Stationary natural-gas turbine systems are rapidly increasing and becoming one of the prime technologies for generating electricity. Lean premixed turbulent combustion of natural gas is a promising method to achieve low pollutant emissions from advanced stationary gas turbines. This study has identified useful reduced kinetic schemes that can be used in comprehensive multi-dimensional gas-turbine combustor models. Reduced mechanisms lessen computational cost and possess the capability to accurately predict the overall flame structure, including gas temperatures and key pollutant species such as CO and NOx. In this study, four-new global mechanisms with 5, 6, 7, and 9 steps based on the full GRI 2.11 mechanism were developed and studied for their potential to model natural-gas chemistry (including NOx chemistry) in gas turbine combustors. These new global mechanisms were optimized to model high pressure and lean conditions found in gas turbines operating in the lean premixed mode.

Based on Perfectly Stirred Reactor (PSR) and premixed code calculations, the 5-step global mechanism was identified as the most promising model that can be used in a multi-dimensional gas-turbine code for modeling lean-premixed, high-pressure turbulent combustion of natural gas. If computational costs due to additional global steps are not severe, the newly developed 9-step global mechanism, which is a little more accurate and provided the least convergence problems, can be used. Future experimental research in gas-turbine combustion will provide more accurate data which will allow the formulation of better full and reduced mechanisms. Also, improvement in computational approaches and capabilities will allow the use of reduced mechanisms with larger global steps.

1. Introduction
2. Objectives and Approach
3. Results and Discussion
   • 5 Step Global Mechanism
   • PSR Code Calculations
   • Premixed Code Calculations
4. Summary and Conclusions
5. References

Introduction

Stationary gas turbines are used in industry for power generation. Current turbines have combined-cycle efficiencies in the range of 50%. There is a worldwide effort to increase these combined-cycle efficiencies, and at the same time decrease pollutant emissions which are harmful. Most of the current and older gas turbines operate with a diffusion flame, resulting in high peak flame temperatures. These gas turbines use air at the exit of the combustor as a coolant, but the high peak flame temperatures result in large NOx emissions (on the order of 200 ppm, corrected to 15% O2). Lean premixed combustion (LPC) of natural gas is a promising method to achieve low emissions from advanced gas turbines.

There are no comprehensive computer models presently available to model LPC of natural gas which include the full set of chemical mechanisms along with turbulent interactions (Correa, 1992). Methods have been presented in literature which include complex chemistry in turbulence models, but these models require enormous computing time and are limited to simpler chemical systems such as H2 combustion (Borghi, 1988). Therefore, simplified kinetic schemes with 4 to 5 global reaction steps are being developed for use in turbulent combustion (Mass and Pope, 1992a, Mass and Pope, 1992b). Many reduced mechanisms based on full mechanisms have been developed and published in literature (Bilger, et al., 1990, Chen and Dibble, 1991, Glarborg, et al., 1992b, Trevino and Mendez, 1992, Wang and Frenklach, 1991, Polifke, et al., 1995).

Full mechanisms are usually validated using experimental measurements. These full mechanisms are then used as benchmarks to evaluate reduced mechanisms. Thus, reduced mechanisms are validated by comparing key species and temperature predictions to predictions of validated full mechanisms. Another approach that is used to validate reduced mechanisms includes comparing predictions of reduced mechanisms to actual data, when the data are obtained in systems that can be easily modeled. However, lack of pertinent experimental data in lean premixed systems at high pressure warrants the use of full mechanisms for validation of reduced mechanisms.

Objectives and Approach

The GRI 2.11 mechanism (Bowman, et al., 1995) developed by the Gas Research Institute (GRI) is one of the best mechanisms, currently available that accurately describes CH4/NOx chemistry for natural gas combustion. The full GRI 2.11 mechanism was reduced using a computer code that eliminated unimportant species for the desired condition and then applied steady-state approximations for short-lived species (Chen, 1988). This resulted in 5-9 step global mechanisms of CH4 combustion coupled with NOx chemistry. The GRI 2.11 mechanism was then used as a bench-mark to test the global mechanisms in the PSR (Glarborg, et al., 1992a) and the premixed codes (Kee, et al., 1992a). Predictions of these global mechanisms were also compared with predictions of CH4 chemistry obtained using the 4-step Seshadri-Peters reduced mechanism (Chen and Dibble, 1991) (for PSR cases), which appeared to be the best available reduced mechanism in literature. The 4-step Seshadri-Peters mechanism was developed to model CH4 chemistry and does not include NOx chemistry (Chen and Dibble, 1991). It is based on a 25-step skeletal mechanism developed by Smooke and Giovangigli (1991).

