VARIATIONS IN THE NOx EMISSION OF GAS TURBINES: EFFECTS OF AIR TEMPERATURE, AIR HUMIDITY AND NATURAL GAS COMPOSITION

Transcription

1 m THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47th St., New York, N.Y The Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published in an ASME Journal. Papers are available from ASME for 15 months after the meeting. Printed in U.S.A. Copyright 1994 by ASME 94-GT- 261 VARIATIONS IN THE NOx EMISSION OF GAS TURBINES: EFFECTS OF AIR TEMPERATURE, AIR HUMIDITY AND NATURAL GAS COMPOSITION B. Martien Visser and Fred C. Bahlmann Gasunie Research N.V. Nederlandse Gasunie Groningen, The Netherlands ABSTRACT The NOx emission, produced by gas turbines varies with ambient conditions and with fuel gas composition. Often, legislation requires that the NOx emissions of gas turbines has to be corrected to standard conditions. The EPA formula may be used for the correction for ambient temperature and humidity. In the Netherlands, the correction for fuel-gas composition is based on the observation that for natural gases, NOx emission varies linearly with the Lower Calorific Value (LCV). It is concluded that both the EPA and LCV correction formulas are equivalent to the following relation between flame temperature and NOx emission: NOxa = NOxb * ( 1.0(J65)Ta-Th where Ta and Th represent characteristic flame temperatures under conditions a and b. In the paper, the utility of the EPA and the LCV correction formulas for gas turbines equipped with modem lean-premixed combustors is discussed. NOMENCLATURE F air fraction to primary air hm water content in air kg/kg dry air Hw Constant in relation (7) - Lo stoichiometric dry air requirement kg air/kg gas P combustor pressure kpa T temperature C y H2O/fuel ratio kg/kg T residence time ms equivalence ratio - subscripts amb ambient fl flame 1. INTRODUCTION Worldwide, stringent legislation with respect to pollutant emissions from gas turbines has been introduced. Often, the legislation follows the "state of the art" of combustion technology. This implies that many gas turbines are operated at emission levels close to the allowed limits, so that relatively small variations in the NOx emission may affect the judgment as to whether a particular installation satisfies the requirements. Variations in the NOx emission of gas turbines as large as 15% may be due to variations in the ambient conditions. Therefore, in many countries, legislation requires the measured NOx emission to be corrected to standard ambient conditions. The "classical" EPA correction formula may be used for this purpose. The NOx emission of a gas turbine is also dependent on the LCV value of the fuel gas. Generally, a high calorific fuel gas produces more NOx (in g/gj) than a fuel gas with a lower calorific value. In the Netherlands, the NOx limit is corrected (within certain limits) for the fuel gas composition using a linear relation between the NOx emission and the LCV value of the gas. Both the EPA formula and the correction for fuel gas have been developed using data from conventional diffusion-type combustors. Their validity for modem premixed combustors is the subject of ongoing discussion between government and industry. The objective of this study is twofold. Firstly, a relation between NOx emission and flame temperature, derived from experimental data from lean-premixed combustion systems is presented. This relation is in remarkable agreement with the existing correlations between the NOx emission of conventional diffusion type combustors and air temperature, air humidity and fuel gas composition. Secondly, the EPA correction formula and the LCV correction for gas composition are evaluated in order to judge their reliability for modem lean-premixed combustors. This paper focuses on natural-gas-fired combustors. So, the effects of fuel-bound nitrogen on the NOx emission are neglected. Presented at the International Gas Turbine and Aeroengine Congress and Exposition The Hague, Netherlands June 13-16, 1994

