BEST AVAILABLE RETROFIT TECHNOLOGY (BART) DETERMINATION AMERICAN ELECTRIC POWER NORTHEASTERN POWER PLANT

Transcription

1 BEST AVAILABLE RETROFIT TECHNOLOGY (BART) DETERMINATION AMERICAN ELECTRIC POWER NORTHEASTERN POWER PLANT Prepared by: N.N. Dharmarajan AEP Eugene Chen, P.E. Senior Consultant Jeremy Townley Consultant TRINITY CONSULTANTS 120 East Sheridan Suite 205 Oklahoma City, OK (405) May 30, 2008 Project

2 TABLE OF CONTENTS 1. EXECUTIVE SUMMARY BACKGROUND BEST AVAILABLE RETROFIT TECHNOLOGY RULE MODELING PROTOCOL BACKGROUND LOCATION OF SOURCES AND RELEVANT CLASS I AREAS CALPUFF MODEL SYSTEM MODEL VERSIONS MODELING DOMAIN CALMET GEOPHYSICAL DATA TERRAIN DATA LAND USE DATA COMPILING TERRAIN AND LAND USE DATA METEOROLOGICAL DATA MESOSCALE MODEL METEOROLOGICAL DATA SURFACE METEOROLOGICAL DATA UPPER AIR METEOROLOGICAL DATA PRECIPITATION METEOROLOGICAL DATA BUOY METEOROLOGICAL DATA CALMET CONTROL PARAMETERS VERTICAL METEOROLOGICAL PROFILE INFLUENCES OF OBSERVATIONS CALPUFF SOURCE EMISSIONS RECEPTOR LOCATIONS BACKGROUND OZONE AND AMMONIA CALPUFF MODEL CONTROL PARAMETERS CALPOST CALPOST LIGHT EXTINCTION ALGORITHM CALPOST PROCESSING METHOD NATURAL BACKGROUND EVALUATING VISIBILITY RESULTS SUMMARY OF CALPOST CONTROL PARAMETERS American Electric Power i Trinity Consultants Northeastern Power Station

3 7. BART-ELIGIBLE EMISSION UNITS DESCRIPTION OF BART-ELIGIBLE SOURCES MODELED STACK PARAMETERS BART DETERMINATION ANALYSIS FOR NO X STEP 1 - IDENTIFY ALL AVAILABLE RETROFIT CONTROL TECHNOLOGIES COMBUSTION MODIFICATIONS Flue Gas Recirculation Low NO X Burners and Overfire Air Reburning / Methane de-no X POST-COMBUSTION CONTROLS Selective Non-Catalytic Reduction Selective Catalytic Reduction STEP 2 ELIMINATE TECHNICALLY INFEASIBLE OPTIONS FLUE GAS RECIRCULATION REBURNING / METHANE DE-NO X STEP 3 EVALUATE CONTROL EFFECTIVENESS OF REMAINING CONTROL TECHNOLOGIES STEP 4 EVALUATE IMPACTS AND DOCUMENT RESULTS COSTS OF COMPLIANCE ENERGY IMPACTS & NON-AIR IMPACTS REMAINING USEFUL LIFE STEP 5 EVALUATE VISIBILITY IMPACTS BART DETERMINATION FOR NO X BART DETERMINATION ANALYSIS FOR SO STEP 1 - IDENTIFY ALL AVAILABLE RETROFIT CONTROL TECHNOLOGIES DRY FLUE GAS DESULFURIZATION WET FLUE GAS DESULFURIZATION STEP 2 ELIMINATE TECHNICALLY INFEASIBLE OPTIONS STEP 3 EVALUATE CONTROL EFFECTIVENESS OF REMAINING CONTROL TECHNOLOGIES STEP 4 EVALUATE IMPACTS AND DOCUMENT RESULTS COST OF COMPLIANCE ENERGY IMPACTS AND NON-AIR QUALITY IMPACTS REMAINING USEFUL LIFE STEP 5 EVALUATE VISIBILITY IMPACTS BART DETERMINATION FOR SO APPENDIX A- METEOROLOGICAL STATIONS APPENDIX B SAMPLE CALMET CONTROL FILE American Electric Power ii Trinity Consultants Northeastern Power Station

4 APPENDIX C SAMPLE CALPUFF CONTROL FILE APPENDIX D SAMPLE CALPOST CONTROL FILE APPENDIX E CONTROL SCENARIO EMISSION ESTIMATES APPENDIX F CONTROL COST CALCULATIONS APPENDIX G CALPUFF SYSTEM MODEL FILES American Electric Power iii Trinity Consultants Northeastern Power Station

5 LIST OF TABLES TABLE 1-1. NORTHEASTERN STATION BART DETERMINATION SUMMARY TABLE 2-1. BART-ELIGIBLE SOURCES TABLE 2-2. DISTANCE (KM) FROM STATION TO SURROUNDING CLASS I AREAS TABLE 3-1 CALPUFF MODELING SYSTEM VERSIONS TABLE 4-1. VERTICAL LAYERS OF THE CALMET METEOROLOGICAL DOMAIN TABLE 6-1. MONTHLY HUMIDITY FACTORS TABLE 6-2. DEFAULT AVERAGE ANNUAL NATURAL BACKGROUND LEVELS TABLE 7-1. SUMMARY OF BART-ELIGIBLE EMISSION UNITS TABLE 7-2. STACK PARAMETERS TABLE 8-1. AVAILABLE NO X CONTROL TECHNOLOGIES TABLE 8-2. RANKING OF CONTROL TECHNOLOGIES TABLE 8-3. NO X CONTROL COSTS SUMMARY TABLE 8-4. NE STATION NO X VISIBILITY RESULTS TABLE 9-1. RANKING OF CONTROL STRATEGIES TABLE 9-2. SO 2 CONTROLS COSTS SUMMARY TABLE 9-3. NE STATION SO 2 VISIBILITY RESULTS TABLE A-1. LIST OF SURFACE METEOROLOGICAL STATIONS... 1 TABLE A-2. LIST OF UPPER AIR METEOROLOGICAL STATIONS... 5 TABLE A-3. LIST OF PRECIPITATION METEOROLOGICAL STATIONS... 6 TABLE A-4. LIST OF OVER WATER METEOROLOGICAL STATIONS American Electric Power iv Trinity Consultants Northeastern Power Station

