BART Analysis for the Navajo Generating Station Units 1-3

Transcription

1 Prepared for: Salt River Project Navajo Generating Station Tempe, AZ BART Analysis for the Navajo Generating Station Units 1-3 ENSR Corporation November 2007 Document No.:

2 Prepared for: Salt River Project Navajo Generating Station Tempe, AZ BART Analysis for the Navajo Generating Station Units 1-3 Prepared By: Olga Kostrova Reviewed By: Robert Paine and Ian Thomson ENSR Corporation November 2007 Document No.:

3 Contents Executive Summary...ES Introduction Source Description Reasonably Attributable BART Analysis BART Requirements Sulfur Dioxide Emission Control Particulate Matter Emission Control Nitrogen Oxides Emission Control Report Outline CALPUFF Modeling Procedures CALMET Processing CALPUFF Modeling Procedures Natural Conditions and Monthly f(rh) at Class I Areas Light Extinction and Haze Impact Calculations BART Applicability Analysis BART-Eligible Requirements Existing Control Equipment and Emission Rates Affected Class I Areas Baseline CALPUFF Modeling Results BART Engineering Analysis for NO x Emissions Identification of Alternative NO x Controls Low-NO x Burners and Separated Overfire Air (LNB/SOFA) Flue Gas Recirculation Selective Non-Catalytic Reduction Selective Catalytic Reduction ECOTUBE ROFA /ROTAMIX Technical Feasibility of Alternative NO x Controls Effectiveness of Technically Feasible NO x Controls Impacts of Alternative NO x Controls Economic Impacts Non-Air Quality Environmental Impacts i November 2007

4 4.4.3 Energy Impacts Remaining Useful Life BART Control Options Modeling Analysis Modeled Control Scenarios CALPUFF Modeling Results for Control Options Cost of BART Control Options BART Recommendations References List of Appendices Appendix A CALMET/CALPUFF Processing Refinements Appendix B Factors Influencing NOx Emissions Effects on Visibility Appendix C Review of Data from the IMPROVE Monitoring Network Appendix D CALPUFF Modeling Results Appendix E BART Cost Information List of Tables Table ES-1: Total Capital and Annual Costs Associated with LNB/SOFA, SCR and SNCR Applied to NGS Units 1 through 3...ES-4 Table ES-2: Annual Costs vs. Visibility Improvements/Degradation (Average of the Eight Class I Areas)... ES-5 Table ES-3: Selected Best Available Retrofit Technology Low NO x Burners / Separated Overfire Air...ES-6 Table 2-1: Annual Average Natural Background Concentrations Table 3-1: Navajo Generating Station Baseline Emissions Table 3-2: Navajo Generating Station Modeling Stack Parameters Table 3-3: Regional Haze Impacts Due to Baseline Emissions Table 4-1: Annual NO x Emission Rates Resulting from Application of Control Options to NGS Units 1 3 (lb/mmbtu) Table 4-2: Total Capital and Annual Costs Associated with LNB/SOFA, SCR and SNCR Applied to NGS Units 1 through Table 4-3: Summary of Non-Air Quality Environmental Impacts Table 4-4: Navajo Generating Station SOFA/LNB Emission Controls (Option 1) ii November 2007

5 Table 4-5: Navajo Generating Station SOFA/LNB Emission Controls (Alternative Option 1) Table 4-6: Navajo Generating Station SOFA/LNB/SNCR Emission Controls (Option 2) Table 4-7: Navajo Generating Station SOFA/LNB/SCR (Option 3) Table 4-8: Navajo Generating Station SOFA/LNB/SCR (Option 4) Table 5-1: Regional Haze Benefit of BART Controls Table 5-2: Regional Haze Results of BART Controls for Each Year and Class I Areas Table 5-3: Annual Costs vs. Visibility Improvements/Degradation (Average of the Eight Class I Areas) Table 6-1: Recommended Best Available Retrofit Technology Low NO x Burners / Separated Overfire Air List of Figures Figure ES-1 8 th Highest Regional Haze Deciview (Includes Background Haze and NGS BART Control Case Emissions)...ES-2 Figure 2-1: NGS CALPUFF Computational Grid in Relation to WRAP Arizona Domain Figure 3-1: Class I Areas within 300 km of the Navajo Generating Station Figure 3-2: 8 th Highest Regional Haze Impacts for Each Modeled Year Due to Baseline Emissions Figure 5-1: 8 th Highest Regional Haze Total Impacts Averaged Over 3 Years for Baseline and BART Control Case Emissions Figure 5-2: Annual Costs vs. Visibility Improvements/Degradation (Average of Eight Class I Areas) Figure 5-3: Annual Costs vs. Visibility Improvements/Degradations at Petrified Forest NP Figure 5-4: Annual Costs vs. Visibility Improvements/Degradations at Canyonlands NP iii November 2007

6 Executive Summary The Salt River Project Agricultural Improvement and Power District (SRP) operates the Navajo Generating Station (NGS), a coal-fired steam electric generating station located on the Navajo Nation, approximately 3 miles east of Page, Arizona. The NGS facility became operational in stages between 1974 and 1976 and consists of three coal-fired units with a combined net power generating capacity of approximately 2,250 MW. The facility is approximately 20 kilometers from the northern boundary of Grand Canyon National Park (GCNP). The Clean Air Act s Regional Haze Rule (RHR) contains a requirement for each State or Tribe to address the Best Available Retrofit Technology (BART) requirement when preparing the State s Regional Haze State Implementation Plan (SIP). Since the Navajo EPA does not have an approved Tribal Implementation Plan, the Environmental Protection Agency is currently responsible for implementing the RHR at NGS. This BART Analysis for the NGS was prepared pursuant to EPA s July 6, 2005 final RHR titled Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations; Final Rule ( BART Guidelines ). The BART Guidelines include presumptive BART requirements for coal-fired electric steam generating sources greater than 750 MW. The EPA has determined that the NGS is a BART-eligible source. Based on air dispersion modeling performed by ENSR, NGS is subject to BART. SRP retained ENSR to perform a BART analysis for all three units at NGS. The BART analysis was performed for two pollutants: oxides of nitrogen (NO x ) and particulate matter (PM). A BART analysis was not performed for SO 2 because the flue gas desulfurization control system at NGS was previously determined by EPA to satisfy BART. A BART review was previously conducted for the Navajo Generating Station in the 1990s as part of the Reasonably Attributable BART Program, involving visibility impacts in GCNP. This review concluded that sulfates were the major contributors to visibility impairment in the Grand Canyon. As a result, NGS installed flue gas desulfurization equipment on all of their units by In a recent Federal Implementation Plan (FIP) for the Four Corners Power Plant, EPA also discussed BART issues for the Navajo Generating Station in the May 7, 2007 issue of the Federal Register, with an excerpt provided here: EPA determined previously that the SO 2 emission limits in the 1991 FIP for the Navajo Generating Station provide for greater reasonable progress toward the national visibility goal than would BART. See 71 FR at and 72 FR at and PM emissions are controlled with hot-side electrostatic precipitators in combination with wet scrubbers. These controls effectively limit emissions to less than 0.03 lb/mmbtu, which is a PM control standard used by EPA in recent New Source Review settlement cases on existing power plants. In addition, PM is not believed to be a substantial contributor to regional haze in regional class I areas, so a BART analysis of further retrofit controls for PM 10 emissions is not included in this report. NO x emissions are currently controlled by good combustion practices and close coupled overfire air (CCOFA). The baseline peak daily NO x emission rate for Navajo ranges from about 0.45 to 0.50 lb/mmbtu, which exceeds the BART presumptive limit of 0.28 lb/mmbtu for dry-bottom, tangentially-fired boilers burning bituminous coal. Therefore, NO x controls were considered in this BART analysis for the Navajo Generating Station. The BART analysis for NO x was conducted in accordance with the procedures contained in the BART Guidelines. Consistent with the BART Guidelines, the five steps for a case-by-case BART analysis were followed. ES-1 November 2007

7 Step 1 Identify all available control technologies, including improvements to existing control equipment or installation of new add-on control equipment. Step 2 Eliminate technically infeasible options considering the commercial availability of the technology, space constraints, operating problems and reliability, and adverse side effects on the rest of the facility. Step 3 Evaluate the control effectiveness of the remaining technologies based on current pollutant concentrations, flue gas properties and composition, control technology performance, etc. Step 4 Evaluate the annual and incremental costs of each feasible option using approved EPA methods, as well as the associated energy and non-air quality environmental impacts of compliance. Step 5 Determine the visibility impairment associated with baseline emissions and the visibility improvements provided by the control technologies considered in the engineering analysis. The alternative NO x control technologies identified as being technically feasible at NGS include: low NO x burners and separated overfire air (LNB/SOFA), selective non-catalytic reduction (SNCR), and selective catalytic reduction (SCR). The results of the visibility modeling for the candidate BART control options are graphically plotted in Figure ES-1. This figure compares the total visibility impairment (expressed in deciviews) for the modeled results with output from the three modeled years averaged for each case and the eight closest Class I areas for ease of review. Figure ES-1 8 th Highest Regional Haze Deciview (Includes Background Haze and NGS BART Control Case Emissions) Baseline Option 1: SOFA/LNB on Units 1-3 (0.24 lb NOx/MMBtu) Option 1-alt: SOFA/LNB on Units 1-3 (0.20 lb NOx/MMBtu) Option 2: SOFA/LNB/SNCR on Units 1-3 Option 3: SOFA/LNB/SCR on Units 1&3 and SOFA/LNB on Unit 2 Option 4: SOFA/LNB/SCR on Units 1-3 8th Highest delta-deciview Bryce Canyon Canyonlands Capitol Reef Grand Canyon Mesa Verde Petrified Forest Sycamore Canyon Class I Area Zion ES-2 November 2007

8 Option #1 evaluated the installation of low NO x burners and separated overfire air (LNB/SOFA) at an emission rate of 0.24 lb/mmbtu. This is an emission rate that is considered to be readily achievable with the application of the technology at NGS. This combustion modification produced visibility improvements across the eight parks averaging 0.28 deciviews. For comparison, a deciview (dv) change of 1.0 is considered the threshold of humanly-perceived changes in visual air quality. Canyonlands National Park demonstrated the greatest visibility improvement with a change of 0.38 deciviews. The hardware capital costs of this retrofit option were estimated at $30 million dollars with negligible operation and maintenance (O&M) costs relative to the other options. An alternative to Option #1 evaluated the installation of low NO x burners and separated overfire air (LNB/SOFA) with a reduced emission rate of 0.20 lb/mmbtu. This lower emission represents an optimal outcome for LNB/SOFA technology at NGS, and may not be achievable. Although many tangentially fired units firing western sub-bituminous coals such as those from the Powder River Basin have achieved rates well below this rate, there is very little experience with LNB/SOFA retrofits with western bituminous fuels comparable to the coal burned at NGS. Lowering the NO x emission rate to 0.20 lb/mmbtu for this combustion modification produced visibility improvements across the eight parks averaging 0.36 deciviews. Canyonlands National Park demonstrated the greatest visibility improvement with a change of 0.51 deciviews. The hardware capital costs of this retrofit option were also estimated at $30 million dollars with negligible O&M costs relative to other options. Option #2 evaluated selective non-catalytic reduction (SNCR) in addition to LNB/SOFA. The injection of ammonia (NH 3 ) or urea ((NH 2 ) 2 CO) into the boiler further lowers NO x emissions but produces the detriment of increasing stack emissions of ammonia salts and sulfuric acid mist. These additional fine particulate emissions negatively impacted visibility at Capitol Reef, Grand Canyon, and Bryce Canyon National Parks. Moreover, on the three year average basis and across the eight parks the visibility impacts did not change from the baseline conditions. Option #3 evaluated the construction of selective catalytic reduction (SCR) systems for Units 1 and 3 with no post-combustion control on Unit 2 (due to very high SCR retrofit costs for that unit) in combination with LNB/SOFA for all units. It is not expected that emission characteristics would change if SCR were applied without LNB/SOFA. But for purposes of this report it is assumed that LNB/SOFA would reduce annual O&M costs associated with SCR and be the better economic decision if SCR were required, and therefore the technologies are shown as being applied together. This option demonstrated visibility improvements across the eight parks averaging 0.40 deciviews. Canyonlands National Park demonstrated the greatest visibility improvement with a change of 0.66 deciviews. Option #4 evaluated SCRs for all three units in combination with LNB/SOFA. The hardware capital costs associated with this option were estimated at $660 million dollars with annual O&M costs of $12.5 million dollars each year. This option produced an average visibility improvement of 0.53 deciviews compared to 0.28 deciviews for Option #1. Similar to Option #3, if SCR were ever required at the plant, more detailed work would be required to determine if the technology would be applied with or without LNB/SOFA. The associated annual costs, derived from capital expenditures and annual O&M costs, are shown in Table ES-1. The modeled visibility improvements and the calculated cost in terms of dollars per deciview are provided in Table ES-2. ES-3 November 2007

