DAQ-2025-001205

PM2.5 SIP Evaluation Report: Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City (Previously Tesoro Refining & Marketing Company LLC) Salt Lake City Nonattainment Area
Utah Division of Air Quality Major New Source Review Section
Originally Adopted July 1, 2018 Revised February 5, 2025

1
PM2.5 SIP EVALUATION REPORT Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Refinery
1.0 Introduction The following is part of the Technical Support Documentation (TSD) for Section IX, Part H.12 of the Utah SIP; to address the Salt Lake County PM2.5 Nonattainment Area. This document specifically serves as an evaluation of the Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Refinery (Marathon).
The revision to this TSD documents how each emission unit that existed at the refinery on January 1, 2019, met BACT/BACM. For any determination that BACT/BACM was met with
existing controls (existing prior to the 2018 BACT determination, required at that time by Federal or state regulation, or permitted prior to the 2018 determination), no new control requirements will be added to the SIP.
Economic and technical feasibility for determining BACT is based upon the 2017 BACT Analyses. Marathon cannot retroactively install equipment to meet the BACT deadline of
January 1, 2019. 40 CFR 51.1011 establishes that control measures must be implemented no later than the beginning of the year containing the applicable attainment date. Thus, for purposes of RFP and SIP credit, the deadline for implementation of all BACT and BACM is January 1, 2019.
Any control measures implemented beyond such date through June 9, 2021 (4 years after the date of reclassification) are instead regarded as “additional feasible measures.” Control measures that can only be implemented after June 9, 2021 are beyond the scope of this SIP. 1.1 Facility Identification Name: Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Address: 474 West 900 North, Salt Lake City, Utah, Salt Lake County Owner/Operator: Tesoro Refining & Marketing Company wholly owned by Marathon
Petroleum Corporation
UTM coordinates: 4,515,950 m Northing, 423,400 m Easting, Zone 12 1.2 Facility Process Summary
The Marathon Refinery is a petroleum refinery capable of processing 57,500 barrels per day of crude oil. The source consists of a FCCU, catalytic reforming unit, hydrotreating units, a sulfur recovery unit, and cogeneration units. The source also has the usual assorted heaters, boilers, cooling towers, storage tanks, flares, and fugitive emissions. The source operates flare gas recovery with compressor availability requirements for the
north and south flares. 1.3 Facility Criteria Air Pollutant Emissions Sources The following is a listing of the main emitting units from the Marathon Refinery:

2
Crude Unit Furnace, with ultra-low NOx burners Ultraformer Unit (UFU) Furnace, with ultra-low NOx burners UFU Regeneration Heater, with low NOx burners
Fluid catalytic Cracking Unit (FCCU), Carbon Monoxide Boiler (Heat Recovery Unit), with electrostatic precipitator (ESP), wet gas scrubber/LoTOx system (WGS) Distillate Desulfurization Unit (DDU) charge heater and rerun boiler equipped with “ultra-ultra” low-NOx burners Hydrogen Compressors (Ultraformer compressors), with catalytic converters
South Flare (Flare covering Crude/UFU Unit/DDU) North Flare (Flare covering FCCU/VRU/Alkylation Unit/GHT) Modular Flare Gas Recovery System (FGR) Sulfur Recovery Unit/Tail Gas Incinerator/Tail Gas Treatment Unit Fuel Gas Desulfurization Unit/Sour Water Stripper (FGDU/SWS) Flare (this unit is
physically integrated with the Sulfur Recovery Unit (SRU)) Emergency/Standby Sources Waste Water Treatment Plant (WWTP) Generator Portable Electrical Generators Plant Air Compressors
Miscellaneous Air Compressors Fire Water Pumps B-1 Air Preheater Temporary Package Boilers Gasoline Hydrotreater (GHT) Unit with process heater
Benzene Saturation Unit (BSU) Two cogeneration turbines (CG1 and CG2), each with SoLoNOx, and heat recovery steam generating units (HRSG) Loading/Unloading Racks
Cooling Towers
Fugitives Tank Farm Storage Vessels Remote Tank Farm Storage Vessels
This is not meant to be a complete listing of all equipment which may be involved or
required during permitting activities at the refinery, rather it is a listing of all significant emission units or emission unit groups (such as the tank farm). Emission units such as a fluidized catalytic cracking unit (FCCU) which may have multiple individual component parts, but which can be treated as a single unit for purposes of BACT analysis and
discussion, will be treated in such a manner. See Appendix A for a more complete listing
of all refinery emission units and emission unit groups. See Appendix B for supporting documentation for all units with CEMs. Please note that the data in Appendix B are presented in raw, unprocessed form and include periods of monitor downtime, quality assurance, calibration, maintenance, out of control periods, malfunctioning CEMs data,
and exempt periods, etc.
1.4 Facility 2016 Baseline Actual Emissions and Current PTE
In 2016, Marathon’s baseline actual emissions were determined to be the following (in tons per

3
year)1: Table 1: Actual Emissions
Pollutant Actual Emissions (Tons/Year)
PM2.5 89.4 SO2 708.3 NOx 358.1 VOC 250.4 NH3 3.8

The current PTE values for Marathon, as established by the most recent AO issued to the source (prior to the beginning of the year containing the applicable attainment date, i.e. January 1, 2019) (DAQE-AN103350075-18)2 are as follows:
Table 2: Current Potential to Emit
Pollutant Potential to Emit (Tons/Year)
PM2.5 154.0 SO2 1637.0 NOx 638.0
VOC 767.9 NH3 6.6*
* NH3 emissions not quantified in the AO, PTE is estimated
2.0 Modeled Emission Values A full explanation of how the modeling inputs are determined can be found elsewhere. However,
a shortened explanation is provided here for context. The base year for all modeling was set as 2016, as this is the most recent year in which a
complete annual emissions inventory was submitted from each source. Each source’s submission was then verified, checking for condensable particulates, ammonia (NH3) emissions, and calculation methodologies. Once the quality-checked 2016 inventory had been prepared, a set of
projection year inventories was generated. Individual inventories were generated for each projection year: 2017, 2019, 2020, 2023, 2024, and 2026. If necessary, the first projection year, 2017, was adjusted to account for any changes in equipment between 2016 and 2017. For new
equipment not previously listed or included in the source’s inventory, actual emissions were assumed to be 90% of its individual PTE.
While some facilities were adjusted by “growing” the 2016 inventory by REMI growth factors; most facilities were held to zero growth. This decision was largely based on source type, and how each source type operates. The refineries have reported to UDAQ as a production group that they are operating at capacity and are not planning any production or major emission increases in the time frame covered by the SIP BACT analysis. For these reasons, UDAQ used zero growth for all projection years beyond the 2016 baseline inventory. In the case of the Marathon Refinery, there were a number of projects that were included both to
complete the 2016 inventory, and to adjust the 2016 inventory for the first projection year of
1 see References: Item #10 2 see References: Item #7

4
2017. These projects are summarized as follows and all include emission reductions to the 2016 inventory.:
• DAQE-AN103350058B-13, although issued in 2013, the project was not completed until late 2014 – early 2015, so the SO2 emission reduction did not appear until the 2016 emission
inventory. This project was the Waxy Crude Processing Project, which included the addition of a TGTU on the SRU.
• DAQE-AN103350065-14, issued in 2014, this project was for flare gas minimization and was included in the 2016 inventory.
• DAQE-AN103350066-16, issued in April of 2016, this project addressed the addition of a
wet gas scrubber to control emissions from the FCCU. The project was only completed in January of 2018 and will be covered extensively throughout this document as BACT-level control. Following the process laid out above (applying no growth, but adjusting the baseline emissions to include permitted changes) for the Marathon Refinery, this yielded the following modeled emission values – summarized in Table 3.
Table 3: Modeled Emission Values
Pollutant Potential to Emit (Tons/Year)
PM2.5 89.4 SO2 544.4
NOx 360.1 VOC 249.3 NH3 3.8

Since a value of zero (0) growth was applied for all projection years, the values listed above (the 2017 corrected values) would then be propagated through for each of the subsequent projection years – 2019, 2020, 2023, 2024 and 2026.
Next, the effects of BACT would be applied during the appropriate projection year. Any controls applied between 2016 and 2017 (such as any RACT or RACM required as a result of the
moderate PM2.5 SIP), was already taken into account during the 2017 adjustment performed previously. Future BACT, meaning those items expected to be coming online between today and the regulatory attainment date (December 31, 2018), would be applied during the 2019 projection
year. Notations in the appropriate projection year table of the emission inventory model input spreadsheet indicate the changes made and the source of those changes. Similarly, Additional Feasible Measures (AFM) or Most Stringent Measures (MSM), which might be applied in future projection years beyond 2019 are similarly marked on the spreadsheet. The effects of those types of controls are applied on the projection year subsequent to the installation of each control – e.g. controls coming online in 2021 would be applied in the 2023 projection year, while controls installed in 2023 would be shown in 2024.
3.0 BACT Selection Methodology
The general procedure for identifying and selecting BACT is through use of a process commonly referred to as the “top-down” BACT analysis. The top-down process consists of five steps which consecutively identify control measures, and gradually eliminate less effective or infeasible
options until only the best option remains. This process is performed for each emission unit and each pollutant of concern. The five steps are as follows:

5
1. Identify All Existing and Potential Emission Control Technologies: UDAQ evaluated various
resources to identify the various controls and emission rates. These include, but are not limited to: federal regulations, Utah regulations, regulations of other states, the RBLC, recently issued permits, and emission unit vendors.
2. Eliminate Technically Infeasible Options: Any control options determined to be technically infeasible are eliminated in this step. This includes eliminating those options with physical or
technological problems that cannot be overcome, as well as eliminating those options that cannot be installed in the projected attainment timeframe. 3. Evaluate Control Effectiveness of Remaining Control Technologies: The remaining control options are ranked in the third step of the BACT analysis. Combinations of various controls are also included. 4. Evaluate Most Effective Controls and Document Results: The fourth step of the BACT
analysis evaluates the economic feasibility of the highest ranked options. This evaluation includes energy, environmental, and economic impacts of the control option.
5. Selection of BACT: The fifth step in the BACT analysis selects the “best” option. This step also includes the necessary justification to support the UDAQ’s decision.
Should a particular step reduce the available options to zero (0), no additional analysis is required. Similarly, if the most effective control option is already installed, no further analysis is needed.
4.0 BACT for the FCCU Regenerator and CO Boiler
The fluidized catalytic cracking unit, or FCCU, is a reactor where pre-heated feedstock is combined with a very hot catalyst in order to “crack” or break the long-chain hydrocarbon molecules making up the feedstock. The long-chain molecules are broken down into shorter, lighter molecular weight hydrocarbons. These lighter materials then rise to the top of the reactor where they are removed and sent elsewhere in the refinery for further processing. The spent catalyst is removed from the recovered material through a series of cyclones and sent to the regenerator section.
The regenerator in most FCCUs is a secondary vessel located alongside (in a side-by-side configuration) the main reactor vessel. The regenerator is used to remove residual carbon buildup from the surface of the catalyst. This residual carbon, also called coke, reduces catalyst
performance simply by adhering and coating the active surfaces of the catalyst. The catalyst is quite hot when it exits the reactor, and simply introducing forced air is enough to cause the coke to combust. The additional heat from this combustion keeps the regenerator operating around
1300ºF. Catalyst coke contains a high amount of entrapped impurities depending on the chemical nature of the feedstock. Sulfur, various nitrogen compounds, trace metals and other compounds may be present. These materials will be released during combustion of the coke and depending
on the design of the regenerator may be altered during the combustion process as well. The regenerator is the primary point of emissions from the FCCU.
Depending on the design of the FCCU, the unit may operate in either complete or partial burn mode. In complete burn units, the FCCU is operated to completely combust the coke and exhaust gases generating primarily CO2. In partial burn units, such as the FCCU at the Marathon

6
Refinery, the unit is operated to produce mainly CO, which is then used as a fuel source for a downstream heat recovery boiler called a CO boiler (or COB). The COB uses the residual heat of
the exhaust gases, plus the heat generated by combustion of the CO in the off-gas to produce steam for use elsewhere in the refinery. The exhaust gases from the COB are then sent through air pollution controls and out the exhaust stack. The air pollutants generated and types of air
pollution controls are the same for both complete and partial burn units. Currently Marathon controls emissions from the FCCU/COB using a combination of electrostatic
precipitation (ESP), wet gas scrubber (WGS) with LoTOx (an ozone treatment system for NOx control), and use of a low SOx catalyst (DeSOx). These controls serve as the baseline for Marathon’s submitted analysis as well as UDAQ’s BACT review. 4.1 PM2.5 4.1.1 Available Control Technology
For control of particulate emissions from a FCCU regenerator, a source can choose from the usual array of options, either high efficiency electrostatic precipitation (ESP) or fabric filtration (baghouse) being the primary choices depending on the electrical resistivity of the coke burn-off
at the particular refinery. Two additional, more recent choices have also emerged: wet gas scrubbing (WGS) and a “flue gas blowback filter” (FGF). The FGF is an in-stack filter that operates in a similar fashion to a fabric filtration system, but on a smaller and faster cleaning
scale. They are designed specifically for use with a FCCU, and have generally not been commercially applied in the U.S. but have seen successful application overseas.
The other control options normally available for combustion related activities, such as fuel switching or “good combustion controls,” are inherently limited by the nature of the process. The chemical nature of the feedstock and the type of cracking catalyst do make some difference in the
resulting particulates generated during the regeneration process, but an individual refinery is rather limited in which feedstocks it can accept based on physical configuration, geographical location, market forces (availability), and regulatory limits (on both the refinery emissions and the allowed final product). Ultimately, feedstock blending and catalyst changes have little to no effect on particulate emissions. 4.1.2 Evaluation of Technical Feasibility of Available Controls
All of the available controls are technically feasible; however the controls are mutually exclusive – they cannot (in most cases) be used together.
4.1.3 Evaluation and Ranking of Technically Feasible Controls In terms of efficiency for control of particulate emissions, the available controls are all
approximately equal. • Pulse jet fabric filter
• FGF • WGS • ESP
Fabric filters have the highest efficiency but are designed only to control particulate emissions. Because of their high efficiency, they suffer from a problem other control options do not have.

7
Particulate emissions from catalytic coke burn-off can be extremely sticky, and the fabric in these baghouses can easily become fouled and lead to blown bags. Higher cost bags can avoid this
problem, but this application leads to higher operating costs. The FGF option has a control efficiency nearly as high as a well-maintained pulse jet fabric filter,
with a higher installation cost than that of a fabric filter. Both the fabric filter and FGF control only the filterable fraction of particulate emissions.
While the WGS system has the added benefit of removing condensable particulates, it is primarily designed as a control device for removal of SO2 emissions. Installation and operation of a WGS is also far more expensive than any of the other options. Wet scrubbing inherently involves water treatment and disposal/discharge, which must be included in the operating cost. WGS, however, has an additional benefit over both of the above options in that it also controls the condensable fraction of particulate emissions – which can often be significantly larger than the filterable fraction.
Use of a high efficiency ESP is the typical default option. Holly used an ESP prior to the installation of the WGS system, and the nearby Chevron refinery still employs an ESP as the final
particulate control system. The Marathon refinery employs both an ESP as the primary particulate control device, and the WGS as a secondary control device – in essence it acts almost as a polishing scrubber in this case.
4.1.4 Further Evaluation of Most Effective Controls
Should Marathon have chosen to use a FGF or fabric filter control, emissions of 0.2 lb/1000 lb of coke burned are possible, although these values are filterable particulate only. WGS is slightly less efficient, with reported values of 0.3-0.5 lb/1000 lb coke burned. The default ESP option is
typically limited to 0.5-0.7 lb/1000 lb coke burned, although this meets the various requirements of both the moderate PM2.5 SIP as well as the emission limitations of 40 CFR 63 Subpart UUU and 40 CFR 60 Subpart Ja (limits are 1.0 lb PM/1000 lb coke burned). Marathon did not provide an economic analysis of the various controls, as it has not removed the existing ESP and is adding a WGS as a secondary control system on the FCCU. As the Marathon refinery is already meeting the existing limitation required of all non-attainment area (NAA) refineries of 1.0 lb PM/1000 lb of coke burned with the existing ESP, and is adding
additional control of a WGS, no additional analysis is necessary. This configuration is the only possible combination of two particulate control technologies that can functionally operate in series on a FCCU.
4.1.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to use the existing ESP and WGS system to control emissions of particulate from the FCCU catalyst regenerator and integrated COB. These practices are required through existing permit requirements. No additional controls are required for BACT;
thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emissions limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 4.2 SO2

8
4.2.1 Available Control Technology
There exist several options for removing sulfur from FCCUs: • Feed hydrotreating removes the sulfur from FCCU feedstocks prior to cracking operations.
• SOx removing (deSOx) catalyst injection prevents the sulfur from forming in the coke so it isn’t burned off during regeneration forming SO2. • WGS allows for normal catalyst use, and then removes the SO2 from the exhaust gases
through wet contact scrubbing. These options, while not necessarily mutually exclusive, do have impacts on the control options for other pollutants. Feed hydrotreating has some positive benefit on NOx formation. Using a SOx reducing catalyst additive creates additional sulfate (condensable PM2.5). The use of WGS prevents the use of fabric filtration for particulate control, but allows for the use of LoTOx, a NOx control option.
4.2.2 Evaluation of Technical Feasibility of Available Controls Save for feed hydrotreating, all of the controls are technically feasible and currently employed at
the Marathon refinery. Feed hydrotreating requires that the residual oil is removed from the feedstock prior to the introduction of hydrogen. This is accomplished through vacuum distillation to separate the vacuum gas oil from the residual oil. Marathon does not operate a vacuum tower
at present, and considers this option to be technically infeasible. Although UDAQ disagrees that the process itself is inherently technically infeasible, it agrees that the addition of such a process would not add any value – emissions would not be further reduced, this would be an additional
unnecessary expense, and the additional engineering, design, construction time and changes in FCCU feedstock from such an undertaking would push such a process change outside of the SIP BACT and even Additional Feasible Measures (AFM) windows.
4.2.3 Evaluation and Ranking of Technically Feasible Controls Some combining of control options is possible. Feed hydrotreating and deSOx catalysts can be used in combination. WGS systems do not gain any additional benefit when combined with either of the other two control methods; however, neither are they harmed through combination with either control method.
The use of WGS technology can achieve the limits required by Subpart Ja: 50/25 ppmv (7-day/annual). As noted above in the summary for particulate control, WGS is a far more expensive option than either feed hydrotreating or DeSOx catalyst. It also has the added
disadvantage of water waste treatment and/or disposal. The use of SOx reducing catalyst, can also meet the Subpart Ja limits. The known disadvantage of
sulfate formation can be treated with effective particulate control systems. As both remaining control options are viable, and have been deemed equally effective at reaching
the required limits under Subpart Ja – further evaluation is required. 4.2.4 Further Evaluation of Most Effective Controls Marathon did not submit an economic analysis for either of these control options as the top control was chosen. However, since both control options have similar control efficiencies and

