Development of an Ultra-Low NOx Gaseous Fuel Burner for OTSG Applications

Paper from the AFRC 2016 conference titled Development of an Ultra-Low NOx Gaseous Fuel Burner for OTSG Applications

The paper presents an overview of the development process used to design improvements to an Ultra-Low NOx gaseous fuel burner for once through steam generator (OTSG) applications. It discusses the use of Computational Fluid Dynamics (CFD) for the prediction of key performance metrics such as NOx emissions and vessel heat flux profiles and simplifications in CFD processing such as the use of steady state solvers and the use of reduced chemistry sets. Results of CFD modeling are compared with data collected from large scale combustion tests.

Development of an Ultra-Low NOx Gaseous Fuel Burner for OTSG Applications American Flame Research Committee 2016 Industrial Combustion Symposium Kevin Anderson, Vladimir Lifshits, and Santhosh Umapathy John Zink Company, LLC | Power and Marine Division Abstract The paper presents an overview of the development process used to design improvements to an Ultra-Low NOx gaseous fuel burner for once through steam generator (OTSG) applications. It discusses the use of Computational Fluid Dynamics (CFD) for the prediction of key performance metrics such as NOx emissions and vessel heat flux profiles and simplifications in CFD processing such as the use of steady state solvers and the use of reduced chemistry sets. Results of CFD modeling are compared with data collected from large scale combustion tests. Introduction The use of the steam assisted gravity drainage (SAGD) process has become an effective and prevalent means for the production of heavy crude oil in the oil sands deposits found in northern Alberta. In this process, high pressure steam is generated and then injected into horizontal wells deep underground to aid in the extraction of heavy crude oil. A brief description of the SAGD process and some details related to heavy oil extraction in Alberta can be found on the Government of Alberta webpage [1]. During the SAGD process, some ninety percent of the water that is injected into the wells as wet steam is recovered with the extracted heavy oil. The recovered water is treated, then returned (as feed water) to steam generating equipment used in the SAGD process. The water treatment process used for SAGD operation is extensive. Yet, it leaves substantial amount of residual contaminants in the form of minerals and suspended hydrocarbons returned to the steam generating equipment. The recovered water quality generally does not meet standards typically needed for conventional saturated steam boilers. Instead, steam used for SAGD is most commonly produced using once through steam generator (OTSG) units which are, by design, more tolerant of the water contaminants typically associated with SAGD operation. Still, mitigating fouling of OTSG heat transfer surfaces continues to be an area of extensive study and research. Pugsley et al. discuss these challenges in some detail in [2]. Generally, significant effort is made to reduce peak heat fluxes on the OTSG heat transfer surfaces. Figure 1 depicts a typical OTSG installation. An OTSG consists of a horizontal cylindrical radiant section coupled with a convection section. The furnace is lined with serpentine steam tubes to absorb heat released by combustion of fuels to generate steam. Since it is usually more cost effective to construct the OTSG units offsite, the size of OTSG units is limited by shipping constraints making for a certain uniformity between units in overall furnace dimensions. Figure 1 - An OTSG Unit during Installation in Northern Alberta Generally, natural gas and "produced" gas (a byproduct of bitumen oil extraction) are used as fuels in OTSG applications. For most units, the heat required to generate steam in an OTSG is supplied through a single burner mounted on the firing end of the cylindrical furnace. As industrial burners go, these burners are generally quite large. Firing capacities in excess of three hundred million Btu per hour are typical. Projects with OTSG firing capacities in excess of four hundred million Btu per hour are in development. 