Evaluating the NOx Performance of a Steam Generator for Heavy Oil Production: Impact of Combustion System Design

Paper from the AFRC 2014 conference titled Evaluating the NOx Performance of a Steam Generator for Heavy Oil Production: Impact of Combustion System Design by J.N. Spinti.

Chevron operates approximately 150 steam generators for heavy oil production in Californiaʼs San Joaquin Valley. To meet increasingly stringent NOx regulations, these steam generators have been retrofitted with the Fives North American GLE combustion system. To better understand the combustion environment where NOx is being formed and to devise new firing schemes that further reduce NOx emissions, researchers at the Institute for Clean and Secure Energy (ICSE) at the University of Utah have teamed with personnel from Chevron U.S.A. Inc., and Fives North American Combustion to apply high-performance computing Large Eddy Simulation (LES) tools to a Chevron steam generator. The high fidelity simulations needed to adequately resolve the length and time scales critical to NOx formation were performed in two parts. In part one, the commercial software package STAR-CCM+ was used to resolve the flow field through the complex geometry of the burner. In part two, this computed flow field at the burner's outlet plane provided the input to a simulation of the front half of the steam generator using ARCHES, an LES code developed by ICSE researchers. ARCHES simulations, with a computational domain of 50 million cells, required 2800 processors for one week to reach a statistical steady state. Though not a formal validation/uncertainty quantification study, simulation data is been compared with field data collected on a GLE-equipped steam generator at various axial and radial locations. The primary quantity of interest is NOx concentration, but comparisons are also made with O2 concen-trations and gas temperature. Simulation and experimental data exhibit similar profiles for both NOx and O2 concentrations. The simulation temperatures are consistently higher than experimentally-measured temperatures, but the temperature sensitivity of the NOx formation rate indicates that the higher temperatures are more probable. Two combustion system configurations were compared in this study: (1) a baseline GLE system, and (2) a GLE system with the addition of Flue Gas Recirculation (FGR) in a Large Scale Recirculation (LSR) arrangement. These two configurations represent contrasting combustion regimes as seen in volume-rendered images and movies created from the ARCHES simulations. With the GLE burner, the injection of secondary fuel creates high-temperature flame fronts where fuel and oxidant mix and react. With the addition of FGR into the secondary fuel, a flameless burning mode results in which maximum flame temperatures are reduced. The implications of these burner designs on NOx emissions will be discussed.

Evaluating the NOx Performance of a Steam Generator for Heavy Oil Production: Impact of Combustion System Design University of Utah, Institute for Clean and Secure Energy J.N. Thornock⇤, J.P. Spinti, M. Hradisky, P.J. Smith Chevron U.S.A. Inc B. Coleman, N. Brancaccio, S. Storslett Fives North American Combustion Inc J. Nowakowski, T. Robertson July 17, 2014 Abstract Chevron operates approximately 150 steam generators for heavy oil production in California's San Joaquin Valley. To meet increasingly strin-gent NOx regulations, these steam generators have been retrofitted with the Fives North American MagnaFlame GLE combustion system (Mag-naFlame GLETM). To better understand the combustion environment where NOx is being formed and to devise new firing schemes that further reduce NOx emissions, researchers at the Institute for Clean and Secure Energy (ICSE) at the University of Utah have teamed with personnel from Chevron U.S.A. Inc., and Fives North American Combustion to ap-ply high-performance computing Large Eddy Simulation (LES) tools to a Chevron steam generator. The high fidelity simulations needed to adequately resolve the length and time scales critical to NOx formation were performed in two parts. In part one, the commercial software package STAR-CCM+ was used to resolve the flow field through the complex geometry of the burner. In part two, this computed flow field at the burner's outlet plane provided the input to a simulation of the front half of the steam generator using Arches, an LES code developed by ICSE researchers. Arches simulations, with a computational domain of 50 million cells, required 2800 processors for one week to reach a statistical steady state. Though not