6 LIST OF FIGURES FIGURE 2-1. PLOT OF SOURCES AND NEAREST CLASS I AREAS FIGURE 3-1. REFINED METEOROLOGICAL MODELING DOMAIN FIGURE 4-1. PLOT OF LAND ELEVATION USING USGS TERRAIN DATA FIGURE 4-2. PLOT OF LAND USE USING USGS LULC DATA FIGURE 4-3. PLOT OF SURFACE STATION LOCATIONS FIGURE 4-4. PLOT OF UPPER AIR STATIONS LOCATIONS FIGURE 4-5. PLOT OF PRECIPITATION METEOROLOGICAL STATIONS FIGURE 4-6. PLOT OF BUOY METEOROLOGICAL STATIONS FIGURE 8-1. PRIMARY SNCR REACTION SEQUENCES American Electric Power v Trinity Consultants Northeastern Power Station

7 1. EXECUTIVE SUMMARY American Electric Power /Public Service Company of Oklahoma (AEP/PSO) operates the Northeastern (NE) Power Station, which is located at Section 4, T22N, R15E, in Rogers County, Oklahoma. The NE Power Station is currently operating in accordance with Oklahoma Department of Environmental Quality (DEQ) Title V Operating Permit, TV, issued in June 30, The NE Power Station is considered eligible for the application of Best Available Retrofit Technology (BART) 1 as part of the Environmental Protection Agency (EPA) Regional Haze Rule. On March 30, 2007, AEP submitted to the Oklahoma DEQ, the BART determination for its BART eligible facilities in Oklahoma, in accordance with OAC 252:100, Subchapter 8, Permits for Part 70 Sources, Part 11, Visibility Protection Standards, effective October 8, Specifically, AEP indicated that it would: For its gas-fired NE Unit 2, meet the BART requirement to satisfy the 40CFR Part 51, Regional Haze regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations; Final Rule, in effect on July 2005, of current combustion control technology for NOx control, as specified in these rules For its solid-fuel fired NE Units 3 & 4, it would meet the presumptive limits for SO 2 and NOx (as applicable to plants greater than 750 MW for the type of fuel and boiler), in accordance with the same 40CFR Part 51 rules cited above. For SO 2 control, the presumptive limit of 0.15 Lbs/MMBtu would be met on a 30-day rolling average using Flue Gas Desulfurization (FGD) and for NOx control, the presumptive limit of 0.15 Lbs/MMBtu would be met on a 30-day rolling average using Low NOx burners/overfire air. The electrostatic precipitators would be BART for PM based on the modeling and engineering analysis performed and submitted to the Oklahoma DEQ. In a letter reply dated December 12, 2007, from Mr. Eddie Terrill, the Oklahoma DEQ requested a formal five-factor analysis be performed in order to determine BART for NO X and SO 2 for NE units 3 & 4. This report summarizes the result of the five factor analysis performed for SO 2 and NOx for the BART-eligible Units 3 & 4 at the Northeastern Station. The message to be gleaned from the discussions in these sections, is that the high implementation costs to gain emissions reductions with minimal 98 th percentile perceptible visibility benefits would seem to suggest that BART for these units be the existing low NOx burners for NOx and low sulfur fuel for SO 2. As summarized in Table 1-1 below, AEP proposes the following measures as BART: 1 40 CFR 51 Appendix Y, Guidelines for BART Determinations Under the Regional Haze Rule American Electric Power 1-1 Trinity Consultants Northeastern Power Station

8 TABLE 1-1. NORTHEASTERN STATION BART DETERMINATION SUMMARY Unit ID NO X BART Determination SO 2 BART Determination Northeastern Unit 3 Low-NO X Burners Low Sulfur coal Northeastern Unit 4 Low- NO X Burners Low Sulfur coal The five factors considered in determining BART described in this report are: Cost of compliance Energy impacts Non-air quality impacts The remaining useful life of the source; and Visibility impacts Evaluation of the fifth factor, visibility impacts, requires the use of the CALPUFF modeling system. A substantial portion of this document (Sections 3-6) is devoted to describing the modeling methodology used in determining potential visibility impacts from the NE Station s BART-eligible units 3 & 4: Section 2 Background Section 3 CALPUFF Modeling System Section 4 CALMET Section 5 CALPUFF Section 6 CALPOST Section 7 BART-eligible Units Section 8 NO X BART Determination Section 9 SO 2 BART Determination The CALMET data set described in Section 4 was developed in adherence to the CALMET processing protocol approved by the Oklahoma DEQ on March 10, 2008, for a similar BARTdetermination related modeling effort. CALPUFF and CALPOST settings described in Section 5 and 6 (sample files of which are included in the Appendices) have not been reviewed by the state but adhere, where practical, to the modeling settings and procedures established in CENRAP s BART Modeling Guidelines. 2 Sections 7-9 address the five-factor analysis for NO X and SO 2, and include an analysis of compliance costs and visibility impacts of various control technologies: 2 Released December American Electric Power 1-2 Trinity Consultants Northeastern Power Station