9
Marathon has elected to employ a WGS, no additional analysis is required. 4.2.5 Selection of BACT Controls UDAQ recommends that Marathon continue to use the WGS as needed to meet the Subpart Ja
FCCU SO2 limits. These limits have already been established in Section IX, Part H.11.g of the SIP and are required through existing permit requirements. Monitoring, recordkeeping and reporting requirements are included as well. The existing limits account for current process
variability while still limiting SO2 emissions from the FCCU. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emissions limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 4.3 NOx
4.3.1 Available Control Technology
The available options for control of NOx from FCCUs are listed below: • Low-NOx promoter catalysts
• Selective non-catalytic reduction (SNCR) • Selective catalytic reduction (SCR) • Feed hydrotreating
• LoTOx in conjunction with WGS Low-NOx promoter catalysts and NOx reducing additives (as found in another BACT analysis)
can be considered the same technology for purposes of this review. Both are catalytic additives (meaning they are not consumed in the process) although they serve slightly different purposes. The promoter catalysts specifically serve as FCC catalysts – providing sites for the cracking of long chain hydrocarbon molecules into shorter ones, but helping prevent the formation of NOx during the regeneration phase. The additives are supposed to prevent nitrogen from being trapped in the coke in the first place so that there is less “fuel-bound” nitrogen to form NOx during the regeneration process.
4.3.2 Evaluation of Technical Feasibility of BACT Controls All control options are technically feasible, strictly speaking. However, Marathon does not
operate a vacuum tower and still considers feed hydrotreating to be technically infeasible. UDAQ has agreed that the addition of feed hydrotreating is outside the scope of SIP BACT, and does not need to be evaluated further.
Although LoTOx requires that a WGS system is simultaneously in use, this does not invalidate its technical feasibility. LoTOx is an add-on control system to the standard WGS where ozone is
generated and injected into the exhaust gas stream prior to contact with the scrubbing liquid. The ozone reacts with the NOx present in the gas stream and generates N2O, which is highly soluble in water (forming N2O5). Marathon, and to some degree the other refineries as well, has extensively investigated the use of NOx reducing additives and determined that they had no effect on NOx emissions. Low-NOx

10
promoter catalysts are useful, and so only the promoter catalysts will be evaluated further.
The use of SNCR or direct ammonia injection into the FCCU regenerator exhaust cannot be used in conjunction with the WGS/LoTOx system because of the rapid cooling provided by the WGS. The use of SCR would also be severely hampered by a WGS/LoTOx system for much the same
reason, although the injection of the ammonia would likely not harm the functionality of the WGS or LoTOx systems. 4.3.3 Evaluation and Ranking of Technically Feasible Controls None of the refineries provided detailed analysis for the evaluation of SNCR beyond stating that no ammonia injection into the FCCU was occurring. Expected control efficiencies would be rather low, based on residence time, exhaust temperatures, and overall emission reductions of SNCR-based systems. The remaining options of SCR, and WGS with LoTOx are approximately equal in terms of
overall control effectiveness. 4.3.4 Further Evaluation of Most Effective Controls
SCR has an additional drawback in the form of ammonia slip. In order to control NOx, ammonia is injected to reduce the NOx to N2 and water. Ideally, a stoichiometric amount of ammonia
would be added – just enough to fully reduce the amount of NOx present in the exhaust stream. However, some amount of ammonia will always pass through the process unreacted; and since the process possesses some degree of variability, a small amount of additional ammonia is added
to account for minor fluctuations. The ammonia which passes through the process unreacted and exits in the exhaust stream is termed “slip” (sometimes “ammonia slip”). The amount varies from facility to facility, but ranges from almost zero to as high as 30 ppm in poorly controlled systems.
In the case of SCR systems, the catalyst also degrades over time, and the degree of slip will gradually increase as increasing amounts of ammonia are needed to maintain NOx reduction performance. WGS systems, with or without LoTOx, generate wastewater which must be treated before discharge or stored before disposal. Systems with LoTOx either have an acidic wastewater (nitric acid generated by N2O5 in the aqueous phase), or one with soluble solids from neutralization of that acid (typically with NaOH or Na2CO)
4.3.5 Selection of BACT Controls
UDAQ does not recommend any changes. Marathon should continue to meet the existing rolling 7-day permit NOx emission limits from the FCCU as required by NSPS Subpart Ja using WGS with LoTOx. The existing limits account for current process variability while still limiting NOx
emissions from the FCCU. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. No additional controls are required for BACT; thus, no additional limits are required to be established for the
SIP. Marathon will comply with any applicable emissions limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 4.4 VOC and Ammonia Considerations

11
UDAQ was unable to locate any additional controls to reduce emissions of VOCs from the FCCU/COB. In 2016, Marathon’s listed VOC emissions from this unit were 0 tons. Marathon
has not tested the emissions from this emitting unit, and thus UDAQ is unable to comment. However, in a review of other refineries, no viable add-on control device or technique was found to further reduce the emissions of VOCs from FCC catalyst regenerators. Typical VOC reduction
controls such as thermal or catalytic oxidation require relatively high VOC concentrations and often additional heat input in the form of fuel burning (negating the controls already achieved for other pollutants). Control techniques such as fuel switching are negated by the nature of the
process – the catalytic coke must be removed to continue the cracking process in the FCCU. Marathon did list a VOC promoter catalyst which was immediately eliminated as being technically infeasible in partial burn units – UDAQ was unable to locate any additional information on this catalyst, and it is not in use at any of the other area refineries. The only remaining technique is simply good combustion practices, which is required by the other control systems already in place. No additional consideration for good combustion practices is required. There are two possible mechanisms for ammonia emissions from the FCCU regenerator. Most
refineries emit some amount of ammonia from the coke burn-off process itself, as trapped ammonia salts present in the coke are released during the regeneration process. These emissions are typically relatively small. The second mechanism is the injection of ammonia for control of
NOx emissions using either SCR or SNCR as a control process. The injection of ammonia is fairly common among refineries in the U.S., but does not occur among the refineries in Utah. None of the refineries located in the Salt Lake City PM2.5 NAA uses ammonia injection for NOx
control. Therefore, UDAQ recommends that no additional BACT limitations be required for these two
pollutants. Good combustion practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP.
5.0 BACT for Refinery Process Heaters and Boilers UDAQ has separated the analysis of process heaters and boilers into two groups. For those heaters and boilers with heat input ratings less than 30 MMBtu/hr; UDAQ has included its analysis in a separate document which addresses similar emission units which are common to many sources – such as small heaters and boilers. Please refer to the PM2.5 Serious SIP - BACT for Small Sources – Section 5 for details of the analysis for these smaller units. The remaining
larger items are covered below. Marathon operates six (6) fired process heaters:
H-101 Crude Unit Furnace F-1 Ultraformer Unit Furnace F-680 DDU Charge Heater
F-681 DDU Rerun Boiler Steam for the refinery is primarily supplied by the COB and two cogeneration units (discussed
elsewhere in this document), and through unfired waste heat boilers located in various process units. On various occasions Marathon also operates temporary rental package boilers when steam is needed because of process unit shut down or when one or both of the cogeneration units may be out of service. The BACT review process for the rental boilers is the same as for the process heaters at the refinery, as these boilers are fired on natural gas and generally subject to the same federal standards. In addition, the boilers are also subject to 40 CFR 60.41b, which limits “time

12
on site” to 180 days or less. 5.1 PM2.5 No add-on controls for particulates were considered by UDAQ for these boilers. Given that these
emission units are fired on gaseous fuels, with inherently low particulate formation, no controls are expected to be cost effective. UDAQ did review the usual particulate control options of good combustion practices, use of low sulfur fuels, wet gas scrubbers, electrostatic precipitators
(ESPs), cyclones, and baghouse/fabric filtration; and determined that only good combustion practices and use of low sulfur gaseous fuels were technically feasible. Both refinery fuel gas and natural gas are low sulfur fuels. UDAQ reviewed one refinery’s economic analysis of switching to using exclusively natural gas as fuel and determined such a switch to not be economically feasible, with a control cost in excess of $2.2 million/ton of particulate removed. Good combustion controls remain the only viable control option. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable
emission limits in Section IX, Part H.11. 5.2 SO2
Generally, SO2 is formed from the combustion of sulfur present in the fuel. By limiting the sulfur content of the fuel, less SO2 will be generated. Emissions of SO2 can also be controlled by post
combustion control devices or processes. 5.2.1 Available Control Technology
By consolidating all process heaters and boilers together into a single group for BACT consideration UDAQ is able to consider controls on some emissions from this group which would
ordinarily be dropped as being insignificant. However, it also limits the available options. In this particular case, only one option is available. The long term Subpart Ja refinery fuel gas H2S limit of 60 ppmv as well as the existing short term Subpart Ja limit of 162 ppmv on a 3-hour average. The normally available options of flue gas desulfurization (FGD) or fuel switching are not available in this case. Fuel switching is not possible given the requirements of eliminating the refinery fuel gas generated during production of gasoline and other petroleum derivatives. The refinery fuel gas cannot be flared, and too much is produced to allow for reforming into heavier
products (the energy losses would negate any positive benefit gained. Desulfurization systems rely on a relatively high concentration of sulfur compounds in the exhaust stream to function effectively and efficiently. By meeting the fuel gas H2S limits in Subparts Ja, the exhaust gas
concentrations of SO2 will naturally fall below the critical concentrations necessary for optimum control.
Refinery fuel gas, while still considered a low sulfur fuel, is higher in sulfur content than natural gas. Pipeline quality natural gas has a very low sulfur content of approximately 4 ppm – typically in the form of mercaptans used as odorants. The sulfur content of refinery fuel gas varies
depending on the performance/removal efficiency of the amine scrubbing unit at the refinery. At the Marathon refinery, the amine unit is designed to produce fuel gas with an average H2S content of 60 ppm on an annual average. However, short term spikes are allowed under the current rules of the SIP (the previously mentioned 162 ppmv on a 3-hour average). By way of comparison, ultra-low sulfur diesel fuel (ULSD) is 15 ppm sulfur.

13
Wet gas scrubbing (WGS) can be considered available technologies. A typical WGS system consists of either a packed bed tower or venturi-type scrubber. The flue gas to be cleaned passes
through the absorber where misting nozzles form a dense curtain of liquid. The liquid reagent helps to cool the flue gas, neutralize the SO2 in the flue gas, as well as trap any particulate matter in the gas. Liquid collects in the bottom of the scrubber where caustic soda (NaOH) is added to
prevent the formation of sulfuric acid (H2SO4). The scrubbed gas continues upward through the vessel passing through filters prior to release into
the atmosphere. Waste collected at the bottom of the scrubber is pumped off for additional treatment. This waste contains sulfites such as NaHSO3 and Na2SO3 along with residual catalyst fines and precipitated solids. Solids removal is done through a clarifier using flocculation to settle out the solids. 5.2.2 Evaluation of Technical Feasibility of Available Controls WGS is available for control of emissions from sources with higher concentrations of SO2 or acid
gases in the exhaust stream, but for these types of sources they are not commercially available. To some degree this can also be viewed as a technical concern, but in either case the end result is the same. As WGS is not commercially available for emission sources of this concentration,
WGS will not be considered further. This leaves only the use of low sulfur fuels and good combustion practices as technically feasible
and available controls. 5.2.3 Evaluation and Ranking of Technically Feasible Controls
No ranking of control techniques is required. 5.2.4 Further Evaluation of Most Effective Controls As mentioned previously, one area refinery did conduct an analysis of switching to running exclusively on pipeline quality natural gas as fuel versus using refinery fuel gas or a combination of refinery fuel gas and natural gas. While this analysis was conducted on particulate emissions and not SO2, the difference in emission totals between particulate and SO2 is about 4 tons (at the Marathon refinery total particulate emissions from heaters and boilers ~ 13.4 tons/year, total emissions of SO from heaters and boilers ~ 9.3 tons/year), the economic analysis result is similar.
Using exclusively natural gas as fuel is not economically feasible, with a control cost in excess of $3.2 million/ton of SO2 removed. 5.2.5 Selection of BACT Controls UDAQ recommends that Marathon continue to use good combustion controls and refinery fuel
gas or natural gas as fuel for control of SO2 emissions from the refinery process heaters and boilers as BACT. Refinery fuel gas is required to meet the existing Subpart Ja fuel gas H2S limits of 60 ppmv on a 365-day rolling average and 162 ppmv on a 3-hour average. These limits are
currently listed as work practice requirements in Section IX, Part H.11.g of the SIP. These practices are required through existing permit requirements. The existing limits account for current process variability while still limiting SO2 emissions from each process heater and boiler. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. For a summary of the potential emissions from each process heater and boiler with a capacity greater than 40 MMBtu/hr, see Appendix C.

14
These calculations depend on the maximum SO2 emissions each heater and boiler could have emitted in 2017, compared to the actual emissions reported in the 2017 inventory. The actual
emissions from 2017 are representative of normal operation throughout the year based on process and operation variability. Marathon will still comply with all existing permit and SIP requirements.
No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX,
Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 5.3 NOx NOx, or oxides of nitrogen, are formed from the combustion of fuel. There are three mechanisms for the formation of NOx: fuel NOx, which is the oxidation of the nitrogen bound in the fuel; thermal NOx, or the oxidation of the nitrogen (N2) present in the combustion air itself; and prompt
NOx, which is formed from the combination of combustion air nitrogen (N2) with various partially-combusted intermediary products derived from the fuel. For combustion within the process heaters and boilers, fuel NOx and thermal NOx are the major contributors, with prompt
NOx contributing slightly only in the initial stages of combustion. All three processes are temperature dependent – combustion temperatures below 2700ºF greatly inhibit NOx formation. 5.3.1 Available Control Technology The following technologies were identified as potential control methodologies for control of NOx
emissions by Marathon3: good combustion practices; low emission combustion (LEC); selective non-catalytic reduction (SNCR), the injection of ammonia or urea directly into the late stages of the combustion zone; selective catalytic reduction (SCR); flue gas recirculation (FGR); and
EMx™ (previously known as SCONOx™). Low Emission Combustion (or LEC) is a summary term given to a host of different combustor designs and pre-combustion controls such as water or steam injection. These can be combined with a similar type of combustion control known as staged air/fuel combustion or overfire air injection. All serve the same general purpose – to reduce NOx formation by lowering the overall flame temperature. The various combustor (burner) designs, ranging from low-NOx, through ultra-low-NOx and up to “next generation” ultra-low-NOx reduce flame temperature through a
combination of flame diffusion, internal flue gas recirculation and some degree of staged combustion design. Water injection uses the inherent high specific heat of the injected water to absorb some of the combustion energy without increasing the ambient gas temperature. Staged
air/fuel combustion limits the total amount of combustion air so that reducing and oxidizing sections are created in the combustion chamber. Combustion happens in stages, with intermediary products needing to physically move between sections before continuing
combustion. Combustion is slowed down, limiting the “flux” (energy output/time) which lowers the total temperature.
Low-NOx Burners (LNB): Typically thought of as an advanced version of a standard burner, the LNB reduces NOx formation through the restriction of oxygen, flame temperature, and/or residence time. There are two main types of LNB: staged fuel and staged air burners. Staged fuel burners divide the combustion zone into two regions, limiting the amount of fuel supplied in the

3 see References: Item #8

15
first zone with the standard amount of combustion air, and then supplying the remainder of the fuel in the second zone to combust with the un-combusted oxygen from the first zone. Staged air
burners reverse this, limiting the combustion air in the first zone then supplying the remainder of the combustion air in the second zone to combust the remaining fuel. Staged fuel LNBs are more suited to natural gas-fired boilers as they are designed to restrict flame temperature.
Ultra-Low-NOx Burners (ULNB): Most commonly a combination of LNB technology with some internal flue gas recirculation. The burner recirculates some of the hot flue gases from the flame
or firebox back into the combustion zone. Since these high temperature flue gases are oxygen depleted, the burner lowers the speed at which fuel can be combusted without reducing the flame temperature below the level needed for optimum combustion efficiency. Reducing oxygen concentrations in the firebox most directly impacts fuel NOx generation. Flue Gas Recirculation (FGR): External FGR involves recycling of flue gas back into the firebox as part of the fuel-air mixture at the burner. Although similar to the concept of ULNB, rather than using burner design features to recirculate gases from within the firebox, FGR uses external
ductwork to route a portion of the exhaust stream back to the inlet side of the boiler and return it into the boiler windbox.
In the SCR process, a reducing agent, such as aqueous ammonia, is introduced into the exhaust, upstream of a metal or ceramic catalyst. As the exhaust gas mixture passes through the catalyst bed, the reducing agent selectively reduces the nitrogen oxide compounds present in the exhaust
to produce elemental nitrogen (N2) and water (H2O). Ammonia is the most commonly used reducing agent. Adequate mixing of ammonia in the exhaust gas and control of the amount of ammonia injected (based on the inlet NOx concentration) are critical to obtaining the required
reduction. For the SCR system to operate properly, the exhaust gas must maintain minimum O2 concentrations and remain within a specified temperature range (typically between 480ºF and 800ºF with the most effective range being between 580°F and 650°F), with the range dictated by
the type of catalyst. Exhaust gas temperatures greater than the upper limit (850°F) will pass the NOx and unreacted ammonia through the catalyst. The most widely used catalysts are vanadium, platinum, titanium, or zeolite compounds impregnated on metallic or ceramic substrates in a plate of honeycomb configuration. The catalyst life expectancy is typically 3 to 6 years, at which time the vendor can recycle the catalyst to minimize waste. One final technology is CETEX, which is a process of descaling firebox steam tubes and then recoating these tubes – improving heat transfer and lowering total fuel consumption for a given
amount of steam output. 5.3.2 Evaluation of Technical Feasibility of Available Controls
Most of the listed controls are technically feasible, although certain control techniques cannot be used in conjunction. For example, ULNB and FGR both use some degree of flue gas
recirculation, making the use of both technologies redundant and counter-productive. FGR can only be applied on mechanical draft heaters/boilers with burners that can accommodate increased gas flows. Secondly, in order to physically connect FGR, a separation of at least three feet must
exist between the windbox and the burner to prevent the accumulation of potentially explosive gas mixtures if a heater tube should fail. Heaters F-1, H-101, F-680, and F-681 are already equipped with ULNB. Both F-15 and F-701 are small units less than 40 MMBtu/hour and have small furnace boxes unable to accommodate the FGR retrofit described above. FGR is eliminated from further consideration.