a formal validation/uncertainty quantification study, sim-ulation data has been compared with field data collected on a MagnaFlame ⇤Corresponding Author ([e-mail]) 1 GLETM-equipped steam generator at various axial and radial locations. The primary quantity of interest is NOx concentration, but comparisons are also made with O2 concentrations and gas temperature. Simulation and experimental data exhibit similar profiles for both NOx and O2 con-centrations. The simulation temperatures are consistently higher than experimentally-measured temperatures, but the temperature sensitivity of the NOx formation rate indicates that the higher temperatures are more probable. Two combustion system configurations were compared in this study: (1) a baseline MagnaFlame GLETMcombustion system, and (2) a Mag-naFlame GLETMcombustion system with the addition of Flue Gas Recir-culation (FGR) in a Large Scale Recirculation (LSR) arrangement. These two configurations represent contrasting combustion regimes as seen in volume-rendered images and movies created from the Arches simulations. With the MagnaFlame GLETM combustion system, the injection of sec-ondary fuel creates high-temperature flame fronts where fuel and oxidant mix and react. With the addition of FGR into the secondary fuel, a flameless burning mode results in which maximum flame temperatures are reduced. The implications of these burner designs on NOx emissions will be discussed. Keywords: Large Eddy Simulation, Multi-scale, Multi-physics, Combustion, Flame-less Combustion, Once-through steam generators, Oil field heaters, NOx 1 Introduction The objective of this work, a three-way partnership among the University of Utah, Chevron U.S.A. Inc ("Chevron"), and Fives North American Combustion Inc, is to demonstrate the coupling of science-based simulations with industrial demonstration-scale experiments to provide predictions of NOx formation in an air-fired combustion system at very low concentrations. Specifically, this work focuses on applying high-fidelity Large Eddy Simulation (LES) to aid assess-ment of current technology and in retrofit of a once-through steam generator (OTSG) used for enhanced oil recovery operations (EOR). The approach taken 1) validates the LES model against measured field-data of total NOx stack emis-sions, and radial profiles of NOx, O2, and temperature profiles from a full-scale oil-field heater, and 2) provides a tool for assessing a retrofit option for total NOx emission reduction. Impending NOx regulations for San Joaquin Valley (SJV) OTSGs motivate the need for rapid assessment of retrofit options to lower overall NOx emissions. We aim to demonstrate that HPC simulation tools are crucial in assessing and deploying new combustion systems to meet evolving emissions regulations. In effort to meet NOx and Ozone ambient air quality standards, the EPA, the California Air Resources Board, and the SJV Air Pollution Control District have continuously adopted more stringent NOx regulations [4] for a range of combustion sources, including OTSGs, which would see the standard lowered 2 from <15ppm to <5ppm1. In general, an EOR operation would employ multiple OTSGs across the oil field such as the SJV's Kern River oil field. Thus, the new regulations pose an economic challenge to oil producers in the valley as complete replacement of all OTSGs is cost-prohibitive and generally not practical. This challenge poses a unique opportunity for simulation to provide rapid evaluation of OTSG retrofit options and, using validation/uncertainty quantification (V/UQ) meth-ods, to inform the uncertainty of the measurements. We employ HPC LES tools, Arches2[7] and STAR-CCM+3, to perform simulations of an existing OTSG with experimental data available for comparison and to evaluate retrofit op-tions. Coupling the simulation and experimental data through more advanced V/UQ methods is left for future work. The simulation of full-scale combustion systems with average concentrations of emission in the single digit ppm range is challenging. To accurately predict NOx concentrations at these levels, significant time and length scales must be explicitly represented using first-principles with little reliance on heuristic or empirical modeling techniques. For the problem at hand, representing enough scales requires significant computing power and scalable algorithms. For exam-ple, important geometric scales in the OTSG range from the ports of the fuel in-jectors (O[1mm]) to the length scales of the radiant section of the boiler (O[1m]). When considering other physical phenomena, such as energy dissipation at the