9 Table ES-1: Total Capital and Annual Costs Associated with LNB/SOFA, SCR and SNCR Applied to NGS Units 1 through 3 a b c Control Option Control Technology Total Capital Cost ($) Fixed Capital Costs ($/yr/unit) a Annual O&M Costs ($/yr) Total Annual Costs ($/yr) 1/alt. 1 LNB/SOFA (Units 1-3) 30,000,000 2,831, ,831,700 2 LNB/SOFA&SNCR (Units 1-3) 60,000,000 5,663,000 9,478,000 c 15,141,000 3 LNB/SOFA (Unit 2); LNB/SOFA&SCR (Units1&3) 396,000,000 b 37,378,440 8,316,000 45,694,440 4 LNB/SOFA&SCR (Units 1-3) 660,000,000 b 62,297,000 12,474,000 74,771,000 Fixed capital costs based on a CRF of , assuming an interest rate of 7% and amortization period of 20 years. Total capital costs include costs associated with outages required for installation of control equipment. Annual O&M costs include the lost revenues resulting from ammonia contamination of fly ash from SNCR. ES-4 November 2007

10 Table ES-2: Annual Costs vs. Visibility Improvements/Degradation (Average of the Eight Class I Areas) Options BART Controls 8th Highest Ave over 3 Years and 8 Class I Areas (total deltadeciview) Annualized Cost ($/Year) Incremental Cost from Previous Control Scenario ($/Year) Incremental Cost Effectiveness Relative to the Previous Control Scenario ($/deciview)) Baseline Option 1 Alternative Option 1 None LNB/SOFA on Units 1-3 LNB/SOFA on Units $0 $0 $ $2,831,700 $2,831,700 $10,031, $2,831,700 $0 $0 Option 2 LNB/SOFA/SNCR on Units $15,141,400 $12,309,700 Not effective; visibility does not improve from Baseline Option 3 LNB/SOFA/SCR on Units 1&3 and SOFA/LNB on Unit $45,694,440 $30,553,040 $77,349,468 Option 4 LNB/SOFA/SCR on Units $74,771,400 $29,076,960 $212,369,763 ENSR and SRP recommend LNB/SOFA for all three units as the Best Available Retrofit Technology alternative for the Navajo Generating Station based on the expected incremental visibility improvement, the cost of compliance, energy impacts, and other non-air quality environmental impacts of compliance. Control technologies using SNCR and SCR were rejected due to the small incremental improvement (and is some cases degradation) in visibility over that resulting from use of LNB/SOFA; increased emissions of ammonia, ammonia salts, and sulfuric acid aerosols, which negatively impact local plume visibility; the high capital and O&M compliance costs for these technologies; increased generation needed; and other considerations associated with the shipping, storage, and injection of anhydrous ammonia. A summary of the BART controls is presented in Table ES-3. ES-5 November 2007

11 Table ES-3: Selected Best Available Retrofit Technology Low NO x Burners / Separated Overfire Air Estimated NO x Emission Rate 0.24 lbs/mmbtu, with possible reduction to 0.20 lb/mmbtu NO x Emission Limit Reduction from Baseline Estimated NO x Reduction (TPY) at 85% utilization 48% at 0.24 lb/mmbtu 22,675 tons per year at 0.24 lb/mmbtu Costs of Compliance $125/ton of NO x removed $30 million capital cost $2.8 million/year annualized cost Energy Impacts Slight increase in fuel consumption associated with reduced boiler efficiency Non-Air Quality Environmental Impacts Increased LOI of fly ash, which could reduce recycling sales Potential for increased corrosion and more frequent replacement of furnace waterwall tubes Modeled Visibility Impacts 0.38 deciview improvement at Canyonlands (largest Class I area improvement) 0.28 deciview improvement (8-park average) ES-6 November 2007

12 1.0 Introduction 1.1 Source Description The Salt River Project Agricultural Improvement and Power District (SRP) operates the Navajo Generating Station (NGS), a coal-fired electric steam generating station located on the Navajo Nation, approximately 3 miles northeast of Page, Arizona. The NGS facility became operational in stages between 1974 and 1976 and consists of three coal-fired units with a combined power generating capacity of 2,250 MW. Each unit has a single stack with a height of 236 meters. The facility is approximately 20 kilometers from the northern boundary of Grand Canyon National Park (GCNP), which is the closest Class I area to the facility. 1.2 Reasonably Attributable BART Analysis A BART review has already been conducted for the Navajo Generating Station as an implementation of the Reasonably Attributable BART Program. In a letter to the Environmental Protection Agency (EPA) dated March 24, 1986, the Department of Interior (DOI) certified the existence of visibility impairment at the GCNP, and further identified the NGS as a probable source of the impairment. The initial certification was based upon photographic evidence. In 1991, E.H. Pechan and Associates, Inc. issued a BART Analysis for NGS under EPA contract # The analysis closely followed the 1980 EPA document Guidelines for Determining Best Available Retrofit Technology for Coal-Fired Power Plants and Other Existing Facilities by addressing six major components: 1) types of emission control systems that are possible; 2) the cost of these control systems; 3) energy requirements of the possible controls; 4) environmental impacts of these controls on such items as water and solid waste disposal; 5) life expectancy of the facility and the possible controls; and 6) the degree of improvement in visibility associated with each possible control. In August of 1991 as a result of negotiations between representatives of NGS, Environmental Defense Fund (EDF), and Grand Canyon Trust (GCT), the parties completed a negotiated memorandum of understanding (MOU) which recommended that EPA adopt a regulatory approach designed to achieve a greater degree of visibility improvement at GCNP at lower cost than the proposal published by EPA in February The recommendation proposed a 90 percent reduction in sulfur dioxide emissions to be installed on each of the three units in stages, with a final completion date in Subsequently, in October 1991, the EPA issued a final rule implementing the recommendations of the MOU. Compliance dates for sulfur dioxide controls at NGS were set by the final rule and required 90 percent control, based on an annual average, for one unit beginning November 19, 1997, two units beginning November 19, 1998, and the third unit beginning August 19, NGS fulfilled the rule requirement with the installation of sulfur dioxide controls on the final unit in BART Requirements Federal regulations under 40 CFR 51, Appendix Y, provide guidance and regulatory authority for conducting a visibility impairment analysis for designated eligible sources. The program requires the application of BART to those existing eligible sources that are believed to cause or contribute to visibility impairment in order to help meet the targets for visibility improvement at designated Class I areas. All three units at the NGS are BART eligible because they meet the following criteria: 1. The units were in existence on August 7, 1977, but had not been in operation for more than 15 years as of that date. 2. The sum of the emissions of any single visibility-impairing pollutant from the affected units is greater than 250 tons/year. 1-1 November 2007

13 3. NGS is a fossil-fuel fired steam electric plant of more than 250 MMBtu/hr heat input, and thus is in one of the categories of sources identified in the regional haze rule. 4. In addition, because the plant has a capacity of greater than 750 MW, it is subject to the presumptive BART requirements identified in 40 CFR Part 51, Appendix Y. The RHR provides that BART-eligible sources that may reasonably be anticipated to cause or contribute to visibility impairment at a federal Class I area are subject to BART. The contribution threshold for visibility impairment is a 0.5 deciview change. ENSR has conducted BART exemption modeling of Units 1-3, and the results indicate that these units are subject to BART review because the predicted visibility impacts with baseline emissions exceed 0.5 delta deciviews in at least one Class I area. 1.4 Sulfur Dioxide Emission Control SO 2 emissions at NGS are controlled by wet limestone forced-oxidation flue gas desulfurization (FGD) scrubbers, in response to the previous Reasonably Attributable BART process. In a recent Federal Implementation Plan (FIP) for the Four Corners Power Plant, EPA discussed BART issues for the Navajo Generating Station, with an excerpt provided here: Moreover, as explained in the preamble to the 2006 proposed FIP, there are only two major sources of SO 2 on the Navajo Reservation that are potentially subject to the BART requirements--navajo Generating Station and FCPP. See 71 FR at and 72 FR at and EPA determined previously that the SO 2 emission limits in the 1991 FIP for the Navajo Generating Station provide for greater reasonable progress toward the national visibility goal than would BART. Based on stack test data, the SO 2 removal efficiencies are on the order of 95%. The existing SO 2 emission limit for Navajo is 0.10 lbs/mmbtu on a 365-day boiler operating day average. This SO 2 emission limit is more restrictive than the presumptive BART limit for tangentially fired boilers burning bituminous coal of 0.15 lb/mmbtu. Consequently, no further analysis of BART controls for SO 2 is presented in this report. 1.5 Particulate Matter Emission Control The use of hot-side ESPs and wet scrubbers effectively limits the PM emissions from Navajo Units 1 through 3 at the Navajo Generating Station. During historic stack tests at the plant, filterable PM is typically less than 0.30 lb/mmbtu. These emissions rates are comparable to emissions limits applied to new coal-fired power plants and limits included in New Source Review settlements for existing coal-fired power plants. Thus, application of additional particulate controls to NGS would not be expected to produce substantial additional reductions in PM emissions, and an evaluation of PM controls is not included in this document. BART for PM is considered to be the current control configuration. 1.6 Nitrogen Oxides Emission Control NO x emissions at NGS are controlled by good combustion practices and CCOFA. Baseline peak daily NO x emissions for NGS range from about 0.45 to 0.50 lb/mmbtu, which exceeds the BART presumptive limit of 0.28 lb/mmbtu for dry-bottom, tangentially-fired boilers burning bituminous coal. Therefore, this BART analysis considers BART control options for NO x. The BART analysis for NO x was conducted in accordance with the procedures contained in the Final BART Guidelines. Consistent with the BART Guidelines, the five steps for a case-by-case BART analysis were followed. Step 1 Identify all available control technologies for the unit, including improvements to existing control equipment or installation of new add-on control equipment. 1-2 November 2007

14 Step 2 Eliminate technically infeasible options considering the commercial availability of the technology, space constraints, operating problems and reliability, and adverse side effects on the rest of the facility. Step 3 Evaluate the control effectiveness of the remaining technologies based on current pollutant concentrations, flue gas properties and composition, control technology performance, and other factors. Step 4 Evaluate the annual and incremental costs of each feasible option in accordance with approved EPA methods, as well as the associated energy and non-air quality environmental impacts of compliance. Step 5 Determine the visibility impairment associated with baseline emissions and the visibility improvements provided by the control technologies considered in the engineering analysis. This report documents the case-by-case BART analysis conducted for NO x emissions from Units 1 through 3 at the Navajo Generating Station. 1.7 Report Outline Section 2 of this report discusses the general CALPUFF modeling approach. It refers to more detailed discussions in appendices attached to this report for meteorological data processing as well as the CALPUFF modeling approach for the visibility impact analysis. A presentation of the baseline emission impacts on the nearby Class I areas for the BART exemption analysis is provided in Section 3. Because this section indicates that the incremental visibility impact of the plant is in excess of the EPA-designated limit of 0.5 delta deciviews for contributing to visibility impairment, a BART engineering analysis was conducted on available NO x control options, as presented in Section 4. A CALPUFF modeling analysis was conducted for these control options; the results are presented in Section 5. A recommendation for BART controls is presented in Section 6, based upon the information presented in Sections 4 and 5. References are provided in Section November 2007