16
Next generation ULNB have proven less reliable than “normal” ULNB. They have been prone to
plugging from impurities present in the fuel source. They also have not been shown to have any better emissions control than normal ULNB.
SCR and SNCR are opposed technologies as well, as SNCR is the use of ammonia injection without the added benefit of a catalyst bed to aid in pollutant reduction. Strictly speaking, SCR is simply the most commonly used catalytic reduction technique. More generally, NSCR (or non-
selective catalytic reduction) represents those catalytic reduction techniques using alternative catalysts also capable of reducing NOx. Often these catalysts reduce NOx in addition to many other compounds and are not specifically designed or optimized for NOx reduction. Efficient operation of these catalysts typically requires that the exhaust gases contain low oxygen concentrations – perhaps as low as 0.5% and no more than 4%. Since this requires the use of lean burn engines, furnaces and boilers, NSCR was eliminated by Marathon as a valid control technique.
SNCR was eliminated based on a lack of load responsiveness. The control technique is somewhat crude by current standards, using only ammonia injection to reduce NOx emissions and relying on turbulence, temperature and residence time to allow the reaction to come to
completion. The technical limitation of this technique is that as load changes NOx emissions can vary and change fairly rapidly, yet the only control mechanism is to change the amount of ammonia being injected into the flue gas. Periods of too little ammonia will be followed by
periods of too much ammonia, both scenarios leading directly to an increase in PM2.5 emissions. SNCR is eliminated from consideration.
The EMx process is highly sensitive to poisoning from sulfur compounds present in the exhaust gas. EMx has never been demonstrated in practice on refinery fuel gas-fired heaters or boilers and is not deemed commercially available for this fuel type. EMx is therefore eliminated from
consideration. Water/steam injection is of limited effectiveness on process heaters and boilers. While some NOx reduction is possible, the benefit gained is minimal on units already equipped with low-NOx burners or better. Water/steam injection is typically employed on turbines where the increased mass of the steam-laden exhaust gas increases the efficiency of the turbine by improving the momentum transfer to the power generator. It is not typically employed on process heaters or boilers. This technology has been eliminated from further consideration.
Low excess air firing (overfire/staged air combustion) was also eliminated based on flame lengthening problems. Reducing the oxygen concentration causes the combustion flame to
lengthen, potentially causing flame impingement – where the flame comes into physical contact with one or more surfaces of the unit. This can cause severe damage to the unit and hazardous safety situations for refinery workers. This technique has also been eliminated from further
consideration. 5.3.3 Evaluation and Ranking of Technically Feasible Controls
The remaining control techniques and their control effectiveness are listed below: Table 5-1 NOx Control Effectiveness Technology Range of Control (%)

17
ULNB + SCR 85-99
LNB + SCR 80-99
SCR 80-90 ULNB 65-75 LNB 50-60
CETEX NA* * Not applicable, see below Based on these control efficiencies, the use of SCR in combination with some form of NOx
controlling burner (either LNB or ULNB) is the top ranked control option. The use of SCR without a burner upgrade can also be applied, but is rarely found in practice. The second ranked option is the use of ULNB alone, followed by LNB. Of the various process heaters at the
Marathon refinery, none are “boilers” in the traditional sense, requiring descaling such as that described under CETEX. Rather, the process unit may be described as a boiler or reboiler and is merely a process vessel with associated direct heating. Operations similar to CETEX could be
performed on the rental package boilers, but would require disassembly of those units – something typically not performed on rental equipment. Thus CETEX is eliminated from further consideration.
5.3.4 Further Evaluation of Most Effective Controls Marathon provided additional analysis for the technically feasible controls4. Installing SCR can have adverse energy and environmental impacts. One potential source of concern with operation of SCR is the generation of ammonia slip. Unreacted ammonia, meaning any ammonia which does not react with the NOx present in the exhaust stream, may react with SO3 to form ammonium sulfate/sulfite. This can occur either in the exhaust stream or in the ambient air. The unreacted
ammonia is referred to as “ammonia slip.” Ammonia slip itself often requires permit limitations as a precursor pollutant.
Another source of concern is handling and disposal of the spent catalyst, which becomes a solid waste product. Operation of the SCR system reduces air flow, requiring additional energy in the form of fan power upgrades. Marathon has not completed an engineering design analysis of
installing SCR on any of the process heaters at the refinery and believes that design, construction, shake-down and testing could not be completed by December 31, 2018 – making such an endeavor technically infeasible. Even without considering the timing considerations, Marathon
completed an economic analysis and determined that no additional controls are economically feasible, with a control cost in excess of $165,000/ton of NOx removed.5 Marathon has also expressed concerns about space availability for locating SCR on many of the process heaters, but SCR was not considered to be technically feasible for H-101 and F-1 because there is no nearby available plot space. Given these concerns SCR has been eliminated as BACT. Installation of LNB or ULNB can change the heating pattern of the furnace or boiler by extending and cooling the flame.
Marathon already operates with ULNB on four of the six process heaters. The remaining two units F-15 and F-701 are the smallest units and operate only with LNB in place. Installation of
ULNB on these two heaters is not technically feasible due to flame impingement5. A review of NOx emissions from these two units shows a total of 4.96 tons of NOx in 2016. Based on the 4 see References: Item #8 5 See References: Item #9

18
difference in expected control efficiencies, perhaps a total of 1.65 tons of NOx could be reduced by switching to ULNB on both units. Given this low amount of possible reduction, and that
ULNB are not technically feasible no additional upgrades are recommended. 5.3.5 Selection of BACT Controls
UDAQ recommends the existing NOx controls remain as BACT. Marathon will comply with any applicable emission limits in Section IX, Part H.11.This section also contains additional
monitoring, recordkeeping and reporting requirements. These practices are required through existing permit requirements. While no additional controls are required for BACT, UDAQ recommends additional stack testing requirements be added to bolster existing monitoring, recordkeeping, and reporting requirements. UDAQ has added additional limits for all process heaters and boilers with a capacity greater than 40 MMBtu/hr. This threshold is based on an established threshold in 40 CFR 60.102a for NOx limitations on process heaters, which was established based on the application of the best system
of emission reduction while taking into consideration costs and impacts. Based on the existing NOx controls, UDAQ has established the following additional emission limits as BACT in Section IX, Part H.12:

• Crude Unit Furnace H-101: 0.054 lb/MMBtu
• UFU Furnace F-1: 0.065 lb/MMBtu See Appendix A for a complete listing of existing permit conditions, established emission limits,
and the basis of those limits. For a summary of the potential emissions from each process heater and boiler with a capacity greater than 40 MMBtu/hr, see Appendix C. These calculations depend on the maximum NOx emissions each heater and boiler could have emitted in 2017 at the maximum of the above limits, compared to the actual emissions reported in the 2017 inventory. The actual emissions from 2017 are representative of normal operation throughout the year based
on process and operation variability, existing stack tests, and established emission factors. Marathon will still comply with all existing permit and SIP requirements. 6.0 BACT for Cogeneration Units The cogeneration units are a pair of gas-fired combustion turbines which produce electricity for
the refinery. Each turbine’s exhaust gas is also sent to a pair of heat recovery steam generators (HRSGs) that use the waste heat to produce steam for use in the refinery. The turbines are fired on natural gas and refinery fuel gas. The HRSGs heat input is supplemented with refinery fuel
gas. Passive NOx control is provided through the design of the turbines – which are SoLoNOx lean pre-mix combustion technology. 6.1 PM2.5 6.1.1 Available Control Technology Controls for particulate emissions fall into one of three groups: pre-combustion controls, which seek to eliminate contaminants in the inlet air prior to the combustion chamber; combustion controls, such as specific burners or combustion design; and post-combustion controls, such as electrostatic precipitators or baghouses.
The identified controls are as follows:

19
Inlet air filters: primarily used to filter out small particulate matter in the inlet air to protect the
combustion turbine. These filters can be static or self-cleaning, with the self-cleaning type requiring less maintenance.
Good combustion practice: this is nothing but properly operating the combustion turbines with the correct ratio of air to fuel in order to maximize combustion and minimize unburned fuel.
Clean burning fuels: includes the use of inherently low emitting fuels like natural gas. Specific burner and/or combustion chamber design: the more efficiently a turbine is able to operate, the less pollution it will generate for a given amount of fuel combusted (or, to be more precise); as less fuel will be required to generate the same amount of power. This option includes both the use of high efficiency turbines, as well as inherently lower emitting burners such as dry low-NOx (DLN) combustors.
Add-on particulate controls: this final option includes traditional “add-on” control systems such as baghouses or electrostatic precipitators. These types of controls would be installed post combustion, and prior to the emissions exiting the stack.
6.1.2 Evaluation of Technical Feasibility of Available Controls
As a refinery, Marathon is limited in the amount of fuel switching it can perform. Although some amount of natural gas is burned in the combustion turbines, refinery fuel gas is generated during the refining process and must be consumed in some fashion. The Marathon refinery was
designed to combust this fuel gas in cogeneration turbines supplementing Marathon’s power needs and limiting the amount of steam generation required by direct fired boilers. Switching the cogeneration units to run on exclusively natural gas would force diversion of the refinery fuel gas
to the refinery flares – which would result in flow rates in excess of the refinery flare cap. This would also not reduce emissions, as the reduction from the turbines would be offset by the increased flaring emissions. Fuel switching is considered technically infeasible. Post-combustion particulate controls such as baghouses and electrostatic precipitators have not been demonstrated in practice for use on combustion turbines. There are multiple factors that combine to eliminate these types of controls from consideration. 1) Combustion turbine particulate emissions have a small aerodynamic diameter – typically on the order of 1 micron or
less – which makes the use of most direct physical capture systems problematic. 2) gaseous fuel-fired turbines generate little in the way of particulate emissions; yet also have high volume exhaust flows. This combination results in a low concentration of PM in the exhaust. 3) Post-
combustion controls have difficulty operating efficiently or effectively in low concentration environments. Baghouse-style filtration systems rely on the buildup of a filter cake of captured particulates to enhance capture efficiency, while scrubbing systems require a reasonable
particulate concentration in order to operate efficiently. Electrostatic precipitators can operate in low concentration conditions, but also suffer efficiency problems. In addition, a search was conducted for the use of ESPs with gaseous fuel-fired turbines and no results were found. A
single article which discussed a natural gas-fired bench-scale experiment was found, but no commercially available results6. The UDAQ was unable to identify any combustion turbines fired on gaseous fuels using post combustion controls for the control of particulate emissions. Post-combustion controls are therefore technically infeasible and removed from additional

6 see supplemental reference material

20
consideration.
All of the remaining control options are considered technically feasible and require additional evaluation. 6.1.3 Evaluation and Ranking of Technically Feasible Controls The remaining control options under consideration are not mutually exclusive. A high-efficiency
combustion turbine can be operated with inlet air filters and using good combustion practices. The turbine can be fired exclusively on gaseous fuel, and use properly designed combustors. Thus, the remaining controls do not need to be ranked – rather they need to be combined and considered as a group. All of the remaining control options can be combined into this group (effectively: “high efficiency combustion”), so no further evaluation under step 3 is required.
6.1.4 Further Evaluation of Most Effective Controls Marathon did not evaluate the use of inlet air filters, but did review the remaining items and
propose that the use of gaseous fuels, good combustion practices and a properly designed combustor should be considered BACT. UDAQ has evaluated the use of inlet air filters in its review of gas-fired turbines installed at power generation facilities (see e.g. the BACT analyses
for PacifiCorp Gadsby, PacifiCorp Lake Side, UAMPS West Valley Power Plant). Inlet air filters function best on large combustion turbines with large volume air flows such as an F-Class turbine. Smaller turbines such as these cogeneration turbines receive little benefit from pre-
filtering the inlet air. 6.1.5 Selection of BACT Controls
UDAQ recommends that retention of the existing control techniques of good combustion practices and use of only gaseous fuel (refinery fuel gas and natural gas) be considered as BACT. As work practice standards, no limitation on emissions is required. These practices are required through existing permit requirements and standards which have been established in Section IX, Part H.11.g. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions,
established emission limits, and the basis of those limits. 6.2 SO2
Emissions of SO2 are directly a function of the amount of sulfur present in the fuel. As the fuel is burned, the fuel-bound sulfur is oxidized to SO2.
6.2.1 Available Control Technology
Most sulfur control technologies require the use of some sort of acid reducing agent such as a lime slurry or limestone injection. This leads to residual solid or liquid waste which requires subsequent disposal. The remaining add-on control techniques rely on the post-combustion control of emissions of particulates and allowing any residual sulfur to be captured with the particulate. A second option would be to reduce the amount of sulfur present in the fuel, thus eliminating the source of the SO2.

21
6.2.2 Evaluation of Technical Feasibility of Available Controls
Neither of the possible control options is technically feasible. The turbines and HRSGs are fired on a combination of natural gas and refinery fuel gas (the HRSGs are fired only on refinery fuel
gas). The refinery fuel gas is treated prior to being used in the refinery and limited in fuel-sulfur content. Although refinery fuel gas is not as low in sulfur content as natural gas it is still considered a low-sulfur fuel. Marathon is required to meet the long-term Subpart Ja refinery fuel
gas H2S limit of 60 ppmvd as well as the existing short term Subpart Ja limit of 162 ppmv on a 3-hour average. Fuel switching to run on natural gas exclusively has been discussed in the PM2.5 section (6.1.2 above) and is not technically feasible nor will it result in any decrease in emissions. Post-combustion desulfurization systems, such as limestone injection or dry-lime scrubbing, are typically designed for exhaust streams with much higher SO2 (and acid gas) concentrations than those found with combustion turbines fired on low-sulfur gaseous fuel. The low concentration leads to lowered control efficiencies. Effective control then requires longer residence times,
longer exhaust stream runs, lowered exhaust temperatures, and worsened emission dispersal upon release. 6.2.3 Evaluation and Ranking of Technically Feasible Controls Only the use of low-sulfur gaseous fuels was determined to be technically feasible. As this fuel is
currently in use at Marathon’s cogeneration units, no further analysis is required. 6.2.4 Further Evaluation of Most Effective Controls
N/A, the only technically effective control is currently in use at the cogeneration units. 6.2.5 Selection of BACT Controls UDAQ recommends that Marathon continue to operate the cogeneration units using low sulfur gaseous fuel meeting the refinery fuel gas requirements of NSPS Ja. These practices are required through existing permit requirements and Section IX, Part H.11.g. of the SIP. The existing limits account for current process variability while still limiting SO2 emissions from the cogeneration units. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. No additional controls are required for
BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits.
6.3 NOx
NOx, or oxides of nitrogen, are formed from the combustion of fuel in the turbine. There are three mechanisms for the formation of NOx: fuel NOx, which is the oxidation of the nitrogen bound in the fuel; thermal NOx, or the oxidation of the nitrogen (N2) present in the combustion air
itself; and prompt NOx, which is formed from the combination of combustion air nitrogen (N2) with various partially-combusted intermediary products derived from the fuel. For combustion within the turbines, fuel NOx and thermal NOx are the major contributors, with prompt NOx contributing slightly only in the initial stages of combustion. All three processes are temperature dependent – combustion temperatures below 2700ºF greatly inhibit NOx formation.

22
6.3.1 Available Control Technology
The following technologies have been identified as potential control methodologies for control of NOx emissions: good combustion practices; low emission combustion (LEC); selective non-catalytic reduction (SNCR), the injection of ammonia or urea directly into the late stages of the
combustion zone; selective catalytic reduction (SCR); and EMx™ (previously known as SCONOx™).
In the SCR process, a reducing agent, such as aqueous ammonia, is introduced into the turbine’s exhaust, upstream of a metal or ceramic catalyst. As the exhaust gas mixture passes through the catalyst bed, the reducing agent selectively reduces the nitrogen oxide compounds present in the exhaust to produce elemental nitrogen (N2) and water (H2O). Ammonia is the most commonly used reducing agent. Adequate mixing of ammonia in the exhaust gas and control of the amount of ammonia injected (based on the inlet NOx concentration) are critical to obtaining the required reduction. For the SCR system to operate properly, the exhaust gas must maintain minimum O2 concentrations and remain within a specified temperature range (typically between 480ºF and
800ºF with the most effective range being between 580°F and 650°F), with the range dictated by the type of catalyst. Exhaust gas temperatures greater than the upper limit (850°F) will pass the NOx and unreacted ammonia through the catalyst. The most widely used catalysts are vanadium,
platinum, titanium, or zeolite compounds impregnated on metallic or ceramic substrates in a plate of honeycomb configuration. The catalyst life expectancy is typically 3 to 6 years, at which time the vendor can recycle the catalyst to minimize waste.
The EMx™ system uses a coated oxidation catalyst installed in the flue gas to remove both NOx and CO without a reagent such as ammonia. The NO emissions are oxidized to NO2 and then
absorbed onto the catalyst. A dilute hydrogen gas is passed through the catalyst periodically to de-absorb the NO2 from the catalyst and reduce it to N2 prior to exit from the stack. EMx™ prefers an operating temperature range between 500°F and 700°F. The catalyst uses a potassium
carbonate coating that reacts to form potassium nitrates and nitrites on the surface of the catalyst. When all of the carbonate absorber coating on the surface of the catalyst has reacted to form nitrogen compounds, NO2 is no longer absorbed, and the catalyst must be regenerated. Dampers are used to isolate a portion of the catalyst for regeneration. The regeneration gas consists of steam, carbon dioxide, and a dilute concentration of hydrogen. The regeneration gas is passed through the isolated portion of the catalyst while the remaining catalyst stays in contact with the flue gas. After the isolated portion has been regenerated, the next set of dampers close to isolate and regenerate the next portion of the catalyst. This cycle repeats continuously. At any one time,
four oxidation/absorption cycles are occurring and one regeneration cycle is occurring. Two additional post-combustion control systems were also identified as being potentially
applicable: Linde’s LoTOx™ technology uses ozone injection to oxidize NO and NO2 to N2O which is
highly soluble and easier to remove through the use of another control device such as a wet scrubber. UDAQ has seen and permitted the application of this technology in combination with a wet gas scrubber for emission control at a petroleum refinery.
Enviroscrub’s Pahlmann™ Process is a sorbent-based control system which functions similarly to a dry scrubber. In this system, Pahlmanite (a manganese dioxide sorbent) is injected into the exhaust stream for NOx removal and then collected in a particulate control device like a baghouse. The sorbent is then regenerated in an aqueous process, filtered and dried, and is then ready for reinjection. The wastewater is sent offsite for disposal.

23
6.3.2 Evaluation of Technical Feasibility of Available Controls
All of the identified control options are potentially technically feasible; however, some additional explanation is warranted:
In the case of LEC, more than one variant of combustor design exists:
• Dry-low-NOx: The modern, dry low-NOx (DLN) combustor is typically a three-staged, lean, premixed design, which utilizes a central diffusion flame for stabilization. The lean, premixed
approach burns a lean fuel-to-air mixture for a lower combustion flame temperature resulting in lower thermal NOx formation. The combustor operates with one of the lean premixed stages and the diffusion pilot at lower loads and the other stages at higher loads. This
provides efficient combustion at lower temperatures, throughout the combustor-loading regime. The dry low-NOx combustor reduces NOx emissions by up to approximately 87 percent over a conventional combustor.
• Catalytic combustors: These combustors use a flameless catalytic combustion module to initiate the combustion process, followed by a more traditional combustion process
downstream of the catalyst. This two-stage process lowers the overall combustion temperature.
• Xonon Cool Combustion®: Catalytica Energy Systems’ Xonon Cool Combustion® System is a specific type of catalytic combustion process, and often mentioned independently in control technology reviews. In practical application, however, it functions similarly to other catalytic
combustors. Along with these types of burner designs, another pre-combustion process – water or steam
injection – can also be used to lower the combustion temperature. Depending on the amount of water or steam used, this process can also increase both the maximum and actual power output of the turbine – by allowing more fuel to be burned without overheating, and by increasing the
density of the exhaust flow through the turbine. However, water and steam injection tend to reduce combustion efficiency, prevent complete combustion leading to an increase in CO and VOC emissions, and are of limited effectiveness in combined cycle systems (turbine/HRSG systems) where the lowered temperature and increased specific heat of the turbine exhaust gas directly results in an increased need for duct firing in the HRSG unit. Neither the LoTOx™ nor Pahlmann™ processes are determined to be technically feasible. While the LoTOx™ system is technically feasible from a mere engineering standpoint, it suffers from two flaws. It has not yet reached the commercial stage for large scale, combined-cycle, combustion turbines; and it requires the use of a second control system, such as a wet gas scrubber, for final removal of the N2O. In the application of LoTOx™ UDAQ has previously
permitted, the system was included as an additional module to a wet gas scrubber designed for removal of SO2 and other acid gases. Achieving additional NOx removal at relatively low cost (on a $/ton basis) was the ideal fit for this technology. However, requiring the addition of another
control system for final pollutant removal, especially where the secondary system does not add to emission reduction of other pollutants, demonstrates that LoTOx™ is not yet technically feasible. Similarly, the Pahlmann™ Process also requires the addition of: a baghouse for particulate
removal (for capture of the sorbent), an aqueous sorbent regeneration process, and a wastewater treatment/disposal process. While the technology does show promise for control of multiple pollutants, it was not intended for control of only the NOx emissions from gas-fired turbines and is not commercially available for such units. Both processes are eliminated from further consideration.