smallest fluid length scale as well as the scalar mixing and combustion, the scale disparity increases. In this scenario, LES with HPC provides a clear advantage over traditional methods such as Reynolds averaging because, as resolution is added, the reliance on the sub-grid modeling is decreased as more time/length scales are resolved directly. Thus the LES may approach direct representation of rate-limiting physics, decreasing the approximations required for representing sub-grid phenomena. This paper presents a loosely coupled approach for simulating the wide range of length and time scales that impact NOx emissions in the OTSG. The fine geometric detail of the burner configuration is represented using the commer-cial software STAR-CCM+. STAR-CCM+ uses an unstructured computational mesh to capture the complex burner geometry and the near-burner flow field. Data from the STAR-CCM+ simulation in the near burner region is filtered and coupled to the far-field, structured mesh Arches LES simulation through a static boundary condition. Arches is built within the Uintah computational frame-work, which provides efficient and scalable data handling for structured meshes and message passing across distributed memory architectures[1, 5]. Arches within Uintah has been shown to scale well to 256K cores[6]. For this work, we employ a combustion model within Arches built on the Rate-Constrained Controlled Equilibrium (RCCE) [3] approach. In short, RCCE segregates slow, grid resolved chemical mechanisms from fast, sub-grid mechanisms to span the 1S.J.V.A.P.C. District. (2008). Rule 4306, boilers, steam generators, and process heaters-phase 3. http://www.valleyair.org/rules/1ruleslist.htm#reg1 2see http://www.icse.utah.edu/sim_tools 3see http://www.cd-adapco.com/products/star-ccm® 3 range of chemical time-scales. The slow time-scales are directly resolved on the LES mesh while the remaining are treated as subgrid. In addition, because of the relatively slow time-scales for NOx formation, we resolve the formation of NOx directly on the mesh. In this manner, several orders of magnitude of scales are represented across the full simulation. 2 Steam generator description The OTSG, shown in Figure 1, is a relatively simple gas-fired combustion unit whose objective is to provide low-quality, high-pressure wet steam (70% vapor, 30% liquid) for EOR operations; one OTSG might service 50-75 wells. Most of the OTSGs operated by Chevron were installed in the 1960s and have been upgraded/retrofitted to accommodate different types of fuels and evolving regu-lations. The typical OTSG has one or two inlet water streams that traverse the convective and radiant sections of the unit. The water is heated to near boiling in the convective section before entering the serpentine length of tubing that lines the walls of the radiant section. In the radiant section, gas combustion from a burner provides the thermal load, operating in a range of about 62.5 MMBtu/h at high fire, and 3:1 turndown (20.8 MMBtu/h) at low fire. Most units have high thermal efficiency (~87% thermal efficiency) while meeting the current NOx standards (<15ppm, dry). 2.1 Current/Base Steam Generator The radiant section of the OTSG considered here has an internal diameter of approximately 3.175 meters or m (10.4 feet) and is approximately 12.2 m (40 feet) in length. The OTSG was fired with natural gas through a low-NOx Mag-naFlame GLETM burner, a schematic of which is shown in Figure 2. Premixed natural gas and air are injected into the primary reaction zone while additional natural gas is introduced into the OTSG through secondary gas injectors. A small quantity of natural gas is also injected through the pipe at the burner center (labeled "Center Gas Inlet" in Figure 2). 2.2 LSR Design In order to attain ultra-low NOx levels (<5ppm), a retrofit option is considered for the existing OTSG described in Section 2.1. This scenario removes the ex-isting secondary gas inlets and replaces them with a premixed stream composed of recycled flue gas and natural gas. The total fuel input remains constant from the base case. The momentum of the secondary is reduced considerably by introducing the flow through a simple open pipe, the opening of which has a much greater area than the secondaries from the base case. The lean primary combustion zone remains the same as the base scenario. This configuration is labeled the LSR design. 4 Figure 1: Steam generator with the burner in the foreground. Figure 2: MagnaFlame GLETMburner used in steam generator tests. 