15 2.0 CALPUFF Modeling Procedures For the refined CALPUFF modeling, SRP followed the WRAP common BART modeling protocol with the exception of the model version and a few refinements to CALMET settings. These differences are discussed below in Section CALMET Processing The Western Regional Air Partnership (WRAP) has developed six 4-km CALMET meteorological databases for three years ( ). The CALMET modeling domains are strategically designed to cover all potential BART eligible sources within WRAP states and all PSD Class I areas within 300 km of those sources. The extents of the six domains are shown in Figure 3-a through Figure 3-1f of the WRAP common BART modeling protocol, available at The BART modeling for NGS was done using the Arizona 4-km domain, as shown in Figure 2-1 of this report. The WRAP CALMET meteorological inputs, technical options, and processing steps are described in Sections 2 and 3 of the WRAP protocol. USGS 3 arc-second Digital Elevation Model (DEM) files were used by WRAP to generate the terrain data at 4-km resolution for input to the six CALMET runs. Likewise, the Composite Theme Grid format (CTG) files using Level I USGS land use categories were used by WRAP to generate the land use data at 4-km resolution for input to the six CALMET runs. See Sections and of the WRAP common BART modeling protocol for more details on the data processing. Three years of 36-km MM5 data ( ) were used by WRAP to generate the 4-km sub-regional meteorological datasets. Section 2 of the WRAP protocol discusses MM5 data extraction. The BART CALPUFF modeling for NGS was done using the Arizona 4-km CALMET database with application-specific modifications described in Appendix A. 2.2 CALPUFF Modeling Procedures SRP and ENSR used the EPA-approved version of CALPUFF (Version 5.8, Level ) that has been posted at Although the WRAP BART protocol mentions the use of CALPUFF version 6, the EPA s Office of Air Quality Planning and Standards has clearly stated that the use of a version other than the official EPA version is a non-guideline application that must obtain regional EPA approval on a case-by-case basis. It is clear from the discussion provided in Appendix A that CALPUFF version 6 is not approvable by EPA at this time without a significant effort to show that it is technically superior. To avoid the need for the justification and documentation required to use a non-guideline version of the model, ENSR used the official EPA version. The area covered by the 4-km WRAP domain for Arizona is shown in Figure 2-1. The BART CALPUFF modeling for NGS was done using a smaller computational grid within the WRAP domain to minimize computation time and output file size. The computational grid domain is also shown in Figure 2-1. This domain includes eleven Class I areas within 300 km of the source, plus a 50-km buffer around each Class I area and a 100-km buffer around the source to assure puffs recirculation. The receptors used for each of the Class I areas are based on the National Park Service database of Class I receptors. For CALPUFF model technical options, inputs, and processing steps, ENSR followed the WRAP common BART protocol with the exception of the model version. Due to the long distance to the nearest Class I area, building downwash effects were not included in the CALPUFF modeling. WRAP has developed hourly ozone measurement files for three years ( ), available at Data collection and processing are described in 2-1 November 2007

16 Section of the WRAP protocol. These ozone data files were used as input to CALPUFF. The monthly ammonia background concentrations selected for use in CALPUFF are discussed in Appendices A and B. 2.3 Natural Conditions and Monthly f(rh) at Class I Areas Eleven Class I areas were modeled for the Navajo Generating Station. For these Class I areas, natural background conditions must be established in order to determine a change in natural conditions related to a source s emissions. For the modeling described in this document, ENSR used the natural background light extinctions shown in Table 2-1, modified as noted below with site-specific considerations, and corresponding to the annual average (EPA 2003, Appendix B), consistent with the July 19, 2006 EPA guidance to Region 4 on this issue ( Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations, Joseph W. Paise/ EPA OAQPS to Kay Prince/Branch Chief). Table 2-1: Annual Average Natural Background Concentrations Class I Area Natural Background Concentrations (deciviews) Natural Background Concentrations (Mm -1 ) Arches National Park Bryce Canyon Canyonlands National Park Capitol Reef National Park Grand Canyon National Park Mesa Verde National Park Mazatzal Wilderness Petrified Forest National Park Sycamore Canyon Pine Mountain Zion To determine the input to CALPOST, it is first necessary to convert the deciviews to extinction using the equation: Extinction (Mm -1 ) = 10 exp(deciviews/10). For example, for Arches, 4.43 deciviews is equivalent to an extinction of 5.57 inverse megameters (Mm -1 ); this extinction excludes the default 10 Mm -1 for Rayleigh scattering. This remaining extinction is due to naturally occurring particles, and is held constant for the entire year s simulation. Therefore, the data provided to CALPOST for Arches would be the total natural background extinction minus 10 (expressed in Mm -1 ), or This is most easily input as a fine soil concentration of 5.57 μg/m 3 in CALPOST, since the extinction efficiency of soil (PM-fine) is 1.0 and there is no f(rh) component. The concentration entries for all other particle constituents would be set to zero, and the fine soil concentration would be kept the same for each month of the year. The monthly values for f(rh) that CALPOST needs were taken from "Guidance for Tracking Progress Under the Regional Haze Rule" (EPA, 2003) Appendix A, Table A November 2007

17 Figure 2-1: NGS CALPUFF Computational Grid in Relation to WRAP Arizona Domain 2-3 November 2007

18 2.4 Light Extinction and Haze Impact Calculations The CALPOST postprocessor was used for the calculation of the impact from the modeled source s primary and secondary particulate matter concentrations on light extinction. The formula that is used is the existing IMPROVE/EPA formula, which is applied to determine a change in light extinction due to increases in the particulate matter component concentrations. Using the notation of CALPOST, the formula is the following: b ext = 3 f(rh) [(NH 4 )2SO 4 ] + 3 f(rh) [NH 4 NO 3 ] + 4[OC] + 1[Soil] + 0.6[Coarse Mass] + 10[EC] + b Ray The concentrations, in square brackets, are in μg/m 3 and b ext is in units of Mm -1. The Rayleigh scattering term (b Ray ) has a default value of 10 Mm -1, as recommended in EPA guidance for tracking reasonable progress (EPA, 2003a). The assessment of visibility impacts at the Class I areas used CALPOST Method 6. Each hour s sourcecaused extinction is calculated by first using the hygroscopic components of the source-caused concentrations, due to ammonium sulfate and nitrate, and monthly Class I area-specific f(rh) values. The contribution to the total source-caused extinction from ammonium sulfate and nitrate is then added to the other, non-hygroscopic components of the particulate concentration (from coarse and fine soil, secondary organic aerosols, and elemental carbon) to yield the total hourly source-caused extinction. 2-4 November 2007

19 3.0 BART Applicability Analysis 3.1 BART-Eligible Requirements The BART-affected emission units at the NGS are Units 1, 2, and 3. The units are BART eligible because they meet the following criteria: 1. The units were in existence on August 7, 1977, but had not been in operation for more than 15 years as of that date. 2. The sum of the emissions of any single visibility-impairing pollutant from the affected units is greater than 250 tons/year. 3. NGS is a fossil-fuel fired steam electric plant of more than 250 MMBtu/hr heat input, and thus is in one of the categories of sources identified in the regional haze rule. In addition, because the plant has a capacity of greater than 750 MW, it is subject to the presumptive BART requirements identified in 40 CFR Part 51, Appendix Y. Because the NGS units are BART-eligible, the next step in the process is to determine whether the BARTeligible emissions contribute to a perceptible visibility impact at any Class I area. The contribution threshold is a modeled change in the natural background visibility of 0.5 delta-deciview. This modeling assessment is described in Section Existing Control Equipment and Emission Rates For purposes of determining BART applicability, the SO 2 and NO x baseline emissions were determined by ENSR using CEMS data compiled in baseline calendar years 2001 through PM filterable baseline emissions were also determined by ENSR based on stack test data and CEMS data collected in 2001 through For purposes of determining BART applicability, the SO 2 and NO x baseline emissions were based on the highest calendar day emission rates for each pollutant and the highest daily heat input rate for each individual unit compiled by the CEMS for baseline calendar years The determination of the highest daily emissions did not consider emissions associated with malfunctions, start-up, and shutdown, consistent with the WRAP BART protocol. Filterable PM baseline emissions were based on stack test data for each individual unit over the period of and the highest daily heat input rate for each individual unit during the period Speciation of the particulate matter emissions into filterable and condensable PM components was determined using the following approach: Filterable PM was subdivided by size category consistent with the default approach cited in AP-42, Table For coal-fired boilers equipped with ESPs, 67% of the filterable PM emissions are filterable PM 10 and 29% of the PM emissions are fine filterable PM 10 emissions (less than 2.5 microns in size). For coal-fired boilers, elemental carbon is expected to be 3.7% of fine PM 10 based on the best estimate for electric utility coal combustion in Table 6 of Catalog of Global Emissions Inventories and Emission Inventory Tools for Black Carbon, William Battye and Kathy Boyer, EPA Contract No. 68-D , January November 2007

20 Condensable inorganic PM 10 emissions, assumed to consist of H 2 SO 4, are based on Estimating Total Sulfuric Acid Emissions from Stationary Power Plants," EPRI, Technical Update, March For coal-fired boilers, H 2 SO 4 emissions are determined by the following relationship: E = (Q)(98.06/64.04)(F1)(F2) where: E is the H 2 SO 2 emission rate (lb/hr), Q is the baseline SO 2 emission rate (lb/hr), F1 is the fuel factor ( for western bituminous coal), and F2 is the control factor (0.63 for a hot-side ESP, 0.56 for an air pre-heater, and 0.40 for a wet FGD). For coal-fired boilers with FGD, the total condensable organic PM 10 emission factor is lb/mmbtu based on AP-42, Table Table 3-1 summarizes the baseline emissions that were used in the modeling of baseline conditions. Table 3-2 presents the stack parameters that were used for baseline conditions, as well as the BART control options. 3-2 November 2007

21 Table 3-1: Navajo Generating Station Baseline Emissions 3-3 November 2007

22 Table 3-2: Navajo Generating Station Modeling Stack Parameters Units Unit 1 Unit 2 Unit 3 Latitude Deg Longitude Deg Stack Height Ft Base Elevation Ft 4,365 4,365 4,365 Diameter Ft Flow ACFM 2,497,502 2,497,502 2,497,502 Gas Exit Velocity ft/min 5,298 5,298 5,298 Stack Gas Exit Temperature Deg F Lambert Conformal X (1) Km Lambert Conformal Y (1) Km Stack Height M Base Elevation M 1, , ,330.5 Diameter M Gas Exit Velocity m/s Stack Gas Exit Temperature Deg K Lambert Conformal coordinate system is based on 40 deg N/97 deg W projection origin, 33 deg N/45 deg N standard parallels and NWS-84 datum. 3.3 Affected Class I Areas Class I areas within 300 km of the facility are shown in Figure 3-1 and include the following eleven Class I areas: 1. Arches National Park 2. Bryce Canyon National Park 3. Canyonlands National Park 4. Capitol Reef National Park 5. Grand Canyon National Park 6. Mazatzal Wilderness 7. Mesa Verde National Park 8. Petrified Forest National Park 9. Pine Mountain Wilderness 10. Sycamore Canyon Wilderness 11. Zion National Park 3-4 November 2007

23 Figure 3-1: Class I Areas within 300 km of the Navajo Generating Station 3-5 November 2007

24 3.4 Baseline CALPUFF Modeling Results CALPUFF modeling results of the baseline emissions at eleven Class I areas are presented in Table 3-3 and graphically plotted in Figure 3-2. Modeling was conducted for all three years of CALMET meteorological data ( ). For each Class I area and year, Table 3-3 lists the 8 th highest delta-deciview, and the total 8 th highest deciview (source contribution plus the natural background). Figure 3-2 shows the total 8 th highest deciview impacts. The figure indicates that the higher visibility impacts generally occur at Grand Canyon National Park, Capitol Reef National Park, and Canyonlands National Park. Higher impacts at these Class I area are due to their proximity to NGS and local meteorological conditions. EPA recommends in its BART Guidelines that the 98 th percentile value of the modeling results should be compared to the threshold of 0.5 deciviews to determine if a source contributes to visibility impairment. The Guidelines also recommend using the 98 th -percentile statistic for comparing visibility improvements due to BART control options. On an annual basis, the 98 th percentile value implies the 8 th highest day at each modeled Class I area. The results of the baseline emissions analysis indicate that the Navajo units have predicted visibility impacts exceeding 0.5 deciviews in at least one Class I area. Therefore, per 40 CFR Part 51, Appendix Y, the NGS is presumed to be subject to BART because its emissions may reasonably be anticipated to cause or contribute to visibility impairment at a relevant Class I area. Candidate BART controls are discussed in Section 4. The results of the visibility improvement modeling for these candidate controls are discussed in Section 5. The final BART recommendations are provided in Section 6. Table 3-3: Regional Haze Impacts Due to Baseline Emissions 3-6 November 2007