24
The other control options (SNCR, SCR, good combustion practices, and burner design) have all
been found to be technically feasible by UDAQ for various combustion turbine installations. However, Marathon supplied counter-arguments for some of these technologies. Concerning water/steam injection, Marathon contacted the manufacturer for a feasibility determination7. The
manufacturer stated that water/steam injection was not available as a control system for that type of combustion turbine. Marathon also argued that the design, installation, shakedown and testing period for SNCR, SCR, or installation of a new combustion turbine/HRSG system was beyond
the December 31, 2018 attainment date, even beyond any extension of that attainment date – were such to be granted. UDAQ took these arguments into consideration. 6.3.3 Evaluation and Ranking of Technically Feasible Controls The combustion turbines installed at the Marathon Refinery are built around the SoLoNOx combustor; Solar Turbines’ specific brand of dry-low-NOx combustor. This particular system is a lean pre-mix burner design, which uses a combination of staged combustion and differing fuel-air
mixing for each combustion stage to both lower the combustion temperature and still allow for complete combustion.
The use of additional pre-combustion controls is not technically feasible, regardless of the type of combustion technology chosen. As discussed in the previous section, the reduction in exhaust temperature and increased specific heat greatly increase the need for duct firing in the HRSG – a
much less efficient process for generating power. Switching to a Xonon-based combustor is not a viable option. Further review shows that the
catalyst in Xonon systems is highly sensitive to sulfur compounds present in the exhaust gas, even a low sulfur fuel like refinery fuel gas contains too many sulfur-based compounds which will lead to poisoning the catalyst. The system was designed for use on pipeline quality natural
gas-fired turbines only, and was never intended to be operated in a refinery setting. In comparing SNCR and SCR, two factors come into play – the hypothetical effectiveness of control, and whether either system will work more effectively in the specific design of Marathon’s cogeneration units. In all of the SNCR designs available, ranging from simple ammonia/urea injection in the main combustion zone through Fuel Tech’s NOxOUT™ process, every one requires an exhaust gas temperature somewhere between 1600ºF and 2100ºF for most effective conversion of NOx to N2. SCR systems, on the other hand, use a catalyst to lower this
effective temperature range down to between 480ºF and 800ºF. The exhaust temperature from the Marathon cogeneration units is approximately 320ºF, below even the effective temperature for SCR systems.
Marathon has not completed an engineering design analysis of installing SCR and believes that design, construction, shake-down and testing could not be completed by December 31, 2018 –
making such an endeavor technically infeasible. Even without considering the timing considerations, Marathon completed an economic analysis and determined that no additional controls are economically feasible, with a control cost in excess of $23,000/ton of NOx removed.8
Combined with Marathon’s feasibility arguments, only the existing burner design and good combustion practices remain as viable options for NOx control.
7 see References: Item #8 8 See References: Item #9

25
6.3.4 Further Evaluation of Most Effective Controls
N/A The only remaining viable controls are both in effect on the cogeneration turbines/HRSG units.
6.3.5 Selection of BACT Controls
UDAQ recommends that the combustion turbines continue to be operated in conjunction with good operating practices. These practices are required through existing permit requirements established as BACT during the cogeneration turbines construction and permit issuance. Testing is already established in existing permit conditions. While no additional controls are required for BACT, UDAQ recommends additional stack testing requirements be added to bolster existing monitoring, recordkeeping, and reporting requirements. UDAQ has included a NOx limit of 32 ppmdv at 15% O2 dry for each cogeneration unit. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits.
6.4 Consideration of VOC and Ammonia
VOC emissions are the result of unburned hydrocarbons formed during incomplete combustion. To some degree the formation of VOCs is dependent on combustion system design, choice of fuel, combustion temperature (itself dependent on equipment design and operating practices), and
operating practices (which can control the air-to-fuel ratio, timing, temperature, and other factors).
Only one type of post combustion control has been identified by UDAQ - the use of oxidation catalysts. An oxidation catalyst is similar in design and operation to a catalytic control system on a passenger vehicle, in that an inline, self-regenerating, catalyst system is placed within the
exhaust stream prior to the final stack, so that emissions of both VOC and CO can be further oxidized to CO2 and water. Oxidation of VOC can approach efficiencies of 70%, depending on initial concentrations and stack characteristics. However, efficiency drops off rapidly as concentration decreases – to the point that catalytic (and thermal) oxidation is ineffective below inlet VOC concentrations of 100 ppm. The VOC concentration from the cogeneration units is estimated at 13 ppm (or less). Oxidation catalysts are deemed technically infeasible. Control techniques such as fuel switching are not helpful since gaseous fuels such as refinery fuel
gas and natural gas are already the best available. The only control technique remaining is the use of good combustion practices (GCP). As GCP are already required or included as a part of the control techniques for the other pollutants listed previously no additional consideration is
required. There are few emissions of ammonia from the cogeneration units naturally (some minor amounts
of ammonia may be generated as part of the combustion process). Ammonia emissions would be more of a concern if SCR or SNCR had been chosen as a viable control option. However, as no ammonia injection is being used, no ammonia slip can result. UDAQ does not recommend
ammonia controls on the cogeneration units at this time. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11. 7.0 BACT for the SRUs

26
7.1 SO2
Marathon operates a single well-controlled sulfur recovery plant meeting the established 95% sulfur recovery required under the PM10 SIP (SIP Section IX, Part H.1). Generically, the sulfur recovery systems at the various refineries located in the PM2.5 non-attainment areas are referred to
as sulfur recovery units or SRUs. For purposes of this review a “well-controlled SRU” is one that is already operating with a tail gas treatment system followed by tail gas incineration.
There are only two pollutants of concern from a well-controlled SRU: SO2 and NOx. The system is designed to remove sulfur (primarily in the form of H2S) from the refinery fuel gas through a combination of catalytic treatment and combustion. Particulate emissions are extremely low as the only emission source from the entire process is the Tail Gas Incinerator. This unit, fired on refinery fuel gas, is similar to other process heaters located at the refinery. The Tail Gas Incinerator oxidizes H2S to SO2, but neither gas is considered a particulate. VOC emissions are also low, as few unburned VOCs are present in the inlet streams passed to the SRU, and the catalytic process does not generate additional VOCs. VOC generation from combustion of
gaseous fuels is minimal. This review will focus on the control of SO2 and NOx from operation of the Marathon SRU.
The SRU at the Marathon refinery operates as follows: The sour water stripper gas and amine acid gas are sent to a burner to convert some of the H2S
into SO2, and all of the ammonia to nitrogen (N2). The heated gas is fed to the first of three reactor stages where the SO2/H2S mixture is converted to sulfur vapor over a catalyst bed – generating extra heat in the process. The sulfur vapor is cooled and separated, while the
remaining mixture is reheated via heat exchanger. This cycle is repeated a total of three times (once time in each of the three reactor stages). After the last stage, the remaining gas, now called tail gas, is sent to the Tail Gas Treating Unit (TGTU). The separated sulfur, now in liquid form,
is drained into a sulfur pit for temporary storage until it can be sold as elemental sulfur. In the TGTU, the tail gas is reduced back to H2S for additional capture by an amine absorber and sent back to the front of the SRU. The outlet stream from the TGTU is sent to a thermal oxidizer (the Tail Gas Incinerator or TGI) which controls the reduced sulfur (H2S) emissions. The TGI uses refinery fuel gas as a fuel source. 7.1.1 Available Control Technology
Three control systems were identified to further control emissions from a well-controlled SRU. For purposes of this review a “well -controlled SRU” is one that is already operating with a tail
gas treatment system followed by tail gas incineration. • LoCat
• WGS • Caustic Scrubbing
LoCat is unusual in that it can serve as both a final treatment following the SRU (both in addition to, or in-lieu of a tail gas unit) or as a fuel gas sulfur removal unit (in case the SRU itself goes down). WGS is a final control option, where the exhaust from the SRU is sent to the WGS in-lieu of tail gas treatment. It can also be utilized following application of a TGTU, but would rarely be

27
implemented following a TGI.
Caustic scrubbing is typically used as a replacement for a SRU, such as a redundant back-up device, but can also be used as a final scrubbing process.
Marathon also identified a fourth option, applicable only to the Marathon refinery, which it termed a “sulfur shedding plan.9” This option is to reduce SO2 emissions by managing H2S generation in the refinery during upset conditions. Marathon has an established limit on SO2
emissions from the SRU at 60 tons per year. This limit was part of the negotiations for the SO2 SIP, resulting in the following limit established in the January 11, 2018 AO:
SO2 emissions from the SRU/TGTU/TGI shall be limited to: A. 1.68 tons per day (tpd) for up to 21 days per rolling 12-month period, and B. 0.69 tpd for the remainder of the rolling 12-month period. Both this limit and the previously established 60 ton limit can be found in the latest NSR permit
issued to the Marathon refinery (DAQE-AN103350075-18) as condition II.B.5.b. 7.1.2 Evaluation of Technical Feasibility of Available Controls
All controls are technically feasible. 7.1.3 Evaluation and Ranking of Technically Feasible Controls Although all three technology options are technically feasible, none is a good option as an add-on
control. Well-controlled SRUs can achieve 99.9% or better sulfur recovery/destruction efficiency rates. Marathon installed a TGTU and TGI as an upgrade to the existing SRU as part of its Waxy Crude Processing Project beginning in 2013 (DAQE-AN103350058B-13). The project was
completed in late 2015/early 2016, and the baseline emissions from 2016 reflect the drop in SO2 emissions from this change. Prior to 2016, actual emissions from the SRU averaged approximately 195 tons of SO2 annually. Beginning in 2016, total SO2 emissions dropped to around 30 tons annually. Marathon’s sulfur shedding plan is considered a work practice standard, which is demonstrated through reduced emissions during upset conditions. Due to being a voluntary work practice standard during upset conditions only, drafting a limitation or enforceability provision will not be
investigated further. Marathon will continue to implement sulfur shedding and report the plan as corrective actions taken during startup, shutdown, and malfunction events. 7.1.4 Further Evaluation of Most Effective Controls None of the control options will effectively reduce emissions below the levels already achieved.
Although any of the control options could be applied in lieu of the existing controls, and either LoCat or WGS could be applied in addition to the existing controls, the costs of these additional add-on measures would be well above $250,000/ton.
7.1.5 Selection of BACT Controls UDAQ recommends that Marathon continue to operate the TGTU and TGI as SO2 control for the

9 see References: Item #8

28
SRU and continue implementing the sulfur shedding plan. Marathon is subject to the PM10 SIP and PM2.5 SIP refinery requirements found at Section IX Parts H.1.g, H.2.k.iv, and H.11.g. These
practices are required through existing permit requirements. These emission limitations were established based on the application of the best system of emission reduction while taking into consideration costs and impacts. The existing limits account for current process variability while
still limiting SO2 emissions from the SRU. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. No additional controls are required for BACT; thus, no additional limits are required to be
established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 7.2 NOx In the traditional Claus SRU process (the process used at Marathon), the reaction furnace is operated in a reducing environment, so that any ammonia in the acid gas feed is reduced to N2.
The SRU feed results in little fuel-based NOx, and the heat within the reaction furnace is typically below the thermal NOx threshold of 1200ºF. Any thermal NOx generated is typically formed at the TGI, and this would be a minimal amount.
7.2.1 Available Control Technology
Marathon identified only a single control technology as being available and technically feasible for control of NOx emissions from its SRU – Good Combustion Practices (GCP). UDAQ’s review has shown that one other control technology is potentially viable as it is already in use at
another local refinery. WGS has been shown to potentially reduce NOx emissions from a SRU when combined with a LoTOx unit. 7.2.2 Evaluation of Technical Feasibility of Available Controls All controls are technically feasible. The use of GCP is the baseline case as it is already included as part of the operational requirements for the SO2 limits on the SRU (see Section 7.1). WGS with LoTOx is technically feasible as it is already in use at another area refinery. However, inclusion of LoTOx requires a large enough concentration of NOx to be operationally viable. Marathon estimates that NOx emissions from the SRU are limited to 0.1 lb/MMBtu. A review of
the 2016 baseline emission inventory10 shows estimated NOx emissions from the SRU of just 0.56 tons annually. These values are below the levels necessary for LoTOx to be considered operationally viable. This technology will not be investigated further.
7.2.3 Evaluation and Ranking of Technically Feasible Controls
With the elimination of WGS, Marathon is operating with the only remaining control option – GCP. No ranking or further evaluation is required. 7.2.4 Further Evaluation of Most Effective Controls N/A The only remaining control option is already in use at the Marathon SRU.

10 see References: Item #10

29
7.2.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to operate the SRU consistent with good operating practices. This is required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon
will comply with any applicable emission limits in Section IX, Part H.11. 8.0 BACT for Fugitives
In this context, fugitives are referring to fugitive VOC emissions. While Marathon does have fugitive dust emissions from items such as roads, spill containment berms, and similar earthworks – particulate emissions from these items have been evaluated separately. Please refer to the PM2.5 Serious SIP - BACT for Small Sources – Section 12 for the evaluation. Fugitive VOC emissions are those emissions that result from the various pipe connections; feedstock, intermediary, and product transfer activities; loading and unloading operations; and
any and all equipment leaks. They do not typically include the VOC emissions from storage vessels (storage tanks), cooling towers, or wastewater treatment. 8.1 VOCs 8.1.1 Available Control Technology
The only available control option is the low-leak LDAR program as outlined in 40 CFR 60 Subpart VVa and incorporated by reference (with some source category modifications) in 40 CFR
60 Subpart GGGa. Each refinery (including Marathon) became subject to the requirements of low-leak LDAR (Subpart GGGa) as part of the requirements of the moderate PM2.5 SIP. 8.1.2 Evaluation of Technical Feasibility of Available Controls N/A Low-leak LDAR is technically feasible, and Marathon became subject to its requirements on January 1, 2017. 8.1.3 Evaluation and Ranking of Technically Feasible Controls N/A Marathon is already implementing the requirements of 40 CFR 60 Subpart GGGa.
8.1.4 Further Evaluation of Most Effective Controls
N/A Marathon is already implementing the requirements of 40 CFR 60 Subpart GGGa. 8.1.5 Selection of RACT Controls
UDAQ recommends that Marathon continue to implement the general refinery SIP requirements regarding Leak Detection and Repair as outlined in Section IX, Part H.11.g. These practices are
required through existing permit and SIP requirements. No additional controls are required for BACT; thus, no additional limits other than those established in H.11.g are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11. 9.0 BACT for Wastewater System

30
9.1 VOC
The wastewater treatment system consists primarily of a system of drains that route runoff water and storm water to the API separator, which separates entrained oils and volatiles from the
wastewater. Marathon currently operates the API separator with a floating roof cover to limit VOC emissions. 9.1.1 Available Control Technology Three control options were identified to reduce VOC emissions from the wastewater system. API separator dual seal floating covers reduce vapor space and minimize fugitive emissions. In addition, the collected vapors from a fixed roof API separator can be routed to a control device for capture or destruction. Carbon canisters reduce emissions by capturing the VOCs using activated carbon filtration. Oxidation, using either thermal treatment or catalytic oxidation systems, is also a viable option for elimination of VOC emissions.
9.1.2 Evaluation of Technical Feasibility of Available Controls
API separator dual-seal covers are technically feasible. Marathon stated that designing, installing, and beginning operation of either carbon adsorption or oxidation technologies could not be completed by the regulatory attainment date of December 31, 201811.
9.1.3 Evaluation and Ranking of Technically Feasible Controls
Installation of either carbon adsorption or oxidation control systems would be equally effective in controlling VOC emissions from the wastewater treatment plant. API separator dual-seal covers
are less effective. 9.1.4 Further Evaluation of Most Effective Controls Marathon’s API floating covers do not lend themselves to the permanent installation of duct work and capture hoods as would a fixed cover. UDAQ agrees that additional add-on controls, such as RTO or the use of carbon canisters, are economically infeasible and are eliminated from further
consideration as BACT. UDAQ recommends that the use of the existing API OWS with floating covers be retained as BACT. The floating covers should be replaced with double wiper seal-style floating covers no later than December 31, 2019, but this date is past the regulatory attainment
date. Only “partial credit” can be taken for this control system – representing those controls in place by December 31, 2018.
9.1.5 Selection of BACT Controls
UDAQ recommends that Marathon install and operate the existing wastewater treatment system with the existing API OWS with floating covers with good operating practices. Operation of the API OWS with floating covers is required in Part H.12.m.vi. No additional controls are required
for BACT; thus, no additional limits other than those established in H.12.m.vi are required to be

11 see References: Item #8

31
established for the SIP. 10.0 BACT for Flares 10.1 Flare Gas Emissions The refinery hydrocarbon flares include only the North and South Flares. These two flares emit PM2.5, SO2, NOx and VOCs, as well as a minor amount of ammonia. However, rather than
evaluate the flares based on the individual pollutant emissions, UDAQ has historically evaluated the emissions from the flares based on the gases sent to the flares. During development of the Moderate 2.5 SIP, UDAQ established that the refineries’ flares were to be used primarily as safety devices and not as process control devices. Therefore, each refinery was required to meet the requirements of Subpart J and Ja for all hydrocarbon flares, and to install and operate a flare gas recovery or minimization process by January 1, 2019. 10.1.1 Available Control Technology
There are two parts to refinery flares, as outlined in the Refinery General RACT Evaluation. The first is setting all refinery hydrocarbon flares as subject to the requirements of 40 CFR 60 Subpart
Ja. The second is requiring all refineries to have a flare gas recovery system in place and operating by January 1, 2019 that meets the flare event limits listed in 40 CFR 60.103a(c). 10.1.2 Evaluation of Technical Feasibility of Available Controls Neither part is technically infeasible.
10.1.3 Evaluation and Ranking of Technically Feasible Controls
The refinery general requirement of subjecting all hydrocarbon flares to the requirements of Subpart Ja has already been accepted by all listed refineries. As discussed in the Refinery General RACT Evaluation, most refineries will begin economic evaluations of flaring events beginning in November of 2015 to determine whether a flare gas recovery program is viable regardless of any imposing of such requirement by DAQ. For its part, Marathon implemented a flare gas recovery system in 2014 (DAQE-AN103350065-14). This system greatly reduced emissions from both the North and South Flares. Marathon also
established a flare minimization plan, which includes limiting waste gas flow rates to these flares. These two changes (flare gas recovery and flare minimization plan) meet the requirements of SIP section IX.H.11.g.v.B.
10.1.4 Further Evaluation of Most Effective Controls
No additional analysis is required. The general requirements on refinery flares found at Section IX Part H.11.g.v.B of the moderate PM2.5 SIP are the only viable techniques for the control of emissions from the refinery’s flares. No additional analysis is required.
10.1.5 Selection of BACT Controls DAQ recommends that Marathon continue to operate its existing flare gas recovery system as outlined in its latest existing refinery AO; as well as implement the general refinery SIP requirement “Requirements on Hydrocarbon Flares” as outlined in the Refinery General RACT

32
Evaluation. There are no expected emission reductions versus the 2016 “true-up” emission inventory as the flare gas recovery system was already included in that inventory. These practices
are required through existing permit requirements. The existing limits account for current process variability while still limiting SO2 emissions from
the refinery flares. Therefore, the existing limits are considered BACT. For a summary of the most recent CEMs data to support these conclusions, see Appendix B. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon
will comply with any applicable emission limits in Section IX, Part H.11. See Appendix A for a complete listing of existing permit conditions, established emission limits, and the basis of those limits. 10.2 SRU Flare The SRU Flare is not subject to 40 CFR 60.18 nor 40 CFR 63.11 and is designed to combust acid gases from the sour water stripper and amine treatment units during startup, shutdown and
malfunction events. There is no routine waste gas venting to the SRU Flare. This flare does not qualify as a hydrocarbon flare, and thus is not subject to the same provisions of SIP Section IX.H.11.g.v as are the North and South Flares.
10.2.1 Available Control Technology
As a true upset-only flare, the SRU Flare is operated as a safety device. There are no add-on controls which can be applied to this unit as it is not feasible to enclose the flare tip of a safety flare. The only available control options are development of a flare management plan, and use of
a clean-burning gaseous fuel for the pilot flame. 10.2.2 Evaluation of Technical Feasibility of Available Controls
Marathon has developed a flare management plan which includes operational plans and contingencies for the SRU Flare and for the units controlled by the SRU Flare – the sour water stripper and amine treatment unit. In addition, the pilot flame is fired on natural gas instead of refinery fuel gas. 10.2.3 Evaluation and Ranking of Technically Feasible Controls
The implementation of a flare management plan and operation of the flare pilot on natural gas are the only viable control options for the SRU Flare. Marathon is currently implementing both, so no further evaluation is required.
10.2.4 Further Evaluation of Most Effective Controls
No additional analysis is required. Marathon has a flare management plan in place which includes the SRU Flare and those processes controlled by the SRU Flare. The flare pilot is also operated on natural gas.
10.2.5 Selection of BACT Controls
DAQ recommends that Marathon continue to implement its existing flame management plan and operate the flare pilot on natural gas to limit emissions from the SRU Flare as needed. There are no expected emission reductions versus the 2016 “true-up” emission inventory. These practices

