an incorrect dilution factor, which is the ratio of sampling loop flow rate to the flow rate of the spiking gas. The loop flow rate was one of the probe operational parameters calculated on the basis of the temperature and pressure differential across a venture. Unlike the B2 system, the venture of the A2 system was not calibrated before conducting the dynamic spiking experiment. It is necessary to consider the effect of the humidity on the flow rate while performing venture calibration. For the B2 system, the higher than target response was likely due to analyzer drifting. It was found that the B2 system tended to respond high at a higher concentration level even after calibration. Because the oxidized Hg vapor was prepared from a M HgCl 2 solution, the moisture content of the Hg vapor increased as the spiking concentration increased. A series of experiments were conducted to evaluate the effect of moisture content in the calibration gas generated by a HovaCal system on the response of the B2 system. The system was first calibrated by performing filter zero to correct Hg(T)

8 H Energy & Fuels, Vol. xxx, No. xx, XXXX Cheng et al. Figure 4. Temporal responses of Hg CEM systems during dynamic spiking. background and filter span to adjust the dilution factor using its own calibration unit. After calibration, three levels of oxidized Hg vapor were prepared from HgCl 2 solutions with 4 HgCl 2 concentrations, i.e., , , , and M. The moisture content in the generated Hg vapor ranged from 2.9 to 36.04%. Figure 6shows the study results. As can be seen, the response of the test CEM system increased as the moisture content in the calibration gas increased. Substitutes for Hg in Stratification Tests. To test for Hg stratification, we must use the CEM system to measure the Hg concentration at the 12 or fewer traverse points located as specified in the EPA Method. However, the current probe design does not allow easy movement of the sampling probe 12 times within the required time. Therefore, a stratification test was carried out in the first field test to investigate the potential of using other gas components, e.g., SO 2,NO x,o 2, and CO 2,or CO, as a substitute. The Hg concentration was measured by a CEM system provided by PS Analytical (PSA) (Deerfield Beech, FL). Other gas components (i.e., NO x, O 2, CO 2, SO 2, and CO) were measured by an integrated continuous, multi-emission monitoring system provided by Teledyne Instruments (San Diego, CA). One 15 ft Apex Instrument Method 5 probe setup was used to collect flue gas from the stack for both Hg and gas analyses. Flue gas was transported to a wet-chemistry speciation module immediately after being extracted from the stack using a 2 ft insulated Teflon tube. The speciation modules sit on a fourwheel cart, which travels along with the probe. A 100 m heated umbilical line (130 C) was used to transport gas samples to the analyzer after the speciation module. Figure 6. Effect of moisture content in oxidized mercury vapor on the response of the Hg CEM system. Table 9. Correlation Coefficients of Hg(T) as a Function of Other Flue Gas Components a SO 2 CO 2 NO x O 2 CO Hg concentration a Unit of µg/dscm. Figure 4 shows the correlation between Hg(T) and five other gas components (i.e., SO 2,NO x,co 2,O 2, and CO) in the stack gas. Each data point represents the averages of the readings collected from one traverse point. Readings were collected at least 20 min before the move to the next traverse point. The correlation coefficients are summarized in Table 9. As can be seen, there is no significant correlation between Hg and other gas components. The highest correlation coefficient was found to be 0.72 for CO 2 followed by O 2 and SO 2 (0.69 and 0.63, respectively). None of the concentrations of the selected flue gas components were found to be well-correlated (with a correlation coefficient greater than 0.8) with the Hg concentration in flue gas (Figure 5). Sorbent Trap for Hg Monitoring. Using a sorbent trap method as an alternative to the Hg CEMS was also evaluated. Both short (1-2 h) and long (5-16 days) term studies were carried out. In the short-term study, paired sorbent trap sampling was simultaneously conducted with the OH method sampling. Comparisons of Hg monitoring results from short-term sorbent trap sampling to the OH method and CEM system can be seen in Table 6. Except for the FS-1 study, the sorbent trap method passed all of the RA tests. In the 11 RA tests, the averages of sorbent trap sampling results were all greater than the results from the OH method, except the second series from the FS-3 study. The average of 12 1-h runs sorbent trap sampling Figure 5. Correlation between the concentrations of Hg and other flue gas components.