9 2. BACKGROUND 2.1 BEST AVAILABLE RETROFIT TECHNOLOGY RULE On July 1, 1999, the U.S. Environmental EPA published the final Regional Haze Rule (RHR). The stated goal of the RHR is to restore visibility in 156 specific areas known as Class I areas, to natural background levels, over a 60-year period, across the United States. The Clean Air Act defines Class I areas as certain national parks (over 6000 acres), wilderness areas (over 5000 acres), national memorial parks (over 5000 acres), and international parks that were in existence on August 7, On July 6, 2005, the EPA published amendments to its 1999 RHR, (also known as the BART rule) that included guidance for making source-specific Best Available Retrofit Technology (BART) determinations. The BART rule defines BART-eligible sources as sources that meet the following criteria: (1) Have potential emissions of at least 250 tons per year of a visibility-impairing pollutant, (2) Began operation between August 7, 1962 and August 7, 1977, and (3) Are listed as one of the 26 listed source categories in the guidance. A BART-eligible source is not automatically subject to BART. Rather, BART-eligible sources are subject-to-bart if the sources are reasonably anticipated to cause or contribute to visibility impairment in any federal mandatory Class I area. EPA has determined that sources are reasonably anticipated to cause or contribute to visibility impairment if the visibility impacts from a source are greater than 0.5 deciviews (dv) when compared against a natural background. Air quality modeling is the tool that is used to determine a source s visibility impacts. States have the authority to exempt certain BART-eligible sources from installing BART controls if the results of the dispersion modeling demonstrate that the source cannot reasonably be anticipated to cause or contribute to visibility impairment in a Class I area. Further, States also have the authority to define the modeling procedures for conducting modeling related to making BART determinations. 2.2 MODELING PROTOCOL BACKGROUND To promote consistency between States in the development of BART modeling protocols and to harmonize the approaches between adjacent RPOs, the Central States Regional Air Planning (CENRAP) organization developed BART Modeling Guidelines (December 15, 2005). The intent of the guidelines was to assist CENRAP States and source operators in the development of statewide and source-specific modeling protocols. American Electric Power 2-1 Trinity Consultants Northeastern Power Station

10 2.3 LOCATION OF SOURCES AND RELEVANT CLASS I AREAS The sources listed in Table 2-1. BART-Eligible Sources are the sources that have been identified by AEP as sources that meet the three criteria for BART-eligible sources at the NE Station. TABLE 2-1. BART-ELIGIBLE SOURCES EPN Unit 2 Unit 3 Unit 4 Description 4,754 MMBtu/hr Gas-fired 4,775 MMBtu/hr Coal Fired Boiler 4,775 MMBtu/hr Coal Fired Boiler As required in CENRAP s BART Modeling Guidelines, Class I areas within 300 km of each station will be included in each analysis. The following tables summarize the distances of the four closest Class I areas to the NE station. As seen from this summary, one Class I area (Wichita Mountains NWR) is more than 300 km from the station, but has been included in the analysis. Note that the distances listed in the tables below are the distances between the stations and the closest border of the Class I areas. TABLE 2-2. DISTANCE (KM) FROM STATION TO SURROUNDING CLASS I AREAS Class I Area Name Distance from Source (km) Caney Creek Wilderness 263 Hercules-Glades Wilderness 244 Upper Buffalo Wilderness 211 Wichita Mountains National Wildlife Refuge 323 A plot of the Class I areas with respect to the station is provided in Figure 2-1. American Electric Power 2-2 Trinity Consultants Northeastern Power Station

11 FIGURE 2-1. PLOT OF SOURCES AND NEAREST CLASS I AREAS WIMO HEGL Northeastern Station UPBU CACR LCC Northing (km) Class I Areas LCC Easting (km) American Electric Power 2-3 Trinity Consultants Northeastern Power Station

12 3. CALPUFF MODEL SYSTEM The main components of the CALPUFF modeling system are CALMET, CALPUFF, and CALPOST. CALMET is the meteorological model that generates hourly three-dimensional meteorological fields such as wind and temperature. CALPUFF simulates the non-steady state transport, dispersion, and chemical transformation of air pollutants emitted from a source in puffs. CALPUFF calculates hourly concentrations of visibility affecting pollutants at each specified receptor in a modeling domain. CALPOST is the post-processor for CALPUFF that computes visibility impacts from a source based on the visibility affecting pollutant concentrations that were output by CALPUFF. 3.1 MODEL VERSIONS The versions of the CALPUFF modeling system programs that were used for conducting AEP s BART modeling are listed in Table 3-1. TABLE 3-1 CALPUFF MODELING SYSTEM VERSIONS Processor Version Level TERREL CTGCOMP CTGPROC MAKEGEO CALMET 5.53a CALPUFF POSTUTIL CALPOST MODELING DOMAIN The CALPUFF modeling system utilizes three modeling grids: the meteorological grid, the computational grid, and the sampling grid. The meteorological grid is the system of grid points at which meteorological fields are developed with CALMET. The computational grid determines the computational area for a CALPUFF run. Puffs are advected and tracked only while within the computational grid. The meteorological grid is defined so that it covers the areas of concern and gives enough marginal buffer area for puff transport and dispersion. A plot of the meteorological modeling domain with respect to the Class I areas being modeled is provided in Figure 3-1. The American Electric Power 3-1 Trinity Consultants Northeastern Power Station

13 computational domain was set to extend at least 50 km in all directions beyond the NE Station and the Class I areas of interest. Note that the map projection for the modeling domain was Lambert Conformal Conic (LCC) and the datum was the World Geodetic System 84 (WGS-84). The reference point for the modeling domain is Latitude 40ºN, Longitude 97ºW. The southwest corner was set to km LCC, km LCC corresponding to Latitude ºN and Longitude ºW. The meteorological grid spacing was 4 km, resulting in 462 grid points in the X direction and 376 grid points in the Y direction. FIGURE 3-1. REFINED METEOROLOGICAL MODELING DOMAIN Meteorological Modeling Domain WIMO HEGL Northeastern Station UPBU CACR LCC Northing (km) Computational Modeling Domain Class I Areas LCC Easting (km) American Electric Power 3-2 Trinity Consultants Northeastern Power Station

