Mercury Removal Review in Utility Working Group

1. Utility MACT Working Group Negative Removals -- A Review Bill Maxwell, OAQPS/ESD

2. Purpose • To present a review of test reports from units showing negative mercury removals during ICR speciated mercury emissions tests • Not to definitively answer all questions

3. Background - All tests • Negative mercury removals were indicated at 31 test sites when calculated either from coal-to-stack or across the tested control device • 12 sites showed negative removals only when calculated from coal-to-stack • Because of differences in test methodologies, for this analysis, this is not a calculation that we feel should be used

4. General theories -- coal to stack • Representativeness of sample • For some sites, coal analyses during testing did not match similar ICR analyses for surrounding period • Method and location of sample acquisition different than that for ICR samples • Timing of coal sampling vs. when that flue gas gets to stack sample train • Does not negate usefulness of coal-to-stack analyses overall

5. Background -- Control device only • 19 remaining test sites indicated negative removals across the tested control device • Reviewed each test report in greater detail • Cannot definitively say “here is the reason” for any site • Have plausible theories for most sites that are felt to reasonably account for, or at least contribute strongly to, the negative removals • Some sites fall under more than one theory

6. General theories - control device • Coal type • Low level of mercury present • Normal sampling distribution around “0” • Use of hot-side ESP • Sampling locations • Test/analytical method problems

7. Coal type • 15 of the 19 sites burned subbituminous or lignite coals • Generally have higher proportion of elemental mercury in flue gas • Not expected to be removed • Subject to loss during sample handling • Ontario-Hydro method has been modified to add stabilizer to permanganate solution that inhibits loss of elemental mercury

8. Low level of mercury present • Ontario-Hydro method validated at 2 ug/m3 • For units having emissions near this level (may vary somewhat between labs), small variations in numerical measurements lead to large variations in percentages • Indicates care must be taken in use of significant figures and in precision of any standard • Also creates problems with non-detect values

9. Normal sampling distribution around “0” • Normal variability in sampling would lead one to expect a distribution of results around any value • When that value is “0” -- normal distribution may indicate negative removals • Normal sampling/analytical variability in each sample exacerbates problem more at low removal levels than at higher removal levels • Would apply to coal-to-stack values also

10. Example for tests with +/- 30% error: Poor mercury control

11. Example for tests with +/- 30% error: Good mercury control

12. Use of hot-side ESP • 5 of 19 units utilized hot-side ESP controls • Not expected to exhibit effective mercury removal due to relatively high temperatures • Adds to problem of normal distribution of data around “0” removal

13. Sampling location • Several sites had multiple inlet and/or scrubber by-pass configurations • Multiple modules for control devices complicate inlet duct configurations and sampling locations • Inlet sites in general were less adequate than outlets • Scrubber by-pass to meet SO2 requirements • May be biasing the inlet data +/- • Indicates that “percent removal” format may not be appropriate if taken across control

14. Test/analytical method problems • Ontario-Hydro, at the time of testing, was a new test method with few testing or analytical firms familiar with its use • Is not an “easy” method and requires strict adherence to procedures, QA/QC, etc. • Some test contractors indicated problems with blank levels, lost samples (dropped impingers, etc.), etc. • Sample hold times sometimes exceeded

15. How the theories fit the units • Following slides fit tested sites to the theories noted earlier • Some sites fit more than one theory • Generally cannot say “Eureka”

16. Cliffside Unit 1 Clifty Creek Unit 6 Columbia Unit 1 Gibson Unit 3 (both sets of tests) Leland Olds Unit 2 Limestone Unit LIM1 Navajo Unit 3 Newton Unit 2 Rawhide Unit 101 Sherburne County Unit 3 Stanton Station Unit 10 Wyodak Unit BW91 Negative removal coal to stack only

17. Big Brown Unit 1 Cholla Unit 2 Cholla Unit 3 Clay Boswell Unit 4 Colstrip Unit 3 Coronado Unit U1B George Neal South Unit 4 GRDA Unit 2 Laramie River Unit 3 Lawrence Unit 4 Monticello Unit 1 Nelson Dewey Unit 1 Platte Unit 1 Sam Seymour Unit 3 Stanton Station Unit 1 Coal type