25 Figure 3-2: 8 th Highest Regional Haze Impacts for Each Modeled Year Due to Baseline Emissions SRP Navajo Generating Station Regional Haze Impacts for each Modeled Year Due to Baseline Emissions Impacts Include Annual Average Background Met Year 2001 Met Year 2002 Met Year th Highest deciview Arches NP Bryce Canyon NP Canyonlands NP Capitol Reef NP Grand Canyon NP Mazatzal W Mesa Verde NP Petrified Forest NP Pine Mountain W Sycamore Canyon W Zion NP Class I Area 3-7 November 2007

26 4.0 BART Engineering Analysis for NO x Emissions Nitrogen oxides (NO x ) formed during the combustion of coal are generally classified as either thermal NO x or fuel-bound NO x. Thermal NO x is formed when elemental nitrogen in the combustion air is oxidized at the high temperatures in the primary combustion zone, yielding nitrogen oxide (NO) and nitrogen dioxide (NO 2 ). The rate of formation of thermal NO x is a function of residence time and free oxygen, and increases exponentially with peak flame temperatures. Thermal NO x from coal combustion can be effectively controlled by techniques that limit available oxygen or reduce peak flame temperatures in the primary combustion zone. Fuel-bound NO x is formed by the oxidation of chemically bound nitrogen in the fuel. The rate of formation of fuel-bound NO x is primarily a function of the coal nitrogen content, but is also affected by gas turbulence and fuel/air mixing. The following sections describe (1) the technologies for reducing NO x emissions from coal fired power plants, (2) the feasibility of applying such controls to NGS, (3) the expected effectiveness of such controls if used at NGS, and (4) the impacts of implementing such controls at NGS. 4.1 Identification of Alternative NO x Controls The alternative NO x control technologies available for limiting NO x emissions from Units 1 through 3 include combustion techniques, such as LNB/SOFA systems and flue gas recirculation, and post-combustion controls, such as selective non-catalytic reduction (SNCR), selective catalytic reduction (SCR), ECOTUBE, and ROFA /ROTAMIX,. These alternative NO x control technologies are evaluated below in terms of their possible application to Unit 1 through Low-NO x Burners and Separated Overfire Air (LNB/SOFA) LNB/SOFA systems in tangentially fired boilers are designed to control fuel and air mixing to reduce peak flame temperatures, resulting in less NO x formation. Combustion, reduction, and burnout are achieved in three stages within a conventional low NO x burner. In the initial stage, combustion occurs in a fuel-rich, oxygendeficient zone where the NO x is formed. A reducing atmosphere follows where hydrocarbons are formed that react with the already formed NO x. In the third stage, internal air staging completes the combustion, but may result in additional NO x formation. This, however, can be minimized by completing the combustion in an airlean environment. Combustion air is separated into primary and secondary flow sections to achieve complete burnout and to encourage the formation of nitrogen, rather than NO x. Primary air (70-90%) is mixed with the fuel, producing a relatively low-temperature, oxygen-deficient, fuel-rich zone and, therefore, reducing the formation of fuel-bound NO x. Secondary air representing 10-30% of the combustion air is injected above the combustion zone through a special wind-box with air introducing ports and/or nozzles mounted above the burners. Combustion is completed at this increased flame volume. Hence, the relatively low-temperature secondary stage limits the production of thermal NO x. The location of the injection ports and mixing of overfire air are critical to maintain efficient combustion. Retrofitting SOFA on an existing boiler involves waterwall tube modifications to create the ports for the secondary air nozzles and the addition of ducts, dampers, and the wind-box. The effectiveness of the technology at reducing NO x is influenced by a number of site specific factors including fuel characteristics (e.g., nitrogen content, ratio of fixed to volatile carbon, heat content), furnace geometry, structural factors that may limit options for locations of SOFA ports, etc Flue Gas Recirculation Flue gas recirculation for NO x control includes gas recirculation into the furnace or into the burner. In this technology, 20 to 30% of the flue gas (at C) is recirculated and mixed with the combustion air. The resulting dilution in the flame decreases the temperature and availability of oxygen, thereby reducing thermal NO x formation. When flue gas recirculation into the burner is used in low NO x burners, the flue gas is usually recirculated subject to the operational constraints of flame stability and impingement. Retrofitting an existing coal-fired unit with flue gas recirculation involves installation of a system to extract the flue gas 4-1 November 2007

27 from the boiler unit, additional ductwork, fan, and a fly ash collecting device. The fly ash control device is necessary in order to clean the flue gas of particulate matter prior to recirculation. Heat distribution in the furnace may be affected due to the increase in throughput. Excessive flue gas recirculation can also result in flame instability problems and increased steam temperatures. Flue gas recirculation alone in coal-fired boilers achieves a low NO x reduction efficiency of less 20%. This is because the ratio of thermal-no x to total NO x emissions is relatively low in coal-fired plants Selective Non-Catalytic Reduction Selective non-catalytic reduction is based on a gas-phase homogeneous reaction that involves the injection of an amine-based compound into the flue gas within an appropriate temperature range for reduction of NO x. An amine-based compound, such as ammonia, NH 3, or urea, (NH 2 ) 2 CO, is used as the NO x reducing agent in SNCR processes. When ammonia or urea is injected into the flue gas stream, it selectively reduces the NO x into molecular nitrogen, N 2, and water, H 2 O. At stoichiometric conditions, when the adequate residence time is reached, the overall reactions that occur may be characterized by: Ammonia 4NO + 4NH N 2 + 6H 2 0 2NO 2 + NH N 2 + 6H 2 0 Urea 2(NH 2 ) 2 CO + 4NO + O 2 4N 2 + 2CO 2 + 4H 2 O In an SNCR system, NO x reduction does not take place in the presence of a catalyst, but rather is driven by the thermal decomposition of ammonia or urea and the subsequent reduction of NO x. Consequently, the SNCR process operates at higher temperatures than the SCR process. Critical to the successful reduction of NO x with SNCR is the temperature of the flue gas at the point where the reagent is injected. For the ammonia injection process, the necessary temperature range is 1,700 to 1,900 F. The other factors affecting SNCR performance are gas mixing, residence time at operating temperatures, and ammonia slip. Because oxygen is present in the flue gas, a portion of the ammonia may oxidize at temperatures greater than 2,000 F. Above 2,000 F, the reaction of ammonia oxidation becomes predominant. Nitrogen monoxide is formed as a product of this reaction of ammonia oxidation. As a result, when the flue gas temperature at reagent injection locations is higher than the appropriate temperature window, the SNCR process results in NO x formation rather than NO x reduction. At temperatures lower than the required temperature window, the NO x reduction reaction rates become lower, and unreacted ammonia may slip through and be emitted to the atmosphere. As with SCR, SNCR equipment vendors suggest a higher ammonia injection rate than is stoichiometrically required to achieve higher NO x control efficiencies. The various SNCR vendors typically will not provide a guarantee against ammonia slip of less than 5 ppm for systems designed for high NO x performance levels. This excess ammonia may react with SO 3 and water vapor to form ammonium bisulfate and ammonium sulfate. Although no SO 2 is oxidized by the SNCR system, naturally occurring SO 3 concentrations are high enough to be a concern with potentially high ammonia slip rates. Ammonium bisulfate may precipitate out at air heater operating temperatures and can ultimately lead to air heater fouling and plugging. Furthermore, the ammonium salts may condense as the flue gases cool and can lead to increased emissions of both PM 10 and PM 2.5. Ammonia slip from SNCR systems occurs either from injection at temperatures too low for effective reaction with NO x or from over-injection of reagent, leading to uneven distribution. Controlling ammonia slip in SNCR systems is difficult as there is no opportunity for effective feedback to control reagent injection. The reagent injection system must be able to place the reagent where it is most effective within the boiler because 4-2 November 2007

28 NO x distribution varies within the cross section. An injection system that has too few injection control points or fails to inject a uniform amount of ammonia across the entire section of the boiler will almost certainly lead to a poor distribution ratio and high ammonia slip. Distribution of the reagent can be especially difficult in larger coal-fired boilers because of the long injection distance required to cover the relatively large cross-section of the boiler. Multiple layers of reagent injection as well as individual injection zones in the cross-section of each injection level are commonly used to follow the temperature changes caused by boiler load changes. On small coal-fired units (i.e., less than 200 MW), SNCR has been demonstrated to achieve NO x reductions ranging from 25 to 50% with acceptable levels of ammonia slip. This variation in NO x reduction depends on site-specific considerations and the amount of ammonia slip considered acceptable. In practical applications, non-uniformities in velocity and temperature at the reagent injection location can pose operational difficulties because of the inherent sensitivity of the process to these parameters. The physical location of the effective temperature range within the boiler depends on operating factors such as unit load, combustion air distribution, and soot blowing cycles. Generally, these factors require the utilization of multiple injection elevations in full-scale systems. For larger boilers (i.e., greater than 300 MW), there are numerous challenges associated with applying SNCR. In particular, such boilers large physical dimensions pose challenges for injecting and mixing the reagent with the flue gas. Another issue with larger units is the fact that the SNCR temperature window often exists within the convective passes. Demonstrations at the Port Jefferson, Morro Bay, and Merrimac plants have shown that injecting in the convective pass can create high ammonia slip due to limited residence time at the operating temperatures of SNCR. 1,2 EPRI sponsored a computational fluid dynamics modeling program to evaluate the performance of SNCR on Southern Company Service s Wansley Unit 1 located in Roopville, Georgia. This 880-MW unit is a tangentially-fired boiler equipped with a low-no X burner and separated overfire air. The modeling results demonstrated that SNCR has the potential to reduce NO x emissions by only 22% with an acceptable ammonia slip of 6 ppm. The firing characteristics of the boiler make achieving higher levels of NO x reduction impractical. The most influential factor is the separated overfire air system, which elevates upper furnace temperatures by causing the combustion process to extend beyond the furnace nose and into the convection section Selective Catalytic Reduction Selective catalytic reduction is a process that involves post-combustion removal of NO x from flue gas utilizing a catalytic reactor. In the SCR process, ammonia injected into the flue gas reacts with nitrogen oxides and oxygen to form nitrogen and water vapor. The SCR process converts NO x to nitrogen and water by the following general reactions: 4NO + 4NH 3 + O 2 4N 2 + 6H 2 O 2NO 2 + 4NH 3 + O 2 3N 2 + 6H 2 O The reactions take place on the surface of a catalyst. The function of the catalyst is to effectively lower the activation energy of the NO x decomposition reaction to about 375 to 750 F, depending on the specific catalyst and other contaminants in the flue gas. The factors affecting SCR performance are catalyst reactor design, optimum operating temperature, sulfur content of the fuel, catalyst deactivation due to aging or poisoning, ammonia slip emissions, and design of the ammonia injection system. 1 Shore, D., et al, "Urea SNCR Demonstration at Long Island Lighting Company's Port Jefferson Station, Unit 3," Proceedings of the EPRI/EPA Joint Symposium on Stationary Combustion NO x Control, May Lin, Chin-I, " Full Scale Tests of SNCR Technology on a Gas-Fired Boiler," EPRI Workshop on NO x Controls for Utility Boilers, July Harmon, A., et al., Evaluation of SNCR Performance on Large-Scale Coal-Fired Boilers, Institute of Clean Air Companies (ICAC) Forum on Cutting NO x Emissions, Durham, NC, March November 2007