14 4. CALMET CALMET is the meteorological processor that compiles meteorological data from raw observations of surface and upper air conditions, precipitation measurements, mesoscale model output, and geophysical parameters into a single hourly, gridded data set for input into CALPUFF. CALMET was used to assimilate data for using National Weather Service (NWS) surface station observations, upper air station observations, precipitation station observations, buoy station observations (for overwater areas), and mesoscale model output to develop the meteorological field. 4.1 GEOPHYSICAL DATA CALMET requires geophysical data to characterize the terrain and land use parameters that potentially affect dispersion. Terrain features affect flows and create turbulence in the atmosphere and are potentially subjected to higher concentrations of elevated puffs. Different land uses exhibit variable characteristics such as surface roughness, albedo, Bowen ratio, and leaf-area index that also effect turbulence and dispersion TERRAIN DATA Terrain data was obtained from the United States Geological Survey (USGS) in 1-degree (1:250,000 scale or approximately 90 meter resolution) digital format. The USGS terrain data was then processed by the TERREL program to generate grid-cell elevation averages across the modeling domain. A plot of the land elevations based on the USGS data for the modeling domain is provided in Figure 4-1. American Electric Power 4-1 Trinity Consultants Northeastern Power Station

15 FIGURE 4-1. PLOT OF LAND ELEVATION USING USGS TERRAIN DATA Northeastern Station LCC Northing (km) LCC Easting(km) 0 Terrain Elevation (m) LAND USE DATA The land use land cover (LULC) data from the USGS North American land cover characteristics data base in the Lambert Azimuthal equal area map projection was used in order to determine the land use within the modeling domain. The LULC data was processed by the CTGPROC program which generated land use for each grid cell across the modeling domain. A plot of the land use based on the USGS data for the modeling domain is provided in Figure 4-2. American Electric Power 4-2 Trinity Consultants Northeastern Power Station

16 FIGURE 4-2. PLOT OF LAND USE USING USGS LULC DATA -200 Northeastern Station ) ( k m r N o t h i n g C L LCC Easting (km) 10 Land Use COMPILING TERRAIN AND LAND USE DATA The terrain data files output by the TERELL program and the LULC files output by the CTGPROC program were uploaded into the MAKEGEO program to create a geophysical data file that was input into CALMET. 4.2 METEOROLOGICAL DATA CALMET was used to assimilate data for 2001, 2002, and 2003 using mesoscale model output and National Weather Service (NWS) surface station observations, upper air station observations, precipitation station observations, and National Oceanic and Atmosphere Administrations (NOAA) buoy station observations to develop the meteorological field MESOSCALE MODEL METEOROLOGICAL DATA Hourly mesoscale data was also used as the initial guess field in developing the CALMET meteorological data. It is AEP s intent to use the following 5 th generation mesoscale model meteorological data sets (or MM5 data) in the analysis: 2001 MM5 data at 12 km resolution generated by the U.S. EPA 2002 MM5 data at 36 km resolution generated by the Iowa DNR American Electric Power 4-3 Trinity Consultants Northeastern Power Station

17 2003 MM5 data set at 36 km resolution generated by the Midwest RPO The specific MM5 data that was used are subsets of the data listed above. As the contractor to CENRAP for developing the meteorological data sets for the BART modeling, Alpine Geophysics extracted three subsets of MM5 data for each year from 2001 to 2003 from the data sets listed above using the CALMM5 extraction program. The three subsets covered the northern, central, and southern portions of CENRAP. AEP is proposing to use the southern set of the extracted MM5 data. The 2001 southern subset of the extracted MM5 data includes 30 files that are broken into 10 to 11 day increments (3 files per month). The 2002 and 2003 southern subsets of extracted MM5 data include 12 files each of which are broken into 30 to 31 day increment files (1 file per month). Note that the 2001 to 2003 MM5 data extracted by Alpine Geophysics was not able to be used directly in the modeling analysis. To run the Alpine Geophysics extracted MM data in the EPA approved CALMET program, each of the MM5 files needed to be adjusted by appending an additional six (6) hours, at a minimum, to the end of each file to account for the shift in time zones from the Greenwich Mean Time (GMT) prepared Alpine Geophysics data to Time Zone 6 for this analysis. No change to the data occurred. The time periods covered by the data in each of the MM5 files extracted by Alpine Geophysics include a specific number of calendar days, where the data starts at Hour 0 in GMT for the first calendar day and ends at Hour 23 in GMT on the last calendar day. In order to run CALMET in the local standard time (LST), which is necessary since the surface meteorological observations are recorded in LST, there must be hours of MM5 data referenced in a CALMET run that match the LST observation hours. Since the LST hours in Central Standard Time (CST) are 6 hours behind GMT, it is necessary to adjust the data in each MM5 file so that the time periods covered in the files match CST. Based on the above discussion, the Alpine Geophysics MM5 data was not used directly. Instead the data files were modified to add 8 additional hours of data to the end of each file from the beginning of the subsequent file. CALMET was then run using the appended MM5 data to generate a contiguous set of CALMET output files. The converted MM5 data files occupy approximately 1.2 terabytes (TB) of hard drive space SURFACE METEOROLOGICAL DATA Parameters affecting turbulent dispersion that are observed hourly at surface stations include wind speed and direction, temperature, cloud cover and ceiling, relative humidity, and precipitation type. AEP used the surface stations listed in Table A-1 of Appendix A. The locations of the surface stations with respect to the modeling domain are shown in Figure 4-3. The stations were selected from the available data inventory to optimize spatial coverage and representation of the domain. Data from the stations was processed for use in CALMET using EPA s SMERGE program. American Electric Power 4-4 Trinity Consultants Northeastern Power Station