2 John Zink Hamworthy Combustion (JZHC) and their affiliates have furnished burners for approximately two hundred (200) OTSG applications located in Alberta. Most of these installations utilize QLNTM burner technology, which was developed by Coen Company, Incorporated in the 1990's as a low NOx burner design for use in industrial boiler applications. The QLNTM burner is particularly well suited for OTSG applications. The patented burner design employs a combination of burner technologies such as partial lean premix and fuel staging to achieve low NOx emissions without the need of external NOx control technologies such as flue gas recirculation (FGR) or selective catalytic reduction (SCR), both of which are generally considered as undesirable for OTSG application in the Canada oil sands areas. Also, the QLNTM burner produces a relatively low intensity flame radiation profile. This is quite well suited to OTSG applications due to the concerns of furnace tube fouling. Figure 2 - QLNTM Burner Used for an OTSG Application, Prior to Shipment Like many areas in the world, emissions requirements for OTSG applications in Alberta have become more stringent over time. Table 1 below shows the history of NOx emissions that are applicable to OTSG units in Alberta. Changes to NOx emissions regulations applicable to OTSG units in Alberta are currently being discussed by Canadian regulatory agencies. There remains some uncertainty as to where these emissions regulations will ultimately end up. Table 1 - Applicable NOx Emission Guidelines for New Alberta Oil Sands OTSG Installations Year 1998 [3] 2007 [4] NOx Compliance Limit (g/GJ) Natural Gas Produced Gas 40 40 26 40 NOx Performance Target (g/GJ) Natural Gas Produced Gas 40 40 7.9 15.8 The first Alberta oil sands OTSG installations utilizing QLNTM burners were commissioned in the late 1990's. These units were designed to meet NOx emissions requirements of 40 g/GJ (HHV) firing natural gas and produced gas fuels. By the mid 2000's, OTSG installations utilizing a QLNTM burners were being 3 commissioned to meet NOx emissions requirements of 26 g/GJ firing natural gas. This represented a thirty five percent reduction in NOx emissions from 1998 compliance limits. Generally, these emissions reductions were easily achieved with QLN burners without any modifications or restrictions on firing modes. Figure 3 - OTSG Furnace Firing End with QLNTM Burner Installed Currently, there remains ambiguity among users as to the required NOx emissions performance for new OTSG applications. NOx emissions requirements of 26 g/GJ (HHV) firing natural gas are still prevalent, yet some users are seeking to meet the Alberta oil sands performance target of 7.9 g/GJ firing natural gas. This would represent an eighty percent reduction in NOx emissions from 1998 compliance limits. Currently, there are OTSG systems with QLNTM burners designed to achieve 7.9 g/GJ NOx emissions using external NOx control methods such as FGR. However, achieving the target emissions of 7.9 g/GJ level without the use of external NOx reduction methods such as FGR represented a formidable challenge. To address the changing NOx emissions requirements, a development effort was undertaken to improve the performance of the QLNTM burner for large OTSG applications. This effort including making design optimizations to the aerodynamic aspects of the burners as well as optimizing the design of the fuel injection components. Analysis of the impact of potential burner changes on the operation of the OTSG system was made to ensure that NOx emissions reductions did not come at the expense of OTSG operational flexibility. As such, parameters such as furnace heat flux profiles, burner operating range, burner ramping capabilities, furnace pulsations, and combustion noise were evaluated as part of the effort. 4 Procedure Development efforts consisted of the design and evaluation of various potential improvements to the QLNTM burner. As discussed, this effort included developing and evaluating aerodynamic changes to the airside of the burner as well as development and evaluation of changes to the burner fuel injection components. Both Computational Fluid Dynamics (CFD) analysis and physical testing of a scaled burner prototype were used to evaluate the proposed design improvements. CFD Analysis CFD software has evolved from a tool to evaluate isothermal flow fields in simple geometries to a powerful and robust tool which can model reacting flow fields in complex geometries. The use of CFD in the design of equipment for various industries has become prevalent as the capabilities of the CFD codes and speed of the solvers have improved. A good general description of CFD applied to industrial combustion applications is presented by Londerville and Baukal [5]. As NOx emissions reduction was a primary objective of this project, a significant effort was undertaken to develop good correlation between NOx emissions