29 The SCR system includes a number of subsystems, including the SCR reactor, ammonia injection system, and ammonia storage and delivery system. The SCR reactor is located downstream of the economizer and electrostatic precipitators, and upstream of the air pre-heater. From the ESP outlet, the flue gas first passes through a low-pressure ammonia/air injection grid designed to provide optimal mixing of ammonia with flue gas. The ammonia-treated flue gas then flows through the catalyst bed and exits to the air pre-heater. The SCR system for a pulverized coal boiler typically uses a fixed bed catalyst in a vertical down-flow, multi-stage reactor. Reduction catalysts are divided into two groups: base metal, primarily vanadium, platinum, or titanium (lower temperature); and zeolite (higher temperature). Both groups exhibit advantages and disadvantages in terms of operating temperature, ammonia-no x ratio, and optimum oxygen concentration. The optimum operating temperature for a vanadium-titanium catalyst system is in the range of 550 to 800 F, which is significantly higher than for platinum catalyst systems. However, the vanadium-titanium catalyst systems begin to break down when continuously operating at temperatures above this range. Operation above the maximum temperature results in oxidation of ammonia to either ammonia sulfate or NO x, thereby actually increasing NO x emissions. To achieve high NO x control efficiencies, the SCR vendors suggest a higher ammonia injection rate than is stoichiometrically required to react all of the NO x in the combustion gases. This results in emissions of unreacted ammonia or ammonia slip. The various SCR vendors typically guarantee ammonia slip of about 2 ppm for systems designed for very high NO x performance levels. This excess ammonia may react with SO 3 and water vapor to form ammonium bisulfate (NH 4 )HSO 4 and ammonium sulfate (NH 4 ) 2 SO 4. Higher levels of ammonia and SO 2 result in higher levels of ammonium bisulfate and ammonia sulfate. These ammonium salts may condense as the flue gases cool and can lead to increased emissions of both PM 10 and PM 2.5. Furthermore, the catalyst promotes the partial oxidation of SO 2 to SO 3, which in turn combines with water, increasing the formation of these ammonia salts and potential emissions of PM 10 and PM 2.5. Some SCR installations have experienced significant air pre-heater plugging and corrosion resulting from the deposition of ammonium bisulfate. The plugging and corrosion can cause reduced boiler efficiency, higher flue gas pressure drop, more frequent air pre-heater cleaning and washing, increased boiler downtime, and increased maintenance cost. The primary factors for controlling the formation and deposition of ammonium bisulfate are the level of ammonia, the level of SO 3, the air pre-heater surface temperature profile, the air pre-heater surface material, and the air pre-heater physical configuration. The temperature window for ammonium bisulfate deposition is as wide as 300 to 425 F. The temperature window for sulfuric acid formation, which is primarily a function of SO 3 concentration, is significantly lower, but must also be evaluated in system design. The SCR system is subject to catalyst deactivation over time. Catalyst deactivation occurs through two primary mechanisms: physical deactivation and chemical poisoning. Physical deactivation generally results from either prolonged exposure to excessive temperatures or masking of the catalyst due to entrainment of particulate from ambient air or internal contaminants. Chemical poisoning is caused by the irreversible reaction of the catalyst with a contaminant in the gas stream and thus is a permanent condition. Catalyst suppliers typically guarantee a limited lifetime for high performance catalyst systems. Fly ash plugging generally results from excessive fly ash carryover to the catalyst or poor catalyst gas flow design. Gas, ash, and ammonia flow distribution modeling is critical in the design phase ECOTUBE The ECOTUBE system is a hybrid NO x reduction system that combines SOFA and SNCR technologies. Retractable lance tubes that penetrate the boiler above the primary burner zone inject high-velocity air, which creates turbulent airflow and increases the residence time for the air/fuel mixture. The water-cooled ECOTUBEs are automatically retracted from the boiler on a regular basis and cleaned to remove layers of soot and other depositions. 4-4 November 2007

30 4.1.6 ROFA /ROTAMIX Mobotec provides a NO x reduction system that combines OFA and SNCR technologies into an integrated system. The system uses a modified OFA system with improved mixing characteristics achieved through adding a rotation to the OFA. This system is called ROFA - Rotating Opposed Firing Air. In addition, ROTAMIX, consisting of adding urea or ammonia injection into the ROFA air nozzles, can be added to the system. The extra mixing produced by combining the ROFA nozzles with the reagent injection results in improved mixing and a more homogeneous temperature profile in the boiler. This NO x reduction system has primarily been applied to smaller units, with the vendor reporting NO x reductions of up to 75%. 4.2 Technical Feasibility of Alternative NO x Controls Of the technologies discussed above, the following are not considered technically feasible for application to Navajo Units 1 through 3: Flue gas recirculation (FGR) has been demonstrated on units of the same scale as NGS Units 1 through 3 but has not been demonstrated to achieve NO x reduction that is more effective than staged combustion techniques such as LNB/SOFA. ECOTUBE has not been demonstrated on units of the same scale and design as NGS Units 1 through 3. The Mobotec ROFA/ROTAMIX system has not been demonstrated on units of the same scale and design as NGS Units 1 through 3. Therefore, the BART analysis is limited to those technologies that have been demonstrated to achieve significant NO x reduction on units of the same scale and design as Navajo Units 1 through 3: LNB/SOFA; LNB/SOFA plus SNCR; and LNB/SOFA plus SCR. 4.3 Effectiveness of Technically Feasible NO x Controls The alternative NO x control technologies, LNB/SOFA, SCR, and SNCR, have been successfully applied to new utility coal-fired boilers, as well as retrofitted to existing utility coal-fired boilers. The effectiveness of these technologies in reducing NO x emissions is dependent primarily on the inlet NO x concentrations, residence time, and operating temperatures. Staged combustion techniques have been demonstrated to achieve up to 50% reduction in uncontrolled NO x emissions from coal-fired utility boilers. SNCR has been demonstrated to achieve NO x control efficiencies ranging from 25% to 50% with inlet NO x concentrations of 300 to 400 ppmvd. If staged combustion is used to reduce inlet NO x concentrations to less than 150 ppmvd, SNCR is capable of achieving NO x control efficiencies of only 10 to 25%. Likewise, SCR can achieve NO x control efficiencies as high as 90% with inlet NO x concentrations in the range of 300 to 400 ppmvd. If inlet NO x concentrations are less than 150 ppmvd, SCR can achieve NO x control efficiencies of only 50 to 70%. Based on information provided by equipment vendors, it is estimated that LNB/SOFA will reduce NO x emissions to an emission level of approximately 0.24 lb/mmbtu. Due to the lower inlet NO x concentrations, it is estimated that the addition of SNCR will reduce NO x emissions by another 15%, which corresponds to a NO x emission level of approximately 0.20 lb/mmbtu. SCR is estimated to reduce NO x emissions to approximately lb/mmbtu. Table 3-1 summarizes the annual NO x emissions resulting from the application of LNB/SOFA, SCR, and SNCR to Units 1, 2, or 3. An alternative Control Option 1 is also evaluated assuming LNB/SOFA can achieve a NO x emission rate of 0.20 lb/mmbtu. This emission rate represents an optimal outcome for LNB/SOFA technology at NGS which may not be realized. The uncertainty associated with this rate is largely based on the very limited practical experience with LNB/SOFA retrofits at plants burning western bituminous fuels comparable to that which is burned at NGS. In addition, the units at NGS are relatively small and do not provide adequate residence time for effective staging of combustion and achieving lower NO x rates. 4-5 November 2007

31 Table 4-1: Annual NO x Emission Rates Resulting from Application of Control Options to NGS Units 1 3 (lb/mmbtu) Control Option Control Technology Unit 1 Unit 2 Unit 3 Baseline Emissions Good Combustion & CCOFA LNB/SOFA (Units 1-3) Alternative LNB/SOFA (Units 1-3) LNB/SOFA&SNCR (Units 1-3) LNB/SOFA (Unit 2); LNB/SOFA&SCR (Units1&3) LNB/SOFA&SCR (Units 1-3) To assess the effectiveness of the alternative NO x control technologies in reducing visibility impairment attributable to Units 1 through 3, we developed the stack and emission parameters for the following control options: Option 1 and 1 Alternative. This option involves the retrofit of LNB/SOFA to Units 1 through 3 to control NO x emissions from all three units to 0.24 lb/mmbtu (0.20 lb/mmbtu for Option 1 Alternative). Option 2. This option involves the retrofit of LNB/ SOFA, and SNCR to Units 1 through 3 to control NO x emissions from all three units to 0.20 lb/mmbtu. Option 3. This option involves the retrofit of LNB/SOFA to Unit 2 to control NO x emissions from this unit to 0.24 lb/mmbtu, and the retrofit of LNB/SOFA and SCR to Units 1 and 3 to control NO x emissions from these two units to 0.08 lb/mmbtu. Option 4. This option involves the retrofit of LNB/SOFA, and SCR to Units 1 through 3 to control NO x emissions from all three units to 0.08 lb/mmbtu. The alternative control technologies will not only affect PM, SO 2, and NO x emission levels from Units 1 through 3, but will also affect the emissions and speciation of PM 10. The PM 10 emissions and speciation were determined using the following approach: Filterable PM was subdivided by size category consistent with the default approach cited in AP-42, Table For coal-fired boilers equipped with ESPs, 67% of the filterable PM emissions are filterable PM 10 and 29% of the PM emissions are fine filterable PM 10 emissions (less than 2.5 microns in size). For coal-fired boilers, elemental carbon is expected to be 3.7% of fine PM 10 based on the best estimate for electric utility coal combustion in Table 6 of Catalog of Global Emissions Inventories and Emission Inventory Tools for Black Carbon, William Battye and Kathy Boyer, EPA Contract No. 68-D , January Condensable inorganic PM 10 emissions, assumed to consist of H 2 SO 4, are based on Estimating Total Sulfuric Acid Emissions from Stationary Power Plants," EPRI, Technical Update, March For coal-fired boilers equipped with SCR, H 2 SO 4 emissions are determined by the following relationship: 4-6 November 2007

32 E = (Q)(98.06/64.04)(F1+S2)(F2) where: E is the H 2 SO 2 emission rate (lb/hr), Q is the baseline SO 2 emission rate (lb/hr), F1 is the fuel factor ( for western bituminous coal), S2 is the SCR catalyst SO 2 oxidation rate (0.01 for PRB coal; which is the closest coal type to the fuel used at NGS) F2 is the control factor (0.63 for a hot-side ESP, 0.56 for an air pre-heater, and 0.60 for wet FGD). Note that, for units not equipped with SCR, the factor, S2, is eliminated from the above relationship. For coal-fired boilers with FGD, the total condensable organic PM 10 emission factor is lb/mmbtu based on AP-42, Table The stack parameters for Units 1, 2, and 3 under all four control options are presented in Table 3-2. The NO x, SO 2, and PM 10 emissions from Units 1, 2, and 3 under the four control options are then summarized in Tables 4-2, 4-3, 4-4, and 4-5 respectively. Option 1 Alternative simply reduces the NO x emission rates by 16.7% from those listed in Table Impacts of Alternative NO x Controls The alternative control technologies available to control NO x emissions from Units 1 through 3 are LNB/SOFA, SCR, and SNCR. This section documents the economic, non-air environmental, and energy impacts associated with applying either LNB/SOFA, SCR or SNCR to Units 1 through Economic Impacts Table 4-2 summarizes the total capital and annual costs associated with applying LNB/SOFA, SCR, and SNCR to Units 1 through 3. Table 4-2: Total Capital and Annual Costs Associated with LNB/SOFA, SCR and SNCR Applied to NGS Units 1 through 3 Control Option Control Technology Total Capital Cost ($) Fixed Capital Costs ($/yr) a Annual O&M Costs ($/yr) Total Annual Costs ($/yr) 1/alt. 1 LNB/SOFA (Units 1-3) 30,000,000 2,831, ,831,700 2 LNB/SOFA&SNCR (Units 1-3) 60,000,000 5,663,000 9,478,000 c 15,141,000 3 LNB/SOFA (Unit 2); LNB/SOFA&SCR (Units1&3) 396,000,000 b 37,378,440 8,316,000 45,694,440 4 LNB/SOFA&SCR (Units 1-3) 660,000,000 b 62,297,000 12,474,000 74,771,000 a b c Fixed capital costs based on a CRF of , assuming an interest rate of 7% and amortization period of 20 years. Total capital costs include costs associated with outages required for installation of control equipment. Annual O&M costs include the lost revenues resulting from ammonia contamination of fly ash from SNCR Non-Air Quality Environmental Impacts One of the most significant impacts of retrofitting SCR or SNCR on the facility is the addition of ammonia or urea storage and handling systems. Anhydrous ammonia and aqueous ammonia above 20 percent are considered dangerous to human health. An accidental release of anhydrous ammonia or 20% or greater aqueous ammonia is reportable to local, state, and federal agencies. In anticipation of such an incident, the site would need to develop, implement, and maintain a Risk Management Plan (RMP) and Process Safety 4-7 November 2007