33
are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with
any applicable emission limits in Section IX, Part H.11. 11.0 BACT for Cooling Towers
There are two main pollutants of concern from cooling towers used in refinery settings. Like all industrial cooling towers, some particulate emissions will result during the evaporation of the
cooling water. For further details on BACT controls for particulate emissions from cooling towers please refer to the PM2.5 Serious SIP - BACT for Small Sources – Section 6 for the analysis. Cooling towers found in refineries have a secondary concern. It is possible for the cooling water to pick up volatile compounds during the heat transfer process, and for these compounds to be released as VOCs. As the levels of VOCs in refinery cooling water can be large enough to deserve their own controls, a separate BACT analysis is provided.
11.1 VOCs 11.1.1 Available Control Technology UDAQ employed the services of a contractor during review of the RACT evaluations for the
moderate PM2.5 SIP12. Only a single control technique was determined to be “available.” During that review, it became apparent that UDAQ’s contractor was making the same recommendation to all of the refineries located in the PM2.5 non-attainment area. Specifically, that each refinery apply
the 40 CFR 63 Subpart CC requirements to all cooling towers servicing heat exchangers with high VOC content streams. These requirements are basically leak detection and repair programs that apply specifically to cooling towers by checking for the presence of VOCs in the cooling
water on a periodic basis. If detected, then service or repair of the relevant heat exchanger is warranted. 11.1.2 Evaluation of Technical Feasibility of Available Controls All the refineries located in the PM2.5 non-attainment area agreed to an application of the MACT CC language which was included in the moderate PM2.5 SIP in Section IX, Part H.11.g.
11.1.3 Evaluation and Ranking of Technically Feasible Controls N/A This has become a refinery general SIP requirement.
11.1.4 Further Evaluation of Most Effective Controls
N/A This has become a refinery general SIP requirement. 11.1.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to follow the general refinery SIP requirements found in Section IX, Part H.11.g. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits beyond

12 see References: Item #4

34
those in Part H.11.g are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11.
12.0 BACT for Loading/Offloading
Marathon submitted an analysis for product loading and unloading operations. UDAQ has covered the analysis of loading and unloading operations in two different locations. For fugitive VOC emissions please refer to Section 8.0 of this document for further details. For more direct
VOC emissions from loading/unloading operations please refer to the PM2.5 Serious SIP - BACT for Small Sources – Section 13B for additional details. 12.1 VOC Loading and offloading operations are a source of VOC emissions. Although Stage I and II vapor recovery have already been covered in UDAQ’s BACT analysis for Small Sources, one additional VOC control option in common use at refinery loading racks is a vapor combustion unit or carbon
capture system to treat the recovered VOC vapors. In this section, UDAQ focuses on BACT for the two transportation loading racks: the TLR and the BCLR and for the two liquefied petroleum gases (LPG) racks: a 6-bay rail loading and offloading rack, and a single-bay truck loading and
offloading rack. 12.1.1 Available Control Technology
Once vapors have been recovered from loading/unloading operations, the recovered vapors need to be treated. While some vapor recovery units (VRUs) can return the recovered vapors back into
the system being controlled (depending on the system), some units require external treatment for final control. The use of regenerative thermal oxidation (RTO), non-regenerative thermal oxidation (flaring), or carbon capture via carbon canisters are all potentially available methods of
controlling the recovered vapors. 12.1.2 Evaluation of Technical Feasibility of Available Controls For destruction/control of the collected vapors, only the use of a RTO, carbon adsorption or flaring have been shown to be technically feasible control methods based on the volume of expected VOC emissions (typically less than 5 tons VOC/year).
All three methods are viable for the two transportation racks. Carbon adsorption is not technically feasible for the LPG loading racks, as the low molecular
weight of the LPG makes capture by activated carbon ineffective. 12.1.3 Evaluation and Ranking of Technically Feasible Controls
The transportation loading racks are subject to MACT CC (40 CFR 63.640 Subpart CC) requirements which establish an emission limit of 10 mg VOC/L of product loaded. Both carbon
adsorption and thermal destruction are able to meet this limit. Since the use of carbon adsorption allows Marathon to recover the vapors from the transportation loading racks as potential product (rather than simply increasing emissions through a thermal process), this is the top feasible
control option. For the LPG loading racks, only thermal destruction remains as a viable control option.

35
12.1.4 Further Evaluation of Most Effective Controls
Not required. Marathon has selected the top control option in each case. No additional analysis is necessary.
12.1.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to control product loading and unloading operations using carbon adsorption on the transport loading racks and thermal oxidation on the LPG loading racks. As the transport loading racks are already subject to MACT CC requirements, no additional limitations are required. The LPG loading racks’ thermal oxidizers shall be included in Marathon’s flare management plan. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11.
13.0 BACT for Diesel- and Natural Gas-fired Emergency Engines
Marathon submitted an analysis for both diesel and natural gas-fired emergency engines13. UDAQ has covered the analysis of both diesel-fired and natural gas fired emergency engines in a separate document. See PM2.5 Serious SIP - BACT for Small Sources – Section 8 for additional
details. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11.
14.0 BACT for the K1 Compressors
Marathon submitted a separate analysis for the K1 compressor engines14. The K1 compressors are two compressors which recycle hydrogen into the UFU desulfurization reactor. They are each powered by a natural gas-fired internal combustion (IC) engine which is controlled by a catalytic convertor to control NOx emissions. 14.1 PM2.5 14.1.1 Available Control Technology
None. UDAQ has been unable to identify any additional controls to further reduce emissions of PM2.5 from natural gas-fired IC engines. Marathon evaluated whether replacement of one
compressor motor with an electric motor might be feasible; however, currently, the engines are not scheduled for turnaround or maintenance until after the December 31, 2018 attainment date. 14.1.2 Evaluation of Technical Feasibility of Available Controls None are technically feasible. As mentioned previously, replacement of one compressor motor
with an electric motor was listed by Marathon as available, but not until after the December 31, 2018 attainment date had elapsed. Replacing the motor is considered technically infeasible.
13 see References: Item #8 14 see References: Item #9

36
14.1.3 Evaluation and Ranking of Technically Feasible Controls
N/A no available options remain to be ranked. 14.1.4 Further Evaluation of Most Effective Controls
N/A no available options remain for further evaluation. Even without considering the technical considerations, Marathon completed an economic analysis and determined that no additional
controls are economically feasible15. 14.1.5 Selection of BACT Controls UDAQ recommends that Marathon continue to operate the K1 compressors in line with good operating practices. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX,
Part H.11. 14.2 SO2
14.2.1 Available Control Technology
None. The compressor engines are fired on natural gas, an inherently low-sulfur fuel. As stated previously, Marathon did offer replacement of engine with an electric motor as available, but not technically feasible. No additional controls exist to further limit the emissions of SO2 from
natural gas-fired IC engines. 14.2.2 Evaluation of Technical Feasibility of Available Controls
None are technically feasible. As mentioned previously, replacement of one compressor motor with an electric motor was listed by Marathon as available, but not until after the December 31, 2018 attainment date had elapsed. Replacing the motor is considered technically infeasible. 14.2.3 Evaluation and Ranking of Technically Feasible Controls N/A no available options remain to be ranked.
14.2.4 Further Evaluation of Most Effective Controls
N/A no available options remain for further evaluation. Even without considering the technical considerations, Marathon completed an economic analysis and determined that no additional controls are economically feasible16.
14.2.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to operate the K1 compressors in line with good operating practices. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be
15 See Reference #9 16 See Reference #9

37
established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11.
14.3 NOx 14.3.1 Available Control Technology The engines are controlled with a catalytic convertor to control emissions of NOx. Larger-size
natural gas-fired IC engines, such as these compressor engines, have also been controlled with SCR, SNCR, and replacement with electric motors. 14.3.2 Evaluation of Technical Feasibility of Available Controls As mentioned previously, replacement of one compressor motor with an electric motor was listed by Marathon as available, but not until after the December 31, 2018 attainment date had elapsed. Replacing the motor is considered technically infeasible.
SCR and SNCR are generally considered technically feasible but suffer from the same problem as direct replacement. Neither option can be designed, installed and tested prior to the regulatory
attainment date of December 31, 2018. Plus the compressor engines cannot be taken offline prior to that date. Both control options are considered technically infeasible. 14.3.3 Evaluation and Ranking of Technically Feasible Controls Continuing to fire the existing engines on natural gas and using a NOx controlling catalytic
convertor remains the only viable control technique for the K1 compressors. 14.3.4 Further Evaluation of Most Effective Controls
Marathon is operating with the only viable control option for the K1 compressors. The engines can emit up to 3.2 lb/hr using the existing controls – such limit is found in the most recent NSR permit issued to the refinery (condition II.B.7.c in DAQE-AN103350075-18). Even without considering the technical considerations, Marathon completed an economic analysis and determined that no additional controls are economically feasible17. 14.3.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to operate the K1 compressors in line with good operating practices. NOx emissions shall be controlled using the existing catalytic convertors
installed on both compressor engines. These practices are required through existing permit requirements. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits
in Section IX, Part H.11. 14.4 VOC and Ammonia
14.4.1 Available Control Technology The existing catalytic convertor controls emissions of VOC. There are no available controls for

17 See Reference #9

38
reducing ammonia emissions from IC engines. 14.4.2 Evaluation of Technical Feasibility of Available Controls Catalytic convertors, also known as oxidation catalysts remain the only viable method for
controlling emissions of VOC from IC engines. 14.4.3 Evaluation and Ranking of Technically Feasible Controls
N/A, Marathon is already using the only viable control technique for control of VOC emissions. 14.4.4 Further Evaluation of Most Effective Controls VOC emissions from the engines are less than 0.03 lb/MMBtu on average. 2016 actual emissions for both engines combined were less than 0.5 tons of VOC annually. The catalytic convertors are already required by NSR permit, and for control of NOx emissions. Imposing an additional
VOC restriction would be of no benefit to limiting overall emissions or demonstrating good maintenance. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits
in Section IX, Part H.11. 14.4.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to operate the K1 compressors in line with good operating practices. VOC emissions shall be controlled using the existing catalytic convertors
installed on both compressor engines. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part H.11.
15.0 BACT for Tanks Although most of UDAQ’s analysis of storage vessels, more commonly referred to as storage tanks (or just “tanks”), can be found in the PM2.5 Serious SIP - BACT for Small Sources – Section 13, the refineries as a group were evaluated for two additional BACT controls beyond the small source controls. First, the refineries have some tanks that are larger than the 30,000 gallon cut-off used in the small source analysis. Second, during development of the moderate PM2.5 SIP, the
refineries were required to implement a tank degassing work practice standard at Section IX, Part H.11.g.vi. 15.1 VOC 15.1.1 Available Control Technology
Although tanks as a group were evaluated for tank degassing, individual tanks were not evaluated for working or breathing losses. While some VOCs are emitted during these periods, these can
only be controlled on a tank by tank basis. Larger tanks are already subject to floating roof and specific seal requirements such as those found in 40 CFR 60 Subpart Kb. Some additional VOC reductions could be gained by including slotted guide poles and geodesic domes, but these gains are relatively minor. In the case of slotted guide poles, such requirements are more easily handled through individual NSR permitting requirements. Individual tanks can

39
also be controlled by vapor recovery, vapor scrubbers, or vapor combustors. Geodesic domes have not been found to be economically or technically feasible.
15.1.2 Evaluation of Technical Feasibility of Available Controls
The use of slotted guide poles and vapor controls are both technically feasible. Tank degassing as a group control is also technically feasible, and was included as a requirement of the moderate PM2.5 SIP.
15.1.3 Evaluation and Ranking of Technically Feasible Controls Tank degassing during tank shutdowns was required as part of the moderate PM2.5 SIP. The remaining controls can be employed in conjunction with tank degassing. The various methods of individual tank vapor control (recovery, scrubbing, and combustion) are all similar in effectiveness and are employed primarily on a tank by tank basis. While some economy of scale could conceivably be achieved by combining the emissions from several tanks, tank vapors are
primarily released during filling or unloading, and nearby tanks are rarely loaded or unloaded at the same time. 15.1.4 Further Evaluation of Most Effective Controls Marathon is already required to follow the tank degassing requirements of Section IX, Part
H.11.g. In fact, Marathon developed most of the requirements which were included in the tank degassing subsection prior to their inclusion in the moderate PM2.5 SIP. The remaining vapor controls were all evaluated by Marathon18 and were found not to be economically feasible, with
cost effectiveness values in excess of $200,000/ton of VOC control. 15.1.5 Selection of BACT Controls
UDAQ recommends that Marathon continue to implement the SIP general refinery requirements on tank degassing as outlined in Section IX, Part H.11.g. No additional controls are required for BACT; thus, no additional limits are required to be established for the SIP. Marathon will comply with any applicable emission limits in Section IX, Part. H.11. 16.0 Additional Feasible Measures and Most Stringent Measures
16.1 Extension of SIP Analysis Timeframe As outlined in 40 CFR 51.1003(b)(2)(iii):
If the state(s) submits to the EPA a request for a Serious area attainment date extension simultaneous with the Serious area attainment plan due under paragraph (b)(1) of this section,
such a plan shall meet the most stringent measure (MSM) requirements set forth at § 51.1010(b) in addition to the BACM and BACT and additional feasible measure requirements set forth at § 51.1010(a).
Thus, with the potential for an extension of the SIP regulatory attainment date from December 31, 2019 to December 31, 2024, the SIP must consider the application of both Additional Feasible Measures (AFM) and Most Stringent Measures (MSM).

18 see References: Item #8

40
16.2 Additional Feasible Measures at Marathon
As defined in Subpart Z, AFM is any control measure that otherwise meets the definition of “best available control measure” (BACM) but can only be implemented in whole or in part beginning 4
years after the date of reclassification of an area as Serious and no later than the statutory attainment date for the area. The Salt Lake City Nonattainment Area was reclassified as Serious on June 9, 2017. Therefore, any viable control measures that could only be implemented in
whole or in part beginning 6/9/2021 (4 years after the date of reclassification) are classified as AFM. After a review of the available control measures described throughout this evaluation report, Marathon identified a number of control techniques which were subsequently eliminated because they could not be implemented by the regulatory attainment date of December 31, 2018. Marathon provided economic feasibility analyses which found none of these controls were economically feasible.
16.3 Most Stringent Measures at Marathon
As defined in Subpart Z, MSM is defined as: … any permanent and enforceable control measure that achieves the most stringent emissions
reductions in direct PM2.5 emissions and/or emissions of PM2.5 plan precursors from among those control measures which are either included in the SIP for any other NAAQS, or have been achieved in practice in any state, and that can feasibly be implemented in the relevant PM2.5
NAAQS nonattainment area. This is further refined and clarified in 40 CFR 51.1010(b), to include the following Steps:
Step 1) The state shall identify the most stringent measures for reducing direct PM2.5 and PM2.5 plan precursors adopted into any SIP or used in practice to control emissions in any state. Step 2) The state shall reconsider and reassess any measures previously rejected by the state during the development of any previous Moderate area or Serious area attainment plan control strategy for the area. Step 3) The state may make a demonstration that a measure identified is not technologically or economically feasible to implement in whole or in part by 5 years after the applicable
attainment date for the area, and may eliminate such whole or partial measure from further consideration. Step 4) Except as provided in Step 3), the state shall adopt and implement all control measures
identified under Steps 1) and 2) that collectively shall achieve attainment as expeditiously as practicable, but no later than 5 years after the applicable attainment date for the area. 16.3.1 Step 1 – Identification of MSM For purposes of this evaluation report UDAQ has identified for consideration the most stringent
methods of control for each emission unit and pollutant of concern (PM2.5 or PM2.5 precursor). A summary is provided in the following table: Table 16-1: Most Stringent Controls by Emission Unit Emission Unit Pollutant Most Stringent Control Method
PM2.5 GCP, fuel type, flue gas filter (FGF) / wet gas scrubber (WGS)

41
FCCU Regenerator SO2 DeSOx catalyst, WGS
NOx GCP, deNOx catalyst, feed hydro-treating, deNOx additive,
LoTOx
Heaters/Boilers NOx ULNB, SCR
Ammonia only if SCR is used, feedback controls
Flares Flare Gas flare minimization program SRU SO2 second tail gas treatment unit (TGTU), WGS NOx WGS
Cooling Towers VOC MACT CC requirements
Fugitives VOC NSPS GGGa LDAR requirements
Tanks VOC tank degassing requirements
Wastewater Treatment VOC IAF/API separator; with carbon canister control / oxidation
The above listed controls represent the most stringent level of control identified from all other state SIPs or permitting actions, but do not necessarily represent the final choice of MSM. That is determined in Step 4.
16.3.2 Step 2 – Reconsideration of Previous SIP Measures
Utah has previously issued a SIP to address the moderate PM2.5 nonattainment areas of Logan, Salt Lake City, and Provo. The SIP was issued in parts: with the section devoted to the Logan nonattainment area being found at SIP Section IX.A.23, Salt Lake City at Section IX.A.21, and
Provo/Orem at Section IX.A.22. Finally, the Emission Limits and Operating Practices for Large Stationary Sources, which includes the application of RACT at those sources, can be found in the SIP at Section IX Part H. Limits and practices specific to PM2.5 may be found in subsections 11,
12, and 13 of Part H. Accompanying Section IX Part H was a TSD that included multiple evaluation reports similar to
this document for each large stationary source identified and listed in each nonattainment area. UDAQ conducted a review of those measures included in each previous evaluation report which contained emitting units which were at all similar to those installed and operating at Marathon. There were several technologies that had been eliminated from further consideration at some point during many of the previous reviews. Some emitting units were considered too small, or emissions too insignificant to merit further consideration at that time. The cost effectiveness considerations may have been set at too low a threshold (a question of cost in RACT versus
BACT). And many cases of technology being technically infeasible for application – such as applying catalyst controls to infrequently used emitting units which may never reach an operating temperature where use of the catalyst becomes viable and effective.
In one particular case, these previously rejected control technologies were already brought forward and re-evaluated using updated information (more recent permits, emission rates and cost
information) by Marathon in its BACT analysis report. This was the deferment of VOC controls for the wastewater treatment systems at four Salt Lake City area refineries. Even without considering the technical considerations, Marathon completed an economic analysis and
determined that no additional controls were economically feasible19. 16.3.3 Step 3 – Demonstration of Feasibility

19 See Reference #9

42
A control technology or control strategy can be eliminated as MSM if the state demonstrates that
it is either technically or economically infeasible. This demonstration of infeasibility must adhere to the criteria outlined under §51.1010(b)(3), in
summary: 1) When evaluating technological feasibility, the state may consider factors including but not
limited to a source's processes and operating procedures, raw materials, plant layout, and potential environmental or energy impacts 2) When evaluating the economic feasibility of a potential control measure, the state may consider capital costs, operating and maintenance costs, and cost effectiveness of the measure. 3) The SIP shall include a detailed written justification for the elimination of any potential control measure on the basis of technological or economic infeasibility.
This evaluation report serves as written justification of technological or economic feasibility/infeasibility for each control measure outlined herein. Where applicable, the most effective control option was selected, unless specifically eliminated for technological or
economical infeasibility. Expanding on the previous table, the following additional information is provided: Table 16-2: Feasibility Determination
Emission Unit Pollutant MSM Previously Identified Is Method Feasible?

