A technology for measuring combustion efficiency of industrial & field flares (integrating measurements and simulations)

Flare research over the past decade has increasingly illustrated that there is likely no one simple operational parameter (or even a few parameters) that will characterize the combustion behavior of flare flames.i Simple correlations are unobtainable because of the complexity of the nonlinear mixing, reaction, and heat transfer present in operating flare flames. This complexity motivates the need to accurately measure combustion efficiency from operating flares so that the effect of different designs and operations can be quantified. However, this same complexity, makes such measurements difficult. To date there is no technology that has been been validated at providing quantitative combustion efficiency measurements from operating, open-air, flare flames. While it is difficult to directly measure combustion efficiency due to the heterogeneity in composition and velocity, high performance computing (HPC) can be coupled with diagnostic measurement methods to provide real-time continuous flare efficiency monitoring. Modern computational simulation science methods allow the use of tens of thousands of computer cores (processors) to work together on a single application to produce unprecedented simulation resolution. For example, historically, the largest challenge in flare simulations has been to model the turbulent mixing and reaction process. However, by using current computer cluster technology, the turbulence and reaction process can be computed directly without the need for turbulence models. The purpose of this paper is to describe a technology for obtaining quantitative measurements of combustion efficiency from operating industrial and field flares by bringing together the latest advancements in diagnostic measurements and computer simulation technologies.

version: Friday, August 12, 2011 T E C H N O L O G Y   F O R   M E A S U R I N G C O M B U S T I O N   E F F I C I E N C Y O F   I N D U S T R I A L   &   F I E L D   F L A R E S A ( i n t e g r a t i n g   m e a s u r e m e n t s   a n d   s i m u l a t i o n s ) Philip  J.  Smith The  Institute  for  Clean  &  Secure  Energy The  University  of  Utah ABSTRACT Flare research over the past decade has increasingly illustrated that there is likely no one simple operational parameter (or even a few parameters) that will characterize the combustion behavior of flare flames.i Simple correlations are unobtainable because of the complexity of the nonlinear mixing, reaction, and heat transfer present in operating flare flames. This complexity motivates the need to accurately measure combustion efficiency from operating flares so that the effect of different designs and operations can be quantified. However, this same complexity, makes such measurements difficult. To date there is no technology that has been been validated at providing quantitative combustion efficiency measurements from operating, open-air, flare flames. While it is difficult to directly measure combustion efficiency due to the heterogeneity in composition and velocity, high performance computing (HPC) can be coupled with diagnostic measurement methods to provide real-time continuous flare efficiency monitoring. Modern computational simulation science methods allow the use of tens of thousands of computer cores (processors) to work together on a single application to produce unprecedented simulation resolution. For example, historically, the largest challenge in flare simulations has been to model the turbulent mixing and reaction process. However, by using current computer cluster technology, the turbulence and reaction process can be computed directly without the need for turbulence models. The purpose of this paper is to describe a technology for obtaining quantitative measurements of combustion efficiency from operating industrial and field flares by bringing together the latest advancements in diagnostic measurements and computer simulation technologies. BACKGROUND Flaring waste streams of combustable material is an important control practice for destroying unwanted hydrocarbons and other combustable material. All studies performed to date have reached the conclusion that the proper operation of such flares produces near complete conversion of the hydrocarbons to combustion products. Thus, the focus of the US EPA in regulating flares has been to define proper operational procedures, assuming that if operated properly, the combustion efficiency would be near 100%. This i The American Flame Research Committee (AFRC) of the International Flame Foundation (IFRF) has carried on a focused series of flare research forums. The papers from these forums are available online at http://www.afrc.net/index.jsp?page=1;&l2nid=6; M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y proper operation has been defined in USEPA's 40CFR60.18 "General Control Device Requirements." ii, iii However, defining proper operating procedures to ensure high combustion efficiency for the wide range of conditions that exist for actual