Assessment of Flare Smokeless Capacity Estimation Techniques

Paper from the AFRC 2016 conference titled Assessment of Flare Smokeless Capacity Estimation Techniques

On December 1, 2015, the United States Environmental Protection Agency (USEPA) finalized regulations concerning refinery flares and have included standards for improving combustion efficiency and managing tip exit velocity. As part of these new regulations, the USEPA has required facilities to designate a single "smokeless capacity" for each affected flare. Although many flare manufacturers specify a smokeless capacity in the design documentation for a flare, this value is typically based on a set of conditions specified by the end user. The actual day-to-day smokeless capacity will depend on a range of factors, including vent gas composition, vent gas flow rate, and the actual amount of assist gas available when needed. This paper will present a variety of approaches for estimating the smokeless capacity of a flare, ranging from available USEPA guidance to recent peer-reviewed literature focusing on the predicting smoke formation based on the composition of the vent gas.

Stuck 1 Assessment of Flare Smokeless Capacity Estimation Techniques Derek Stuck, Sage ATC Environmental Consulting, LLC Eric Boley, Sage ATC Environmental Consulting, LLC American Flame Research Council 2016 Industrial Combustion Symposium Kauai, Hawaii September 13, 2016 The United States Environmental Protection Agency (USEPA) recently finalized regulations concerning refinery flares and have included standards for improving combustion efficiency and managing tip exit velocity. As part of these new regulations, the USEPA has required facilities to designate a single "smokeless capacity" for each affected flare. Regulatory Background On December 1, 2015, the USEPA published amendments to the Refinery Sector Rule (RSR). The RSR includes both the refinery Maximum Achievable Control Technology (MACT) 1 and 2 rules, which are found in the National Emission Standards for Hazardous Air Pollutants (NESHAPs) Part 63 Subpart CC and Subpart UUU (referred to herein as MACT CC and MACT UUU, respectively). Source categories within both MACT CC and MACT UUU that utilize a flare as a control device are subject to the flare requirements found in §63.670 of MACT CC, which require refineries with affected flares to meet specific operating limits related to: • • • • • Pilot flare monitoring and presence, Flare tip exit velocity, Visible emissions, Net heating value in the combustion zone (NHV ), and Net heating value dilution parameter (NHV ) for flares which receive perimeter assist air. cz dil The requirements related to pilot flame monitoring and presence, visible emissions, and flare tip velocity were previously found in §63.11 of MACT Subpart A. Provided a refinery flare is in compliance with the amended requirements in MACT CC, that flare is no longer required to comply with the requirements of §63.11 of MACT Subpart A or §60.18 of New Source Performance Standard (NSPS) Subpart A. Refinery flares subject to the provisions of MACT CC are required to be in compliance with all flaring provisions no later than January 30, 2019. In addition to specifying limits for the aforementioned parameters, the emergency flaring provisions of MACT CC require refineries to establish a single value for the design smokeless capacity of each affected flare [§63.670(o)(1)(iii)(B)] if that flare has the potential to operate above its respective design smokeless capacity. The design smokeless capacity will be listed in Stuck 2 the Flare Management Plan (FMP) required by MACT CC and any change to the smokeless capacity will require the FMP to be resubmitted to the USEPA. The specified design smokeless capacity will be used to determine if a root cause analysis (RCA) and a corrective action analysis are required. If either of the following conditions are met, the refinery will be required to prepare a RCA and corrective action analysis within 45 days of the event. • The vent gas flow rate exceeds the smokeless capacity of the flare and visible emissions are present from the flare for more than five (5) minutes during any two (2) consecutive hours during the release event. The vent gas flow rate exceeds the smokeless capacity of the flare and the 15-minute block average flare tip velocity exceeds the maximum flare tip velocity determined using equation (1) [§63.670(d)(2)]. • Ultimately, the goal of these regulations is to ensure that flares are operated with a combustion efficiency of at least 96.5% (or a destruction efficiency of 98%). Previous guidance and testing has shown that flares achieve the highest combustion efficiency at the incipient smoke point, which is typically thought of as the operating condition at which slight, barely visible wisps of smoke are present. So taking this guidance along with the limits established in the aforementioned regulations, we now have been instructed to operate with faint wisps of smoke and without visible emissions for more than five (5) minutes in a two (2) hour period. This paper will seek to establish a balance between the requirements to both operate a flare as efficiently as possible, while at the same time minimizing visible emissions as much as possible. In doing so, the paper will provide an overview of historical and developing a smokeless capacity value for industrial flares. 