Abstract

Since installing continuous emissions monitoring systems (CEMS) as required by the Acid Rain Rule (40 CFR Part 75), many utilities have noted that the CEMS are recording consistently higher heat input and SO₂ emissions than conventional methods (input/output and output loss). The apparent CEMS bias is causing utilities to report more heat input and SO₂ and CO₂ emissions than are believed to be justified. Many believe that the major problem is Method 2, the Environmental Protection Agency (EPA) standard measurement method for stack volumetric flow rate. It has been clearly shown by previous work in this project that Method 2 will be biased high in the presence of "swirling stack flow" and, since all of the stack volumetric flow monitors are calibrated to Method 2 measurements, a high bias in Method 2 will be directly transferred to the flow monitors. In addition to the potential high bias in stack flow measurement, any high bias in SO₂ or CO₂ stack gas concentration measurement will also bias the overall mass (lb/hr) emission rate.

In order to better understand the problem, the Electric Power Research Institute (EPRI) has initiated a project to identify the cause(s) for the high heat input and SO₂/CO₂ emissions measurements. This project has three basic objectives: (1) to understand the potential error in Method 2 under stack flow conditions by the review of existing literature and data, (2) to demonstrate the validity of the literature assessments by conducting flow measurements in a specially designed "swirl tunnel," and, (3) to verify the flow and heat input errors, and identify the cause(s) in full scale field tests.

This paper describes the results of the full scale field measurements portion of this project. The field tests were conducted under tightly controlled conditions so that the error sources can be identified and quantified. Almost every possible source of heat rate discrepancy was simultaneously evaluated. Flue gas flow measurements were made using different 2-D and 3-D pitots. Independent gas concentration measurements were made. Unit heat rate was determined using the conventional input/output method. Multiple fuel sampling and analysis approaches were used. The results of this simultaneous, multiple methodology approach helped shed light on the sources of heat input error.

Introduction

Virtually all electric utility power plants were required to install CEMS as a result of the Acid Rain Program mandated by the Clean Air Act of 1990. Monitors were required for SO₂, NO_X, CO₂ and stack volumetric flow rate. In addition to measuring emissions, CEMS have been used (using fuel F-factors, CO₂ concentration and the volumetric flow rate) to obtain boiler heat inputs and, subsequently, unit heat rates. Since the installation of the CEMS, many plants have found that the heat input/rate as determined by the CEMS was higher (by 5-25%) than determined by conventional heat rate methods (input/output or output loss). This discrepancy is disconcerting since all of the methods should give equivalent results. The heat input value from the CEMS was immediately suspect since there is more than a 50-year history with the conventional methods and, in many cases, the heat input from the CEMS was simply thermodynamically improbable. The individual component of the heat input measurement that was suspected to be the major cause of the problem was the volumetric flow measurement. The flow measurement instruments were new and unproven while the CO₂ instruments and F-factors had been used for a number of years.

If, in fact, the flow measurement was in error, the problem became more than disconcerting; it became a matter of money. The flow measurement is a fundamental component of the SO₂ tonnage emission calculation and, if high, results in excess SO₂ allowances being used. A SO₂ allowance permits a utility to emit one ton of SO₂ and the allowances can be bought and sold (~ $75 per ton) on the open market. Therefore, using excess allowances can have a multimillion dollar impact on a large utility. In addition, many utility boilers have operating permits that contain heat input and SO₂ tonnage limits and, in the past, compliance with these limits has been demonstrated with fuel analysis. Some state agencies and EPA regional offices have begun using the CEMS data to evaluate compliance with the permit limits and have started applying pressure on utilities that are showing "excess" emissions and heat input.

The problem of high heat input was so pervasive throughout the utility industry and the costs of excess allowances were so great that the Electric Power Research Institute (EPRI) initiated a project with RMB Consulting & Research, Inc. to better understand the high heat input measurement problem. The objectives of the project are; (1) to identify the source(s) of the heat input discrepancy, (2) to quantify the errors, and (3) to suggest ways to reduce the error. In order to accomplish these objectives a three-task project was developed. Since flow measurement errors were suspected as the primary problem, the first task was to conduct a literature/technology survey on flow measurement methods and potential errors. This work has been completed and a report has been published (EPRI TR-106698, available to EPRI members only). The results of Task 1 were also reported in a paper presented at the May 1996 EPRI CEM Users Group Meeting in Kansas City. Task 2 was to construct a "swirl tunnel" and to test various pitot configurations under controlled yaw swirl conditions. This work has been completed and is reported in another paper at this conference. Task 3 was to confirm the results of the swirl tunnel work and to define and quantify error sources in a series of tightly controlled field tests at power plants that were experiencing CEMS heat input measurement errors. This paper reports those field test results.