flare operations is surely not only difficult (maybe impossible) but unwise iv. Since the regulatory goal is to achieve the highest possible combustion efficiency, it would seem wise to encourage creative technical solutions by promoting changes in both design and operations to achieve ever increasing combustion efficiency under an ever widening range of operating conditions. Flare research over the past decade has increasingly illustrated that there is likely no one simple operational parameter (or even a few parameters) that will characterize the combustion behavior of flare flames.v Simple correlations are unobtainable because of the comFigure 1: Heterogeneity in flare combustion effiplexity of the nonlinear mixing, ciency as shown in these volume rendered images reaction, and heat transfer present of a large eddy simulation (LES) of a field flare in operating flare flames. This operating under two different crosswind condicomplexity motivates the need to tions. Hot colors (red) represent high combustion accurately measure combustion inefficiency. efficiency from operating flares so that the effect of different designs and operations can be quantified. However, this same complexity, makes such measurements difficult. To date there is no technology that has been been validated at providing quantitative combustion efficiency measurements from operating, open-air, flare flames. This paper describes an approach to obtaining quantitative measurements of combustion efficiency from operating industrial and field flares by bringing together the latest advancements in diagnostic measurements and computer simulation technologies. ii 51 Federal Register (FR) 2701, Jan. 21, 1986, as amended at 63 FR 24444, May 4, 1998; 65 FR 61752, Oct. 17, 2000; 73 FR 78209, Dec. 22, 2008 iii made available online by the National Archives and Records Administration at http://ecfr.gpoaccess.gov/ iv For a further discussion of current EPA regulations see http://home.earthlink.net/~jim.seebold/id20.html . v The American Flame Research Committee (AFRC) of the International Flame Foundation (IFRF) has carried on a focused series of flare research forums. The papers from these forums are available online at http://www.afrc.net/index.jsp?page=1;&l2nid=6; page 2 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y H E T E R O G E N E I T Y   O F   F L A R E   C O M B U S T I O N   E F F I C I E N C Y One of the reasons why measurement of combustion efficiency from flare flames is so elusive is because of the heterogeneity of the combustion efficiency in the flare. The complex turbulent mixing in buoyancy-driven plumes characteristic of flare flames produces local temporal and spatial combustion (and non-ignition) events, as well as local extinction events. This characteristic of flare flames has been known for some time. Johnson et al.vi called this process ‘fuel stripping' and quantified its effect on small tests in wind tunnels by speciated mass spectroscopic / gas chromatographic analysis of the unburned sampled hydrocarbons gases. The ratio of compounds in the fuel (i.e. for natural gas, the ratio of CH4, C2H6, and C3H8) quite closely matched the ratios in the unburned hydrocarbons from the flare. Also, by using a fast flame ionization detector the authors were able to track "bursts" of hydrocarbons around the flame. Johnson and Kostiakvii performed laser sheet visualization experiments in a wind tunnel in which they show fuel being stripped from the flame. The present authors have shown this heterogeneity in combustion efficiency with large eddy computer simulations viii. Example volume rendered images from these computations illustrating this heterogeneity for a field flare under two different crosswind conditions are shown in Figure 1. The consequences of these local heterogeneities on flare efficiency is that it is necessary to perform a surface integral over the control surface enclosing the plume volume to obtain quantitative combustion efficiencies. For a continuous feed rate of fuel into the flare the combustion efficiency of the fuel is defined as (1) where ηfuel  is the combustion efficiency based on a mass balance of the fuel entering and leaving the control volume (sometimes called the destruction efficiency or fuel efficiency), where ρC(fuel)  is the local mass concentration of all fuel species, v is the local velocity vector, and n is the unit normal vector. Alternatively, the combustion efficiency can be obtained by measuring the concentration of combustion products: (2) and is often called the conversion efficiency. Other mass balances are also possible to obtain the combustion efficiency, but in all cases, for an inhomogeneous mixture, a surface vi Johnson, Wilson & Kostiuk, "Fuel Stripping Mechanism for Wake Stabilized Jet Diffusion Flames in Crossflow," Combust. Sci. andTech., 169: 155-174, 2001. vii Johnson & Kostiuk, "Visualization of the Fuel Stripping Mechanism for Wake-Stabilized Diffusion Flames in a Crossflow," Published in IUTAM Symposium on Turbulent Mixing and Combustion, A. Pollard and S. Candel, editors, pp. 295-303 (2002). viii