1 Formation of Visible Emissions Visible emissions, or smoke, from flares are formed when there is insufficient oxygen in the combustion zone to complete the oxidation of hydrocarbon molecules to carbon dioxide. Instead of bonding to oxygen, the carbon molecules bond to each other, creating chains of carbon, or soot. The visible flame of a flare is actually the result of these soot particles glowing inside the flame . As the particles cool, they appear black, resulting in the characteristic black smoke which the regulations are attempting to limit. 2 Generally speaking, the more carbon a hydrocarbon contains, the more likely it is to smoke. This tendency for a compound to smoke can be defined by a specific characteristic called the Allen, D., Torres, V., et al. TCEQ 2010 Flare Study Final Report. University of Texas at Austin - Center for Energy and Environmental Resources. https://www.tceq.texas.gov/assets/public/implementation/air/rules/Flare/2010flarestudy/2010-flare-study-finalLowest Highest Smoke Points report.pdfSmoke Points Baukal, C., Schwartz, C., Schwartz, R. & John Zink Company. (2001). The John Zink Combustion Handbook (Industrial Combustion series). Boca Raton, FL: CRC Press. 1 2 Stuck 3 smoke point (SP), which can be defined as the height of a flame burning only that compound as fuel. As the simplest hydrocarbon, methane is thought to have the longest SP, while aromatic compounds, such as benzene and toluene, have small SPs. The general trend for a compound's tendency to soot follows the scheme : 3 Parrafins •Ethane •Proane •Butane Olefins Diolefins •Ethlyne •Propylene •Butene Isomers •1,3-Butadiene •Propadiene Acetylenes •Acetylene •Propyne Aromatics •Benzene •Toluene •Styrene SP also tends to decrease as molecular weight increases within each of the groups presented above. Conceptually, this makes sense as the increased molecular weight is the result of additional carbons being added to the molecule. Li and Sunderland provide a list of the SPs for over 100 hydrocarbons . 4 Prevention of Visible Emissions The first smokeless flare was built by the John Zink Company in 1952 . The John Zink Combustion Handbook describes smokeless flare operation as a function of momentum and divides flares into two groups: high and low pressure. High pressure (between 5 and 10 psig) flares are able to take advantage of that pressure to induce mixing at the flare tip. Low pressure flares, however, need additional momentum to be added to induce sufficient mixing in the combustion zone. As such, visible emissions from these low pressure flares are controlled through the addition of assist gas. Typically, either steam or air (and in some cases both) are added to the combustion zone. In the case of steam, oxygen from the surrounding ambient air is entrained in the steam as it is injected, thus aerating the combustion zone of the flare. Assist air directly adds additional oxygen to the combustion zone. 2 Steam is typically added through one or more of three (3) injection points: center, upper ring, and lower ring. Center steam is not typically thought to contribute to the prevention of visible emissions. As the diameter of a flare tip increases, upper ring steam becomes less effective at entraining air into the combustion zone. As such, larger steam assisted flares may also include the addition of a steam/air mixture through tubes located within the perimeter of the flare tip. Air assisted flares feature a variety of designs within the perimeter of the flare tip. Similar to lower steam addition, the internal flare geometry is designed to encourage efficient mixing between the flare vent gas and the assist air. Glassman, I. 1989. Soot formation in combustion processes. Proc. Combust. Inst., 22, 295-311. Li, L., and Sunderland, P. "An Improved Method of Smoke Point Normalization." Combustion Science and Technology 184.6 (2012): 829-41. http://terpconnect.umd.edu/~pbs/2012-Li-and-Sunderland-CST.pdf 3 4 Stuck 4 The use of steam or air to prevent is a proven method of preventing visible emissions. Over the last 30 years, the prevention of visible emissions has been the overriding concern regarding flares subject to federal environmental regulations. So much so that operators at many facilities learned that a surefire way of keeping the flare from smoking was to add significant quantities of assist gas. Unfortunately, as both industry and the USEPA have come to realize, the addition of excess assist gas was not without its consequences. Specifically, over-assisted flares failed to achieve the 98% destruction efficiency many facilities claimed, potentially resulting in excess emissions. This new awareness of over-assisting has driven both the USEPA Flare Enforcement Initiative and the development of new regulations over the last decade. With the focus on ensuring flares are not over-assisted, additional attention must also be given to the control of assist gas going to the flare. Assist Gas Control Typically, steam control should be comprised of a configuration with at least three (3) levels of control for upper ring and lower ring steam additions. Center steam additions typically require less variable control as this addition point is primarily utilized to prevent burn back into the flare tip and not to reduce smoke formation. The first level should include a steam line that is equipped with a restriction orifice (RO) to provide the manufacturer's recommended minimum