obtained from experimental results (i.e. known behavior) and NOx emissions predicted using CFD analysis. At the same time, it was imperative that cycle time for CFD analysis was minimized in order to allow for efficient use of resources to evaluate a substantial number of potential burner design modifications. Steady state CFD analysis was used to predict the NOx emissions for the considered burner configurations. A description of how steady state CFD modeling techniques can be used to predict NOx emissions from gaseous fuel burner systems is presented by Anderson and Londerville [6]. ANSYS Fluent version 16.1 CFD software was used to develop, execute, and process all of the CFD simulations. A burner with heat release of 350 million Btu per hour (HHV) was used for the CFD simulations. Methane gas was used to simulate the fuel gas in the simulations. A two step reaction mechanism was used to model methane combustion. A finite rate model with proprietary kinetics coupled with the eddydissipation model based on the work of Magnussen and Hjertager [7] was used to model the reactions. Turbulence in the simulations was modeled using the realizable k-epsilon model. As radiation is the primary mode of heat transfer from the flame to other surfaces, it was believed that accurate radiation modeling would be critical for realistic a NOx prediction. The Weighted Sum of Gray Gases model (WSGGM) with proprietary parameters was used to approximate the radiation properties for the combustion products. A one-step soot formation model was included in the simulations to improve the accuracy of flame radiation computations. Radiation calculations themselves were modeled using the discrete ordinates (DO) model. The NOx emissions models included thermal NOx and prompt NOx formulations. Thermal NOx emissions were computed using the extended Zeldovich mechanism. The significant reactions were as follows: 𝑂 + 𝑁2 ⇌ 𝑁 + 𝑁𝑂 𝑁 + 𝑂2 ⇌ 𝑂 + 𝑁𝑂 𝑁 + 𝑂𝐻 ⇌ 𝐻 + 𝑁𝑂 Equation 1 Equation 2 Equation 3 Some simplifications were made to the thermal NOx calculation. First, Equation 3 was neglected since it is generally only significant for fuel rich combustion. Second, the quasi-steady assumption for [N] was 5 used which allows for equations 1 and 2 to be combined. Finally, reverse reactions were neglected since resultant NO concentrations were generally very low in all computational cells. The [O] concentration assumed at equilibrium was computed via post processor based on predicted [O2] concentration and predicted temperature using a correlation derived from interpolation of JANAF thermochemical tables which are available from NIST [8]. The kinetic rate constant proposed by Baulch et al. [9] for Equation 1 was used to compute predicted thermal NOx. Prompt NOx formation was also considered. The chemistry of Prompt NOx formation is complex and the species and kinetics driving Prompt NOx formation were not included in the simplified steady state combustion simulations. Instead, Prompt NOx formation was approximated for the combustion simulations using the model proposed by De Soete [10]. Fluctuations in the steady state solution due to turbulence were approximated using a beta probability function (PDF) as described by Missaghi et al [11]. Fluctuations in species were neglected in the PDF formulation. Figure 4 depicts the geometry and grid used for one of the simulations. For each of the cases evaluated, a one eighth segment of the furnace and burner was modeled. Periodic boundary conditions were assumed at the segment boundaries. Typical computational domain size used for each model was approximately five million cells. Figure 4 - CFD Model Geometry/Computational Domain used for Steady State Simulations (Typical) Several burner configurations were evaluated. First, a baseline case of a conventional QLNTM burner was modeled and validated with known experimental results to establish a benchmark to compare other cases. Next a case was evaluated to predict the effect of airside burner changes on NOx emissions. Finally, 6 several cases were evaluated to predict the effectiveness of various fuel injector design changes. Each of the evaluated cases shared a common OTSG furnace design. Physical Testing of Scaled Burner Prototype A prototype QLNTM burner was designed and fabricated in order to test various changes to the burner design. The burner was designed to fire natural gas at 63.0 million Btu per hour (HHV), which represented eighteen percent of the full scale burner capacity. Figure 5 shows the burner installed at the John Zink Hamworthy Combustion