5 2.3 Description of experimental data Experimental data were collected over the course of two days from the steam generator. Data available for the overall process include the ambient air flow rate and temperature and the fuel flow rate, temperature, and composition. Detailed data collection in the radiant section occurred at four longitudinal locations (0.46 m, 1.37 m, 2.29 m, and 5.94 m) with radial measurements taken every 7.6 cm (3 inches) from the wall to the center of the steam generator. At each location, extractive sampling was used to measure O2, CO, and NO, and NO2 concentrations as well as temperature and pressure. The extracted sample was analyzed with the Testo 350 Emission Analyzer. The stated accuracy of the analyzer is ± 2 ppm for NO and ± 5 ppm for NO2. Other sources of sampling error were not analyzed. Infrared thermographic images were also taken of the entire steam generator exterior and of the internals of the radiant section. 3 Simulation and modeling approaches The problem of interest here spans several scales. Geometrically speaking, the finest scales fuel injectors (~1mm) located within the fuel-oxidizer mixing tubes. The largest being the width of the radiant section (~3m). Given these ranges of scales, two simulation tools are used: the unstructured STAR-CCM+ LES solver, and the structured Arches LES solver. Given its ability to represent the geometry accurately, the STAR-CCM+ code is used to represent the fine geometric detail in the near-burner and secondary fuel injection regions. Arches, on the other hand, is used to simulate the primary combustion chamber and the radiant section of the boiler. The details of each approach and the coupling are described next. 3.1 Near burner/injector STAR-CCM+ simulations There are two principle components of the MagnaFlame GLETM combustion system considered for the STAR-CCM+ computations; the mixer tube from which the premixed fuel/air stream enters the primary reaction zone and the secondary gas injector from which natural gas is injected into the radiant section downstream of the primary reaction zone. The mixer tube simulation required 30 million cells and employed the Wall-Adaptive Local-Eddy viscosity (WALE) LES model. The computational mesh of the mixer tube is in shown in Figure 3a. Because the simulation included flow of gas through the fuel distributor and through all nozzles, the simulation time step was on the order of 1E-5 seconds. The simulation was run on 600 cores for four days to reach statistical steady state (SSS). After reaching SSS, we began collecting samples of all velocity components, the density distribution, the enthalpy, and the fuel/air mass ratio at the exit plane of the mixer tube for an additional three days. In total, this simulation required about 100K CPU hours to capture a few seconds of mixing occurring inside the mixer tube. The output was then time-averaged over the period of SSS. 6 For the secondary fuel injector, we also used the WALE LES model with a time step of the same order of magnitude as for the mixer tube simulation. This simulation required approximately three million cells (see Figure 3b) and took two days of computer time on 600 cores to reach SSS and to collect time-averaged samples for all velocity components, density, enthalpy, and the fuel/air mass ratio. In total, this simulation required on the order of 30K CPU hours to obtain the low NOx injector profiles. As with the mixer tube, this data was filtered as described next. A spatial filter was applied to the time-averaged output from the STAR-CCM+ to coarsen it to the resolution of the Arches structured mesh. The spatial filter was applied at a plane just downstream of the mixer tube or low NOx injector tip and involved averaging the fine-scale data over the filter width of the Arches computational mesh (7.6 mm). Figure 4a shows the time-averaged output from the mixer tube simulation, including the velocity components, the density (e.g., mass flux), the fuel/air mass ratio (e.g., mixture fraction and eta), and the enthalpy. Figure 4b displays the same set of data after filtering to the Arches resolution. The same process was applied to the time-averaged output from the STAR-CCM+ simulation of the secondary fuel injector. The time-averaged output is shown in Figure 5a with spatially-filtered data displayed in Figure 5b. While some fine scale information is lost during the filtering operations, the images in both Figure 4a and 5b appear to capture the overall distribution of the various velocity components and of the other variables. 