steam flow to the tip at all times. The second level of control would be a smaller steam control valve that should be utilized to respond to the initial increases in vent gas flow and would provide the ability to fine tune the steam flow during low flow flaring events. The third level of control would be a larger control valve, which should be utilized to respond to high flow flaring events that cannot be controlled by the smaller steam line. Although not required by regulations, an automated steam-to-vent gas ratio control system is also commonly used to ensure sufficient steam is added to a flare during the start of an event and (perhaps more importantly) removed as the event subsides. Air assisted flares are typically equipped with one or more blowers at the base of the flare stack, although some designs allow for the injection of compressed air through nozzles similar to those used for steam injection. For blower-equipped flares, the amount of assist air supplied to the combustion zone is a function of the fan speed of the blower, which may be equipped with a single-speed, dual-speed, or variable frequency motor. Using a variable frequency motor provides the best control over the amount of assist air added to the combustion zone. Ultimately, a facility's ability to control the amount of assist gas provided to a flare is of critical importance in determining the smokeless capacity of that flare. If the assist gas cannot be provided accurately or reliably, flare performance may be impacted as a result of either visible emissions or over-assisting the flare. Design Smokeless Capacity Stuck 5 Although many flare manufacturers specify a smokeless capacity (typically 10-20% of the maximum hydraulic capacity of the flare) in the design documentation for a flare, this value is typically based on a set of conditions specified by the end user. The actual day-to-day smokeless capacity will depend on a range of factors, including flare tip design, vent gas composition, vent gas flow rate, and the actual amount of assist gas available when needed. Although the USEPA has required a single value for the smokeless capacity of a given flare, in actuality, smokeless capacity at a given time is a variable value which will depend primarily on two parameters: • • The amount of assist gas the refinery is able to provide, and The composition of the vent gas. The requirement to specify the design smokeless capacity leads to a number of concerns. First, the amount of assist gas (primarily steam or air) a refinery is able to provide to a flare is likely now a different amount than when the flare was purchased and designed. Apart from an overall change in steam capacity at the refinery, a number of other scenarios can affect this. For steam assisted flares, during a process upset, a boiler may go down, reducing the amount of steam available to be provided to the flare. Also, flares are typically located a significant distance apart from other process units, meaning that even though the boilers may produce a certain amount of steam, pressure drops and condensation losses may prevent that amount from being provided to the flare. For air assisted flares, the blower(s) used to provide the assist air require electricity, meaning that during a power outage scenario, those blowers will be incapable of providing assist gas. For both steam and air assisted flares, the aforementioned scenarios could potentially result in visible emissions (i.e., a smoking flare). Like MACT Subpart A, MACT CC specifies that flares are prohibited from operating with visible emissions for more than five (5) minutes in any two (2) hour period. As such, visible emissions from any of the discussed scenarios could potentially result in a deviation of the standard if the conditions persist for more than the allowed five (5) minutes. The actual smokeless capacity of a flare is also highly dependent on the composition of the vent gas going to that flare. When flare tip manufacturers list the smokeless capacity of a flare, that capacity is typically provided with a molecular weight which represents an estimated relief scenario at the refinery. Assuming the maximum amount of assist gas is available during the specified relief scenario, the design smokeless capacity and the actual smokeless capacity would converge. However, the specified relief scenario is only one of potentially hundreds of scenarios (each potentially with its own unique composition profile) which could result a smoking flare. Assessment of Smokeless Capacity Estimation Techniques Historically, there has been little to no USEPA guidance regarding how to establish a smokeless capacity for a flare. As such, flare manufacturers, industry, and more recently, academic Stuck 6 institutions, have led the way in developing novel methods of determining the smokeless capacity of a flare. Up until recent academic research, a common thread in many of these techniques was the tendency to establish a single value for the smokeless capacity of a flare. As previously described, a single value cannot be expected to account for the wide range of potential operating conditions an industrial flare is expected to encounter. Synthetic Organic Chemical Manufacturing Industry Many recent studies of the smokeless capacity for industrial flares state that hydrocarbons with a carbon-to-hydrogen ratio of greater 0.35 are more likely to result in visible