18 Missing surface data was filled using procedures recommended by U.S. EPA. 3 Missing data periods of 5 hours or less were replaced using these procedures. For periods greater than 5 hours, data was left either unfilled or was not used in CALMET processing. A large enough quantity of surface stations was included in the domain that overlapping areas of influence allowed data from an alternate station to be used. FIGURE 4-3. PLOT OF SURFACE STATION LOCATIONS HEGL Northeastern Station -600 WIMO CACR UPBU LCC Northing (km) Class I Areas Surface Stations LCC Easting (km) UPPER AIR METEOROLOGICAL DATA Observations of meteorological conditions in the upper atmosphere provide a profile of turbulence from the surface through the depth of the boundary layer in which dispersion occurs. Upper air data are collected by balloons launched simultaneously across the observation network at 0000 Greenwich Mean Time (GMT) (6:00 PM CST in Oklahoma) and 1200 GMT (6:00 AM CST in Oklahoma). Sensors observe pressure, wind speed and direction, and temperature (among other parameters) as the balloon rises through the 3 Procedures for Substituting Values for Missing NWS Meteorological Data for Use in Regulatory Air Quality Models, Dennis Atkinson and Russell F. Lee, July 7, 1992, American Electric Power 4-5 Trinity Consultants Northeastern Power Station

19 atmosphere. The upper air observation network is less dense than surface observation points since upper air conditions vary less and are generally not as affected by local effects (e.g., terrain or water bodies). The upper air stations used for this analysis are listed in Table A-2 of Appendix A. The locations of the upper air stations with respect to the modeling domain are shown in Figure 4-4. These stations were selected from the available data inventory to optimize spatial coverage and representation of the domain. Data from the stations was processed for use in CALMET using EPA s READ62 program. Missing upper air data was replaced using a persistence method- the assumption that data from the preceding or following hours are representative of the missing period. Data from either the preceding or following hours were extrapolated to fill the missing hour. FIGURE 4-4. PLOT OF UPPER AIR STATIONS LOCATIONS HEGL Northeastern Station -600 WIMO CACR UPBU LCC Northing (km) Class I Areas Upper Air Stations LCC Easting (km) PRECIPITATION METEOROLOGICAL DATA The effects of chemical transformation and deposition processes on ambient pollutant concentrations were considered in this analysis. Therefore, it was necessary to include observations of precipitation in the CALMET analysis. The precipitation stations that were used for this analysis are listed in Table A-3 of Appendix A. The locations of the precipitation stations with respect to the modeling domain are shown in Figure 4-5. These American Electric Power 4-6 Trinity Consultants Northeastern Power Station

20 stations were selected from the available data inventory to optimize spatial coverage and representation of the domain. Data from the stations was processed for use in CALMET using EPA s PMERGE program. FIGURE 4-5. PLOT OF PRECIPITATION METEOROLOGICAL STATIONS HEGL Northeastern Station -600 WIMO CACR UPBU LCC Northing (km) Class I Areas Precipitation Stations LCC Easting (km) BUOY METEOROLOGICAL DATA The effects of land/sea breeze on ambient pollutant concentrations were considered in this analysis. Therefore, it is necessary to include observations of buoy stations in the CALMET analysis. The buoy stations that were used for this analysis are listed in Table A-4 of Appendix A. The locations of the buoy stations with respect to the modeling domain are shown in Figure 4-6. These stations were selected from the available data inventory to optimize spatial coverage and representation of the domain along the coastline. Data from the stations was prepared by filling missing hour records with the CALMET missing parameter value (9999). No adjustments to the data occurred. American Electric Power 4-7 Trinity Consultants Northeastern Power Station

21 FIGURE 4-6. PLOT OF BUOY METEOROLOGICAL STATIONS HEGL Northeastern Station -600 WIMO CACR UPBU LCC Northing (km) Class I Areas Buoy Stations LCC Easting (km) 4.3 CALMET CONTROL PARAMETERS Appendix B provides a sample CALMET input file used in AEP s modeling analysis. A few details of the CALMET model setup for sensitive parameters are also discussed below VERTICAL METEOROLOGICAL PROFILE The height of the top vertical layer was set to 3,500 meters. This height corresponds to the top sounding pressure level for which upper air observation data was relied upon. The vertical dimension of the domain were divided into 12 layers with the maximum elevations for each layer shown in Table 4-1. The vertical dimensions are weighted towards the surface to resolve the mixing layer while using a somewhat coarser resolution for the layers aloft. American Electric Power 4-8 Trinity Consultants Northeastern Power Station

22 TABLE 4-1. VERTICAL LAYERS OF THE CALMET METEOROLOGICAL DOMAIN Layer Elevation (m) CALMET allows for a bias value to be applied to each of the vertical layers. The bias settings for each vertical layer determine the relative weight given to the vertically extrapolated surface and upper air wind and temperature observations. The initial guess fields are computed with an inverse distance weighting (1/r 2 ) of the surface and upper air data. The initial guess fields may be modified by a layer dependent bias factor. Values for the bias factor may range from -1 to +1. A bias of -1 eliminates upper-air observations in the 1/r 2 interpolations used to initialize the vertical wind fields. Conversely, a bias of +1 eliminates the surface observations in the interpolations for this layer. Normally, bias is set to zero (0) for each vertical layer, such that the upper air and surface observations are given equal weight in the 1/r 2 interpolations. The biases for each layer of the proposed modeling domain was set to zero. CALMET allows for vertical extrapolation of surface wind observations to layers aloft to be skipped if the surface station is close to the upper air station. Alternatively, CALMET allows data from all surface stations to be extrapolated. The CALMET parameter that controls this setting is IEXTRP. Setting IEXTRP to a value less than zero (0) means that layer 1 data from upper air soundings is ignored in any vertical extrapolations. IEXTRP was set to -4 for this analysis (i.e., the similarity theory is used to extrapolate the surface winds into the layers aloft, which provides more information on observed local effects to the upper layers) INFLUENCES OF OBSERVATIONS Step 1 wind fields were based on an initial guess using MM5 data and refined to reflect terrain affects. Step 2 wind fields adjusted the Step 1 wind field by incorporating the influence of local observations. An inverse distance method is used to determine the influence of observations to the Step 1 wind field. RMAX1 and RMAX2 define the radius of influence for data from surface stations to land in the surface layer and data from upper air stations to land in the layers aloft. In general, RMAX1 and RMAX2 are used to exclude observations from being inappropriately included in the development of the Step 2 American Electric Power 4-9 Trinity Consultants Northeastern Power Station