test facility located in Tulsa, OK. Figure 5 - Prototype Burner Installed in the JZHC Test Facility located in Tulsa, OK The impact of geometric scaling on NOx emissions, concerning both on the burner and on the test furnace, was of obvious interest. This topic, specifically applied to scaling industrial sized burners, is presented in some detail by Bollettini et al in [12]. Ultimately, a proprietary scaling method was used to scale the burner. The scaling effects of the test furnace were considered using a combination of CFD simulations and proprietary analytical methods. Performance test data were collected for several burner test configurations. The initial tests evaluated performance of the conventionally designed burner (baseline tests) and fuel injectors. Subsequent testing efforts focused on evaluation of alternative fuel injector configurations which showed promise in CFD simulations. Each design was evaluated with respect to the effect on the NOx emissions, burner excess air, and combustion stability as characterized by the intensity of pressure fluctuations in the furnace and combustion noise. In addition to the NOx emissions and O2 in the furnace exhaust, CO emissions, combustibles emissions, and heat flux profiles along the furnace side wall centerline were recorded. The heat flux intensity was measured using a calibrated water cooled Schmidt-Boelter thermopile radiometer. The device measured radiation intensity with a sixty degree (60°) view angle probe inserted into the furnace through a series ports spaced two feet apart. 7 Results and Discussion CFD Analysis Figure 6 depicts the CFD results showing predicted NOx emissions for the baseline configuration with different excess air levels. The predicted NOx values were normalized to the baseline NOx emissions operating with a typical excess level. Known behavior derived from field data was included in the chart for comparison. NOx Emissions Relative NOx Emissions 1.10 1.00 0.90 0.80 Known Behavior, Baseline Configuration 0.70 CFD Prediction, Baseline Configuration 0.60 0.50 Burner Excess Air Figure 6 - Predicted NOx Emissions for the Baseline Configuration (Known Behavior) Good correlation between the predicted NOx at different excess air levels and with known behavior was observed. The correlation of the CFD predictions with known behavior suggested that the CFD analysis could be effective in accurately predicting changes in NOx emissions due to changes to the burner design. Figures 7, 8, and 9 depict contours of predicted temperature, predicted mole fraction of O2, and total NOx formation rate for the baseline configuration operating with design excess air. These figures and other CFD results for the baseline configuration were analyzed in detail in order to better understand and address the mechanisms controlling NOx formation. 8 Figure 7 - Predicted Gas Temperature, Baseline Configuration Figure 8 - Predicted Mole Fraction of O2, Baseline Configuration Figure 9 - Predicted Mole Fraction of NOx, Baseline Configuration 9 Figure 10 shows the predicted heat flux profile for the baseline configuration. The data was presented as the heat flux absorbed by the cylindrical furnace wall (averaged circumferentially) plotted by distance from the furnace front wall. The distance was evaluated as the dimensionless number of burner diameters such that the data could more readily be compared against the scale model test data later. Note that the peak predicted heat flux was approximately 44,000 Btu/hr-ft2 at 4.6 burner diameters from the furnace front wall. Predicted Wall Heat Flux 50,000 Wall Heat Flux (Btu/hr-ft2) 45,000 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 0 2 4 6 8 10 12 14 16 Axial Distance From Furnace Front Wall (Burner Diameters) Figure 10 - Predicted Furnace Wall Heat Flux, Baseline Configuration For subsequent cases, adjustments were made to the fuel injection components in attempts to mitigate NOx formation. Each of these configurations was modeled and analyzed in a similar fashion as the baseline case. Analysis of the CFD results ultimately led to the design of several modified fuel injection configurations. Predicted NOx emissions and heat flux profiles were evaluated against the baseline configuration. Figure 11 shows the relative NOx emissions as predicted by the CFD simulations for the baseline configuration and the modified fuel injection component configurations. The data points were sorted from the highest predicted NOx emissions to lowest predicted NOx emissions. Note that Case 1 (colored red) represented the baseline