emissions. The source of this number is unclear as the ratio appears to have originated in Background Information for Proposed Standards for Reactor Processes in Synthetic Organic Chemical Manufacturing Industry (SOCMI) . The Background Information states the following: 5 "Poor mixing at the flare tip or poor flare maintenance can cause smoking (particulate). Fuels with high carbon to hydrogen ratios (greater than 0.35) have a greater tendency to smoke and require better mixing if they are to be burned smokelessly." From a strictly intuitive standpoint, this appears to match expectations. Compounds like methane and ethane are not typically expected to smoke and their carbon-to-hydrogen ratios are 0.25 and 0.33, respectively. Single double-bonded olefins, such as ethylene and propylene, share a carbon-to-hydrogen ratio of 0.5 and are generally thought of as more likely to result in visible emissions. However, no matter how intuitive the ratio may seem, actual test data or theoretical calculations are needed to confirm the statement, and no source or method for how this ratio was arrived at is provided within the SOCMI Background Information. This value is cited in other USEPA guidance and academic research, but ultimately the citations refer back to the SOCMI Background Information. 6 API 521 - Pressure-relieving and Depressuring Systems Although there does not appear to be smokeless capacity guidance available from the American Petroleum Institute (API) or the American Fuel and Petrochemical Manufacturers (AFPM), the API does provide recommended steam-to-vent gas ratios for smokeless combustion of 14 compounds in API Standard 521 - Pressure-relieving and Depressuring Systems . The table does not take into account the potential effects of the different compounds being mixed together in the 7 Reactor Processes in Synthetic Organic Chemical Manufacturing Industry - Background Information for Proposed Standards. EPA-450/3-90-016a. U.S. Environmental Protection Agency - Office of Air and Radiation. June 1990. Air Pollution Control Cost Manual - Section 3 - VOC Controls. EPA/452/B-02-001. U.S. Environmental Protection Agency - Office of Air Quality Planning and Standards. September 2000 American Petroleum Institute Standard 521 - Pressure-relieving and Depressuring Systems. 5 Edition. January 2007. 5 6 7 th Stuck 7 flare vent gas and, like the SOCMI guidelines, the API standard does not present a source for the prescribed ratios. Although the standard does not specify how to determine the smokeless capacity of a flare, the referenced table presents the possibility that it could be used to estimate the amount of steam required for a specific gas mixture, using composition data being reported by an analyzer such as a gas chromatograph or a mass spectrometer. However, recent studies and tests have indicated that the values recommended in the table are inconsistent with real-world observations. As shown in Figure 1, API 521 appears to show a correlation between the steam-to-vent gas ratios and the molecular weight of the listed compounds. The groups of hydrocarbons (paraffins, olefins, aromatics, etc.) all have varying steam to vent gas ratios based on their molecular weight. The relationship for each group is linear. Suggest Steam vs MW Suggested Steam Ratio lb/lb 1.4 Paraffins 1.2 1 Olefins 0.8 Diolefins and Acetylenes Aromatics 0.6 0.4 0.2 0 0 20 40 60 80 100 MW, g/mol Figure 1 - Suggested S/VG Ratios vs. Molecular Weight 120 Stuck 8 Based on solely on Figure 1, molecular weight would appear to be a good general guideline for steam use (bigger molecules have more tendency to smoke), but it is lacking in that it does not take into account aromaticity, double or triple bonds, or the carbon to hydrogen ratio. Further problems arise when one tries to take the API 521 suggested steam ratio for a mixture. In a flare study done at Shell Deer Park Refinery , comparing the molecular weight of the vent gas with the calculated suggested steam ratio produced no discernible pattern. Instead, the steam ratio was impossible to determine based solely on the molecular weight. This was due to the variability of the vent gas during the test and the fact that the average molecular weight of the vent gas is not representative of a single compound. This comparison (as shown in test plan) is shown below in Figure 2. 8 Figure 2 - API 521 Suggested Steam Ratio vs. Average Molecular Weight It may be possible to use composition data (instead of average molecular weight) to determine the correct amount of steam to supply to a flare in order to maintain efficient combustion and prevent the flare from smoking; however, at this time the necessary data is not publicly available. Although API 521 does not address the issue of determining a smokeless capacity value, with more testing, it does present the possibility of improving a facility's ability to better control the addition of steam. John Zink Combustion Handbook According to the John Zink Combustion Manual , the smokeless capacities of many early flares were based on the "vendor performance testing, field tests, or ‘rules of thumb' based on 2 8 Shell Deer Park Refining LP Deer Park Refinery East Property Flare Performance Test Report. April 1, 2011. Stuck 9 operating experience." Attempts to estimate the smokeless capacity were limited in part by a lack of monitoring information related to flares, and perhaps