33 Measures (PSM) Program. Risk communication to the general public typically includes a worst-case analysis with potential impacts possible at up to a mile from the facility. Even the storage of less than 20% anhydrous ammonia is subject to the general duty clause of the RMP Program. Because the nearest rail access is about 150 miles from the NGS, the anhydrous ammonia required for the SCR option would have to be delivered to the site via tanker trucks. To achieve the projected NO x reduction of 53 tons per day, approximately 21 tons of anhydrous ammonia would be required each day in the three units at the NGS. Assuming each tanker truck has a capacity of 6,000 gallons, this would require one to two truck deliveries of anhydrous ammonia each day to the NGS. This truck traffic would add to the atmospheric burden of air pollution in the region. Theoretically, one mole of ammonia will react with one mole of NO x, forming elemental nitrogen and water in both SCR and SNCR. In reality, not all the injected reagent will react due to imperfect mixing, uneven temperature distribution, and insufficient residence time. These physical limitations may be compensated for by injecting a larger amount of ammonia than stoichiometrically required and essentially achieving lower NO x emissions at the expense of ammonia slip. Typically, the ammonia slip associated with SCR is approximately 2 ppm at 6% O 2, while that associated with SNCR is approximately 5 ppm at 6% O 2. This excess ammonia may react with SO 3 or H 2 SO 4 in the flue to form ammonium bisulfate ((NH 4 )HSO 4 ) and ammonium sulfate ((NH 4 ) 2 SO 4 ). These ammonium salts may condense as the flue gases cool and can lead to increased emissions of both PM 10 and PM 2.5. In SCR systems, the catalyst promotes the partial oxidation of SO 2 to SO 3, which in turn increases the formation of ammonium bisulfate and ammonium sulfate. The enhanced formation of these ammonia salts further increases the potential emissions of PM 10 and PM 2.5. Ammonia associated with fly ash has the potential to present several problems with the disposal and/or the use of the fly ash. Once the fly ash is exposed to the SNCR process, there will be a significant quantity of soluble salts associated with the fly ash. These salts are expected to be NH 4 HSO 4 and (NH 4 ) 2 SO 4. Ash buyers have expressed concerns about the presence of ammonia on ash. The issues appear to be perception and odor, rather than actual impacts on product quality (e.g., when used for concrete). The level of concern may be regional in nature (i.e., less concern in a non-freezing climate) and the ultimate market for the ash (i.e., commercial versus residential use). The tendency of fly ash to adsorb ammonia is a function of many factors in addition to the amount of ammonia slip. Ash characteristics such as ph, alkali mineral content, and volatile sulfur and chlorine content help to determine whether or not ammonia will be adsorbed readily by the fly ash. Elevated ph will cause ammonia release to the air and resulting odor. In some applications, properly designed SNCR systems can keep the ammonia slip levels low enough so that the salability of the ash is unaffected. Dry disposal of fly ash can cause the leachate and/or runoff water to contain increased concentrations of ammonia. If and when these salts are contacted by water, they will most likely be dissolved and the resulting aqueous concentration of nitrogen-containing compounds can increase in the waters associated with the ash. Table 4-3 summarizes the non-air quality environmental impacts associated with the proposed BART control options. 4-8 November 2007

34 Table 4-3: Summary of Non-Air Quality Environmental Impacts Control Alternative Summary of Non-Air Quality Environmental Impacts LNB/SOFA Potential to the increase LOI of flyash which could reduce recycling sales Slight increase in CO 2 emissions/kwh associated with reduced boiler efficiency Potential for poor combustion and increased CO emissions Potential for increased corrosion and more frequent replacement of furnace waterwall tubes SNCR Addition of ammonia or urea storage and handling systems present higher risk to NGS workers and the community. Truck deliveries of ammonia/urea via long haul from nearest source Anhydrous ammonia and aqueous ammonia above 20 percent are considered dangerous to human health, and accidental releases are reportable to local, state, and federal agencies. The site must develop, implement, and maintain a Risk Management Plan (RMP) and Process Safety Measures (PSM) Program. Excess ammonia injected into the boiler reacts with SO 3 or H 2 SO 4 in the flue to form ammonium bisulfate (NH 4 )HSO 4 and ammonium sulfate (NH 4 ) 2 SO 4. Sulfuric acid in the flue gas can cause various power plant operating and maintenance problems. Condensation of sulfuric acid has a significant detrimental effect on downstream pollution control equipment, including fouling and corrosion of heat transfer surfaces in the air pre-heater. Ammonia associated with fly ash has the potential to present several problems with the disposal and/or the use of the fly ash. Dry disposal of fly ash can cause leachate and/or runoff water to contain increased concentrations of ammonia and/or nitrogen-containing compounds. SCR Addition of ammonia handling systems. Anhydrous ammonia and aqueous ammonia above 20 percent are considered dangerous to human health, and accidental releases are reportable to local, state, and federal agencies. The site must develop, implement, and maintain a Risk Management Plan (RMP) and Process Safety Measures (PSM) Program Truck deliveries of ammonia via long haul from nearest source 4-9 November 2007

35 4.4.3 Energy Impacts SCR would consume significantly more electrical energy than SNCR. The higher electrical energy consumption for SCR relative to SNCR primarily is due to the power required for the increased fan static pressure required to overcome the pressure drop across the catalyst bed, as well as for pumps and evaporator blower. Assuming a pressure drop of 14 inches W.G. across the catalyst bed, SCR applied to all three units would consume approximately 22,500 kwh more electrical power than SNCR (approximately 1% of the total power generation of NGS). The increased emissions of criteria pollutants required to maintain the net electrical output have not been incorporated into the visibility modeling, but the reviewer should be aware that any reported visibility improvements due to SCR operation do not consider potential negative visibility impacts of the resulting increase in air emissions Remaining Useful Life A 20-year amortization period for NGS was used for purposes of the BART analysis November 2007

36 Table 4-4: Navajo Generating Station SOFA/LNB Emission Controls (Option 1) 4-11 November 2007

37 Table 4-5: Navajo Generating Station SOFA/LNB Emission Controls (Alternative Option 1) 4-12 November 2007

38 Table 4-6: Navajo Generating Station SOFA/LNB/SNCR Emission Controls (Option 2) 4-13 November 2007

39 Table 4-7: Navajo Generating Station SOFA/LNB/SCR (Option 3) 4-14 November 2007

40 Table 4-8: Navajo Generating Station SOFA/LNB/SCR (Option 4) 4-15 November 2007

41 5.0 BART Control Options Modeling Analysis This section provides a summary of the modeled visibility improvement as a result of installing BART control options on Navajo Units Modeled Control Scenarios Four BART control scenarios were modeled for each meteorological year ( ) and for all 11 Class I areas within 300 km. However, 3 of the 11 Class I areas are beyond others that are closer to NGS in the same direction. To summarize the impacts over the nearest Class I areas in several directions from the plant, we report average impacts over these closest 8 areas in some of the summary statistics. Emission rates that were used in modeling the four BART control options are listed in Tables 4-4 through 4-8. These control scenarios, which are more fully discussed in Section 4, are: Options 1 and 1 Alternative: Advanced combustion controls (LNB/SOFA) on Units 1-3. Option 2: Advanced combustion controls with SNCR on Units 1-3. Option 3: Advanced combustion controls with SCR on Units 1 and 3 and advanced combustion controls (LNB/SOFA) on Unit 2. Option 4: Advanced combustion controls with SCR on Units CALPUFF Modeling Results for Control Options The results of the BART control options modeling are presented in Tables 5-1 and 5-2 and graphically plotted in Figure 5-1. Table 5-1 is an overall summary, averaged over the eight closest Class I areas and the three modeled years, of the predicted visibility changes due to installation of the candidate BART controls on the Navajo units. Table 5-2 shows detailed information on the visibility impacts of the BART control options for each modeled Class I area and meteorological year. Figure 5-1 compares the total visibility impairment (expressed in deciviews) for the modeled results with output from the three modeled years averaged for each case and Class I area. The tables and the figure indicate that some of the BART emission controls are predicted to result in visibility benefits, while others would result in visibility degradation (shown in red). Results for each candidate BART control option are discussed in more detail below. Option 1: The results show that the averaged regional haze impacts may improve visibility by about 0.28 delta-dv (relative to the baseline case) with the installation of LNB/SOFA controls. These controls result in a NO x emission rate that is below the presumptive limit, thus complying with the requirements of the EPA BART rule for large electric generating plants. The results indicate that for the two most heavily impacted Class I areas, Grand Canyon and Capitol Reef National Parks, this option produces the best result for visibility improvement, and addition of post-combustion controls for NO x may actually degrade visibility in some cases, especially in the closest Class I areas. For the other Class I areas, the visibility improvements associated with additional control options are relatively minor. Option 1 Alternative: The results show that the averaged regional haze impacts may improve visibility by about 0.36 delta-dv (relative to the baseline case) with the installation of LNB/SOFA controls and optimization over time. This level of improvement is somewhat higher than that from Option 1. These controls also result in a NO x emission rate that is well below the presumptive limit, thus complying with the requirements of the EPA BART rule for large electric generating plants. The results indicate that for the two most heavily impacted Class I areas, Grand Canyon and Capitol Reef National Parks, this option also results in the best result for 5-1 November 2007

42 visibility improvement, and addition of post-combustion controls for NO x may actually degrade visibility, as noted for Option 1. For the other Class I areas, the visibility improvements associated with additional control options are relatively minor. Option 2: Addition of SNCR on all three units will negatively impact visibility at Capitol Reef, Grand Canyon, and Bryce Canyon National Parks. Moreover, on average, across the eight parks the visibility impacts will not change from the baseline conditions. Therefore, BART control Option 2 is not effective in improving visibility and is not recommended. Option 3: Installation of SCR on Units 1 and 3 in addition to the Option 1 NO x controls may improve visibility by 0.40 delta-dv from the baseline, and only 0.12 delta-dv from Option 1 (0.04 delta-dv from Option 1a). It is noteworthy that the installation of SCR would create new emissions of primary sulfates (H 2 SO 4 ) and excess ammonia, and would lead to increased gross generation to run the SCR equipment. Therefore, NO x emission controls involving SCR are relatively ineffective in this case, especially taking into account the high cost of the controls. Option 4: Addition of SCR on all three units in addition to the Option 1 controls may improve visibility by 0.53 delta-dv from the baseline case and by 0.25 delta-dv from control Option 1. The relatively small incremental improvement in visibility is due in part to the small role that nitrates play in the total regional haze contribution. Moreover, for the same reasons described for Option 3 (addition of primary sulfates, excess ammonia, increased generation, and high cost), the installation of SCR is relatively ineffective. 5.3 Cost of BART Control Options Table 5-3 summarizes the annualized control cost that is the product of the $/ton removed and the number of tons of NO x removed by each control strategy, as well as an incremental computation of each control option s visibility improvement/degradation effectiveness and cost. The visibility results in Table 5-3 are based on the average of the three years and the eight closest modeled Class I areas. Figure 5-2 shows a graph of visibility improvements/degradation as a function of the cost for each control option. BART options associated with incremental improvements in visibility relative to a previous beneficial control option are connected with a blue line. Table 5-3 and Figure 5-2 indicate that the visibility would, on average, not improve in the eight Class I areas if the Option 2 controls (using SNCR) were adopted. A similar table and figure were prepared for Class I areas with the highest and lowest modeled visibility impacts. Figure 5-1 indicates that the Petrified Forest National Park is among the least impacted Class I areas. Figure 5-3 and Table D-1 (in Appendix D) show the annual cost compared to the visibility improvements/degradation at Petrified Forest National Park. Capitol Reef and Grand Canyon National Parks experience the highest impacts, with the impacts at Canyonlands slightly lower. After control Option 1 (or 1-alternative), the visibility at both Capitol Reef and Grand Canyon either degrades or shows a minimal benefit with all other control options. The results at Canyonlands National Park are plotted in Figure 5-4 and tabulated in Table D-2. Even though the overall impacts at Petrified Forest and Canyonlands are different in magnitude, the visibility improvement results respond similarly to the proposed BART control options. Specifically, the visibility is expected to improve with Option 1 controls and degrade with Option 2 controls. Options 3 and 4 controls are predicted to improve visibility only slightly and at a very high cost. Moreover, those two options also would result in NO x emission rates far below the EPA-established presumptive NO x emission rate limit for the NGS units of 0.28 lb/mmbtu. Unlike Options 1 and 1-alternative -- which, as discussed above, would also meet the presumptive NO x limit -- Options 2, 3, and 4 would all require use of post-combustion controls that EPA specifically determined would not be used as the basis for presumptive NO x limits for coal-fired utility boilers (other than cyclone units). See 70 Fed. Reg. at 39134, (July 6, 2005). 5-2 November 2007