Description of Units Tested

Field testing was performed at two large coal-fired units. The first series of tests was performed at Wisconsin Power & Light's Columbia Unit 2. Columbia Unit 2 is a conventional 560 MW (gross) unit that burns Wyoming low-sulfur, sub-bituminous coal. The CEMS is a typical dilution extractive system equipped with an ultrasonic flow monitor. The CEMS and test ports are located in the stack approximately 2.5 diameters up from the entrance of two opposed entry ducts. The stack diameter at the test location is 21 feet. The unit is equipped with gravimetric coal feeders/scales using the latest load cell-based technology. It also has an on-line heat rate monitoring system.

The second series of tests was performed at Cooperative Power's Coal Creek Unit 2. Coal Creek Unit 2 is a 560 MW (gross) that burns lignite from an on-site mine (mine-mouth). The unit is equipped with a lime-based SO₂ scrubber. Lignite is supplied to the unit by six gravimetric feeders/scales which were a mix of load cell-based and an older technology. The CEMS is a typical dilution extractive system equipped with an ultrasonic flow monitor. The CEMS and test ports are located in the stack approximately 7.7 diameters up from the single entry duct. The stack diameter at the test location is 25.6 feet. This unit is also equipped with an on-line heat rate monitoring system.

Columbia Unit 2 Field Test

Five tests were conducted at Columbia Unit 2. These tests were based on preliminary findings on Unit 2 that revealed flow with moderate yaw and pitch components. Also, since the yaw components varied from port to port, four additional test ports were added at a 45-degree offset from the original ports. The five tests included runs designed to compare the results of S-type and 3-D pitot measurements and to evaluate the effect of test port variation. During each test CEMS, coal flow and reference method measurements were taken. Tests 1 and 2 were conducted with the unit operating at about 520 MW (gross); Tests 3-5 were conducted with the unit operating at about 550 MW (gross).

• Test 1:  S-type/3-D comparison, No Test Port Variation
• Test 2:  S-type Only, No Test Port Variation
• Test 3:  S-type/3-D comparison w/Test Port Variation (Full Load)
• Test 4:  S-type/3-D comparison, No Test Port Variation (Full Load)
• Test 5:  S-type/3-D comparison w/Test Port Variation (Full Load)

Figure 1 and Table 1 below show comparisons of stack flow data for the tests performed at full load. The S-type pitot measurements averaged 4.1% higher (2.9%-5.2%) than the 3-D probe. This small difference demonstrated the effect of the slight yaw swirl on the S-type pitot.

Table 2 below also shows another effect of the swirl variation from port-to-port on the S-type pitot. As the effective yaw angle increases, the bias seen in the S-type measurements (as compared with the 3-D probe values) increases. [It should also be noted that the CEMS flowmeter measured slightly lower than the 3-D probe and considerably lower than the S-type pitot. The reason for this is unclear.]

To illustrate the bias effects in terms of unit heat rate, the results of the full load tests (Tests 3-5) are shown below in Figure 2.

For consistency, only flow traverse values taken from the original test ports are included. Except as indicated by the 3-D corrected bar, the data in Figure 2 are uncorrected and based on manual Delta P readings. The standard Fc-factor of 1800 for subbituminous coal was used and no corrections were made for wall effects. The unit heat rate based on S-type pitot flow measurements and reference method CO₂ values was an average of 10.8% higher and the heat rate based on CEMS data was an average of 9.8% higher than the heat rate calculated using the input/output method. The uncorrected heat rate based on 3-D probe flow measurements and reference method CO₂ values was an average of 7.1% higher than the input/output method. Figure 2 also shows that, with the proper corrections, the 3-D-based heat rate agrees well (within 1.9%) with the input/output method. (Discussions of the corrections are included in a following subsection -- See Heat Rate Bias Components)

Curves depicting the yaw/pitch variation measured during the Columbia tests are shown below in Figures 3 and 4.

The "d" traverse points are closest to the wall and the "a" points are closest to the stack center. Data for two different 3-D probes are shown. The DAT probe is a commercially available probe that is widely used by stack testing firms. The MS5 probe is a spherical probe that was custom fabricated for this test program. The MS5 probe has the advantage of significantly higher Delta P output relative to the DAT probe. While there was some variability in the angles measured from test-to-test and fairly significant differences from port-to-port, the pitch and yaw patterns remained relatively consistent during the tests. This is not to suggest that these "swirl" patterns are consistent over longer time periods.