23 wind field if the distance from an observation station to a grid point exceeds the maximum radius of influence. If the distance from an observation station to a grid point was less than the value set for RMAX, the observation data was used in the development of the Step 2 wind field. R1 represents the distance from a surface observation station at which the surface observation and the Step 1 wind field are weighted equally. R2 represents the comparable distance for winds aloft. R1 and R2 were used to weight the observation data with respect to the MM5 data that was used to generate the Step 1 wind field. Large values for R1 and R2 give more weight to the observations, where as small values give more weight to the MM5 data. In this BART modeling analysis, RMAX 1 was set to 20 km, and R1 was set to 10 km. This limited the influence of the surface observation data from all surface stations to 20 km from each station, and equally weighed the MM5 and observation data at 10 km. RMAX2 was set to 50 km, and R2 was set to 25 km. This limited the influence of the upper air observation data from all surface stations to 50 km from each station, and equally weighed the MM5 and observation data at 25 km. These settings of radius of influence allowed for adequate weighting of the MM5 data and the observation data across the modeling domain due to the vast domain to be modeled. RAMX 3 was set to 500 km. American Electric Power 4-10 Trinity Consultants Northeastern Power Station

24 5. CALPUFF The CALPUFF model uses the output file from CALMET together with source, receptor, and chemical reaction information to predict hourly concentration impacts. A three-year CALPUFF analysis was conducted using data and model settings as described below. 5.1 SOURCE EMISSIONS Baseline (pre-bart) emission data is based upon CEMS data collected by AEP over the time frame. In accordance with CENRAP guidelines, the emission rate over the highest calendar day (24-hr average) was used to establish baseline emissions. Emission estimates for various control scenarios were developed based on AEP s experience. The effectiveness of a number of different control technologies for NO X, and SO 2 were examined. Emission estimates for these various scenarios are included in Appendix E. 4 The cost effectiveness has been evaluated on a facility-wide basis (as opposed to a unit-by-unit basis) because of the common stack shared by NE Units 3 & 4, with the intent to apply any applicable final selected control technology on each of the affected units at the facility. 5.2 RECEPTOR LOCATIONS The National Park Service (NPS) has electronic files available on their website that include the discrete locations and elevations of receptors to be evaluated in Class I area analyses. These receptor sets were used in the CALPUFF model. 5.3 BACKGROUND OZONE AND AMMONIA Background ozone concentrations are required in order to model the photochemical conversion of SO 2 and NO X to sulfates (SO 4 ) and nitrates (NO 3 ). CALPUFF can use either a single background value representative of an area or hourly ozone data from one or more ozone monitoring stations. Hourly ozone data files were used in the CALPUFF simulation. As provided by the Oklahoma DEQ, hourly ozone data from the Oklahoma City, Glenpool, and Lawton monitors over the time frame were used. Background concentrations for ammonia were assumed to be temporally and spatially invariant and were set to 3 ppb, as described in the CENRAP protocol. 5.4 CALPUFF MODEL CONTROL PARAMETERS Appendix C provides a sample CALPUFF input file that was used for the AEP refined modeling analyses. Please note that puff splitting is a generally accepted option in refined modeling analyses over large model domains for assessing impacts on Class I areas; however, this option would require significant computer resources and longer runtime. Based upon previous model runs performed on 4 CENRAP BART Modeling Guidelines, Appendix B, 15 December American Electric Power 5-1 Trinity Consultants Northeastern Power Station

25 domains (and restricted computational grids) of the size described in this report, it is expected that runtimes could increase by a factor of 4 to 5 with the inclusion of puff-splitting. Due to this, the use of this option was evaluated during the modeling analysis and it was felt that it was not necessary to obtain representative concentrations at the individual Class I areas. American Electric Power 5-2 Trinity Consultants Northeastern Power Station

26 6. CALPOST A three-year CALPOST analysis was conducted to determine the visibility change in deciview (dv) caused by AEP s BART-eligible sources when compared to a natural background. 6.1 CALPOST LIGHT EXTINCTION ALGORITHM The algorithm was used to calculate the daily light extinction attributable to AEP s BART-eligible sources and light extinction attributable to a natural background. The change in deciviews based on the source and background light extinctions was evaluated using the equation below. dv = 10*ln b ext, background b + b ext, background ext, source EPA s currently approved algorithm for assessing light extinction and the updated light extinction calculation algorithms developed by the Interagency Monitoring of Protected Visual Environments (IMPROVE) workgroup were used to assess visibility impacts from the NE Station. The background extinction coefficient b ext, background is affected by various chemical species and the Rayleigh scattering phenomenon. The original equation for the background extinction coefficient in the FLM s FLAG guidance is as follows: ext, background 1 ( Mm ) = bso + bno + boc + bsoil + bcoarse + bap bray b where, b b b b b b b SO4 NO3 OC Soil Coarse ap Ray = = = 10 3[ ( NH 4 ) SO 4 ] f ( RH ) 2 3[ NH 4 NO 3 ] f ( RH ) [ ] [ ] 0.6[ Coarse Mass] [ EC] = 4 OC = 1 Soil = = Rayleigh Scattering f ( RH ) = 3 [] = Concentration in µg m 1 ( 10 Mm by default) Relative Humidity Function [( NH4 ) SO ] 2 4 [ NH4NO3] [ OC] denotes [ Soil] denotes [ Coarse Mass] [ EC] denotes denotes the ammonium sulfate concentration denotes the ammonium nitrate concentration the concentration of organic carbon the concentration of fine soils denotes the concentration of coarse dusts the concentration of elemental carbon Rayleigh Scattering is scattering due to air molecules American Electric Power 6-1 Trinity Consultants Northeastern Power Station