configuration, and that all results were normalized against this case. 10 Predicted NOx Emissions Baseline Relative NOx Emissions 1.40 1.20 1.00 0.80 0.60 0.40 0.20 Case 9 Case 7 Case 6 Case 4 Case 1 Case 3 Case 2 Case 5 Case 8 0.00 Fuel Injection Configuration Figure 11 - CFD Predictions of NOx Emissions for Various Fuel Injection Configurations Some of the cases that were evaluated showed predicted NOx emissions in excess of the baseline configuration (Cases 8, 5, 2, and 3). These cases were not studied further. The cases that showed reductions in NOx emissions relative to the baseline configuration (Cases 4, 6, 7, and 9) were of interest. Analysis of Case 9 results suggested that NOx emissions reductions of approximately thirty percent or more were possible. Predicted heat flux profiles for the cases showing a reduction in predicted NOx emissions (Cases 4, 6, 7, and 9) are shown in Figure 12. The profile for the baseline case (Case 1) was included in the figure for reference. The CFD results showed moderate increases in peak predicted heat flux for Cases 6 and 9. The other profiles were similar to the baseline case. Predicted Wall Heat Flux Wall Heat Flux (Btu/hr-ft2) 60,000 50,000 40,000 Case 1 Case 4 30,000 Case 6 20,000 Case 7 10,000 Case 9 0 0 2 4 6 8 10 12 14 16 Axial Distance From Furnace Front Wall (Burner Diameters) Figure 12 - Predicted Furnace Wall Heat Flux, Reduced NOx Configurations 11 Physical Testing of Scaled Burner Prototype The prototype burner was outfitted with components to replicate the firing scenarios which showed favorable predicted performance in the CFD simulations. Burner components for Cases 1 (baseline), 4, 6, 7, 9, and 10 were designed and fabricated. For each case, fuel injection components were installed and the burner was fired in the test furnace. Figure 13 shows the test burner firing at capacity with Case 1 (baseline) components. Figure 13 - Prototype Burner Firing at Capacity in the JZHC Test Facility located in Tulsa, OK Relative NOx Emissions 1.20 Baseline NOx emissions data measured for each of the firing cases and is shown in Figure 14 below. Also included are the corresponding CFD predictions of NOx performance. Generally, close agreement between CFD predictions and prototype test results were obtained. In fact, when furnace scaling was considered, CFD predictions of NOx emissions matched test data within +11%/-4% for all cases evaluated. NOx Emissions Comparison 1.00 0.80 0.60 CFD Results 0.40 Test Data 0.20 Case 9 Case 7 Case 6 Case 4 Case 1 0.00 Fuel Injection Configuration Figure 14 - Comparison of Measured NOx Emissions to CFD Predictions Case 9 showed the lowest NOx emissions during testing of the prototype burner. However, operation of the burner with the Case 9 fuel injection configuration exhibited undesirable transient behavior which manifested as low intensity pulsations and excessive combustion noise. It was not possible to discern this 12 behavior from the steady state CFD simulations. Ultimately, the Case 9 configuration was abandoned due to this undesirable behavior. The remaining prototype configurations (Cases 4, 6, and 7) showed comparable performance relative to the baseline configuration in terms of combustion stability, sensitivity to excess air, burner turndown, and ramping capabilities. Of these three cases, Case 4 showed the most favorable performance in terms of NOx emissions, furnace pulsations, and combustion noise. The tests showed that NOx emissions reductions of approximately twenty one percent were attainable using the Case 4 configuration. Figure 15 shows normalized measured wall heat flux intensity for the baseline configuration and for the Case 4 configuration. The CFD predictions were included in the figure for reference. Note that actual measurements showed peak heat flux intensities farther down the furnace than the CFD predictions. This was likely due to the simplifications to the reaction mechanisms used in the CFD simulations which were selected deliberately in order to simplify the CFD modeling efforts. Radiation intensity measurements of the Case 4 configuration showed a reduction in peak wall heat flux intensity of approximately seven percent compared to the baseline configuration. The CFD simulations showed a similar trend. The CFD simulation of the Case 4 configuration showed a reduction in peak wall heat flux intensity of approximately three percent compared to the baseline configuration. Wall Heat Flux (Normalized to Baseline) Wall Heat Flux Comparison 1.20 1.00 0.80 Case 1 - CFD 0.60 Case 1 - Measured 0.40 