Coal Creek Unit 2 Field Test

Seven tests were conducted at Coal Creek Unit 2. During each test run, separate but simultaneous measurements were taken by test teams from FERCo, RMB's stack test subcontractor and Climax, EPA's stack test subcontractor. (While many measurements were duplicated, not all measurements for all tests were performed by both test teams given the physical limitations of the test site.) All tests were conducted with the unit operating at full load, 560-565 MW (gross). During each test CEMS, coal flow and reference method measurements were taken.

• Test 1:  FERCo S-type and 3-D; Climax S-type and 3-D
• Test 2:  FERCo 3-D; Climax S-type
• Test 3:  FERCo S-type and 3-D; Climax S-type and 3-D
• Test 4:  FERCo 3-D; Climax S-type and 3-D
• Test 5:  FERCo S-type and 3-D; Climax S-type and 3-D
• Test 6:  FERCo 3-D; Climax S-type and 3-D (w/near wall measurements)
• Test 7:  FERCo 3-D; Climax S-type and 3-D

For simplification, various aspects of the tests are included in the preceding list. A "French" probe was also tested as a potential alternative to the S-type pitot and, while favorable, those results are not included in this paper. Yaw nulling techniques were also investigated but the results are not included herein.

Table 3 below shows a comparison of simultaneous stack flow measurements recorded by FERCo and Climax.

The measurements showed excellent agreement with an average S-type flow value disagreement of less than 1.2% and an average 3-D measurement disagreement of less than 0.8%. (Since such good agreement was found, the average of any duplicate measurements taken by FERCo and Climax was used in any subsequent analysis contained in this paper.)

Figure 5 and Table 4 below show a comparison of simultaneous stack flow measurements recorded during the seven tests. The S-type pitot measurements were an average of 15.0% higher (12.3%-19.9%) than the 3-D probe. This large difference demonstrated the effect of the significant amount of swirl (the average yaw angle was ~ 21 ) on the S-type. The CEMS flowmeter reported values an average of 6.6% higher (2.4%-9.9%) than the 3-D probe.

To illustrate the bias effects in terms of heat rate, the results of the full load tests are shown in Figure 6 below. The data in Figure 6 are uncorrected and based on manual readings. Except as indicated by the 3-D corrected bar, the standard Fc-factor of 1910 for lignite was used and no corrections were made for wall effects.

The heat rates based on S-type pitot flow measurements and reference method CO₂ values were an average of 23.7% higher than heat rate values calculated using the input/output method and values based on CEMS data were an average of 18.6% higher. The heat rates based on uncorrected 3-D probe flow measurements and reference method CO₂ values were an average of 8.1% higher. Figure 6 also shows that, with proper corrections, the 3-D-based heat rate agrees well (within 3.4%) with the input/output method.

Curves depicting the yaw/pitch variation measured during the Coal Creek tests are shown below in Figures 7 and 8. While there was some variability in the angles measured from test-to-test and fairly significant difference from port-to-port, the pitch and yaw patterns remained relatively consistent during the tests. This is not to suggest that these "swirl" patterns are consistent over long time periods.

Heat Rate Bias Components

Figures 9 and 10 below illustrate the discrepancies observed between heat rates based on 3-D measurements and the input/output method at Columbia and Coal Creek. As the figures show, once a few biases were corrected, there was excellent agreement between the two measurements. The corrected 3-D-based heat rate agreed within 1.9% of the input/output-based heat rate at Columbia and within 3.4% at Coal Creek.

Corrected biases to the 3-D-based heat rate values included:

• Fc-factor. At Columbia, the Fc-factor calculated based on the average coal percent as-fired carbon and average gross calorific value as determined from as-fired samples collected during the tests was 1839 scf CO₂/mmBtu. Using the standard Fc-factor of 1800 would result in an overestimation of the heat rate by approximately 2.2%. At Coal Creek, the Fc-factor determined from the as-fired coal analysis was 1942 scf CO₂/mmBtu. Using the standard Fc-factor of 1910 for lignite would result in an overestimation of the heat rate by approximately 1.6%.