3.2 Far-field large eddy simulations Simulations of the far-field domain (i.e., beyond the exit of the primary fuel/oxidizer injectors for the primary combustion chamber and beyond the end of the sec-ondary fuel injectors in the radiant section) were simulated using the Arches LES tool. The Arches LES tool was developed over a 15+ year partnership with the Department of Energy and the University of Utah. The Arches code solves the finite-volume formulation of Favre-filtered mass, momentum and en-ergy equations on a structured Cartesian mesh. Velocities are staggered midway between cell-centers. Mass continuity is enforced via a pressure-projection step. For momentum, convection and diffusion are discretized using second order, central schemes. The scalar balances use a superbee flux limiter for convection and central schemes for diffusive terms. Momentum closure is accomplished with the dynamic Smagorinsky model. Scalar turbulent closure assumes con-stant turbulent Prandtl and Schmidt numbers for enthalpy and mass transport respectively. All transport equations are integrated explicitly in time with a strong-stability preserving second order Runge-Kutta scheme. For the energy balance, the radiative intensity equation is solved using a discrete-ordinate ap-proach. Convective and conductive heat transfer modes are neglected. The combustion state-space is defined as = (⇢, T, "1,"2, ...,"n1) (1) where ⇢ is the fluid density, T is the fluid temperature, and "i represents the 7 (a) (b) Figure 3: STAR-CCM+ representation for the (a) primary mixer tube and (b) secondary fuel injector in MagnaFlame GLETM combustion system. 8 (a) (b) Figure 4: Plots for the exit plane of the primary mixing tube at the hand-off boundary condition. Plot (a) shows the time-filtered STAR-CCM+ results while plot (b) shows the spatially filtered Arches representation. 9 (a) (b) Figure 5: Plots for the exit plane of the secondary mixing tube at the hand-off boundary condition. Plot (a) shows the time-filtered STAR-CCM+ results while plot (b) shows the spatially filtered Arches representation. 10 one to (n 1) independent chemical species. Here, is represented using a rate-controlled constrained-equilibrium (RCCE) technique[3]. RCCE assumes that the equilibrium state is controlled by a one or more rate-controlling steps, which typically is a representative mechanism of slower rate processes. That is, the instantaneous state-space, which is at equilibrium as determined by a minimization of Gibbs free energy, is constrained by the state of the resolved rate controlling steps. In the LES context, these rate controlling steps are grid resolved because they occur at time-scales commensurate with the LES time-scales. The equilibrium reactions, therefore, occur instantaneously and at the sub-grid scales. In this manner, the timescales of the combustion chemistry are partitioned into slow (resolved) and fast (unresolved) processes. To further constrain our model, we make use of flammability limits as de-scribed by Zabetakis [8]. That is, homogeneous fuel-oxidizer mixtures can only propagate flames from an ignition source within a somewhat narrow range of concentrations. The presence of an inert can further alter the flammability region as a function of the inert concentration. The resulting flammability di-agrams as reported are often described as the Zabetakis "nose-plots". Here we incorporate the flammability diagrams by allowing the slow reaction to pro-ceed only with mixtures falling within the fuel-specific flammability envelope. Mixtures outside of these limits are not allowed to proceed. This constraint is determined on a computational cell-by-cell basis and can be viewed as a mixing constraint on the combustion model. While we anticipate that the LES will resolve a larger portion of the scalar energy spectrum, we do recognize that the use of the flammability diagram assumes a homogenous mixture and thus an assumption of a perfectly mixed mixture at the sub-grid scale. The NOx formation, like the RCCE model just described, also exploits the captured time-scales of LES. Here, we assume prompt and fuel NOx sources are insignificant. Because the formation and destruction rates of thermal NOx are relatively slow and given that they can be considered independent of the combustion process (reference), we use grid-resolved representation of the NOx mass fraction. This mass fraction is transported on the LES mesh and is tied to the local chemistry state-space via a source term representing an overall Zeldovich mechanism [2], d[NOx] dt = A T1/2 exp(E/RT)[N2][O2]1/2. (2) Here T represents the local gas temperature, N2 and O2 are the local species concentrations, A is the pre-exponential factor, and E is the