Results and Discussion

This work was focused at providing a reduced mechanism with minimum possible steps for modeling key essentials of natural gas turbine combustion. A set of four new global mechanisms with 5, 6, 7, and 9 steps were developed and optimized to agree with PSR calculations using the GRI 2.11 mechanism at 30 atm and equivalence ratios ranging from phi = 0.4 to 0.6. These newly-developed global mechanisms contain both CH4 chemistry (including C2 chemistry) and NOx chemistry that is present in lean premixed turbulent combustion of natural gas. Only a brief description of the 5-step global mechanism is presented here. Further details of these mechanisms are provided by Mallampalli (1996). Emphasis was placed on the 5-step global mechanism, since this small mechanism seems to be more compatible with comprehensive 3-dimensional gas turbine combustor codes (e.g., Cannon, 1996).

5 Step Global Mechanism: The global rates of the above 5 global steps were derived from a skeletal mechanism based on the 276-step full GRI 2.11 mechanism (Bowman, et al., 1995) using the computer-assisted reduction mechanism (CARM) code (Chen, 1988) developed by Professor J.Y. Chen at U.C. Berkeley. CARM is an interactive program that runs in conjunction with CHEMKIN (Kee, et al., 1992b) and flame codes such as the perfectly stirred reactor (PSR) code (Glarborg, et al., 1992a) and the premixed code (Kee, et al., 1992a). The PSR program is more commonly used due to its simplicity, and also because it provides solutions to flame problems more quickly. Figure 1 shows the schematic diagram that explains the interaction between the CARM code, CHEMKIN, and a flame code such as the PSR code (Chang, 1995). The 5-step global mechanism is shown below:

R1: 3H2 + O2 + CO2 = 3H2O + CO (1)

R2: H2 + 2OH = 2H2O (2)

R3: 3H2 + CO = H2O + CH4 (3)

R4: H2 + CO2 = H2O + CO (4)

R5: 3H2 + CO2 + 2NO = 3H2O + CO + N2 (5)

The global rates of the above 5 reactions are functions of the rate constants and species present in the full GRI 2.11 mechanism. These global reaction rates were calculated assuming that the species C2H, H2CN, HCNN, C2H3, C2H5, C, CH3O, CH, CN, N, C2H6, NH, HCCO, NNH, CH2OH, NH2, HCCOH, CH2(S), NCO, C2H2, HOCN, C2H4, HNO, HCO, CH2CO, NO2, CH2, CH3OH, H2O2, HCNO, HO2, HNCO, HCN, CH2O, CH3, N2O, H, and O are in steady state. Here, the R5 global step incorporates the total NOx formed and includes thermal NOx, prompt NOx, and NOx formed via the nitrous oxide pathway.

PSR Code Calculations. The four newly-developed global mechanisms were evaluated according to the test matrix shown in Table 1. 108 test cases were performed using the full GRI 2.11 mechanism, Seshadri-Peters global mechanism and each of the four newly-developed reduced mechanisms (648 total cases) in order to examine the effects of pressure, equivalence ratio, inlet temperature, and residence time. The inlet temperature of 1800 K shown in Table 1 is unrealistically high, but was used to allow testing of the mechanisms at extreme temperatures. The full GRI 2.11 mechanism did not converge to a solution for 3 test cases (i.e., at 30 atm, phi=0.4, Tinlet = 600K and different residence times of 2 ms, 5 ms, and 10 ms), since the lean blow-out limit was reached at these conditions. Consequently, the four newly-developed global mechanisms with 5, 6, 7, and 9 steps which were based on the full GRI 2.11 mechanism also did not converge to a solution at these conditions. The conditions for these calculations were selected to be non-equilibrium conditions in order to emphasize the importance of chemical kinetics.

Figures 2a and 2b show predictions of temperature in a PSR as a function of equivalence ratio for pressures of 1 and 30 atmospheres (Tinlet = 600 K, tau = 2 ms). It is seen from Figs. 2a and 2b that the four newly-developed global mechanisms show excellent agreement with the full GRI 2.11 mechanism at both pressures (max. relative error ~1.6 %). The 4-step Seshadri-Peters global mechanism predictions also show good agreement. The Seshadri-Peters mechanism deviates from the full mechanism by less than 50 K at high pressures and high equivalence ratios (max. relative error ~3 %).