27 6.2 CALPOST PROCESSING METHOD CALPOST Method 6, which calculates hourly light extinction impacts for the source and background using monthly average relative humidity adjustment factors was used in the refined BART analysis. Monthly Class I area-specific relative humidity adjustment factors based on the centroid of the Class I areas as included in Table A-3 of EPA s Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Program were used. The factors for the Class I areas listed that were evaluated in the analysis are provided in Table 6-1. TABLE 6-1. MONTHLY HUMIDITY FACTORS Class I Area Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Caney Creek Hercules-Glades Upper Buffalo Wichita Mountains NATURAL BACKGROUND EPA s default average annual aerosol concentrations for the U.S. that are included in Table 2-1 of EPA s Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Program were used. The annual average concentrations are provided in Table 6-2 TABLE 6-2. DEFAULT AVERAGE ANNUAL NATURAL BACKGROUND LEVELS Class I Area Region SO 4 NO 3 OC EC Soil Coarse Mass Caney Creek WEST Hercules-Glades EAST Upper Buffalo EAST Wichita Mountains WEST EVALUATING VISIBILITY RESULTS When evaluating cost-control effectiveness of the various control scenarios, the 98 th percentile of the daily dv values output by CALPOST was examined. Peak 24-hr impact values have been included for reference. 6.5 SUMMARY OF CALPOST CONTROL PARAMETERS Appendix E provides a sample CALPOST input file that was used for the modeling analysis. Variable values that differ from the CENRAP protocol are generally the result of data input/output handling issues (e.g., types of output, receptor numbers, etc.). American Electric Power 6-2 Trinity Consultants Northeastern Power Station

28 7. BART-ELIGIBLE EMISSION UNITS The BART Guidelines define the following three steps for determining which emission units at a facility are BART-eligible: 1. Identify the emission units in the BART source categories, 2. Identify the start-up dates of those units, and 3. Compare potential emissions to the 250 ton/yr cutoff. Electric Generating Units (EGU) are listed source categories under BART. AEP has determined that the NE coal-fired Units 3 and 4 are BART-eligible sources at the NE Generating Station because the units, were in existence on August 7, 1977 and began operation after August 7, Units 3 and 4 both have the potential to emit NO X, SO 2, and PM 10 emissions greater than 250 tpy. A CALPUFF dispersion model was run to determine the visibility impact associated with emissions from these units. A summary of the parameters used in the BART modeling eligibility criteria for each emission unit is provided in Table 7-1. TABLE 7-1. SUMMARY OF BART-ELIGIBLE EMISSION UNITS Emission Unit BART Source Category Year of Completion of Construction or Reconstruction Maximum (24-hr average) SO 2 Emissions (lb/hr) Maximum (24-hr average) NO X Emissions (lb/hr) Maximum (24-hr average) PM/PM 10 Emissions (lb/hr) Unit 3 Boiler a , , Unit 4 Boiler a , , a Fossil-fuel boilers of more than 250 million BTUs per hour heat input. 7.1 DESCRIPTION OF BART-ELIGIBLE SOURCES NE Units 3 and 4 are tangentially-fired steam generating boilers burning western sub-bituminous coal, with natural gas or Number 2 fuel oil, used as secondary fuels for flame stabilization at low loads. Emissions from both units discharge through a common stack 600 feet in height and 27 feet in diameter. Particulate matter emissions are controlled by electrostatic precipitators (ESP) with 99.7% and 99.5% efficiencies for Units 3 and 4 respectively. Unit 3 has the capability of burning 266 tons per hour (tph) or 2,330,000 tons per year (tpy) of coal while unit 4 has the capability of burning 279 tph or 2,450,000 tpy of coal. The boilers are subject to NSPS Subpart D which requires the following: (a) PM emissions shall not exceed 0.10 lb/mmbtu. (b) Opacity shall not exceed 20%, except for one six-minute period per hour of not more than 27%. The boilers are currently meeting these standards with the electrostatic precipitator American Electric Power 7-1 Trinity Consultants Northeastern Power Station

29 7.2 MODELED STACK PARAMETERS Actual stack parameters were input to the CALPUFF model to represent each emissions point. The location of each point was represented using the Lambert Conformal Coordinate (LCC) system. Table 7-2 summarizes the stack parameters modeled for the BART-eligible emission units at AEP s NE Station. TABLE 7-2. STACK PARAMETERS Emission Unit LCC East (km) LCC North (km) Base Elevation (m) Stack Height (m) Stack Diameter (m) Exhaust Temperature (K) Exhaust Velocity (m/s) Northeastern Unit Northeastern Unit As described further in this report, various retrofit control technologies were evaluated for these two units. Many of these technologies have the potential to alter source parameters from their existing profile. A summary of source parameters associated with each control technology are included in Appendix E. American Electric Power 7-2 Trinity Consultants Northeastern Power Station

30 8. BART DETERMINATION ANALYSIS FOR NO X In general, BART is determined for each eligible emissions unit using the following five (5) steps from Section IV.D of the BART Guidelines: Step 1 Identify all available retrofit control technologies, Step 2 Eliminate technically infeasible options, Step 3 Evaluate control effectiveness of remaining control technologies, Step 4 Evaluate impacts and document the results, and Step 5 Evaluate visibility impacts. The BART determinations for each visibility impairing pollutant are presented separately in the following sections in a step by step approach. Each required step of the BART determination analysis for emissions of NO X from Unit 3 and Unit 4 is presented below. Whichever control is selected in the following BART determination will be applied to both units at the station. 8.1 STEP 1 - IDENTIFY ALL AVAILABLE RETROFIT CONTROL TECHNOLOGIES The BART Guidelines require the consideration of all control technologies with a practical potential for application to the emissions unit and the regulated pollutant under evaluation. The list of available control options should include the most stringent option and a reasonable set of options for analysis [, but] it is not necessary to list all permutations of available control levels that exist for a given technology the list is complete if it includes the maximum level of control each technology is capable of achieving. Per the BART Guidelines, the BART determination analysis must take into account technology transfer of controls that have been applied to similar source categories and gas streams [in addition to] existing controls for the source category in question. However, technologies which have not yet been applied to (or permitted for) full scale operations need not be considered as available; [the U.S. EPA does] not expect the source owner to purchase or construct a process or control device that has not already been demonstrated in practice. The BART Guidelines provides the following additional considerations for preparing the list of potential control options: One of the control options should reflect the level of control equivalent to any applicable NSPS, Source redesign should not be considered, Fuel switching should not be considered, and For emission units with existing control measures or devices, one of the control options should involve improvements to the existing controls. American Electric Power 8-1 Trinity Consultants Northeastern Power Station