activation energy. In this work, we assumed A = 6.0E17 and E/R = 69090. Note that this ap-proach allows for only a one-way coupling of the NOx formation from the RCCE model. Given the low levels of NOx, this approximation seems reasonable. Results are presented in Section 4 from two coupled simulations; one rep-resenting the existing base OTSG operation and another representing the pro-posed LSR design. The Arches domain begins from the entrance of the primary fuel/oxidizer mixing tubes into the primary combustion zone and extends down the radiant section of the boiler to about 5m (16.5ft). This length does not 11 represent the entire length of the steam generator, but does cover the first three measurement traversals in the radial direction and the NOx formation zones. For the base case, the boundaries for the primary and secondary injection points were defined using the time-averaged information from the STAR-CCM+ simu-lation. The wall boundary conditions for the primary combustion chamber and the burner face were modeled with an adiabatic condition. The tubes of the radiant section were not explicitly represented. The temperature of the walls in the radiant section were assumed constant and equal to T = 854oF. The entire domain was initialized with a fully-mixed, burnt condition. In order to ensure that the primary injectors ignited, the RCCE rate for the slow reaction was raised artificially on every other injector directly above the entrance into the primary combustion zone. The area corresponded to a cylindrical pancake region with the same diameter of the nozzle exit. This ignitor/pilot region per-sisted for t = 0.1sec of the simulation and thereafter was removed. Interactions between the injectors caused ignition for all inlet ports. The simulations were run until a statistical steady value of NOx was measured at the outlet plane of the simulation. Examples of the temperature, NOx, and O2 field are shown for the base case in Figure 6. Equivalent images for the proposed LSR design are shown in Figure 7. 4 Results For the full-scale experiments, a water-cooled suction probe was used to sample the gas stream at various axial and radial locations. At each sampling point, the measurement was recorded after steady state was reached. Simulation data along lines of sight across the steam generator diameter were averaged in time in an effort to replicate experimental methodology. These simulation data lines were extracted at the appropriate downstream locations to correspond with the experimental measurement. The time-averaging was performed over a statis-tically steady region to produce the simulation data points. Figures showing the location of the simulated probe are shown in Figure 8. Samples were taken across the direct downstream path of the secondaries and also midway between the downstream paths (labeled with a "_diag"). Given that the actual radial trajectory of the experimental measurement is uncertain, these simulated tra-jectories represent the two limiting radial trajectories across the steam generator diameter. Figure 9 shows the radial NOx profiles measured 1.5, 4.5 and 7.5 feet from the burner face for the experimental and simulation data. The LSR data are plotted on the left and the current steam generator data are plotted on the right. Note that the BASE and LSR data are mirrored to keep all BASE data to the right and all LSR data to the left. Note that no experimental data was collected on the LSR. The error bars on the experimental data are only the instrument error as reported in the Testo 350 product literature (Testo, 2011). As seen in Figure 9, the circumferential location of the sampling point is an important parameter, especially in the near burner region. A location directly 12 (a) (b) (c) Figure 6: Snapshot slices of (a) temperature [K] (b) NOx [ppm dry] and (c) O2 [mass fraction] fields for the base design. 13 (a) (b) (c) Figure 7: Snapshot slices of (a) temperature [K] (b) NOx [ppm dry] and (c) O2 [mass fraction] fields for the base design. 14 (a) (b) Figure 8: Cartoon of the cross-section view (normal to the primary flow direc-tion) of the steam generator downstream from the burner face. Shown are the locations of the lines of simulation data used to compare with the experimental measurements. 15 (a) (b) (c) Figure 9: Radial NOx profiles (ppm, dry) across the steam generator diameter measured 1.5 feet (top), 4.5 feet (middle), and 7.5 feet (bottom) from the burner face. Data to the left of centerline (0) is from the LSR, data to the right of centerline is from from the current steam generator (BASE). 