9 EValuation of Mercury Emission Monitoring Systems Energy & Fuels, Vol. xxx, No. xx, XXXX I Table 10. MDL of RP-M324 Hg Analyzer run number IR (ng) standard deviation, S 1.12 MDL ) 3.143S 3.5 MDL ) t(n - 1, 1 -R)0.99) (S) was 0.04 µg/dscm, which is much less than the OH method (0.1 µg/dscm) and Hg CEM results (0.34 µg/dscm). It was likely due to a low Hg concentration level in the flue gas. Short sampling duration resulted in the amount of Hg adsorbed in the sorbent trap being very close to the analytical detection limit. The detection limit of the sorbent trap method during the RA tests can be estimated by conducting the method detection limit (MDL) measurement for the RP-M324 (Ohio Lumex, OH) analyzer, which was used for sorbent trap Hg analysis in the tests. MDL is defined as the minimum concentration of a substance that can be measured and reported with 99% confidence that the Hg concentration is greater than zero and is determined from analysis of a sample in a given matrix containing Hg. To calculate MDL, the RP-M324 analyzer measured seven prespiked sorbent traps, which all had an Hg concentration level 5 times higher than the estimated MDL. The estimated MDL, which was 4 ng in the test, was 5 times that of the instrument noise. The instrument MDL was then determined by multiplying the standard deviation of the seven measurement results by a t test value, which was 3.141, with a sample size of seven. The MDL of the analyzer was determined to be 4 ng (Table 10). For a given 1 h sorbent trap sampling run, approximately m 3 of dry flue gas was collected. As the result, the detection limit of the sorbent trap method for a1hrunwas0.2 µg/dscm, which was higher than the results obtained from the OH method. Long-term (5-16 days) sampling was carried out to evaluate the operational feasibility of using the sorbent trap method. Results from the sorbent trap method were compared to the averages of the Hg data from Hg CEM systems collected during the same sorbent trap sampling period. Detailed sampling duration and results can be seen in Table 11. For a given sampling event, two sorbent traps were installed at the tip of the sampling probe. Two types of sampling devices [i.e., regularand inertial-type probe (modified by ICSET)] were used to investigate the effect of moisture and fly ash on sampling operations. As shown in Table 11, the results from the sorbent trap were consistently about 1 order of magnitude lower than the CEM average readings in the FS-2 study. Results from the sorbent trap method seem to be more reliable compared to the results from the Hg CEM system. In a 7 day sampling, the sorbent trap collected approximately 4 dscm of stack gas. More than 100 ng of Hg was collected in a sorbent trap after each sampling, which is well above the detection limit of the RP-M234 analyzer. In the case of the Hg CEM B1, after dilution, the Hg concentration in the sample gas delivered to the analyzer was lower than 20 ng/dscm, which is very close to the background noise of the instrument (about 4 ng/dscm). Using a regular probe in FS-3, it was found that the sorbent trap results were consistently higher than the Hg CEM results. Unlike the FS-2 study, deposition of ash at the sorbent tip was found. The deposition caused high vacuum during sampling and interrupted the sampling several times. To eliminate the potential ash effect, an inertial-type probe was used in the FS-4. The sorbent trap results were found to be either slightly less (<0.3 µg/dscm) or slightly higher (<0.02 µg/dscm) than the Hg CEM results. No deposition or operation difficulty was found. A side-by-side comparison of the results from regular- and inertial-type sampling setups was carried out at the FS-5 study. No significant difference was found between the results from the sorbent trap and the A1 Hg CEM system, with the exception of the first series. One of the two sorbent traps showed a much higher value, resulting in a higher relative deviation (>21%). In the other two series, the sorbent trap results were greater than the results from the A1 Hg CEM system (<0.32 µg/dscm). No ash deposition was found in any of the traps after a weeklong sampling. The relative deviation (RD, %) was calculated by dividing the absolute value of the difference between the results from Table 11. Comparison of Long-Term Hg Monitoring Results from the Sorbent Trap Method and Hg CEM Systems field study 2 a,b 3 a,b 4 a,c Hg CEM unit B1 A1 B1 series sampling duration (days) sorbent trap RD d (%) 3.5 (pass) 0.3 (pass) 5.0 (pass) 1.2 (pass) 5.0 (pass) 1.1 (pass) 5.8 (pass) 12 (pass) 6.6 (pass) 11.9 (pass) 35.9 (pass e ) spike recovery 6 (%) 48.4 (fail) 4.0 (fail) 19.2 (fail) 93.7 (pass) 93 (pass) 86.8 (pass) 93.7 (pass) 84.8 (fail) NA f NA f NA f CEMS data g d h CEMS field study Hg CEM unit A1 series sampling duration (days) CEMS data e sampling device regular inertial probe regular inertial probe regular inertial probe sorbent trap 0.28 (pass e ) 1.85 (pass) 0.53 (pass) 0.67 (pass) 0.51 (pass) 0.65 (pass) RD d (%) spike recovery (%) 86.8 (pass) 87.3 (pass) 98.7 (pass) 97.9 (pass) 96.0 (pass) (pass) d f CEMS a Unit of µg/dscm. b Regular sorbent trap probe was used for sampling. c Ash-free inertial probe was used. d RD (%) ) [( C trap 1 - C trap 2 )/(C trap 1 + C trap 2)] 100%. e The difference between the results from OH and sorbent trap methods was less than 1.0 µg/dscm. 6 Spike recovery ) [C measured/ C spiked] 100%. f A spike test was not performed. g Averages of monitoring results during sorbent trap sampling. h CEMS average - ST average. 5 a-c