43 Table 5-1: Regional Haze Benefit of BART Controls 5-3 November 2007

44 Figure 5-1: 8 th Highest Regional Haze Total Impacts Averaged Over 3 Years for Baseline and BART Control Case Emissions Baseline Option 1: SOFA/LNB on Units 1-3 (0.24 lb NOx/MMBtu) Option 1-alt: SOFA/LNB on Units 1-3 (0.20 lb NOx/MMBtu) Option 2: SOFA/LNB/SNCR on Units 1-3 Option 3: SOFA/LNB/SCR on Units 1&3 and SOFA/LNB on Unit 2 Option 4: SOFA/LNB/SCR on Units 1-3 8th Highest delta-deciview Bryce Canyon Canyonlands Capitol Reef Grand Canyon Mesa Verde Petrified Forest Sycamore Canyon Class I Area Zion 5-4 November 2007

45 Table 5-2: Regional Haze Results of BART Controls for Each Year and Class I Areas 5-5 November 2007

46 Table 5-3: Annual Costs vs. Visibility Improvements/Degradation (Average of the Eight Class I Areas) Options BART Controls 8th Highest Ave over 3 Years and 8 Class I Areas (total deltadeciview) Annualized Cost ($/Year) Incremental Cost from Previous Control Scenario ($/Year) Incremental Cost Effectiveness Relative to the Previous Control Scenario ($/deciview)) Baseline Option 1 Alternative Option 1 None LNB/SOFA on Units 1-3 Optimal LNB/SOFA on Units $0 $0 $ $2,831,700 $2,831,700 $10,031, $2,831,700 $0 $0 Option 2 LNB/SOFA/SNCR on Units $15,141,400 $12,309,700 Not effective; visibility does not improve from baseline Option 3 LNB/SOFA/SCR on Units 1&3 and SOFA/LNB on Unit $45,694,440 $30,553,040 $77,349,468 Option 4 LNB/SOFA/SCR on Units $74,771,400 $29,076,960 $212,369, November 2007

47 Figure 5-2: Annual Costs vs. Visibility Improvements/Degradation (Average of Eight Class I Areas) $80,000,000 $70,000,000 Option 4 $60,000,000 Cost ($/year) $50,000,000 $40,000,000 Option 3 $30,000,000 $20,000,000 Option 2 $10,000,000 Baseline $ Option 1 Option 1-alt th Highest delta-dv (Average of 3-Years and 8 Class I Areas) November 2007

48 Figure 5-3: Annual Costs vs. Visibility Improvements/Degradations at Petrified Forest NP $80,000,000 $70,000,000 Option 4 $60,000,000 $50,000,000 Cost ($/year) $40,000,000 $30,000,000 Option 3 $20,000,000 Option 2 $10,000,000 $ Baseline Option Option 1-alt th Highest delta-dv (3-Year Average) 5-8 November 2007

49 Figure 5-4: Annual Costs vs. Visibility Improvements/Degradations at Canyonlands NP $80,000,000 $70,000,000 Option 4 $60,000,000 $50,000,000 Cost ($/year) $40,000,000 $30,000,000 Option 3 $20,000,000 Option 2 $10,000,000 $ Baseline Option Option 1-alt th Highest delta-dv (3-Year Average) 5-9 November 2007

50 6.0 BART Recommendations ENSR and SRP recommend LNB/SOFA for all three units as the Best Available Retrofit Technology alternative for the Navajo Generating Station based on the expected incremental visibility improvement, the cost of compliance, energy impacts, and non-air quality environmental impacts. SNCR and SCR were rejected due to the small (if any) incremental improvement in visibility over that resulting from use of LNB/SOFA; increased emissions of ammonia, ammonia salts, and sulfuric acid aerosols, which negatively impact local plume visibility; the high capital and O&M compliance costs for these technologies; increased generation requirements; and other considerations associated with the shipping, storage, and injection of anhydrous or aqueous ammonia. A summary of the BART controls is presented in Table 6-1. Table 6-1: Recommended Best Available Retrofit Technology Low NO x Burners / Separated Overfire Air Estimated NO x Emission Rate 0.24 lbs/mmbtu, with possible reduction to 0.20 lb/mmbtu with experience and optimization of operations NO x Emission Limit Reduction from Baseline Estimated NO x Reduction (TPY) at 85% utilization 48% at 0.24 lb/mmbtu 22,675 tons per year at 0.24 lb/mmbtu Costs of Compliance $125/ton of NO x removed $30 million capital cost $2.8 million/year annualized cost Energy Impacts Slight increase in fuel consumption associated with reduced boiler efficiency Non-Air Quality Environmental Impacts Increased LOI of fly ash, which could reduce recycling sales Slight increase in CO 2 emissions/kwh associated with reduced unit efficiency Potential for poor combustion and increased CO emissions Potential for increased corrosion and more frequent replacement of furnace waterwall tubes Modeled Visibility Impacts 0.38 deciview improvement at Canyonlands (largest Class I area improvement) 0.28 deciview improvement (8-park average) ENSR conducted CALPUFF modeling of the baseline emissions and four NO x control emission scenarios, all involving NO x emission reductions and any associated increased emissions of H 2 SO 4 and ammonia. The BART control options that were analyzed include the retrofit of advanced post-combustion NO x controls, i.e., SNCR and SCR, even though EPA determined that these kinds of controls are more stringent than those that may be needed to meet the BART presumptive limits for boilers of the type at NGS. Each control 6-1 November 2007

51 option was modeled to assess visibility changes at eleven Class I areas within 300 km of NGS, while the primary visibility improvement/degradation results are reported for the closest eight Class I areas. The 98 th percentile prediction results indicate that the installation of the post-combustion NO x controls in addition to LNB/OFA would have a detrimental, or at best an insignificant, effect on visibility at the two closest and most heavily impacted Class I areas: Grand Canyon and Capitol Reef National Parks. Indeed, at the closest Class I area -- Grand Canyon National Park -- the selected option produces decidedly better visibility than would Options 2, 3, or 4; Option 2 would worsen visibility at that park, and Options 3 and 4 would produce virtually no change from baseline conditions. At other Class I areas, the improvement in visibility (i.e., from Option 1a to Option 3 or Option 4) as depicted by the 98 th percentile result is always less than the 0.5-deciview contribution (half the perceptible limit) that EPA describes in the BART rule. It is also evident from the cost information reported in Section 5 that the $/deciview cost for post-combustion controls is extremely high (many tens of millions of dollars per deciview) for those areas for which any improvement in visibility is projected. Therefore, post-combustion NO x emission controls are relatively ineffective in this case, especially taking into account the extraordinarily high cost of those controls. Consequently, none of the post-combustion-control options can be recommended as BART for the NGS units. Finally, as noted above, the use of combustion controls for Navajo Units 1-3 is estimated to produce a NO x emission rate (0.24 lb/mmbtu), which exceeds the requirements of the BART presumptive limit. 6-2 November 2007

52 7.0 References Atkinson, D. and T. Fox Dispersion Coefficients for Regulatory Air Quality Modeling in CLAPUFF. Memorandum from U.S. EPA/OAQPS to Kay T. Prince, EPA Region 4. March 16. Battye, W. and K. Boyer Catalog of Global Emissions Inventories and Emission Inventory Tools for Black Carbon, Table 6, USEPA Contract No. 68-D , available at _resource_1.pdf. E.H. Pechan and Associates, Inc Best Available Retrofit (BART) Analysis for the Navajo Generating Station in Page, Arizona, United States Environmental Protection Agency. EPA Contract # , January 31, Electric Power Research Institute Estimating Total Sulfuric Acid Emissions from Stationary Power Plants. Technical Update, March, EPA CFR Part 51, FRL , RIN: 2060-AJ31, Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations (Appendix Y), updated June 24, 2005 EPA. 2003a. Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Program, EPA-454/B September EPA. 2003b. Guidance for Tracking Progress Under the Regional Haze Rule, EPA-454/B , Appendix A, Table A-3, September EPA Interagency Workgroup on Air Quality Modeling (IWAQM) Phase 2 Summary Report and Recommendations for Modeling Long Range Transport Impacts, EPA-454/R , page 14, December EPA AP 42, Fifth Edition, Compilation of Air Pollutant Emission Factors, Volume 1: Stationary Point and Area Sources, January 1995 National Research Council, Haze in the Canyon: An Evaluation of the Winter Haze Intensive Tracer Experiment, Committee on Haze in National Parks and Wilderness Areas, National Research Council, National Academy Press, Washington, D.C., October Paise, J.W. 2006a. Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations. Memorandum to Kay Prince, Branch Chief EPA Region 4. Attachment A to April 20, 2006 DC Circuit Court document UARG vs. EPA, No Paise, J.W. 2006b. Letter to Mel S. Schulze, Esq., Hunton and Williams representing the Utility Air Regulatory Group (UARG). Attachment B to April 20, 2006 DC Circuit Court document UARG vs. EPA, No Personal correspondence between John Notar of the NPS and Bob Paine of ENSR, November 28, Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations; Final Rule (FR Vol. 70, No. 128 published July 6, 2005). Regional Haze Regulations; Revisions to Provisions Governing Alternative to Source-Specific Best Available Retrofit Technology (BART) Determinations; Final Rule (FR Vol. 71, NO. 198 published October 13, 2006). Western Regional Air Partnership (WRAP) CALMET/CALPUFF Protocol for BART Exemption Screening Analysis for Class I Areas in the Western United States. August 15, November 2007

53 Appendix A CALMET/CALPUFF Processing Refinements November 2007

54 CALMET meteorological inputs, technical options, and processing steps used in this BART analysis were identical to those specified in the WRAP common BART modeling protocol with the exception of only R1, R2, and RMAX1, and the model version. These differences are illustrated in Figures A-1 through A-3 and listed in Table A-1, and are further discussed below. Figure A-1 shows the CALMET/CALPUFF modeling domain established by the WRAP for Arizona. Figure A-1: WRAP CALMET Modeling Domain for Arizona Enhancements to the WRAP CALMET Database ENSR made two refinements to the 4-km Arizona CALMET WRAP database. They are as follows: 1. Weighting Factors for Modifying the Step 1 Wind Field. The 4-km Arizona CALMET database has been produced by ENSR using the downloaded CALMET inputs from the WRAP website ENSR initially ran CALMET with the setting suggested in the WRAP BART modeling protocol. As part of ENSR s internal quality assurance procedure, we displayed and examined the 4-km Arizona WRAP CALMET wind fields in the visualization software CALDESK. Figure A-2 graphically shows wind fields with the WRAP settings for a typical hour. Arrows represent wind direction and wind speed for that hour at 10 meter height. Circular areas in these figures with common winds and abrupt transitions at the edge of the circles indicate a radius of influence of surface stations, R1, which was set to 100 km, as suggested in the WRAP BART protocol. The R1 value was coupled with R1MAX = 50 km so that the influence of the surface stations is established out to 50 km and then it abruptly ends beyond that distance. Setting R1 and R1MAX to such high values is not recommended by the model developer and Federal Land Managers, especially with MM5 data resolution of 36 km with areas of complex terrain. Typically, R1 is set to a fairly small value, generally not exceeding half of the MM5 data resolution (18 km), according to recent guidance on multiple PSD projects involving CALPUFF modeling in the WRAP region from John Notar of the November 2007