16 in the flow path of a low NOx injector will exhibit a much different profile than a location situated between the flow paths of two injectors. In Figure 9(top), the sharp increase in NOx beginning 10 inches from the centerline (CL) followed by a sharp decrease in NOx 30 inches from CL, seen in several radial profiles, indicates a radial profile in the flow path of a secondary fuel injector. At the flame front (~ 20 inches from centerline), temperatures are high enough for thermal NOx to form. Further from the centerline, the fuel concentration downstream of the injector is so high that the mixture is outside its flammability limits and no NOx is present. The radial paths situated between two injectors (labeled "_diag" in the figure) appear to be more representative of the radial path measured experimentally. Depending on the radial path selected, the simulation data generally follow the trend of the experimental data for the current steam generator configuration. It is difficult to make any statements about the consistency of experimental and simulation data for two reasons. One, no parametric simulation studies were conducted. Two, only the instrument error of the measured NOx data is known; the actual experimental uncertainty may be much higher than is indicated by the error bars in these plots. However, it is clear that by coupling our thermal NOx formation mechanism to the resolved time and length scales in the steam generator, the low NOx levels achieved by these burners in the field are well represented. The impact of the LSR design on NOx profiles is clear from these plots. Centerline NOx values are relatively unchanged because the primary reaction zone remained the same for the two simulations. The NOx formation in the region downstream of the secondary injectors, however, show significant differ-ences. The well-mixed stream of natural gas and FGR that is fed through the injectors at high velocity creates a flameless-like combustion zone where heat release is distributed, resulting in temperatures that are too low for thermal NOx formation. In Figure 10, similar plots are shown for radial O2 profiles at the same axial locations as above. The error bars on the experimental data are again only the Testo 350 instrument error. The unique radial trend of the data at each axial location is matched very well by the range of lines generated from the simulation data. For the current steam generator, the O2 profiles are not nearly as sensitive to the radial sampling location as the NOx profiles. The presence of O2 in the FGR stream is noted in the LSR radial profiles that pass through the middle of injector jets. The fact that the O2 profiles are similar for the two burner designs shows that O2 concentration is of secondary importance in thermal NOx formation. The local gas temperature, which is resolved at the very fine length and time scales in the Arches LES computations, has the strongest impact on NOx formation. Radial temperature plots at the three axial location are shown in Figure 11. Experimental temperature data were measured using a suction pyrometer with a shielded tip and a small-diameter thermocouple; data are plotted "as measured" and do not include any corrections for radiation losses. No associated temperature error was available, so error bars of +/-180ºF (100 K) are used 17 (a) (b) (c) Figure 10: Radial O2 profiles (dry) across the steam generator diameter mea-sured 1.5 feet (top), 4.5 feet (middle), and 7.5 feet (bottom) from the burner face. Data to the left of centerline (0) is from the LSR, data to the right of centerline is from from the current steam generator (BASE). 18 (a) (b) (c) Figure 11: Radial temperature profiles across the steam generator diameter measured 1.5 feet (top), 4.5 feet (middle), and 7.5 feet (bottom) from the burner face. Data to the left of centerline (0) is from the LSR, data to the right of centerline is from from the current steam generator (BASE). 