FCCU Regenerator
PM2.5 GCP, fuel type, FGF/WGS See below
SO2 deSOx catalyst, WGS See below
NOx GCP, deNOx catalyst, feed hydro-treating, deNOx additive,
LoTOx
See below
Heaters/Boilers NOx ULNB, SCR See below
Ammonia NH3 feedback See below
Flares Flare Gas flare minimization program Yes
SRU SO2 TGTU or WGS No, high cost NOx WGS No, high cost Cooling Towers VOC MACT CC Yes
Fugitives VOC LDAR Yes
Tanks VOC tank degassing Yes
WW Treatment VOC carbon canister / oxidation No, high cost

Most of the entries in the above table were determined to be feasible on a technological basis. However, in several cases two distinct paths exist that are mutually exclusive. Two control techniques could serve equally as BACT/BACM or MSM, but they are not simply
interchangeable. Once a source has elected to follow a particular path for emission control, needing to change over to the alternative control path would involve considerable expense as well as complete removal of the existing system(s). In many cases this would also involve shutting
down operation of the source for an extended period of time – posing additional economic burden on the source.
One particular example of this is the application (or lack) of WGS. Wet gas scrubbing has the capability of removing both particulates and acid gases (SO2 and derivatives) and, if the LoTOx

43
option has been pursued, NOx as well. However, this control system is not compatible with other control systems such as fabric filtration (baghouses or FGF), catalytic controls (SCR), or tail gas
treatment (as these are also catalytic controls). If the WGS is installed secondary to the existing controls, these would render the use of WGS redundant and extremely cost ineffective (the inlet concentrations would simply be too low to be viable). Alternatively, the WGS would be installed
as the primary control, creating a similar situation for the “existing” controls, but with an additional problem of a now water saturated exhaust stream and a greatly lowered exhaust temperature. Removing the existing controls to swap to the new control option is often millions
of dollars above and beyond the millions already spent on the primary BACT level control. Marathon employed WGS as the solution in one situation (the FCCU) but elected to install a more traditional TGTU in the second situation (the SRU). Both control techniques are viable, and reached BACT level of control. The costs for WGS or a second TGTU on the SRU do not currently justify including either of these controls as MSM.
17.0 New PM2.5 SIP – General Requirements The general requirements for all listed sources are found in SIP Subsection IX.H.11. These serve
as a means of consolidating all commonly used and often repeated requirements into a central location for consistency and ease of reference. As specifically stated in subsection IX.H.11.a below, these general requirements apply to all sources subsequently listed in either IX.H.12 (Salt
Lake City) or IX.H.13 (Provo/Orem), and are in addition to (and in most cases supplemental to) any source-specific requirements found within those two subsections.
IX.H.11.a. This paragraph states that the terms and conditions of Subsection IX.H.11 apply to all sources subsequently addressed in the following subsections IX.H.12 and IX.H.13. It also clarifies that should any inconsistency exist between the general requirements and the source
specific requirements, then the source specific requirements take precedence. IX.H.11.b Paragraph i: States that the definitions found in State Rule 307-101-2, Definitions, apply to SIP Section IX.H. Since this is stated for the Section (IX.H), it applies equally to IX.H.11, IX.H.12 and IX.H.13. A second paragraph (ii), includes a new definition for natural gas curtailment for those sources in IX.H.12 and IX.H.13 that reference it. IX.H.11.c This is a recordkeeping provision. Information used to determine compliance shall
be recorded for all periods the source is in operation, maintained for a minimum period of five (5) years, and made available to the Director upon request. As the general recordkeeping requirement of Section IX.H, it will often be referred to and/or discussed as part of the
compliance demonstration provisions for other general or source specific conditions. It also includes provisions referring to the reporting of emission inventories (paragraph ii) and reporting deviations (paragraph iii).
IX.H.11.d Statement that emission limitations apply at all times that the source or emitting unit is in operation, unless otherwise specified in the source specific conditions listed in IX.H.12
or IX.H.13. It also clarifies that particulate emissions consist of both the filterable and condensable fractions unless otherwise specified in IX.H.12 or IX.H.13.
This is the definitive statement that emission limits apply at all times – including periods of startup or shutdown. It may be that specific sources have separate defined limits that apply during alternate operating periods (such as during startup or shutdown), and these limits will be

44
defined in the source specific conditions of either IX.H.12 or IX.H.13.
Conditions 1.a, 1.b and 1.d are declaratory statements, and have little in the way of compliance provisions. Rather, they define the framework of the other SIP conditions. As condition 1.c is the primary recordkeeping requirement, it shall be further discussed under item 4.2 below.
IX.H.11.e This is the main stack testing condition, and outlines the specific requirements for demonstrating compliance through stack testing. Several subsections detailing Sample Location,
Volumetric Flow Rate, Calculation Methodologies and Stack Test Protocols are all included – as well as those which list the specific accepted test methods for each emitted pollutant species (PM10, NOx, or SO2). Finally, this subsection also discusses the need to test at an acceptable production rate, and that production is limited to a set ratio of the tested rate. IX.H.11.f This condition covers the use of CEMs and opacity monitoring. While it specifically details the rules governing the use of continuous monitors (both emission monitors and opacity monitors), it also covers visible opacity observations through the use of EPA
reference method 9. Both conditions 11.e and 11.f serve as the mechanism through which sources conduct monitoring
for the verification of compliance with a particular emission limitation. 17.1 Monitoring, Recordkeeping and Reporting
As stated above, the general requirements IX.H.11.a through IX.H.11.f primarily serve as declaratory or clarifying conditions, and do not impose compliance provisions themselves.
Rather, they outline the scope of the conditions which follow in the source specific requirements of IX.H.12 and IX.H.13.
For example, most of the conditions in those subsections include some form of short-term emission limit. This limitation also includes a compliance demonstration methodology – stack test, CEM, visible opacity reading, etc. In order to ensure consistency in compliance demonstrations and avoid unnecessary repetition, all common monitoring language has been consolidated under IX.H.11.e and IX.H.11.f. Similarly, all common recordkeeping and reporting provisions have been consolidated under IX.H.11.c. 18.0 Revised PM2.5 SIP – General Refinery Requirements
The revised PM2.5 SIP incorporates several new requirements that apply specifically to those petroleum refineries listed in Section IX.H.12 of the SIP. Some subsections of IX.H.11.g also
apply more broadly and could affect additional petroleum refineries in addition to those listed in IX.H.12. Where this greater applicability exists for a particular condition or limitation, such will be noted in the discussion for that requirement.
IX.H.11.g.i.A This condition covers SO2 emissions from fluidized catalytic cracking units (FCCUs). The limit is 50 ppmvd @ 0% excess air on a 7-day rolling average
basis, as well as 25 ppmvd @ 0% excess air on a 365-day rolling average basis. The condition is based on 40 CFR 60 Subpart Ja, and includes the same limitation found in that subpart. Compliance is demonstrated by CEM, as outlined in 40 CFR 60.105a(g) – also from Subpart Ja.

45
IX.H.11.g.i.B This condition addresses PM emissions from FCCUs. The limit is 1.0 lb PM per 1000 lb coke burned. The emission limit applies on a 3-hour average basis.
The emission limit is derived from 40 CFR 60 Subpart Ja, although Subpart Ja does not specifically state that the limit applies on a 3-hour average. Instead it states that compliance will
be demonstrated via a performance test using Method 5, 5b or 5f, using an average of three 60-minute (minimum) test runs.
Compliance is demonstrated by stack test as outlined in 40 CFR 60.106(b). This stack testing procedure is from Subpart J, rather than Subpart Ja. The equations utilized and reference methods involved are identical between the two subparts; however, the protocol to follow for testing is much more direct and straightforward in §60.106(b). The condition also requires the installation of a continuous parameter monitoring system (CPMS) to monitor and record operating parameters for determination of source-wide PM10 emissions.
IX.H.11.g.ii This condition limits the H2S content of gases burned within any refinery located within (or affecting) an area of PM2.5 nonattainment. The limit is 60 ppm H2S or less as described in 40 CFR 60.102a on a rolling average of 365 days.
Compliance is demonstrated through continuous H2S monitoring, as outlined in 40 CFR 60.107a. Both the limitation and the compliance methodology are based on 40 CFR 60 Subpart Ja.
IX.H.11.g.iii This condition places limits on heat exchangers in VOC service.
The condition requires that all heat exchangers in VOC service meet the requirements of 40 CFR 63.654, which requires use of the “Modified El Paso Method” for calculation of VOC emissions. Sources are allowed an option to use another EPA-approved method if allowed by the Director.
An exemption is also given for heat exchangers that meet specific criteria that are outlined within the condition language. IX.H.11.g.iv Leak Detection and Repair Requirements. This condition subjects each source to the requirements of 40 CFR 60 Subpart GGGa – also known as Enhanced LDAR. The Sustainable Skip Period provisions of that subpart have also been retained.
IX.H.11.g.v This condition establishes new requirements on hydrocarbon flares.
First, it states that all hydrocarbon flares (defined as all non-dedicated SRU flare and header systems and all non-HF flare and header systems) are subject to Subpart Ja as of January 1, 2018 if not previously subject.
Second it requires that each major source refinery either: 1) install a flare gas recovery system designed to limit hydrocarbon flaring from each affected flare during normal operations below
the values listed in Subpart Ja (specifically 40 CFR 60.103a(c)), or 2) limit flaring during normal operations to 500,000 scfd or less for each affected flare.
This requirement is based on Subpart Ja, and is designed to incorporate the flare gas recovery requirements of that subpart ahead of the normal implementation schedule. The refineries located in, or impacting, the nonattainment areas are relatively small. As a consequence, the possibility

46
that they would trigger the flare gas recovery provisions of Subpart Ja in the near term (5-10 years) was very small. Although one of the refineries had elected to install a flare gas recovery
system voluntarily, the system only addressed a part of the refinery’s total flaring capacity, and was not originally designed to Subpart Ja specifications.
IX.H.11.g.vi This condition requires that vapor control devices be employed during tank degassing operations. Some provisions are made for connecting and disconnecting degassing equipment. Notification must also be made to the
Director prior to performing such an operation – unless an emergency situation is at play. This condition applies to sources beyond just refineries – any owner/operator of any stationary tank meeting the outlined criteria must fulfill the requirements of this condition. IX.H.11.g.vii No Burning of Liquid Fuel Oil in Stationary Sources – Establishes that no petroleum refineries in or affecting any PM nonattainment or maintenance area
shall be allowed to burn liquid fuel oil in stationary sources except during natural gas curtailments or as specified in the individual subsections of Section IX, Part H. The use of diesel fuel meeting the specifications of 40 CFR 80.510 in standby
or emergency equipment is exempt from this requirement. This requirement addresses a provision of the original PM10 SIP, which prevented the refineries
from burning liquid fuel oil in any capacity – including in emergency or standby equipment. This brings forward the original intent, maintains consistency with the PM10 maintenance plan provisions of IX.H.1.g, and addresses the problem of emergency and standby equipment.
IX.H.11.i This condition requires that good combustion practices will be followed.
This condition applies to all combustion units and sets a general work practice that good combustion practices and maintenance will be in line with manufacturer’s recommendations, to ensure equipment stays in good working order. IX.H.11.j This condition requires additional recordkeeping and reporting requirements specific to the refineries. This condition applies to the refineries until such time that a Title V operating permit is issued.
This condition ensures all applicable recordkeeping and reporting requirements are being followed. 18.1 Monitoring, Recordkeeping and Reporting The new petroleum refinery requirements establish several specific emission limitations.
Primarily these limits are monitored continuously – such as the SO2 CEM on the FCCU or the H2S monitor on fuel gas. Where continuous monitoring is used, the requirements of IX.H.11.f apply, which incorporates by reference R307-170, 40 CFR 60.13 and 40 CFR 60, Appendix B –
Performance Specifications. Under R307-170, paragraph 170-8 addresses Recordkeeping, while 170-9 addresses Reporting. The FCCU PM limit is demonstrated by stack test. This stack test requirement is subject to the requirements of IX.H.11.e. In addition, any source with a direct stack emission limitation is

47
subject to the requirements of R307-165.
These conditions are also subject to the general recordkeeping and reporting requirements of IX.H.11.c. 19.0 Revised PM2.5 SIP – Marathon Specific Requirements The Marathon specific conditions in Section IX.H.12 address those limitations and requirements
that apply only to the Marathon Refinery in particular. The following controls were determined as necessary for the PM2.5 SIP to satisfy BACT. IX.H.12.m.i This condition establishes NOx emission limits for two combustion units at Marathon. These emission limitations were determined as necessary for BACT. This condition requires initial and ongoing stack testing to ensure emission limitations and existing control requirements
are being met. IX.H.12.m.iii This condition establishes NOx emission limits for the two cogeneration turbines
with heat recovery steam generation. These emission limitations were determined as necessary for BACT. This condition requires
initial and ongoing stack testing to ensure emission limitations and existing control requirements are being met.
IX.H.2.m.vi Listing of required emission controls. Shown in table format, all emission controls which were determined to be BACT are listed in this condition. 19.1 Monitoring, Recordkeeping and Reporting Monitoring for all conditions is addressed through a variety of methods, depending on the emission point in question. Stack testing, CEMs, parameter monitoring – all are viable options, and have been included in the language of IX.H.12.m.i through IX.H.12.m.v. As appropriate, these monitoring requirements are complemented by the general provisions of IX.H: specifically, 11.e for stack testing, 11.f for CEMs and other continuous monitors, and 11.c for recordkeeping and reporting.
Where necessary, additional monitoring, recordkeeping and/or reporting requirements have been directly included in the language of IX.H.12.m to address specific concerns.
20.0 References
1. Tesoro, PM2.5 SIP Major Point Source RACT Documentation – Salt Lake City Refinery 2. Tesoro – response to information request, dated June 11, 2014 3. DAQE-AN103350058B-13, DAQE-AN103350059-12
4. UDSHW Contract No. 12601, Work Assignment No. 7, Utah PM2.5 SIP RACT Support – TechLaw Inc. 5. DAQE-AN103350058B-13 6. DAQE-AN103350065-14 7. DAQE-AN103350075-18

48
8. Tesoro Refining & Marketing Company LLC – Best Available Control Technology (BACT) Assessment for Tesoro Refining & Marketing Company LLC and Tesoro Logistics
Operations LLC Salt Lake City Utah, dated May 5, 2017 9. Tesoro Refining & Marketing Company LLC – Best Available Control Technology (BACT) Assessment for Tesoro Refining & Marketing Company LLC and Tesoro Logistics Operations LLC Salt Lake City Utah, follow up dated December 8, 2017 10. Final Tesoro Refining and Marketing 10335 PM2.5 SIP BACT.xls, dated June 26, 2018 Additional references reviewed during UDAQ BACT research: 3-2-1-2.pdf. (n.d.). Retrieved from https://www.netl.doe.gov/File%20Library/Research/Coal/energy%20systems/turbines/handbook/3-2-1-2.pdf
5ce1d8028599a7954783ca08d5489afbb8b8.pdf. (n.d.). Retrieved from https://pdfs.semanticscholar.org/9722/5ce1d8028599a7954783ca08d5489afbb8b8.pdf
7FA Gas Turbine DLN 2.6 Gas Fuel Control System.pdf | Servomechanism | Valve. (n.d.).
Retrieved August 30, 2017, from https://www.scribd.com/doc/308100183/7FA-Gas-Turbine-DLN-2-6-Gas-Fuel-Control-System-pdf
2017-Jan6-001.pdf. (n.d.). Retrieved from http://www.aqmd.gov/docs/default-
source/Agendas/Governing-Board/2017/2017-Jan6-001.pdf
30464f.pdf. (n.d.). Retrieved from http://permits.air.idem.in.gov/30464f.pdf
Annealing (metallurgy). (2017, December 4). In Wikipedia. Retrieved from
https://en.wikipedia.org/w/index.php?title=Annealing_(metallurgy)&oldid=813625590
ASME99-Presentation.pdf. (n.d.). Retrieved from http://www.nsittech.com/images/ASME99-Presentation.pdf
BACT / TBACT Workbook. (n.d.-b). Retrieved from http://www.baaqmd.gov/permits/permitting-manuals/bact-tbact-workbook
BACT Clearinghouse (Searchable). (n.d.). Retrieved from http://www.valleyair.org/busind/pto/bact/bactLoader.htm
bact_boilheatfurn.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_boilheatfurn.pdf
bact_bulkgasterm.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_bulkgasterm.pdf
bact_cooltow.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_cooltow.pdf
bact_engine.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_engine.pdf
bact_fccu.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_fccu.pdf

49
bact_flares.pdf. (n.d.). Retrieved from https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_fl
ares.pdf
bact_fugitives.pdf. (n.d.). Retrieved from https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_f
ugitives.pdf
bact_loading.pdf. (n.d.-a). Retrieved from https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_l
oading.pdf
bact_loading.pdf. (n.d.-b). Retrieved from https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_loading.pdf
bact_processfurn.pdf. (n.d.). Retrieved from https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_processfurn.pdf
bact_sru.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_sru.pdf
bact_turbines.pdf. (n.d.). Retrieved from
https://www.tceq.texas.gov/assets/public/permitting/air/Guidance/NewSourceReview/bact/bact_turbines.pdf
C302-4-(BMK750-1000-Alt-Derate).pdf. (n.d.). Retrieved from
http://aerco.com/sites/default/files/document/document/C302-4-%28BMK750-1000-Alt-Derate%29.pdf
chapter7.pdf. (n.d.). Retrieved from http://www.valleyair.org/busind/pto/bact/chapter7.pdf
Current Rules. (n.d.). Retrieved from http://www.baaqmd.gov/rules-and-compliance/current-rules
Development, O. of R. &. (n.d.). ENVIRONMENTAL TECHNOLOGY VERIFICATION REPORT, NOX CONTROL TECHNOLOGIES, CATALYTICA COMBUSTION SYSTEMS, INC., XONON FLAMELESS COMBUSTION SYSTEM. Retrieved October 17, 2017, from https://cfpub.epa.gov/si/si_public_record_Report.cfm?dirEntryID=86205
DPFs and DOCs: The new components at the heart of your Tier 4 engine | Equipment World | Construction Equipment, News and Information | Heavy Construction Equipment. (n.d.). Retrieved December 6, 2017, from https://www.equipmentworld.com/maintenance-24/
Electric Power Monthly - U.S. Energy Information Administration. (n.d.). Retrieved March 7, 2018, from https://www.eia.gov/electricity/monthly/
epm.pdf. (n.d.). Retrieved from https://www.eia.gov/electricity/monthly/current_month/epm.pdf
ffdg.pdf. (n.d.). Retrieved from https://www3.epa.gov/ttncatc1/dir1/ffdg.pdf
fnoxdoc.pdf. (n.d.). Retrieved from https://www3.epa.gov/ttncatc1/dir1/fnoxdoc.pdf
iron. (n.d.). Retrieved from https://www.chemguide.co.uk/inorganic/transition/iron.html
LEC_Appendix 5.1C_BACT Analysis.pdf. (n.d.). Retrieved from http://www.energy.ca.gov/sitingcases/lodi/documents/applicant/afc/Volume_II/LEC_Appendix%205.1C_BACT%20Analysis.pdf