• Wall effects. The S-type and the 3-D flow values were based on equal area traverses which do not take into account the fact that the stack velocity goes to zero at the stack wall. To account for this effect, near wall measurements were taken at Coal Creek. Based on numerical integration, not taking into account the wall effects introduces a bias of 1.9%. Similar effects were also seen in the "swirl" tunnel tests. (Although near wall flow measurements were not taken at Columbia, it is assumed that the wall effect at Columbia is approximately equal to that seen at Coal Creek based on the similarity of the stacks.)

• Manual pressure reading bias. During the field tests, pitot Delta P readings were made both manually using calibrated magnehelics and automatically using a data logger equipped with precision pressure transducers. Subsequent analysis revealed a consistent bias in the manual readings when compared to the readings collected automatically using the data logger. This bias appears to be related to a tendency of individuals to overestimate when doing "eyeball averaging" of fluctuating readings. At Columbia, the manual readings resulted in flow values 1.1% higher than those based on the automatic readings. At Coal Creek, the manual readings resulted in flow values 0.9% higher than those based on the automatic readings.

With S-type pitot measurements, in addition to the bias introduced by the Fc-factor, wall effects and manual pressure readings, bias is also introduced by non-axial flow:

• Non-axial flow. The difference between the S-type and the 3-D flow values is related to the non-axial components of the flow that are erroneously included in the velocity head of the S-type measurement. At Columbia where only small non-axial flow components were found, the S-type pitot yielded full-load flow values that were 4.1% higher than the 3-D measurements. Thus, the S-type measurements, and subsequently the CEMS flowmeter data, were biased 4.1% high due to non-axial flow conditions. At Coal Creek where significant non-axial flow components were found a 15.0% high bias due to non-axial flow conditions was seen.

Since CEMS flowmeters are calibrated and certified using S-type pitot reference method flow measurements, any bias in the reference method would be passed on to the certified flowmeter. These "calibration bias" effects include wall effects, non-axial flow effects and manual pressure reading bias. In addition to "calibration bias" and Fc-factor bias, any bias in the CEMS CO₂ measurement would also be transferred to the CEMS-based heat rate value:

• CO₂ Discrepancies. Tables 5 and 6 below show comparisons of average CEMS and reference method CO₂ values for Columbia and Coal Creek, respectively. A consistent bias is seen in the CEMS values when compared with the reference method.

At Columbia, the CEMS values were an average of 3.6% higher than the reference method values. At Coal Creek, the CEMS values were an average of 3.3% higher than the reference method values. Any bias in the CEMS CO₂ measurements would result in a corresponding bias in the CEMS-based heat rate. While some of the CEMS CO₂ error may be attributed to calibration drift, we believe there may be a fundamental measurement difference relative to the reference method. Small errors in CO₂ are not generally considered to be significant; however, at a nominal 10,000 Btu/kWh unit heat rate an absolute 0.1% CO₂ error is equivalent to about 100 Btu/kWh error in the heat rate. Further study of CO₂ error sources may be desirable.

Conclusions

This paper has evaluated the results of carefully controlled stack tests at two sites to determine unit heat rate in comparison with conventional input/output heat rate methodology. Closure within 1.9% was achieved at one site and within 3.4% at the other. Error sources impacting CEMS heat input/heat rate measurements were identified and quantified.

It is clear that the EPA Reference Method 2 is biased high in the presence of yaw swirl in the stack. The amount of bias is related to the amount of yaw swirl--the greater the swirl, the higher the bias. This results in a high bias in stack volumetric flow monitors (and all emissions calculated using the flow monitors) because all of the monitors are presently "calibrated" to Method 2. There is also a positive bias from the stack wall effect that is approximately 2%. EPA should allow the use of 3-D pitots to eliminate the yaw bias and should also allow a wall effect correction.

At the two sites tested there was also a small bias (average 1.9%) in the fuel Fc-factor and the present Acid Rain rules allow for this correction. "Eyeball" manual readings of Delta P also appear to have a slight (~ 1%) positive bias relative to computerized instrumental readings.

Both field test sites also showed a positive CO₂ CEM bias that was partially due to "within specification" calibration drift; however, there was some indication of an inherent positive bias. This apparent problem should be investigated in more detail.

It is clear from these field tests that well controlled stack tests, using precise test methods, can be made to produce heat rate measurements that agree (within experimental error) with conventional heat rate methodology. Without equivalent EPA Reference Method accuracy and precision, as well as appropriate real, physical corrections, many utilities will continue to report inaccurate, high-biased emissions under the Acid Rain Program.

Agora Environmental Consulting
Last Revised: February 25, 2008