Smith, Thornock, Hradisky, Isaac & Jatale, "What can we learn about flare combustion efficiency from large eddy simulations?," International Flame Research Foundation (IFRF), TOTeM36, September 2010. page 3 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y integral is required and both the velocity and concentration of selected species are required simultaneously to obtain mass fluxes in order to close the mass balance of the species of interest to obtain quantitative combustion efficiency. The heterogeneities in the flare occur in both the local and temporally evolving composition field and in the velocity field. Both heterogeneities are persistent in the atmospheric boundary layer and become self-preserving over extremely large distances.ix The need to perform integration over the flare flame to obtain realistic combustion efficiencies was noted in the US EPA's 1983-86 investigation of the combustion efficiency of industrial flaresx. The archival data produced by the agency-sanctioned extractivesampling protocols demonstrated that to obtain the so-called "global" combustion efficiency, which is to say a combustion efficiency that is accurately reflective of a flare's overall emission control performance, "requires detailed integration over the flare plume both radially and axially." Over the last decade, pilot-scale flare tips have been tested in wind-tunnels where all of the products of combustion of the flare plume can be accurately sampled by extraction in a well-mixed (homogeneous) region of the stack. While this provides an alternative means of accounting safely, accurately and reliably for the inhomogeneous distribution of combustion efficiency in flare plumes, it will not work for operating flares. This is paramount to burning the waste stream in a furnace, obviating the need for a flare. M E A S U R E M E N T   M E T H O D S These inhomogeneities in full-scale flare flames make direct measurement of combustion efficiency difficult. Modern measurement methods have been deployed for identifying (and to some extent quantifying) some specific species from which combustion efficiency can be inferred. Current methods being explored for performing these types of measurements include optical methods like: • Fourier transformed infra-red (FTIR) spectroscopy and passive FTIR (PFTIR)xi, • Differential Absorption Lidar (DIAL)xii, • Tunable Diode Laser Absorption Spectroscopy (TDLAS), ix R. Sangras, Z. Dai, and G. M. Faeth, "Mixing Structure of Plane Self-Preserving Buoyant Turbulent Plumes," J. Heat Transfer 120, 1033 (1998), DOI:10.1115/1.2825887 x "Evaluation of the Efficiency of Industrial Flares," United States Environmental Protection Agency, Office of Air Quality Planning and Standards, EPA-600/2-84-095, May 1984; EPA-600/2-85-106, September 1985; EPA-600/2-85-106, September 1985 xi "Performance Test of a Steam-Assisted Elevated Flare With Passive FTIR," FINAL REPORT, Prepared by Clean Air Engineering, Inc. Project No: 10810 for Marathon Petroleum Company, LLC Texas Refining Division, May 2010, Marathon Petroleum Company, LLC Texas Refining Division 502 10th Street South Texas City, Texas 77590. xii Chambers, Strosher, Wootton, Moncrieff & McCready, "DIAL Measurements of Fugitive Emissions from Natural Gas Plants and the Comparison with Emission Factor Estimates," US EPA 15th International Emission Inventory Conference, New Orleans, May 15 - 18, 2006. Available online http://www.epa.gov/ttn/chief/conference/ei15/session14/chambers.pdf page 4 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y • Image Multi-Spectral Sensing (IMSS, also called Sherlock)xiii, that uses diffractive optics and image processing algorithms to detect spectral information about objects in the scene of the camera. These optical methods have two weaknesses when they are used alone to produce combustion efficiency estimates.: 1. They collect information along a line of sight over a small solid angle; thus, they collect ‘line of site' or small angle volume information not surface information. 2. They collect concentration information, they do not collect velocity data; thus, mass fluxes are not available for the combustion efficiency computation of equations 1 & 2. Combustion efficiencies obtained from these measurement tools typically assume complete homogeneity in composition and velocity at the measurement location. There are methods being developed for emission measurements that recognize this need for flux data. For example one such method is the optical remote sensing with radial plume mapping (ORS-RPM)xiv. This method uses an optical emission detector such as open-path Fourier transform infrared spectroscopy (FTIR), ultraviolet differential absorption spectroscopy (UV-DOAS), or open-path tunable diode laser absorption spectroscopy (OP-TDLAS); coupled with radial plume mapping software that processes pathintegrated emission concentration data and meteorological data to yield an estimate of uncontrolled emissions. This method has been developed for landfill emissions. Characterizing the heterogeneous meteorological conditions created by the flare plume itself has not yet