50
May 2008 BACT cost effectiveness threshold update staff report.pdf. (n.d.). Retrieved from http://www.valleyair.org/busind/pto/bact/May%202008%20BACT%20cost%20effectiveness%20
threshold%20update%20staff%20report.pdf
m-g-util-tceq-non-ghg-updated052014.pdf. (n.d.). Retrieved from https://archive.epa.gov/region6/6pd/air/pd-r/ghg/web/pdf/m-g-util-tceq-non-ghg-
updated052014.pdf
nitrogen-oxides | netl.doe.gov. (n.d.). Retrieved from https://www.netl.doe.gov/research/Coal/energy-systems/gasification/gasifipedia/nitrogen-oxides
NSR Guidance for Boilers. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/boilers/nsr_fac_boilheat.html
NSR Guidance for Cooling Towers. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/cooling/nsr_fac_cooltow.html
NSR Guidance for Equipment Leak Fugitives. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/fugitives/nsr_fac_eqfug.ht
ml
NSR Guidance for Flares and Vapor Combustors. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/flares/nsr_fac_flares.html
NSR Guidance for Fluid Catalytic Cracking Units (FCCU). (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/cracking/nsr_fac_fccu.html
NSR Guidance for Internal Combustion Engines. (n.d.). Retrieved from
https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/engine/nsr_fac_engine.html
NSR Guidance for Loading Operations. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/loading/nsr_fac_loading.ht
ml
NSR Guidance for Storage Tanks. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/tanks/nsr_fac_tanks.html
NSR Guidance for Sulfur Recovery Units (SRU). (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/sulfur/nsr_fac_sru.html
NSR Guidance for Turbines. (n.d.). Retrieved from https://www.tceq.texas.gov/permitting/air/guidance/newsourcereview/turbines/nsr_fac_turb.html
part-b-an538706-12-2-2016.pdf. (n.d.). Retrieved from http://www.aqmd.gov/docs/default-
source/bact/laer-bact-determinations/aqmd-laer-bact/part-b-an538706-12-2-2016.pdf
reactions of aqua ions with ammonia solution. (n.d.). Retrieved December 13, 2017, from https://www.chemguide.co.uk/inorganic/complexions/aquanh3.html#top
sec6.pdf. (n.d.). Retrieved from https://epd.georgia.gov/air/sites/epd.georgia.gov.air/files/related_files/document/sec6.pdf
Section I - SCAQMD LAER/BACT Determinations. (n.d.). Retrieved from
http://www.aqmd.gov/home/permits/bact/guidelines/i---scaqmd-laer-bact
TN 49920 01-28-09 ECM Technology White Paper.pdf. (n.d.). Retrieved from http://docketpublic.energy.ca.gov/PublicDocuments/Regulatory/Non%20Active%20AFC’s/07-
AFC-9%20Canyon/2009/January/TN%2049920%2001-28-

51
09%20ECM%20Technology%20White%20Paper.pdf
Urea SNCR Systems - NOxOUT® and HERT™ - Fuel Tech Inc. (n.d.). Retrieved March 29,
2018, from https://www.ftek.com/en-US/products/productssubapc/urea-sncr
US EPA, O. (2006, August 30). CICA Technical Resources. Retrieved from https://www3.epa.gov/ttn/catc/cica/atech_e.html
US EPA, O. (2016, August 11). Cost Reports and Guidance for Air Pollution Regulations [Data and Tools]. Retrieved from https://www.epa.gov/economic-and-cost-analysis-air-pollution-regulations/cost-reports-and-guidance-air-pollution
US EPA, O. (n.d.-a). RBLC Search Results. Retrieved from https://cfpub.epa.gov/rblc/index.cfm?action=Results.PermitSearchResults
US EPA, O. (n.d.-b). RBLC Search Results. Retrieved from https://cfpub.epa.gov/rblc/index.cfm?action=Results.PermitSearchResults
US EPA, O. (n.d.-c). RBLC Search Results. Retrieved from https://cfpub.epa.gov/rblc/index.cfm?action=Results.PermitSearchResults
USBROctober2014.pdf. (n.d.). Retrieved from http://47ced92haata3bor58143cb74.wpengine.netdna-cdn.com/wp-
content/uploads/2014/10/USBROctober2014.pdf