55 National Park Service (personal correspondence between John Notar of the NPS and Bob Paine of ENSR). A large R1 value results in wind fields surrounding surface stations that overwrite the MM5 wind fields, which do have terrain influences incorporated into them. In many instances, the extended extrapolation of the surface station data with an abrupt transition at 50 km produces opposing wind directions in adjacent grid squares at the 50 km distance. Figure A-2: CALMET Windfields with WRAP Settings To avoid this problematic wind field result, ENSR used a smaller R1 value of 18 km and R1MAX value of 30 km. The resulting wind fields for the same hour and height are depicted in Figure A-3. The adjusted R1 and R1MAX values blend the surface observations into the MM5 observations much better, creating a more uniform wind field throughout the domain. Therefore, ENSR used the smaller R1 and R1MAX values to be more consistent with FLM guidance and due to the better performance in the wind field depiction associated with the smaller values. November 2007

56 Figure A-3: CALMET Windfields with ENSR Settings 2. Official EPA CALPUFF Version. When rerunning CALMET, ENSR used the latest EPA-approved version of the CALPUFF modeling system CALMET (Version 5.8, Level ) instead of Version that was used by WRAP, available at CALPUFF version 6 is basically equivalent to the VISTAS version of CALPUFF, Version At the time of the WRAP BART protocol development process, the VISTAS version and Version 6 were generally acknowledged to be the latest and best versions available. However, EPA s deliberate attempt to review the nature of the changes between the previous official version (5.711a) and the VISTAS version (and Version 6) uncovered a number of issues that were of concern to EPA. These issues were discussed in a presentation by Mr. Dennis Atkinson of EPA s Office of Air Quality Planning and Standards at the 2007 annual modelers workshop (see CALPUFF_status_update.pdf ). The basic issues of concern with the VISTAS version (and equivalent Version 6) are as follows: There were unexplained and unresolved large differences between Versions 5.711a and Incomplete model documentation has been a problem with the last model users guides now 7 years old. November 2007

57 The VISTAS code changes went beyond just fixing coding errors in Version 5.711a, contrary to what TRC, the model developer, asserted. EPA s annotated in-code documentation identified several categories of changes, including: Bug fixes Non-optional technical enhancements Optional technical enhancements Non-technical enhancements Enhancement adjustments Coordinate conversion fixes EPA had serious technical concerns regarding how the optional technical enhancements, e.g., for mixing height, were implemented in CALMET The new approved Version 5.8 disables some of the VISTAS optional technical enhancements. Therefore, use of Version or Version 6 of CALPUFF would appear to be inconsistent with the current EPA approved version. Default values of technical options specified in the newly approved version are adopted by ENSR. Table A-1: CALMET Options Comparison Variable Description WRAP Value ENSR Value RMAX1 R1 R2 Maximum radius of influence over land in the surface layer Relative weighting of the first-guess field and observations in the surface layer Relative weighting of the first-guess field and observations in the layers aloft Background Ammonia Values The POSTUTIL utility program was used to repartition HNO 3 and NO 3 using appropriate ammonia background values that were approved by the Federal Land managers for the nearby Desert Rock Energy Facility (DREF) PSD permit application. For that project, located nearby in northwestern New Mexico, it was realized that the likely over-prediction by CALPUFF of nitrates in winter can be partially addressed by using a monthly variation of background ammonia concentrations. The default value of 1.0 ppb for arid lands as referenced in the IWAQM Phase 2 document is valid at 20 deg C, but the same document cites a strong dependence with ambient temperature, with variations of a factor of 3-4. This same dependence is seen at the CASTNET monitor at Bondville, Illinois (see page 5 at In addition, a study of light-affecting particles in SW Wyoming indicated that nitrates were over-predicted by a factor of 3 for a constant ammonia concentration of 1.0 ppb, and by a factor of 2 for an ammonia concentration of 0.5 ppb (see slide 57 at Since there are no large sources of ammonia due to agricultural activities near the Class I areas being analyzed (see Figure 1 in it is appropriate to introduce a monthly varying ammonia background concentration to the CALPUFF modeling. Table A-2 lists the values that were used in CALPUFF and have been agreed to by the National Park Service for DREF and other PSD submittals. Note that these values were used only for modeling the baseline and BART Option 1 (and November 2007

58 alternate Option 1) emissions. A refined set of ammonia background values was developed for modeling BART Option 2, 3, and 4 and further discussed in Appendix B. These proposed values are consistent with the CMAQ modeled values provided in Appendix A of Table A-2: Ambient Ammonia Background Concentration Month January February March April May September October November December Ambient Ammonia Background Concentration (ppb) November 2007

59 Appendix B Factors Influencing NOx Emissions Effects on Visibility - November 2007

60 Secondary pollutants such as nitrates and sulfates are significant contributors to the visibility extinction in Class I areas. The CALPUFF model was used to determine the effect of these pollutants on Class I areas, associated with BART control options. CALPUFF uses the EPA-approved MESOPUFF II chemical reaction mechanism to convert SO 2 and NOx emissions to secondary sulfates and nitrates. The discussion below describes how the secondary pollutants are formed and the factors affecting their formation. Formation of Sulfates The rate of transformation of gaseous SO 2 to ammonium sulfate (NH 4 ) 2 SO 4 aerosol is dependent upon solar radiation, ambient ozone concentration, atmospheric stability, and relative humidity, as shown in Figure B-1 (taken from the CALPUFF users guide, 2000). Homogeneous gas phase reaction is the dominant SO 2 oxidation pathway during clear, dry conditions (Calvert et al., 1978). CALPUFF assumes that the sulfate reacts preferentially with ammonia (NH 3 ) to form ammonium sulfate and that any remaining ammonia is available to form ammonium nitrate (NH 4 NO 3 ). Figure B-1: MESOPUFF II SO 2 Oxidation Formation of Nitrates The oxidation of NOx to nitric acid (HNO 3 ) depends on the NOx concentration, ambient ozone concentration, and atmospheric stability. Some of the nitric acid is then combined with available ammonia in the atmosphere to form ammonium nitrate aerosol in an equilibrium state that is a function of temperature, relative humidity, and ambient ammonia concentration, as shown in Figure B-2 (from the CALPUFF users guide). Figure B-2 MESOPUFF II NOx Oxidation In CALPUFF, total nitrate (TNO 3 =HNO 3 + NO 3 ) is partitioned into each species according to the equilibrium relationship between gaseous HNO 3 and NO 3 aerosol. This equilibrium is a function of ambient temperature and relative humidity. Moreover, the formation of nitrate strongly depends on availability of NH 3 to form ammonium November 2007

61 nitrate, as shown in Figure B-3 (from CALPUFF courses given by TRC). The figure on the left shows that with 1 ppb of available ammonia and fixed temperature and humidity (for example, 275 deg K and 80% humidity), only 50% of the total nitrate forms particulate matter. When the available ammonia is increased to 2 ppb, as shown in the figure on the right, as much as 80% of the total nitrate is in the particulate form. Figure B-3 also shows that colder temperatures and higher relative humidity significantly favor nitrate formation and vice versa. A summary of the conditions affecting nitrate formation are listed below: Colder temperature and higher relative humidity create favorable conditions to form nitrate particulate matter, and therefore more ammonium nitrate is formed; Warm temperatures and lower relative humidity create less favorable conditions to form nitrate particulate matter, and therefore less ammonium nitrate is formed; Sulfate preferentially scavenges ammonia over nitrates. In areas where sulfate concentrations are high and ambient ammonia concentrations are low, there is less ammonia available to react with nitrate, and therefore less ammonium nitrate is formed. For this BART analysis, the effects of temperature and background ammonia concentrations on the nitrate formation are the key to understanding the effects of various NOx control options. For parts of the country where sulfate concentrations are relatively high and ammonia emissions are quite low, the atmosphere is likely to be in an ammonia-limited regime relative to nitrate formation. Therefore, NOx emission controls are not very effective in improving regional haze, especially if there is very little ambient ammonia available. Figure B-3: NO 3 /HNO 3 Equilibrium Dependency on Temperature and Humidity Refined Ambient Ammonia Background Concentrations As discussed above, the formation of nitrate is highly sensitive to availability of ammonia to form ammonium nitrate. Ammonium nitrate is a visibility-degrading pollutant. For the purpose of evaluating NOx emissions control options, the ambient ammonia background concentrations were refined to factor in excess ammonia emission increases associated with SNCR and SCR operations. Moreover, the installation of SCR creates primary sulfate emissions (H 2 SO 4 ) that are also visibility-degrading. Excess ammonia emissions associated with SNCR and SCR operations were modeled in CALPUFF to determine the maximum 24-hour ammonia concentration at Grand Canyon National Park as well as the other Class I areas. Predicted excess ammonia concentrations associated with SNCR and SCR operation are listed in Table B-1. For simplicity in the post-processing, the predicted values of additional ambient ammonia concentrations were allocated to several values representing typical values covering the range of the CALPUFF predictions. November 2007

62 The resultant ammonia concentrations for the peak daily impact at the Class I areas (corresponding to a peak regional haze event) produced in μg/m 3 were converted to ppb and then added to the monthly ambient background values, as shown in Table B-1. Then POSTUTIL program (CALPUFF post-processor) was used to re-compute regional haze impacts with the adjusted ammonia background at each Class I areas. Table B-1: Refined Ambient Ammonia Background Concentration November 2007

63 Appendix C Review of Data from the IMPROVE Monitoring Network November 2007

64 The Visibility Information Exchange Web System (VIEWS) is an online database of air quality data designed to understand the effects of air pollution on visibility and to support the Regional Haze Rule enacted by the USEPA to reduce regional haze and improve visibility in national parks and wilderness areas ( The VIEWS database contains annual summary of Class I area-specific charts of visibility-degrading pollutants. Bar charts depict seasonal pattern of pollution and pie charts show the average composition for the 20% best and 20% worst pollution days. An example of a bar and pie chart for Canyonlands National Park (which also represents Arches National Park) is shown in Figure C-1. Bar and pie charts for the modeled eleven Class I areas for year 2002 are presented in this appendix. Year 2002 was chosen because it is the year for which WRAP has established the baseline emissions inventory. Figure C-1: Plot of Measured Visibility-Degrading Pollutants in Arches and Canyonlands NP, Year 2002 Figure C-1 is typical of the composition of visibility-affecting particulate that is shown in the plots for other Class I areas provided below. The figure shows that for 2002, organic aerosols (probably associated with forest fires for November 2007

65 peak impacts) contributed about 40% and coarse particulate matter (due to wind-blown dust) contributed about 22% on the worst 20% days to the visibility extinction at Canyonlands National Park. On the other hand, ammonium nitrate contributed only 8% and ammonium sulfate contributed 14% to the visibility extinction at the park. It is important to note that the nitrate impacts were virtually nonexistent during the warm period of April-October (during the period of the heaviest park visitation), while sulfate impacts were generally present throughout the entire year. In fact, very few of the worst 20% days (marked with a W ) have substantial nitrate contributions. Therefore, reduction of NOx emissions would do very little to improve the visibility for this set of days that is specifically targeted by the Regional Haze Rule. On the other hand, several of the 20% worst days have some sulfate component, which would be increased by certain NOx controls, such as SCR, due to the collateral increases in sulfate and ammonia emissions associated with these controls. This overall pattern is generally present in all of the nearby Class I areas, as can be seen in the composition plots shown below. Figure C-2: Plot of Measured Visibility-Degrading Pollutants in Bryce Canyon NP, Year 2002 November 2007

66 Figure C-3: Plot of Measured Visibility-Degrading Pollutants in Capitol Reef NP, Year 2002 Pie chart for Capitol Reef NP was not available. November 2007

67 Figure C-4: Plot of Measured Visibility-Degrading Pollutants in Grand Canyon NP, Year 2002 November 2007

68 Figure C-5: Plot of Measured Visibility-Degrading Pollutants in Mazatzal W and Pine Mountain W, Year 2002 November 2007

69 Figure C-6 Plot of Measured Visibility-Degrading Pollutants in Mesa Verde NP, Year 2002 November 2007

70 Figure C-7: Plot of Measured Visibility-Degrading Pollutants in Petrified Forest NP, Year 2002 November 2007

71 Figure C-8: Plot of Measured Visibility-Degrading Pollutants in Sycamore Canyon W, Year 2002 November 2007

72 Figure C-9: Plot of Measured Visibility-Degrading Pollutants in Zion NP, Year 2002 November 2007