been attempted. C O M B I N I N G   M E A S U R E M E N T   W I T H   S I M U L A T I O N While it is difficult to directly measure combustion efficiency due to the heterogeneity in composition and velocity, high performance computing (HPC) can be coupled with diagnostic measurement methods to provide real-time continuous flare efficiency monitoring. Modern computational simulation science methods allow the use of tens of thousands of computer cores (processors) to work together on a single application to produce unprecedented simulation resolution. For example, historically, the largest challenge in flare simulations has been to model the turbulent mixing and reaction process. However, by using current computer cluster technology, the turbulence and reaction process can be computed directly without the need for turbulence models. The flare simulations shown in Figure 1 were performed on 1600 cores utilizing simulation software developed over the past ten years by a team of computer scientist and engineers at the University of Utah under DOE NNSA Advanced Simulation and Computing (ASC) sponsorship xv. As illustrated in Figure 1, combustion simulations provide the opportunity to dynamically compute local heterogeneity to obtain quantitative combustion efficiency. However, the accuracy of these predictive simulations is directly affected by the uncertainty in the operating conditions of the flare. This uncertainty can only be quantified and rexiii Michele Hinnrichs, "Remote Sensing and Analysis of Unburned Gases from Stacks and Flares Using Imaging Spectroscopy," International Flame Research Foundation, International Members Conference, Boston, June 2009. xiv Evaluation of Fugitive Emissions Using Ground-Based Optical Remote Sensing Technology, Office of Research and Development, U. S. Environmental Protection Agency, Washington, DC, (EPA/600/R-07/032), February 2007. xv http://www.sandia.gov/NNSA/ASC/univ/univ.html page 5 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y duced by coupling modern measurement and monitoring technologies with advanced computational simulation tools to root the prediction in actual measurements and operating conditions. To accomplish this task the simulation must be used to solve the inverse problem instead of the forward problem. For example, heat flux is routinely measured from operating flares. One would like to use the HPC simulations to answer the question: if this is the measured heat flux from this flare what is the combustion efficiency? Using a predictive simulation tool to solve this inverse problem has historically been a difficult task. This dynamic coupled analysis approach to learning about a systems and moving it forward has been called a dynamic data driven application system (DDDAS)xvi. It draws on the ability to incorporate additional measurement data into a simulation application and/or use simulation data to steer the application dynamically. The National Science Foundation has argued that DDDAS has the potential to transform the way science and engineering are done, and induce a major impact in the way many functions in our society are conducted (i.e. weather forecasting, oil exploration, etc…)xvii. Maximum flare combustion efficiency could be achieved dynamically by using the ‘point and shoot' optical methods described above (as well as other more traditional measurements such as heat flux gauges, flow meters, thermocouples, and the like) together with advanced simulation tools such as those illustrated by the computations of Figure 1 in a dynamic data driven flare combustion efficiency application system. Such a system can be accomplished by using Bayesian inference methods to solve the inverse problem with uncertainty quantificationxviii. For example, the DDDAS algorithm for flare efficiency might proceed as follows: 1. Obtain field information for the expected operating range of the flare to be measured. This operating range must include all operating parameters that would affect combustion efficiency even if not directly measurable, i.e. expected operating ranges for fuel compositions, flow rates, crosswind conditions, steam assist rates, air assist rates, etc. 2. Prepare a design space for conditions that cover the range of operating conditions identified in (1) and perform HPC simulations that cover this design space. 3. Develop a surrogate model that spans the design space of (2). The surrogate must be able to be executed in real time for steps (6) and (7) below. 4. From the operating flare system collect current data for the subset of operating conditions of (1) that can be measured. 