10 J Energy & Fuels, Vol. xxx, No. xx, XXXX Cheng et al. the two traps by the sum of the results from the two traps. On the basis of the criteria, the RD should be less than 10% to validate the sampling results. As can be seen in Table 10, in a total of 14 sampling runs, 5 did not pass the criteria. Another data QA/QC parameter specified in the Appendix K procedures is spike recovery. For field sample analyses, the recovery is required to be within 25% of the spiked concentration. It was found that the recovery was far below the satisfactory criteria with the exception of the FS-4 study. No spike test was performed. Conclusions Results from a series of calibration gas comparison tests showed that, with the same tag values, the elemental Hg gases provided by the calibration units of the Hg CEM systems from manufacturer A were approximately 20% higher than the elemental Hg gases from the calibration units of manufacturer B. The discrepancy of calibration gases resulted in the disagreements observed between the Hg monitoring results from the Hg CEM systems of the two leading manufacturers when both measured the same Hg source. The observations from the FS-1 and FS-2 studies indicate that the evaluated Hg systems provided by the two manufacturers were still in the development stage during the testing period. Many parts had not yet been standardized or well-qualitycontrolled, which resulted in many part replacements and system modifications. After replacing broken parts and performing needed services, the Hg CEM systems provided by the two manufacturers were able to be operated without serious maintenance (FS-3, FS-4, and FS-5). The most challenging procedure in certifying an Hg CEM system was conducting a system integrity check. During the five field studies, a HovaCal system was used to provide oxidized Hg vapors. The oxidized Hg gas setup was not available for the tested Hg CEM systems provided by the two manufacturers. Because the HovaCal system was not integrated into the testing systems, extra cautions were taken while performing the test. For example, the whole oxidized Hg vapor line was kept above 180 C to prevent the formation of cold spots. Glass-coated connection unions were used for gas line connections. The concentrations of the HgCl 2 solutions used for producing oxidized Hg vapors were verified by cold vapor atomic adsorption spectroscopy before use. The delivering Hg concentrations were corrected for temperature before being compared to the CEM responses. In summary, the Hg CEM system from manufacturer A needed more attention and supplies to carry out the monitoring task (e.g., supplies of ultra high purity argon and DI water, peristaltic pump tubings, cleaning and replacement of gold traps, and frequent lamp voltage adjustments). In the case of the Hg CEM system of manufacturer B, the analyzer encountered a noticeable drifting problem. It also had a less stable calibration unit and user-friendly interface to operate the system. When using a Hg CEM system as an instrumental reference method for a RA test, dynamic spiking was the most crucial and difficult step to certify the system. Factors such as dilution factor measurement, variation of native concentration, loop flow measurement, and stability of the instrument can greatly affect the recovery of the dynamic spiking. The criteria of passing the IRM depend upon the selection of the calibration span, which depends upon the stack gas Hg concentration (i.e., the native concentration). The higher the calibration span, the larger the error allowed. In the case of the FS-1 study, concentrations up to 40 µg/n m 3 can be selected as the calibration span, which allows 0.8 and 2 µg/dscm of error for the 3-pt system Hg 0 calibration error and 3-pt system Hg 2+ calibration error checks, respectively. For utilities in the other four field studies, which all had native concentrations less than 5 µg/n m 3, the differences between the reference and measured values were be less than 0.2 or 0.5 µg/n m 3 for the 3-pt system Hg 0 calibration error and 3-pt system Hg 2+ calibration error checks, respectively. Using SO 2, CO, CO 2,orNO x as the substitute for in the Hg stratification test was not feasible. There was no strong correlation between the concentrations of Hg and the other four selected gas components in the duct. Results from sorbent trap studies showed promise for using this technique as an alternative to the Hg CEM for coal combustion utilities. However, deposition of ash at the tip of the sorbent traps was found in the FS-3 study, which interfered with the sampling operation and possibly resulted in the bias results. Further systematic investigation is desired to further explore the feasibility of using this sorbent trap method with respect to different sample digestion and analytical methods, sampling devices, long-term operations, and comparison of longterm monitoring results with Hg CEM. Acknowledgment. This report was prepared by Chin-Min Cheng and ICSET of Western Kentucky University with support, in part, by grants made possible by the Illinois Department of Commerce and Economic Opportunity through the Office of Coal Development and the Illinois Clean Coal Institute. Neither Chin-Min Chen and ICSET of Western Kentucky University, nor any of its subcontractors, nor the Illinois Department of Commerce and Economics Opportunity, Office of Coal Development, the Illinois Clen Coal Institute, nor any person acting on behalf of either (A) makes any warranty of representation, express or implied, with respect to the accuracy, completeness, or usefulness of the information contained in this report, or that the use of any information, apparatus, methods, or process desclosed in this report may not infringe privately owned rights; or (B) assumes any liabilities with respect to the use of, or for damages resulting from the use of, any information, apparatus, method or process disclosed in this report. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not nesessarily constitute or imply its endorsement, recommendation, or favoring; nor do the views and opinions of authors expressed herein necessarily state or reflect those of the Illinois Department of Commerce and Economic Opportunity, Office of Coal Development, or the Illonois Clean Coal Institute. The authors acknowledge valued assistance from Mr. Martin Cohron during the field studies. The authors thank Spectrum Gases for providing Hg calibration gas cylinders. EF