31 Potential NO X control technologies and resulting emission control quantities for the NE Station s Unit 3 and Unit 4 were identified from a combination of control equipment vendor information, publiclyavailable air permits and applications, and technical literature published by the U.S. EPA, and the Regional Planning Organizations (RPOs). Each NO X control option identified as potentially applicable to either unit is listed below and explained in detail in the following subsections. Selective Non-Catalytic Reduction (SNCR) Low NO X Burners (LNB) and Over-fire Air (OFA) Selective Catalytic Reduction (SCR) Reburning / Methane de-no X (MdN) Flue Gas Recirculation (FGR) (Internal and External) Combustion-related NO X emissions are formed by two mechanisms. NO X formed from oxidation of molecular nitrogen (N 2 ) in combustion air is referred to as thermal NO X and is dependent on high temperatures (approximately 2,800 F) and an excess of combustion air. NO X formed by oxidation of nitrogen compounds in fuel is referred to as fuel NO X. The NO X formed from coal combustion is primarily fuel NO X. 5 The possible NO X emissions control technologies generally fit into one of two categories: combustion modifications, which are often associated with improving boiler performance, or flue gas treatment (i.e., post-combustion controls). Pre-combustion techniques to reduce fuel NO X have shown little promise. 6 Combustion modifications are the most common, commercially available means of controlling NO X emissions from fossil fuel-fired boilers COMBUSTION MODIFICATIONS FLUE GAS RECIRCULATION Generally, FGR involves extracting a portion (15 to 30 percent) of the flue gas and readmitting it to the furnace through the burner window. When the flue gas is extracted from the economizer or air heater outlet, a separate fan/blower is needed to withdraw the flue gas. This setup is referred to as external or forced FGR. Internal or induced FGR refers to the setup where the flue gas is extracted from upstream of the stack using the forced draft (FD) fan instead of a separate FGR fan. In either setup, the recirculated flue gas acts as a thermal diluent (i.e., heat sink) to reduce combustion temperatures. It also dilutes the combustion reactants and reduces the excess air requirements thereby reducing the concentration of oxygen in the combustion zone. Thus, thermal NO X formation is inhibited. 8 The onset of thermal NO X occurs around 2,800 F, and NO X generation increases exponentially with temperatures beyond 2,800 F. As only thermal NO X can be 5 MACTEC, Midwest RPO Boiler BART Engineering Analysis, March 30, NCASI, NO X Control in Forest Products Industry Boilers: A Review of Technologies, Costs and Industry Experience, Special Report Ibid. 8 U.S. EPA, Clean Air Technology Center, Nitrogen Oxides (NO X ), Why and How They Are Controlled. Research Triangle Park, North Carolina, EPA-456/F R, November American Electric Power 8-2 Trinity Consultants Northeastern Power Station

32 controlled by this technique, it is especially effective only in oil and gas-fired units LOW NO X BURNERS AND OVERFIRE AIR LNB technology utilizes advanced burner design to reduce NO X formation through the restriction of oxygen, flame temperature, and/or residence time. A LNB is a staged combustion process that is designed to split fuel combustion into two zones, primary combustion and secondary combustion. Two general types of LNB exist: staged fuel and staged air. Lower emission rates can be achieved with a staged fuel burner than with a staged air burner. Staged fuel LNB separate the combustion zone into two regions. The first region is a lean primary combustion region where the total quantity of combustion air is supplied with a fraction of the fuel. Combustion in the primary region (first stage) takes place in the presence of a large excess of oxygen at substantially lower temperatures than a standard burner. In the second region, the remaining fuel is injected and combusted with any oxygen left over from the primary region. The remaining fuel is introduced in the second stage outside of the primary combustion zone so that the fuel/oxygen are mixed diffusively (rather than turbulently), which maximizes the reducing conditions. This technique inhibits the formation of thermal NO X, but has little effect on fuel NO X. By increasing residence times staged air LNB provide reducing conditions, which have a greater impact on fuel NO X than staged fuel burners. The estimated NO X control efficiency for LNB in high temperature applications is 25 percent. 10 Combustion modification with LNB is used in both gas/oil-fired and coal-fired units. 11 In OFA, about 10 to 20 percent of the combustion air flow is directed to separate air ports located downstream of the burners. OFA works by reducing the excess air in the burner zone, thereby enhancing the combustion staging effect and theoretically reducing NO X emissions. Residual unburned material, such as CO and unburned carbon, which inevitably escapes the main burner zone, is oxidized as the OFA is admixed later. 12 LNB with OFA as a combined control technology can reduce emissions from each boiler across the full load. Based on this, LNB with OFA will continue to be reviewed as a viable BART control technology REBURNING / METHANE DE-NO X In reburning, also known as off-stoichiometric combustion or fuel staging, a fraction (5 to 25 percent) of the total fuel heat input is diverted to a second 9 NCASI, NO X Control in Forest Products Industry Boilers: A Review of Technologies, Costs and Industry Experience, Special Report MACTEC, Midwest RPO Boiler BART Engineering Analysis, March 30, NCASI, NO X Control in Forest Products Industry Boilers: A Review of Technologies, Costs and Industry Experience, Special Report Ibid. American Electric Power 8-3 Trinity Consultants Northeastern Power Station