PM2.5 SIP Evaluation Report: Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Salt Lake City Nonattainment Area Utah Division of Air Quality Major New Source Review Section Appendix A
Marathon Refinery Emission Unit Monitoring
Emission Unit Capacity Controls
AO
Conditions[1]SIP Conditions Monitoring Established Emission Limit Basis of Limit
Crude Unit Furnace H-101 174 MMBtu/hr (limited) ULNB II.B.3 NOx Limit (new) Stack Test 0.054 lb/MMBtu NOx (3-hour average) Established in AO through UDAQ BACT ProcessUltraformer Unit (UFU) Furnace F-1 140 MMBtu/hr (design duty) ULNB II.B.7.a NOx Limit (new) Stack Test 0.065 lb/MMBtu NOx (3-hour average) Source supplied based on refinery consent decree
UFU Regeneration Heater F-15 12.8 MMBtu/hr (design duty) LNB II.B.7.a --
Annual Boiler
Tune-up --
No limit in new requirements - BACT included in small source BACT document. MACT Subpart DDDDD requires annual boiler tuning for NOx, CO,
and O2.
FCCU/CO Boiler --
ESP; WGS; LoTOx; CO NOx (CO Boiler)II.B.4 Yes CEMs/Stack Test
10 ppmvd NOx (0% O2 on 365-day rolling average)20 ppmvd NOx (0% O2 on 7-day rolling average)10 ppmvd SO2 (0% O2 on 365-day rolling average)
18 ppmvd SO2 (0% O2 on 7-day rolling average)
25 ppmvd SO2 (0% excess air on 365-day rolling average)50 ppmvd SO2 (0% excess air on 7-day rolling average)1.0 lbs PM/1000 lbs coke burned (3-hour average)
1.0 kg PM10/Mg
2016 refinery consent decree and NSPS Subpart Ja
Distillate Desulfurization (DDU) Charge Heater F-680 37.8 MMBtu/hr (combined) ULNB II.B.7.a -- Stack Test -- No limit in new requirements - BACT included in Marathon TSDDDU Rerun Boiler F-681 37.8 MMBtu/hr (combined) ULNB II.B.7.a -- Stack Test -- No limit in new requirements - BACT included in Marathon TSD
Hydrogen Compressors K1s --
Catalytic Converters w/ Air-
to-Fuel Ratio Controller II.B.7.a & II.B.7.c -- Stack Test 3.20 lb/hr NOx or 933 ppmdv @ 10% O2 No limit in new requirements - BACT included in Marathon TSD
South Flare w/ flare gas recovery (FGR) -- FGR System II.B.1.f SO2 Limit CEMs 162 ppmdv H2S (3-hour rolling average) NSPS Subpart Ja
North Flare w/ FGR -- FGR System II.B.1.f SO2 Limit CEMs 162 ppmdv H2S (3-hour rolling average) NSPS Subpart Ja
SRU/Tail Gas Incinerator/Tail Gas Treatment Unit -- Flare/Incinerator Unit II.B.2[2] & II.B.5[1]SO2 Limit CEMs 1.68 tons/day SO2 (24-hour average) Established in AO through UDAQ BACT Process
Emergency/Standby Sources Varies Varies -- -- -- -- Equipment subject to various federal and general regulations
Gasoline Hydrotreater Process Heater F-701 8.0 MMBtu/hr -- II.B.7.a --
Stack Test and Bi-Annual Boiler Tune-up --
No limit in new requirements - BACT included in small source BACT document. MACT Subpart DDDDD requires annual boiler tuning for NOx, CO,
and O2.
Benzene Saturation Unit N/A Not an emission unit N/A -- N/A N/A Closed system - not an emission unit
Cogeneration Turbines w/ HRSG CG1 & CG2 11.8 MW turbine
157.8 MMBtu/hr HRSG SoLoNOx II.B.7.a NOx Limit (new) CEMs/Stack Test
162 ppmdv H2S (3-hour rolling average)
32 ppm NOx @ 15% O2
SOLAR performance guarantee submitted as part of PM2.5 Serious SIP BACT
analysis
Loading/Unloading Racks Varies N/A -- -- -- -- --
Blending Component Loading Rack (BCLR) 1,000 gpm VRU w/ Carbon Adsorption II.B.2 -- CEMs/Stack Test 10 mg/L TOC NSPS Subpart XX and MACT Subpart CC. Has not been used for several years
Cooling Towers -- Drift Eliminators -- -- -- -- --Fugitive Emissions N/A Federal Regulations -- -- -- Federal Regulations NSPS Subpart GGGa and Enhanced LDAR program included in AO.
Tank Farm Storage Tanks Varies Various based on type II.B.9 -- Varies Various requirements based on federal regulation applicability
V917 Fuel Gas System N/A Not an emission unit II.B.6.g CEMs CEMs
162 ppm H2S (3-hour rolling average)60 ppm H2S (365-day rolling average)NSPS Subpart Ja
[1] DAQE-AN103350075-18
[2] DAQE-AN103350081A-21
[3] Applicable federal regulations:
NSPS Subpart A: General Provisions
NSPS Subpart Db: Industrial-Commercial-Institutional Steam Generating Units
NSPS Subpart J: Petroleum Refineries
NSPS Subpart Ja: Petroleum Refineries after 5/14/07
NSPS Subpart K: Storage Vessels 6/11/73-5/19/78
NSPS Subpart Ka: Storage Vessels for Petroleum Liquids 5/18/78-7/23/84
NSPS Subpart Kb: Storage Vessels for Petroleum Liquids after 7/23/84
NSPS Subpart GG: Stationary Gas Turbines
NSPS Subpart XX: Bulk Gasoline Terminals
NSPS Subpart GGGa: VOC Equipment Leaks in Petroleum Refineries after 11/7/06
NSPS Subpart NNN: VOC Emissions from SOCMI
NSPS Subpart QQQ: VOC Emissions from Petroleum Refinery WWTP
NESHAP Subpart A: General Provisions
NESHAP Subpart M: Asbestos
NESHAP Subpart FF: Benzene Waste Operations
MACT Subpart A: General Provisions
MACT Subpart CC: Petroleum Refineries
MACT Subpart UUU: Petroleum Refineries: Unit Specific
MACT Subpart DDDDD: Industrial, Commercial, Institutional Boilers and Heaters
PM2.5 SIP Evaluation Report: Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Salt Lake City Nonattainment Area Utah Division of Air Quality Major New Source Review Section Appendix B
Note: All data in this document is in raw, unprocessed form and includes periods of monitor downtime,
quality assurance, calibration, maintenance, out of control periods, potential malfunctioning CEMs data,
and exempt periods
UDAQ 2023 Data Request - UDAQ Analysis and Summary
SRU Incinerator SO2 Emissions Daily Rate
Marathon Refinery
Total Data Entries 1,826 Min (lbs) 0.00 Min (lbs) 0.001
Total Invalid Day Entries 0 Max (lbs) 1.39 Max (lbs) 1.39
% Total Invalid Hour Entries 0.00% Average (lbs) 0.03 Average (lbs) 0.03
% Un-Matched Data 31.87% Standard Deviation 0.08 Standard Deviation 0.08
%Un-Matched Bad Data 4.55%
Limit (tons SO2/day) 1.68 10th 0.01 10th 0.01
Total Data Entries <= 0 30 20th 0.01 20th 0.01
% Total Data Entries = 0 1.64% 30th 0.02 30th 0.02
Total Data Entries > Limit 0 40th 0.02 40th 0.02
% Total Data Entries > Limit 0.00% 50th 0.02 50th 0.02
60th 0.02 60th 0.02
70th 0.02 70th 0.03
80th 0.03 80th 0.03
90th 0.03 90th 0.03
97th 0.14 97th 0.14
99th 0.44 99th 0.45
Percentage of Limit Value (tons/day) Total Data Points % of Total Data Percentage of Limit Value (tons/day) Total Data Points % of Total Data
<10% of Limit 0.17 1,784 97.70% <10% of Limit 0.17 1,754 97.66%
<20% of Limit 0.34 1,799 98.52% <20% of Limit 0.34 1,769 98.50%
<30% of Limit 0.50 1,810 99.12% <30% of Limit 0.50 1,780 99.11%
<40% of Limit 0.67 1,816 99.45% <40% of Limit 0.67 1,786 99.44%
<50% of Limit 0.84 1,821 99.73% <50% of Limit 0.84 1,791 99.72%
<60% of Limit 1.01 1,824 99.89% <60% of Limit 1.01 1,794 99.89%
<70% of Limit 1.18 1,824 99.89% <70% of Limit 1.18 1,794 99.89%
<80% of Limit 1.34 1,825 99.95% <80% of Limit 1.34 1,795 99.94%
<90% of Limit 1.51 1,826 100.00% <90% of Limit 1.51 1,796 100.00%
<=100% of Limit 1.68 1,826 100.00% <=100% of Limit 1.68 1,796 100.00%
Data Verification Data Analysis - All Data Included Data Analysis - Excluding All Data = 0
Percentiles (lbs):Percentiles (lbs):
UDAQ 2023 Data Request - UDAQ Analysis and Summary
V917 H2S - Rolling 3-Hour Average & Rolling 365-Day Average
Marathon Oil Refinery
Total Data Entries 43,824 Min (ppm) 0 Min (ppm) 1 Min (ppm) 0 Min (ppm) 1
Total Invalid Hour Entries 546 Max (ppm) 300 Max (ppm) 300 Max (ppm) 160 Max (ppm) 160
% Total Invalid Hour Entries 1.25% Average (ppm) 10.19 Average (ppm) 12.38 Average (ppm) 10.05 Average (ppm) 12.20
% Un-Matched Data 45.27% Standard Deviation 11.17 Standard Deviation 11.16 Standard Deviation 9.59 Standard Deviation 9.24
%Un-Matched Bad Data 2.99%
Limit (ppm H2S) 162 10th 0 10th 3 10th 0 10th 3
Total Data Entries <= 0 7,609 20th 1 20th 5 20th 1 20th 5
% Total Data Entries <= 0 17.58% 30th 4 30th 7 30th 4 30th 7
Total Data Entries > Limit 28 40th 7 40th 9 40th 7 40th 9
% Total Data Entries > Limit 0.06% 50th 9 50th 11 50th 9 50th 11
60th 11 60th 13 60th 11 60th 13
70th 14 70th 15 70th 13 70th 15
Total Data Entries 1,826 80th 16 80th 17 80th 16 80th 17
Total Invalid Hour Entries 0 90th 20 90th 22 90th 20 90th 22
% Total Invalid Hour Entries 0.00% 97th 29 97th 31 97th 29 97th 31
% Un-Matched Data 3.07% 99th 41 99th 43 99th 40 99th 43
%Un-Matched Bad Data 0.71%Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
Limit (ppm H2S) 60 <10% of Limit 16 34,976 80.82% <10% of Limit 16 27,367 76.72% <10% of Limit 16 34,976 80.87% <10% of Limit 16 27,367 76.79%
Total Data Entries = 0 0 <20% of Limit 32 42,203 97.52% <20% of Limit 32 34,594 96.99% <20% of Limit 32 42,203 97.58% <20% of Limit 32 34,594 97.06%
% Total Data Entries = 0 0.00% <30% of Limit 49 42,903 99.13% <30% of Limit 49 35,294 98.95% <30% of Limit 49 42,903 99.20% <30% of Limit 49 35,294 99.03%
Total Data Entries > Limit 0 <40% of Limit 65 42,988 99.33% <40% of Limit 65 35,379 99.19% <40% of Limit 65 42,988 99.39% <40% of Limit 65 35,379 99.26%
% Total Data Entries > Limit 0.00% <50% of Limit 81 43,023 99.41% <50% of Limit 81 35,414 99.29% <50% of Limit 81 43,023 99.48% <50% of Limit 81 35,414 99.36%
<60% of Limit 97 43,042 99.45% <60% of Limit 97 35,433 99.34% <60% of Limit 97 43,042 99.52% <60% of Limit 97 35,433 99.42%
<70% of Limit 113 43,063 99.50% <70% of Limit 113 35,454 99.40% <70% of Limit 113 43,063 99.57% <70% of Limit 113 35,454 99.48%
<80% of Limit 130 43,071 99.52% <80% of Limit 130 35,462 99.42% <80% of Limit 130 43,071 99.59% <80% of Limit 130 35,462 99.50%
<90% of Limit 146 43,083 99.55% <90% of Limit 146 35,474 99.45% <90% of Limit 146 43,083 99.61% <90% of Limit 146 35,474 99.53%
<=100% of Limit 162 43,088 99.56% <=100% of Limit 162 35,479 99.47% <=100% of Limit 162 43,088 99.63% <=100% of Limit 162 35,479 99.55%
Min (ppm) 3
Max (ppm) 16
Average (ppm) 10.72
Standard Deviation 2.93
10th 8
20th 8
30th 9
40th 9
50th 9
60th 11
70th 12
80th 15
90th 16
97th 16
99th 16
Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 6 13 0.71%
<20% of Limit 12 1,356 74.26%
<30% of Limit 18 1,826 100.00%
<40% of Limit 24 1,826 100.00%
<50% of Limit 30 1,826 100.00%
<60% of Limit 36 1,826 100.00%
<70% of Limit 42 1,826 100.00%
<80% of Limit 48 1,826 100.00%
<90% of Limit 54 1,826 100.00%
<=100% of Limit 60 1,826 100.00%
Data Analysis - Excluding All Data <= 0 (3-Hr Average) Data Analysis - Excluding All Data > 162 (3-Hr Average) Data Analysis - Excluding All Data = 0 and > 162 (3-Hr Average)
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification - Daily/365-Day Average
Data Analysis - All Data Included (365-Day Averages)
(No Data = 0 or > 60)
Percentiles (ppm):
Data Verification - Hourly/3-Hr Average Data Analysis - All Data Included (3-Hr Average)
UDAQ 2023 Data Request - UDAQ Analysis and Summary
V917 H2S - Rolling 3-Hour Average & Rolling 365-Day Average
Marathon Oil Refinery
Total Data Entries 43,537 Min (ppm) 0 Min (ppm) 1 Min (ppm) 0 Min (ppm) 1
Total Invalid Hour Entries 130 Max (ppm) 185 Max (ppm) 185 Max (ppm) 151 Max (ppm) 151
% Total Invalid Hour Entries 0.30% Average (ppm) 5.01 Average (ppm) 11.41 Average (ppm) 5.00 Average (ppm) 11.39
% Un-Matched Data 51.87% Standard Deviation 8.61 Standard Deviation 9.79 Standard Deviation 8.53 Standard Deviation 9.63
%Un-Matched Bad Data 27.33%
Limit (ppm H2S) 162 10th 0 10th 2 10th 0 10th 2
Total Data Entries <= 0 24,340 20th 0 20th 3 20th 0 20th 3
% Total Data Entries <= 0 56.07% 30th 0 30th 4 30th 0 30th 4
Total Data Entries > Limit 2 40th 0 40th 7 40th 0 40th 7
% Total Data Entries > Limit 0.005% 50th 0 50th 8 50th 0 50th 8
60th 1 60th 11 60th 1 60th 11
70th 5 70th 15 70th 5 70th 15
Total Data Entries 1,635 80th 10 80th 21 80th 10 80th 21
Total Invalid Hour Entries 0 90th 20 90th 25 90th 20 90th 25
% Total Invalid Hour Entries 0.00% 97th 26 97th 29 97th 26 97th 29
% Un-Matched Data 6.97% 99th 30 99th 33 99th 30 99th 33
%Un-Matched Bad Data 0.31%Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
Limit (ppm H2S) 60 <10% of Limit 16 38,000 87.54% <10% of Limit 16 13,660 71.64% <10% of Limit 16 38,000 87.55% <10% of Limit 16 13,660 71.65%
Total Data Entries = 0 0 <20% of Limit 32 43,167 99.45% <20% of Limit 32 18,827 98.74% <20% of Limit 32 43,167 99.45% <20% of Limit 32 18,827 98.75%
% Total Data Entries = 0 0.00% <30% of Limit 49 43,345 99.86% <30% of Limit 49 19,005 99.67% <30% of Limit 49 43,345 99.86% <30% of Limit 49 19,005 99.69%
Total Data Entries > Limit 0 <40% of Limit 65 43,380 99.94% <40% of Limit 65 19,040 99.86% <40% of Limit 65 43,380 99.94% <40% of Limit 65 19,040 99.87%
% Total Data Entries > Limit 0.00% <50% of Limit 81 43,387 99.95% <50% of Limit 81 19,047 99.90% <50% of Limit 81 43,387 99.96% <50% of Limit 81 19,047 99.91%
<60% of Limit 97 43,394 99.97% <60% of Limit 97 19,054 99.93% <60% of Limit 97 43,394 99.97% <60% of Limit 97 19,054 99.94%
<70% of Limit 113 43,399 99.98% <70% of Limit 113 19,059 99.96% <70% of Limit 113 43,399 99.99% <70% of Limit 113 19,059 99.97%
<80% of Limit 130 43,402 99.99% <80% of Limit 130 19,062 99.97% <80% of Limit 130 43,402 99.99% <80% of Limit 130 19,062 99.98%
<90% of Limit 146 43,402 99.99% <90% of Limit 146 19,062 99.97% <90% of Limit 146 43,402 99.99% <90% of Limit 146 19,062 99.98%
<=100% of Limit 162 43,405 100.00% <=100% of Limit 162 19,065 99.99% <=100% of Limit 162 43,405 100.00% <=100% of Limit 162 19,065 100.00%
Min (ppm) 1
Max (ppm) 10
Average (ppm) 4.51
Standard Deviation 1.98
10th 2
20th 3
30th 4
40th 4
50th 4
60th 4
70th 5
80th 6
90th 7
97th 9
99th 10
Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 6 1,433 87.65%
<20% of Limit 12 1,635 100.00%
<30% of Limit 18 1,635 100.00%
<40% of Limit 24 1,635 100.00%
<50% of Limit 30 1,635 100.00%
<60% of Limit 36 1,635 100.00%
<70% of Limit 42 1,635 100.00%
<80% of Limit 48 1,635 100.00%
<90% of Limit 54 1,635 100.00%
<=100% of Limit 60 1,635 100.00%
Data Analysis - Excluding All Data <= 0 (3-Hr Average) Data Analysis - Excluding All Data > 162 (3-Hr Average) Data Analysis - Excluding All Data = 0 and > 162 (3-Hr Average)
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification - Daily/365-Day Average
Data Analysis - All Data Included (365-Day Averages)
(No Data = 0 or > 60)
Percentiles (ppm):
Data Verification - Hourly/3-Hr Average Data Analysis - All Data Included (3-Hr Average)
UDAQ 2023 Data Request - UDAQ Analysis and Summary
FCCU NOx - Rolling 7-Day Averages
Marathon Refinery
Total Data Entries 1,998 Min (ppm) 0 Min (ppm) 1 Min (ppm) 0 Min (ppm) 1
Total Invalid Day Entries 0 Max (ppm) 180 Max (ppm) 180 Max (ppm) 19 Max (ppm) 19
% Total Invalid Day Entries 0.00% Average (ppm) 6.80 Average (ppm) 6.82 Average (ppm) 5.82 Average (ppm) 5.84
% Un-Matched Data 73.12% Standard Deviation 8.58 Standard Deviation 8.58 Standard Deviation 2.30 Standard Deviation 2.29
%Un-Matched Bad Data 5.71%
Limit (ppm NOx) 20 10th 4 10th 4 10th 4 10th 4
Total Data Entries <= 0 4 20th 4 20th 4 20th 4 20th 4
% Total Data Entries <= 0 0.20% 30th 5 30th 5 30th 5 30th 5
Total Data Entries > Limit 42 40th 5 40th 5 40th 5 40th 5
% Total Data Entries > Limit 2.10% 50th 5 50th 5 50th 5 50th 5
60th 6 60th 6 60th 6 60th 6
70th 6 70th 6 70th 6 70th 6
80th 7 80th 7 80th 7 80th 7
90th 9 90th 9 90th 8 90th 8
97th 16 97th 16 97th 12 97th 12
99th 39 99th 39.2 99th 16 99th 16
Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 2 26 1.30% <10% of Limit 2 22 1.10% <10% of Limit 2 26 1.33% <10% of Limit 2 22 1.13%
<20% of Limit 4 439 21.97% <20% of Limit 4 435 21.82% <20% of Limit 4 439 22.44% <20% of Limit 4 435 22.28%
<30% of Limit 6 1,545 77.33% <30% of Limit 6 1,541 77.28% <30% of Limit 6 1,545 78.99% <30% of Limit 6 1,541 78.94%
<40% of Limit 8 1,778 88.99% <40% of Limit 8 1,774 88.97% <40% of Limit 8 1,778 90.90% <40% of Limit 8 1,774 90.88%
<50% of Limit 10 1,866 93.39% <50% of Limit 10 1,862 93.38% <50% of Limit 10 1,866 95.40% <50% of Limit 10 1,862 95.39%
<60% of Limit 12 1,901 95.15% <60% of Limit 12 1,897 95.14% <60% of Limit 12 1,901 97.19% <60% of Limit 12 1,897 97.18%
<70% of Limit 14 1,925 96.35% <70% of Limit 14 1,921 96.34% <70% of Limit 14 1,925 98.42% <70% of Limit 14 1,921 98.41%
<80% of Limit 16 1,943 97.25% <80% of Limit 16 1,939 97.24% <80% of Limit 16 1,943 99.34% <80% of Limit 16 1,939 99.33%
<90% of Limit 18 1,954 97.80% <90% of Limit 18 1,950 97.79% <90% of Limit 18 1,954 99.90% <90% of Limit 18 1,950 99.90%
<=100% of Limit 20 1,956 97.90% <=100% of Limit 20 1,952 97.89% <=100% of Limit 20 1,956 100.00% <=100% of Limit 20 1,952 100.00%
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification - NOx Data Analysis - All Data Included (NOx) Data Analysis - Excluding All Data <= 0 (NOx) Data Analysis - Excluding All Data > 20 (NOx) Data Analysis - Excluding All Data <= 0 and > 50 (NOx)
UDAQ 2023 Data Request - UDAQ Analysis and Summary
FCCU SO2 - Rolling 7-Day Averages
Marathon Refinery
Total Data Entries 2,057 Min (ppm) -2 Min (ppm) 1 Min (ppm) -2 Min (ppm) 1
Total Invalid Day Entries 59 Max (ppm) 51 Max (ppm) 51 Max (ppm) 18 Max (ppm) 18
% Total Invalid Day Entries 2.87% Average (ppm) 2.51 Average (ppm) 4.59 Average (ppm) 2.27 Average (ppm) 4.19
% Un-Matched Data 79.24% Standard Deviation 4.63 Standard Deviation 5.44 Standard Deviation 3.66 Standard Deviation 4.09
%Un-Matched Bad Data 2.28%
Limit (ppm SO2) 18 10th 0 10th 1 10th 0 10th 1
Total Data Entries <= 0 907 20th 0 20th 1 20th 0 20th 1
% Total Data Entries <= 0 45.40% 30th 0 30th 1 30th 0 30th 1
Total Data Entries > Limit 16 40th 0 40th 1 40th 0 40th 1
% Total Data Entries > Limit 0.80% 50th 1 50th 2 50th 1 50th 2
60th 1 60th 3 60th 1 60th 3
70th 2 70th 6 70th 2 70th 6
80th 4 80th 9 80th 4 80th 8
90th 9 90th 11 90th 9 90th 11
97th 13 97th 15 97th 12 97th 13
99th 16 99th 21 99th 14.17 99th 15
Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 2 1,526 76.38% <10% of Limit 2 619 56.74% <10% of Limit 2 1,526 76.99% <10% of Limit 2 619 57.58%
<20% of Limit 4 1,601 80.13% <20% of Limit 4 694 63.61% <20% of Limit 4 1,601 80.78% <20% of Limit 4 694 64.56%
<30% of Limit 5 1,641 82.13% <30% of Limit 5 734 67.28% <30% of Limit 5 1,641 82.80% <30% of Limit 5 734 68.28%
<40% of Limit 7 1,718 85.99% <40% of Limit 7 811 74.34% <40% of Limit 7 1,718 86.68% <40% of Limit 7 811 75.44%
<50% of Limit 9 1,813 90.74% <50% of Limit 9 906 83.04% <50% of Limit 9 1,813 91.47% <50% of Limit 9 906 84.28%
<60% of Limit 11 1,909 95.55% <60% of Limit 11 1,002 91.84% <60% of Limit 11 1,909 96.32% <60% of Limit 11 1,002 93.21%
<70% of Limit 13 1,952 97.70% <70% of Limit 13 1,085 99.45% <70% of Limit 13 1,952 98.49% <70% of Limit 13 1,045 97.21%
<80% of Limit 14 1,963 98.25% <80% of Limit 14 1,056 96.79% <80% of Limit 14 1,963 99.04% <80% of Limit 14 1,056 98.23%
<90% of Limit 16 1,979 99.05% <90% of Limit 16 1,072 98.26% <90% of Limit 16 1,979 99.85% <90% of Limit 16 1,072 99.72%
<=100% of Limit 18 1,982 99.20% <=100% of Limit 18 1,075 98.53% <=100% of Limit 18 1,982 100.00% <=100% of Limit 18 1,075 100.00%
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification - SO2 Data Analysis - All Data Included (SO2) Data Analysis - Excluding All Data <= 0 (SO2) Data Analysis - Excluding All Data > 18 (SO2) Data Analysis - Excluding All Data <= 0 and > 50 (SO2)
UDAQ 2023 Data Request - UDAQ Analysis and Summary
North Flare H2S - Rolling 3-Hour Average
Marathon
Total Data Entries 69,143 Min (ppm) 0 Min (ppm) 1 Min (ppm) 0 Min (ppm) 1
Total Invalid Hour Entries 1,132 Max (ppm) 300 Max (ppm) 300 Max (ppm) 162 Max (ppm) 162
% Total Invalid Hour Entries 1.64% Average (ppm) 3.67 Average (ppm) 10.27 Average (ppm) 2.68 Average (ppm) 7.56
% Un-Matched Data 22.58% Standard Deviation 18.22 Standard Deviation 29.34 Standard Deviation 8.75 Standard Deviation 13.40
%Un-Matched Bad Data 3.68%
Limit (ppm H2S) 162 10th 0 10th 1 10th 0 10th 1
Total Data Entries = 0 43,698 20th 0 20th 1 20th 0 20th 1
% Total Data Entries = 0 64.25% 30th 0 30th 2 30th 0 30th 2
Total Data Entries > Limit 268 40th 0 40th 3 40th 0 40th 3
% Total Data Entries > Limit 0.39% 50th 0 50th 4 50th 0 50th 4
60th 0 60th 6 60th 0 60th 6
70th 1 70th 8 70th 1 70th 8
80th 3 80th 11 80th 3 80th 10
90th 8 90th 15 90th 8 90th 14
97th 16 97th 44 97th 15 97th 30
99th 49.88 99th 180 99th 32 99th 78
Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 16 66,203 97.34% <10% of Limit 16 22,505 92.56% <10% of Limit 16 66,203 97.73% <10% of Limit 16 22,505 93.60%
<20% of Limit 32 67,078 98.63% <20% of Limit 32 23,380 96.16% <20% of Limit 32 67,078 99.02% <20% of Limit 32 23,380 97.23%
<30% of Limit 49 67,331 99.00% <30% of Limit 49 23,633 97.20% <30% of Limit 49 67,331 99.39% <30% of Limit 49 23,633 98.29%
<40% of Limit 65 67,441 99.16% <40% of Limit 65 23,743 97.66% <40% of Limit 65 67,441 99.55% <40% of Limit 65 23,743 98.74%
<50% of Limit 81 67,516 99.27% <50% of Limit 81 23,818 97.96% <50% of Limit 81 67,516 99.66% <50% of Limit 81 23,818 99.06%
<60% of Limit 97 67,588 99.38% <60% of Limit 97 23,890 98.26% <60% of Limit 97 67,588 99.77% <60% of Limit 97 23,890 99.36%
<70% of Limit 113 67,632 99.44% <70% of Limit 113 23,934 98.44% <70% of Limit 113 67,632 99.84% <70% of Limit 113 23,934 99.54%
<80% of Limit 130 67,675 99.51% <80% of Limit 130 23,977 98.62% <80% of Limit 130 67,675 99.90% <80% of Limit 130 23,977 99.72%
<90% of Limit 146 67,712 99.56% <90% of Limit 146 24,014 98.77% <90% of Limit 146 67,712 99.95% <90% of Limit 146 24,014 99.87%
<=100% of Limit 162 67,743 99.61% <=100% of Limit 162 24,045 98.90% <=100% of Limit 162 67,743 100.00% <=100% of Limit 162 24,045 100.00%
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification Data Analysis - All Data Included Data Analysis - Excluding All Data = 0 Data Analysis - Excluding All Data > 162 Data Analysis - Excluding All Data = 0 and > 162
UDAQ 2023 Data Request - UDAQ Analysis and Summary
South Flare H2S - Rolling 3-Hour Average
Marathon
Total Data Entries 69,143 Min (ppm) 0 Min (ppm) 1 Min (ppm) 0 Min (ppm) 1
Total Invalid Hour Entries 26 Max (ppm) 300 Max (ppm) 300 Max (ppm) 162 Max (ppm) 162
% Total Invalid Hour Entries 0.04% Average (ppm) 10.81 Average (ppm) 14.15 Average (ppm) 9.68 Average (ppm) 12.69
% Un-Matched Data 48.79% Standard Deviation 23.86 Standard Deviation 26.42 Standard Deviation 16.82 Standard Deviation 18.24
%Un-Matched Bad Data 13.45%
Limit (ppm H2S) 162 10th 0 10th 2 10th 0 10th 2
Total Data Entries = 0 16,308 20th 0 20th 2 20th 0 20th 2
% Total Data Entries = 0 23.59% 30th 2 30th 3 30th 2 30th 3
Total Data Entries > Limit 318 40th 2 40th 4 40th 2 40th 4
% Total Data Entries > Limit 0.46% 50th 3 50th 5 50th 3 50th 5
60th 4 60th 7 60th 4 60th 7
70th 7 70th 11 70th 7 70th 11
80th 13 80th 19 80th 13 80th 19
90th 31 90th 39 90th 30 90th 37
97th 60 97th 67 97th 57 97th 62
99th 90 99th 103 99th 79 99th 83
Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data Percentage of Limit Value (ppm) Total Data Points % of Total Data
<10% of Limit 16 57,341 82.96% <10% of Limit 16 41,033 77.70% <10% of Limit 16 57,341 83.35% <10% of Limit 16 41,033 78.17%
<20% of Limit 32 62,456 90.36% <20% of Limit 32 46,148 87.39% <20% of Limit 32 62,456 90.78% <20% of Limit 32 46,148 87.92%
<30% of Limit 49 65,674 95.02% <30% of Limit 49 49,366 93.48% <30% of Limit 49 65,674 95.46% <30% of Limit 49 49,366 94.05%
<40% of Limit 65 67,467 97.61% <40% of Limit 65 51,159 96.88% <40% of Limit 65 67,467 98.06% <40% of Limit 65 51,159 97.46%
<50% of Limit 81 68,195 98.67% <50% of Limit 81 51,887 98.25% <50% of Limit 81 68,195 99.12% <50% of Limit 81 51,887 98.85%
<60% of Limit 97 68,539 99.16% <60% of Limit 97 52,231 98.91% <60% of Limit 97 68,539 99.62% <60% of Limit 97 52,231 99.50%
<70% of Limit 113 68,659 99.34% <70% of Limit 113 52,351 99.13% <70% of Limit 113 68,659 99.80% <70% of Limit 113 52,351 99.73%
<80% of Limit 130 68,726 99.43% <80% of Limit 130 52,418 99.26% <80% of Limit 130 68,726 99.89% <80% of Limit 130 52,418 99.86%
<90% of Limit 146 68,764 99.49% <90% of Limit 146 52,456 99.33% <90% of Limit 146 68,764 99.95% <90% of Limit 146 52,456 99.93%
<=100% of Limit 162 68,799 99.54% <=100% of Limit 162 52,491 99.40% <=100% of Limit 162 68,799 100.00% <=100% of Limit 162 52,491 100.00%
Percentiles (ppm): Percentiles (ppm): Percentiles (ppm): Percentiles (ppm):
Data Verification Data Analysis - All Data Included Data Analysis - Excluding All Data = 0 Data Analysis - Excluding All Data > 162 Data Analysis - Excluding All Data = 0 and > 162
PM2.5 SIP Evaluation Report: Tesoro Refining & Marketing Company LLC d/b/a Marathon Salt Lake City Salt Lake City Nonattainment Area Utah Division of Air Quality Major New Source Review Section Appendix C
Marathon Emission Calculations - Check
Gas-Fired Combustion Units with Proposed Limits
1,020
8,760
7.60
Cogeneration Turbines PM2.5 Emission Factor (lb/MMBtu)[3]0.018
0.01
Emission Unit
Heat Input
Capacity
(MMBtu/hr)
Proposed
NOx Limit
(lb/MMBtu)[5]
2017
Hours of Operation
(hrs/yr)
NOx
Emissions
(tons/yr)
2017 NOx
Inventory
(tons/yr)
PM2.5
Emissions
(tons/yr)
2017 PM2.5
Inventory
(tons/yr)
SO2
Emissions
(tons/yr)
2017 SO2
Inventory
(tons/yr)
Crude Unit Furnace H-101 174.0 0.054 8,197 41.15 31.52 5.68 5.34 7.62 1.38
UFU Furnace F-1 140.0 0.065 8,465 39.86 11.10 4.57 1.67 6.13 0.43
Cogeneration Turbines w/ HRSG 198.1 0.12 8,565 104.12 50.00 15.62 12.82 8.68 0.65
Cogeneration Turbines w/ HRSG 198.1 0.12 8,504 104.12 50.39 15.62 12.64 8.68 0.59
289.26 143.01 41.48 32.47 31.11 3.05
[1] AP-42 Section 1.4.1
[2] AP-42 Section 1.4
[3]Emission factor from 2017 Inventory
[4] Based on H2S limit from NSPS Subpart Ja 365-day average at 0% O2: 60 ppm H2S, assuming full conversion to SO2
[5]Cogeneration turbines NOx limit from proposed Part H Section IX.H.12 limit of 32 ppm NOx @ 15% O2 converted to lb/MMBtu
Source-wide PM2.5 limit adopted by AQB July 1, 2018: 179.0 tons/yr
Source-wide NOx Limit adopted by AQB July 1, 2018: 475.0 tons/yr
Source-wide SO2 Limit adopted by AQB July 1, 2018: 300.0 tons/yr
Total Emissions
Assumed: Refinery gas is equivalent to natural gas
Heating Value of Refinery Gas (Btu/scf)[1]
Maximum Hour of Operation (hrs/yr)
PM2.5 Emission Factor (lb/MMscf)[2]
SO2 Emission Factor (lb/MMBtu)[4]