5. Use the ‘point and shoot' optical methods to dynamically monitor key species in the line of site measurement and/or use other measurement methods to collect data on the output emissions of the operating flare. xvi Darema, "Dynamic Data Driven Applications Systems: A New Paradigm for Application Simulations and Measurements.". International Conference on Computational Science.: pp. 662-669 (2004). xvii The term DDDAS was formalized by Frederica Darema around the time of a National Science Foundation (NSF) workshop in March 2000. xviii an example framework for performing a Bayesian analysis of measurements and simulations is described by M.J. Bayarri, J.O. Berger, R. Paulo, J. Sacks (National Institute of Statistical Sciences), J.A. Cafeo, J. Cavendish, C.H. Lin, J. Tu (General Motors), "A Framework for Validation of Computer Models," Technical Report Number 162, April 2005, National Institute of Statistical Sciences, 19 T. W. Alexander Drive, PO Box 14006, Research Triangle Park, NC 27709-4006, http://www.niss.org page 6 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y 6. Given the measured operating conditions of (4) extract from the surrogate model of (3) the predicted results of the output quantities measured in (5). 7. Using a consistency analysis find the operating parameters not measured in (4) that are consistent with the predictions of (6) and the measurements of (5). Use all accumulated data for this consistency analysis. 8. If a consistent set is obtained, integrate the predictions of (7) for the combustion efficiency. This is the desired output, the ‘measured' combustion efficiency with uncertainty bounds that span the range of operating parameters consistent with both the simulations and the measurements. 9. If a consistent set is not obtained, alter the parameter set and range of parameters, return to step (1) and repeat. 10. Using the surrogate model of (3) and the resulting combustion efficiency of (8) steer the controllable operational parameters to settings that would produce an improved combustion efficiency if possible. Return to step 4 and repeat. This DDDAS algorithm for ‘measuring' flare efficiency has several features: • It recognizes the heterogeneity of the composition and velocity field and includes these effects in the combustion efficiency. • It produces combustion efficiencies with quantified uncertainty bounds that are consistent with all data and all simulations. (A wide range of methods for performing this consistency analysisxix and uncertainty quantificationxx are available and growing under active research.xxi ) • It uses all available measurements to maximum advantage without unnecessary assumption (like homogeneity). • It uses specific measurements to continuously validate the simulations which are then used to obtain combustion efficiencies. • It uses Bayesian inference to perform the inverse problem so that a measurement of one measurable quantity can be used to produce the associated value of a previously ‘unmeasurable' quantity. • It reduces uncertainty and obtains posterior information that improves on the prior measurement and simulation uncertainty by requiring consistency between all data (measured and predicted). • It uses HPC simulations while performing in real time by using a surrogate model that is updated by the HPC simulations ‘off-line' on an ‘as needed' basis. • It uses simulation and measurements together to produce a ‘measured' combustion efficiency. • It steers the operation of the flare to an optimal operating location under actual operating conditions that maximize combustion efficiency for those conditions. xix Feeley, Seiler, Packard & Frenklach, "Consistency of a Reaction Dataset," Journal of Physical Chemistry A, 108, 9573-9583 (2004). xx Russi, Packard, Frenklach, "Uncertainty Quantification: Making Predictions of Complex Reaction Systems Reliable," Chem. Phys. Lett. (2010), doi:10.1016/j.cplett.2010.09.009 xxi see DOE NNSA Predictive Science Academic Alliance Program (PSAAP) as an example of one national research program on Uncertainty Quantification: http://www.sandia.gov/NNSA/ASC/univ/psaap.html page 7 of 8 M e a s u r i n g   F l a r e   C o m b u s t i o n   E f f i c i e n c y SUMMARY Currently flare systems are operated by ‘a priori' governmental specification of operating conditions for the flarexxii. The assumption being that flares that are operated in accordance with these specifications will have very high destruction efficiencies. However, defining proper operating procedures to ensure high combustion efficiency for the nearly infinite set of conditions that exist for actual flare operations is surely not only difficult (maybe impossible) but unwise. One would more wisely prefer to dynamically adjust the operating conditions in order to maximize a measured combustion efficiency. This later approach has historically not been used because to date combustion efficiency from flares has not been measurable. It is possible to develop a dynamic data driven application system (DDAS) to ‘measure' combustion efficiency from flares. This system requires the integration of: • modern high performance computer (HPC) simulations, • new formal uncertainty quantification methodologies, and • measurements of other measurable variables and operating conditions. This DDAS would provide: • a continuous, real-time combustion efficiency ‘measurement', • and could also steer flare operation to continuously maximize combustion efficiency under ever changing operating constraints. xxii in accordance with United States Federal Law 40 CFR 60.18(b) through (d) and 40 CFR 63.11(b) page 8 of 8