Figures 3a and 3b show predictions of CH4 concentrations in a PSR for pressures of 1 and 30 atmospheres (Tinlet = 600 K, tau = 2 ms). It is again seen from Fig. 3a that the four newly-developed global mechanisms give very good predictions for mole% CH4 at fuel-lean mixtures and atmospheric pressures (max. relative error ~15 %). Comparatively at 1 atm pressure, the 4-step Seshadri-Peters global mechanism predictions agree within 26% (on a relative basis). At 30 atm, Fig. 3b shows that the 5 and 6-step mechanisms give identical predictions for all conditions (max. relative error 3 %), but the Seshadri-Peters mechanism deviates (max. relative error ~650 %). At high pressures, C2 chemistry seems to be vital for predicting flame chemistry, and thus the newly-developed global mechanisms predictions are more accurate than the Seshadri-Peters mechanism.

Figures 4a and 4b show predictions of mole% CO in a PSR for pressures of 1 and 30 atmospheres (Tinlet = 600 K, tau = 2 ms). Figure 4a shows that at atmospheric pressure, the full GRI 2.11 mechanism and four newly-developed global mechanisms are in good agreement (max. relative error ~7 %), while comparatively, the 4-step Seshadri-Peters global mechanism (obtained from literature) is in slightly more disagreement (max. relative error ~12 %). Figure 4b shows that at 30 atm pressure, the newly-developed global mechanisms perform very well at all equivalence ratios (max. relative error ~2 %), but the Seshadri-Peters mechanism significantly differs from the full GRI 2.11 mechanism predictions (relative max. error 65 %). As expected, the deviation of the Seshadri-Peters mechanism with the full GRI 2.11 mechanism is more pronounced at high pressures (relative max. error 65 %) than at low pressures (max. relative error ~12 %).

Figures 5a and 5b show the predicted NO concentrations in a PSR as a function of equivalence ratio at 1 and 30 atmospheres (Tinlet = 600 K, tau = 2 ms). It is seen that although the predictions of all four newly-developed global mechanisms are extremely good at high pressures (Fig. 5b) for all equivalence ratios (max. relative error ~5 %), the 5, 6, and 7-step global mechanisms are less accurate for atmospheric pressure (Fig. 5a) at high equivalence ratios (max. relative error ~47 %). This is probably because these 5, 6, and 7-step global mechanisms were optimized for high pressure and low equivalence ratio conditions found in lean premixed gas turbine combustion. At these and all other conditions, Fig. 5a and 5b show that the 9-step global mechanism is the most accurate reduced mechanism. Since, the 4-step Seshadri-Peters mechanism does not contain NOx chemistry, it could not be used for this comparison.

Premixed Code Calculations. Premixed flames are effectively one-dimensional and can be made very steady so that detailed temperature and species concentrations can be measured. The four newly-developed global mechanisms were further tested using the premixed code which was used to predict species profiles using the burner-stabilized flame option and a given temperature profile. Some convergence problems were encountered, and test cases showed that the problems were probably due to the inability of the premixed code to handle significant amounts of intermediate species present at low temperatures (Cannon, 1996). This problem was resolved by assuming all intermediates to be zero below 1200 K for atmospheric cases and below 1450 K for 30 atm cases.

Comparatively, the 9-step global mechanism converged to solutions more easily using the premixed code. Although, the 9-step global mechanism also contains a large number of algebraic expressions for intermediate species, the algebraic relationships are less complicated in the 9-step global mechanism than in the 5, 6 or 7-step global mechanisms. The test matrix shown in Table 2 was used to evaluate the performance of the 5- and 9-step global mechanisms in comparison to the full GRI 2.11 mechanism. These premixed calculations were performed until most major species reached equilibrium conditions.

Figure 6 shows predictions of CH4 concentration along the length of the premixed flow reactor for pressures of 1 and 30 atm for phi = 0.6 and 1.0 (Tinlet = 300 K). The combustion rate predicted by the 9-step and 5-step global mechanisms are similar to the full GRI 2.11 mechanism at all pressures and equivalence ratios. Further, increases in pressure for a given equivalence ratio increase the combustion rate.

Figure 7 shows the predicted CO and CO2 concentrations as a function of distance in the premixed flow reactor for pressure of both 1 and 30 atm at equivalence ratios of 0.6 and 1.0 (Tinlet = 300 K). At 1 atm, the predicted peak CO concentration using the 9-step and 5-step global mechanisms are similar to the full GRI 2.11 mechanism for both phi = 0.6 and phi = 1.0 (max. relative error ~2 %). At 30 atm, the peak CO concentrations predicted using the global mechanisms are higher than the full GRI 2.11 mechanism predictions (max. relative error ~30 %). The CO2 formation rate is similar for the reduced and full mechanisms at all conditions shown.