19 on the plot for reference. The temperatures computed in the simulation are much higher than those measured experimentally at all three axial locations. In order to have significant thermal NOx formation (>1 ppm), however, the local temperature must exceed 2550ºF. The measured temperature then would have produced no thermal NOx. Therefore, it is likely that the temperatures computed in the simulation are closer to the actual temperatures. In comparing the profiles in Figure 11 (middle) and (bottom), the LSR design eliminates the high temperature combustion zone in the region where the injector jets mix with the surrounding air (~20 inches from CL). There are no spikes in thermal NOx formation in this region, indicating a region of flameless-like combustion. Also of note is the effect of the relatively cool premixed jets of FGR and natural gas. When the radial profile crosses the path of these jets, large decreases in temperature are observed. Despite the lower temperatures, the fuel combustion efficiency is maintained. While understanding where NOx is being formed aids with burner design/ modification, the variable that is used to monitor the steam generator during normal operations is the NOx concentration measured in the flue gas. There are no NOx reduction mechanisms occurring in the steam generator downstream of the computational domain used in the simulation, so the average NOx measured across the exit plane of the simulation is equivalent to the NOx that would measured in the flue gas. The time- and spatially-averaged NOx value at the simulation exit plane is 9.5 ppm for the current steam generator and 4.5 ppm for the LSR case. The average, measured by stack pitot probe as a function of probe insertion distance, ranged from 10.3 - 10.7 ppm NOx. The simulation value is well within the instrument uncertainty of +/- 5 ppm. While the 4.5 ppm value cannot be considered a prediction with known uncertainty, it does indicate that the LSR design will have a strongly positive effect in reducing NOx emissions to meet new emissions standards. 5 Conclusions Two high performance LES tools were coupled allowing for representation of a wide range of length and time scales in an oil-field steam generator. As a result, the simulation showed good agreement with NOx, O2, and temperature measurements in a full-scale system. A proposed retrofit option was evaluated to address impending NOx regulations from the current <15ppm rule to <5ppm. The flue-gas in the retrofit LSR design was removed directly from the stack and perfectly mixed with the fuel stream before it was introduced into the secondary combustion zone through a simple open pipe. The result showed a significant decrease in combustion temperatures and lower overall NOx values. While more evaluation is required, the use of HPC with LES in evaluating current OTSG performance and important design changes was successfully demonstrated. 20 6 Acknowledgments This research was sponsored by the National Nuclear Security Administration under the Accelerating Development of Retrofitable CO2 Capture Technolo-gies through Predictivity program through DOE Cooperative Agreement DE-NA0000740. 7 Disclaimer All information in this paper is provided "as is" for informational purposes only, generated from a research collaboration effort between Chevron U.S.A. Inc. ("Chevron"), The University of Utah ("Utah"), and Fives North American Com-bustion, Inc. ("FNAC") to develop a CFD modeling of steam generators. It is not intended, or should not be construed, to grant any license under any patent or other intellectual property right of any of Chevron, Utah, and FNAC. It is not intended to be and shall not be interpreted to be suggestions, advice or recommendations for the use of any product or application of any process in a manner that infringes any patent or other intellectual property of a third party. References [1] M. Berzins, J. Luitjens, Q. Meng, T. Harman, C.A. Wight, and J.R. Pe-terson. Uintah - a scalable framework for hazard analysis. In TG '10: Proceedings of the 2010 TeraGrid Conference, 2010. [2] C.T. Bowman. Kinetics of pollutant formation and destruction in combus-tion. Prog. Energy Combust. Sci., 1:33-45, 1975. [3] James C. Keck. Rate-controlled constrained-equilibrium theory of chemical reactions in complex systems. Prog. Energy Combust. Sci., 16:125-154, 1990. [4] John Nowakowski, Tom Roberston, Beverly K. Coleman, Stein Storslett, and Nicholas Brancaccio. The road to single digit NOx for oilfield once-through steam generators. AFRC conference, Koloa Hawaii, 2013. [5] S.G. Parker. A component-based architecture for parallel multi-physics pde simulation. Future Generation Comput. Sys., 22:204-126, 2006. [6] John Schmidt, Martin Berzins, Jeremy Thornock, Tony Saad, and James Sutherland. Large scale parallel solution of incompressible flow problems using uintah and hypre. In 2013 13th IEEE/ACM International Symposium on Cluster, Cloud, and Grid Computing, 2013. [7] J.P. Spinti, J.N. Thornock, E. Eddings, P.J. Smith, and A. Sarofim. Heat transfer to objects in pool fires. In Transport Phenomena in Fires. WIT Press, Southampton, UK, 2008. 21 [8] M.G. Zabetakis. Flammability characteristics of combustible gases and va-pors. Technical Report 627, Bureau of Mines, 1965. 22