18. Bay Front Unit 5 . . . . . Cholla Unit 3. . . . . . . . Clay Boswell Unit 4 . . . Coronado Unit U1B . . . Laramie River Unit 3 . . Lawrence Unit 4 . . . . . Nelson Dewey Unit 1 . Presque Isle Unit 9 . . . Valley Unit 2. . . . . . . . 2.5 - 4 ug/m3 1 - 2 ug/m3 5 - 6 ug/m3 2.5 - 3 ug/m3 2.5 - 3.5 ug/m3 ~5 ug/m3 (+ non detects) 2 - 3 ug/m3 1 - 2 ug/m3 1 - 2 ug/m3 Low level of mercury present

19. GRDA Unit 2 . . . . . . . . . . . . . . . . . Lawrence Unit 4. . . . . . . . . . . . . . . Platte Unit 1 . . . . . . . . . . . . . . . . . Big Brown Unit 1 . . . . . . . . . . . . . . Cholla Unit 2 . . . . . . . . . . . . . . . . . Coronado Unit U1B. . . . . . . . . . . . . Nelson Dewey Unit 1. . . . . . . . . . . . Presque Isle Unit 9 . . . . . . . . . . . . . Stanton Station Unit 1. . . . . . . . . . . Valley Unit 2. . . . . . . . . . . . . . . . . . +18 to - 24% - 2 to - 34% +27 to - 33% + 1 to - 13% +23 to - 33% - 11 to - 21% + 2 to - 23% + 0 to - 6% + 2 to - 9% + 6 to - 25% Normal sampling distribution around “0”

20. Use of hot-side ESP • Cholla Unit 3 • Gaston Unit 1 • Nelson Dewey Unit 1 • Platte Unit 2 • Presque Isle Unit 9

21. Big Brown Unit 1 Clay Boswell Unit 4 Colstrip Unit 3 Coronado Unit U1B (scrubber by-pass) George Neal South Unit 4 Laramie River Unit 3 Lawrence Unit 4 Sam Seymour Unit 3 (also scrubber by-pass) Sampling locations

22. Bay Front Unit 5 . . . . . . . . Cholla Unit 2. . . . . . . . . . . Cholla Unit 3. . . . . . . . . . . Gaston Unit 1 . . . . . . . . . . George Neal South Unit 4. . Laramie River Unit 3 . . . . . Reported “suspect” elemental data for 2 of 3 of outlet runs Numerous problems Numerous problems Possible sample hold problem Possible bad test run Possible sample hold problem Probable bad test run Contractor/method problems

23. Conclusions • Negative removals are expected for some tests on units with poor mercury removal • Negative removals are primarily a combined function of • Actual low removal efficiencies • Inherent variability in test and analytical procedures • Other factors may also be involved but believed to be of lessor importance

24. Conclusions - More • Negative removals are not expected for tests on units with good mercury removal • Test data with good mercury removal used for setting emission standards • Additional testing would not impact final MACT levels

25. So where does this leave us... • Ontario-Hydro method is not inherently flawed • Nothing in the data to so indicate • Modifications have been made to stabilize elemental mercury in permanganate solution • Data do indicate caution in setting level of standard • Truly positive removal numbers are valid • Mercury is not being generated

26. Examples • Laramie River Unit 3 • Coronado Unit U1B

27. Laramie River Unit 3 • Uses subbituminous coal • Sample storage conditions (~20 °C) and holding times (> 45 days) may have been violated prior to analysis • Low mercury value • Multiple inlet ducts • One inlet run considerably different from 5 other runs at same or similar site

28. Laramie River Unit 3 -- cont. Laramie Units 1 and 3 are same type of boiler, burn the same coal, and in similar quantities

29. Coronado Unit U1B • Uses subbituminous coal • Low mercury value • Uses scrubber by-pass as part of SO2 control • Sampled inlet upstream of by-pass duct • Sampled outlet upstream of by-pass duct • Actual impact on percent removal calculations is unknown

30. List of Units Exhibiting Negative Removals Across Control or Coal-to-Stack