Case 4 - CFD Case 4 - Measured 0.20 0.00 0 2 4 6 8 10 12 14 16 Axial Distance From Furnace Front Wall (Burner Diameters) Figure 15 - Comparison of Measured Wall Heat Flux to CFD Predictions, Baseline and Case 4 Configurations 13 Conclusions The CFD analysis offered significant value to the lower NOx design optimization of the QLNTM burner. Even with significant simplifications such as the use of steady state solvers and greatly simplified reaction mechanisms, it was possible to discern the performance of candidate designs relative to the performance of a baseline configuration with known performance characteristics. For instance, it was possible to evaluate predicted changes in NOx performance relative to the baseline configuration with a reasonable amount of accuracy. Also, it was possible to evaluate predicted furnace heat flux profiles for the various configurations with reasonable accuracy. However, it is still important to compliment the simplified CFD process with prototype physical testing. When scaling effects were considered, the test results showed discrepancies as high as eleven percent between predicted NOx emissions and actual test data. Also, the simplified CFD analysis approach is still unable to predict undesirable transient behavior such as furnace pulsations and/or combustion noise. Still, combining CFD with prototype testing was quite effective for the process of evaluating the burner design improvements. The use of CFD in the design process allowed for the elimination of approximately half of the proposed candidate fuel injector designs, which reduced the number of configurations which physical testing was deemed warranted. As such, physical testing efforts for promising cases were more readily expanded. The results from this development effort were quite promising. Using this process, it was ultimately possible to find simple design modifications promising to achieve a reduction in NOx emissions of some twenty percent without degradation in other burner performance metrics such furnace wall heat flux, burner operating range, and combustion stability. The next step would be to apply the proposed modifications to a large OTSG unit in order to validate the performance at full scale conditions. 14 Bibliography [1] "Talk About SAGD," November 2013. [Online]. Available: http://www.energy.alberta.ca/OilSands/pdfs/FS_SAGD.pdf. [Accessed May 2016]. [2] T. Pugsley, D. Pernitsky, J. Grundler and E. E. Johnsen, "Fouling of Heat Transfer Surfaces in a Steam Assisted Gravity Drainage (SAGD) In Situ Facility for the Recovery of Oil Sands Bitumen," in International Conference on Heat Exchanger Fouling and Cleaning, Budapest, 2013. [3] Canadian Council of Ministers of the Environment, "National Emission Guideline for Commercial/Industrial Boilers and Heaters," Winnipeg, 1998. [4] Oil Sands Environmental Management Division, "Emissions Guidelines for Oxides of Nitrogen," Alberta, 2007. [5] S. B. L. Editor and C. E. B. Editor, "CFD-Based Combustion Modeling," in The Coen & Hamworthy Combustion Handboox, Boca Raton, CRC Press, 2013, pp. 183-210. [6] K. M. Anderson and S. B. Londerville, "Practical Use of Computational Fluid Dynamics to Industrial Combustion Applications," in American Flame Research Committee Combustion Symposium, Kauai, 2013. [7] B. F. Magnussen and B. H. Hjertager, "On mathematical models of turbulent combustion with special emphasis on soot formation and combustion," in 16th Symposium on Combustion, Cambridge, 1976. [8] NIST, "NIST-JANAF Thermochemical Tables," National Institute of Standards and Technology, September 1982. [Online]. Available: http://kinetics.nist.gov/janaf/html/O-001.html. [Accessed 1 May 2013]. [9] D. L. Baulch, C. J. Cobos, R. A. Cox, P. Frank, G. Hayman, T. Just, J. A. Kerr, T. Murrells, M. J. Pilling, J. Troe, R. W. Walker and J. Warnatz, "Summary Table of Evaluated Kinetic Data for Combustion Modeling: Supplement 1," Combustion and Flame, no. 98, pp. 59-79, 1994. [10] G. G. De Soete, "Overall Reaction of NO and N2 Formation from Fuel Nitrogen," in 15th Symposium on Combustion, 1975. [11] M. Missaghi, M. Pourkashanian, A. Williams and T. L. Yap, "Prediction of NOx Emissions from Oxygen Enriched Burners," in International Conference on Environmental Control of Combustion Processes, Honolulu, 1991. [12] U. Bollettini, F. N. Breussin and R. Weber, "A Study on Scaling of Natural Gas Burners," IFRF Combustion Journal, vol. Article 200006, pp. 1-24, 2000. © 2016 John Zink Company LLC. For information on patents and trademarks see johnzinkhamworthy.com/legal-notices 15