Equilibrium conditions were reached for all conditions shown in Fig. 7 except for predictions shown in Fig. 7a (CO reached ~9.3 ppm). For CO, equilibrium values vary with both pressure and equivalence ratio. At low equivalence ratios of phi = 0.6 (Figs. 7a and 7c), the equilibrium CO concentration is < 5 ppm for both atmospheric and 30 atm pressures. However, at stoichiometric conditions (Figs. 7b and 7d), equilibrium values decrease with increase in pressure. At 1 atm, the equilibrium concentration of CO is ~14,300 ppm, while at 30 atm, the equilibrium concentration of CO is ~7,200 ppm.

Figure 8 shows predicted NO concentrations as a function of distance along the premixed flow reactor (Tinlet = 300 K). The NO formation rate is largely dependent on the conditions within the reactor. Fig. 8a shows that, at 1 atm and phi = 0.6, the peak NO concentration is ~3 ppm, which is predicted reasonably well by both the 9-step and 5-step global mechanism. The increase in equivalence ratio from phi = 0.6 (Fig. 8a) to phi = 1.0 (Fig 8b) increases NO. Increasing the flame equivalence ratio raises the gas temperatures, which directly increases the thermal NOx and hence increases the total NOx formed. Figure 8c shows that at low equivalence ratios (phi = 0.6), increases in pressure (compared to Fig. 8a) increases NOx emissions. Figure 8d shows that predicted NOx values are high at high pressures and phi = 1.0.

Equilibrium conditions were not reached for NO in any of the predictions shown in Fig. 8. Also for NO, equilibrium values vary with both pressure and equivalence ratio. At low equivalence ratios of phi = 0.6 (Figs. 8a and 8c), the equilibrium NO concentration is 500 ppm for both atmospheric (not reached with tau = 0.45 sec) and 30 atm pressures (not reached with tau = 3.85 sec). However, at stoichiometric conditions of phi = 1.0 (Figs. 8b and 8d), equilibrium values decrease with increased pressure. At atmospheric pressure, the equilibrium concentration of NO is 3200 ppm (not reached with tau = 0.13 sec), while at 30 atm, the equilibrium concentration of NO is 2500 ppm (not reached with tau = 0.95 sec).

Summary and Conclusions

A major focus of this project was to develop reduced mechanisms of CH4 combustion and NOx formation that describe lean premixed turbulent combustion of natural gas. This included identifying a useful comprehensive mechanism and then comparing the reduced mechanisms to the full mechanism in idealized codes at practical experimental conditions. The potential reduced mechanisms tested were expected to predict gas temperature and concentrations of key species such as NOx and CO.

Efforts were made to develop and evaluate suitable new 5- to 9-step global mechanisms of CH4 and NOx using the recently-released GRI 2.11 mechanism, which incorporates detailed NOx kinetics for combustion of natural gas along with detailed CH4 chemistry. The newly-developed 5- to 9-step global mechanisms were evaluated by comparing key species and gas temperature predictions in these mechanisms to the full GRI 2.11 mechanism. The 4-step Seshadri-Peters global mechanism found in literature was also examined. Using the full GRI 2.11 mechanism as a standard, predictions made using the 5-step global mechanism were more accurate than the 4-step Seshadri-Peters global mechanism.

Based on the PSR and premixed code calculations performed in this study, it is thus concluded that the 5-step global mechanism appears to be a promising reduced mechanism that can be used in multi-dimensional codes for modeling lean premixed turbulent combustion of natural gas. The 5-step global mechanism is reasonably accurate for both NOx and CH4 chemistry, based on comparisons with the validated full GRI 2.11 mechanism for pressures from 1 to 30 atm and equivalence ratios from 0.4 to 1.0. If computational costs due to additional global steps are not severe, the newly-developed 9-step global mechanism, which was a little more accurate and provided the least convergence problems, could be used. The 4-step Seshadri-Peters global mechanism is useful at low equivalence ratios and atmospheric pressures, and should be used if computational time due to the additional NOx step in the 5-step global mechanism is a vital factor. The 4-step Seshadri-Peters reduced mechanism lacks C2 chemistry and is thus less accurate at high pressures.