unsurprisingly, these methods often failed to reliably determine the smokeless capacity of industrial flares. However, over the last 30 years, monitoring instrumentation has been installed on flares throughout the world, providing a data set from which a more reliable method of estimating smokeless. In order to address the failings of the previous estimation techniques, engineers at John Zink developed new methods for estimating the smokeless capacity of new flares. The John Zink Flare Performance Model is a proprietary method for estimating the smokeless capacity of a given flare which "scales the smoke evolution from turbulent diffusion rates by taking the ratio of carbon burn-off, represented by flame length scaling, to the soot formation rate, which is represented by an appropriate set of scaling parameter requirements." The Model allows John Zink to predict a given flare's tendency to smoke at a specific set of initial conditions. However, the applicability of the stated smokeless capacity outside of those initial conditions (particularly as steam availability and composition change) is unclear. 2 Recent Peer-Reviewed Research Over the last decade, researchers at several academic institutions appear to have taken a greater interest in deriving a method to estimate the smoke formation potential from industrial flares. Much of the research has focused on the aforementioned SPs of different compounds. At this time, there still appears to be no established mathematical relationship for calculating the SP of a mixture. Although some experimental data exists for gas mixtures, the data is typically for mixtures containing at most four (4) components, and flare vent gas is typically composed of a much greater number of components. Markstein proposes a method (Equation 1) for calculating the SP volumetric flow rate of any mixture. 9 𝑄"#,%&' = )* 𝑋& . 01 𝐸𝑞. 1 𝑄,-,& Where Q = Smoke point volumetric flow rate of the mixture X = Component mole fraction Q = Component volumetric smoke point SP,mix i sp,i Equation 1 has been tested using mixtures up to four fuel components with results typically in agreement with the predictions. However, flare vent gas is expected to contain significantly more than four components. Additional testing is necessary to determine the effect of adding more components. G.H. Markstein, "Radiant Emission and Smoke Points for Laminar Diffusion Flames of Fuel Mixtures", Proceedings of the Combustion Institute 21 (1986) 1107-1114. 9 Stuck 10 Research at Carleton University and Lamar University is currently underway to develop mechanisms and models to predict the amount of soot, and, as a result, smoke, formation for specific fuel mixtures . Lab scale testing at Carleton and computational fluid dynamic modeling at Lamar have led to the development of a chemical reaction mechanism referred to as LU3.0.1. The mechanism predicts the formation of soot, in addition to volatile organic compound emissions, from flares and "has been validated successfully against experimental performance indicators." Although work continues on LU3.0.1, the development is encouraging as the mechanism has the potential for use as a predictor of real-time smokeless capacity. 10 Potential Novel Approaches to Smokeless Capacity Despite the development of LU3.0.1, other novel approaches to smokeless capacity still need to be investigated. Many flares subject to MACT CC are expected to install gas composition analyzers over the next two years and the data provided by those instruments may lead to a wave of changes in how flares are designed and how capacities are estimated. In particular, the composition data may allow for the previously mentioned single value for the design smokeless capacity to effectively vary with changes in composition. Consider the possibility of defining the smokeless capacity as a molar flow rate instead of a volumetric or mass flow rate. If the smokeless capacity is expressed in terms of moles of carbon, this effectively grants flows of smaller hydrocarbons such as methane and ethane a higher smokeless capacity. Conclusions Given the challenges of choosing a single value for the design smokeless capacity of a flare, a method of determining a variable smokeless capacity is desirable, although the regulatory requirement to designate a single value for the smokeless presents a problem. However, MACT CC does not designate the units or averaging period for the single value of the design smokeless capacity (it is implied to be a volumetric flow rate, but this is never explicitly stated). In order to comply with the requirements of the Rule and take into account the aforementioned considerations upon which the smokeless capacity is dependent, the ideal technique would present a single value as required which would remain constant, yet would respond to changes in assist gas availability and vent gas composition. That ideal technique is not yet available, but recent research at Lamar and Carleton Universities points toward the possibility that one will be developed in the near future. Chen, D., Lou, H., Li, X., Richmond, P., and Johnson, M. "Flare Performance Optimization: DRE/CE cs. Soot - Phase I Final Report." Texas Commission on Environmental Quality. Project # 582-10-94307-FY14-06. August 2014. https://www.tceq.texas.gov/assets/public/implementation/air/am/contracts/reports/oth/5821094307FY140620140831-lamaru-Flare_DRE_CE_Soot.pdf 10 Stuck 11