Further research in this area will result in better detailed mechanisms for modeling CH4 and NOx chemistry in lean premixed turbulent natural gas combustion, and this will subsequently result in better reduced mechanisms. Also, with the advancement in computer technology, faster computers will become available that will perhaps allow the use of larger elementary reaction schemes in practical geometries. Until then, methods such as the computer reduction methods will gain popularity, and will be used for modeling natural gas combustion.

References

Bilger, R. W., S. H. Stårner and R. J. Kee, "On Reduced Mechanisms for Methane-Air Combustion in Nonpremixed Flames," Combustion and Flame, 80, pp 135-149 (1990).

Borghi, R., "Turbulent Combustion Modelling," Progress in Energy and Combustion Science, 14, pp 245-292 (1988).

Bowman, C. T., R. K. Hanson, D. F. Davidson, J. William C. Gardiner, V. Lassianski, G. P. Smith, D. M. Golden, M. Frenklach, H. Wang and M. Goldenberg, in http://www.gri.org (1995).

Cannon, S., Personal Communication (1996).

Chen, J. Y., "A General Procedure for Constructing Reduced Reaction Mechanisms with Given Independent Reactions," Combustion Science and Technology, 57, pp 89-94 (1988).

Chang, W. C., "Modeling of NOx Formation in Turbulent Flames - Development of Reduced Mechanisms and Mixing Models," Ph.D. Thesis, Mechanical Engineering Department, University of California, Berkeley, CA (1995).

Chen, J. Y. and R. W. Dibble, "Applications of Reduced Chemical Mechanisms for the Prediction of Turbulent Nonpremixed Methane Jet Flames," In Reduced Chemical Mechanisms and Asymptotic Approximations for Methane-Air Flames, M. D. Smooke, Ed., Springer-Verlag, New York, Vol. 384, pp 193-226 (1991).

Correa, S. M., "A Review of NOx Formation Under Gas-Turbine Combustion Conditions," Combustion Science and Technology, 87, pp 329-362 (1992).

Glarborg, P., R. J. Kee, J. F. Grcar and J. A. Miller "PSR: A Fortran Program for Modeling Well-Stirred Reactors," Sandia National Laboratories, SAND86-8209.UC-4 (1992a).

Glarborg, P., N. I. Lilleheie, S. Byggstøyl, B. F. Magnussen, P. Kilpinen and M. Hupa, "A Reduced Mechanism for Nitrogen Chemistry in Methane Combustion," Twenty-Fourth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp 889-898 (1992b).

Kee, R. J., J. F. Grcar, M. D. Smooke and J. A. Miller "A Fortran Program for Modeling Steady Laminar One-Dimensional Premixed Flames," Sandia National Laboratories, SAND85-8240.UC-401 (1992a).

Kee, R. J., F. M. Rupley and J. A. Miller "Chemkin-II: A Fortran Chemical Kinetics Package for the Analysis of Gas-Phase Chemical Kinetics," Sandia National Laboratories, SAND89-8009B.UC706 (1992b).

Mallampalli, H.P., "Evaluation of CH4/NOx Global Mechanisms Used for Modeling Lean Premixed Turbulent Combustion of Natural Gas," M.S. Thesis, Chemical Engineering Department, Brigham Young University, Provo, UT (1996).

Mass, U. and S. B. Pope, "Implementation of Simplified Chemical Kinetics based on Intrinsic Low-Dimensional Manifolds," Twenty-Fourth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp 103-112 (1992a).

Mass, U. and S. B. Pope, "Simplifying Chemical Kinetics: Intrinsic Low-Dimensional Manifolds in Composition Space," Combustion and Flame, 88, pp 239-264 (1992b).

Polifke, W., K. Dobbeling, T. Sattlemayer, D. G. Nicol and P. C. Malte, "A NOx Prediction Scheme for Lean Premixed Gas Turbine Combustion based on Detailed Chemical Kinetics," International Gas Turbine and Aeroengine Congress and Exposition; ASME, Houston, TX, (1995).

Smooke, M.D. and V. Giovangigli, "Formulation of the Premixed and Nonpremixed Test Problems," In Reduced Chemical Mechanisms and Asymptotic Approximations for Methane-Air Flames, M. D. Smooke, Ed., Springer-Verlag, New York, Vol. 384, pp 1-28 (1991).

Trevino, C. and F. Mendez, "Reduced Kinetic Mechanism for Methane Ignition,"Twenty-Fourth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp 121-127 (1992).

Wang, H. and M. Frenklach, "Detailed Reduction of Reaction Mechanisms for Flame Modeling," Combustion and Flame, 87, 365-370 (1991).


Hemant's HomePage

This document is best viewed using Netscape 3.0

---