2 The outline of the paper is as follows: in section 2, the relation between flame temperature and NOx emission is introduced. In section 3, this relation will be compared with the EPA correction formula for diffusion type combustors. It will be shown that this EPA formula is equivalent with this relation between NOx emission and flame temperature. The validity of the EPA formula with respect to lean-premixed combustion systems is discussed in section 4. Effects of fuel gas composition on NOx emission are discussed in section 5. Conclusions drawn from this study are presented in section NOx EMISSION VERSUS FLAME TEMPERATURE The NOx emission of gas turbine combustors is affected by a range of parameters including temperature, oxygen concentration, fuel composition, residence time and pressure. In the recent decades, significant progress has been made in the understanding and modeling of NOx formation mechanisms. Nevertheless, the prediction of the NOx emission of gas turbine combustors is still a difficult task. To assist designers in the short term, engineering correlations have been formulated. These are mainly based on the knowledge that NOx emission is strongly dependent on flame temperature. Correlations between flame temperature and NOx emission have been given by (amongst others) Lewis (1981), Lefebvre (1984), Leonard and Stegmaier (1993) and Visser and Levinsky (1993). NOx emission correlations in the literature The relation presented by Lewis (1981) has been derived using experimental data from Anderson (1975) with premixed propane combustion under 5.5 bar and an air factor > 1.25 (0 < 0.8): NOx (ppm, as measured)= e ( (rfl + 273)) (1) Lefebvre (1984) has used experimental data of diffusion type combustors. The relation between NOx emission and primary zone flame temperature equals: NOx (g/kg) = P0-25 F r e (0.01('rfl + 273)) (2) Recently, relation (2) has been successfully applied and extended to include the effects of evaporation and mixing by Rizk and Mongia (1992). Leonard and Stegmaier (1993) present data of premixed combustion with operating pressures, in the range 1 to 30 bar and flame temperatures between 1430 and 1650 C. From these data, the following relation may be deduced between NOx emission and flame temperature. NOx (ppm, 15% 0 2) = a {8.o 10-3 (Tfl + 273)) (3) The relation by Visser and Levinsky (1993), very similar to that obtained by Lewis and Leonard and Stegmaier, is: NOx (ppm, 0% 02) = e( (Tfl + 273)) (4) Relation (4) has been obtained using experimental data from various sources and applications of lean-premixed natural gas combustion. Discussion The factor in the exponent presented by Lefebvre (= 0.01) is larger compared to the factors given by the other authors. However, Lefebvre's correlation also includes residence time (r), fraction of air in the primary zone (F) and system pressure (P). These factors are also affected by the flame temperature. For instance, the residence time (r) is dependent on the flame temperature and on the ignition distance from the fuel injector. Thus, Lefebvre's relation between NOx emission and flame temperature is more complex than it appears on first sight. Therefore, it is difficult to compare the exponential factors presented by Lefebvre (1984) and the other authors (1993). The coefficient in the exponent ( ) as presented by Lewis (1981) is approximately 25% higher compared to that obtained by Visser and Levinsky (1993). Lewis used the NOx "as measured" instead of the NOx, corrected to 0% 0 2, but this can not explain the relative large difference in this coefficient. Instead, it is believed that this difference is due to the effect of pressure on the data used by Lewis. In his analyses in 1981, Lewis has assumed a square root relation between NOx and pressure. At present, there is evidence that the effect of pressure on NOx in premixed combustion systems varies with air factor (Correa 1991). Thus, the data used by Lewis are unequally affected by pressure, leading to an overestimation of the coefficient in the exponent in relation (1). Figure 1 contains the correlations by Leonard and Stegmaier (1993) and Visser and Levinsky (1993). The figure contains experimental data from a variety of experiments employing leanpremixed combustion which were used by Visser and Levinsky (1993) to derive their correlation. It may be concluded that the correlation by Leonard and Stegmaier is well within the large experimental data base considered by Visser and Levinsky. Relation (4) will be used in this paper. It will be shown that this relation is in good agreement with the correction terms for humidity and ambient air temperature as expressed in the EPA correction formula. The relation may be rewritten as: NOxa = NOxb * (1.0065) ( F a-th) (5) where Ta and Th represent the flame temperatures under conditions "a" and "b". Relation (5) implies that the NOx emission of a flame increases with 0.65% when the flame temperature increases with 1 C. Reference flame temperature In premixed combustion, there is a fixed fuel/air ratio in the whole flame area. Thus, the adiabatic flame temperature may be calculated using an enthalpy balance for the fuel/air ratio considered. In diffusion flames, combustion occurs in a variety of mixtures of air, fuel and combustion products and consequently, a distribution of "flame temperatures" exist. The properties of this distribution are generally not known, but, it may be assumed that (small) changes in air composition (humidity), air temperature and fuel gas LCV will not significantly alter this distribution. In that case, a reference flame temperature may be used. The authors believe that in diffusion flames, the majority of the NOx is formed in flame regions with a local stoichiometry close to unity. In premixed combustion systems, the maximal NOx

3 NOx (ppm, 0% 02) os o^^ o... o o Visser a Levinsky Leonard: & 8tegmaier Correa Altemark & Knauber Visser & Levinsky Adiabatic flame temperature ( C) FIGURE 1: NOx EMISSION VERSUS FLAME TEMPERATURE (PREMIXED COMBUSTION) emission occurs at an air factor of approximately In the case of diffusion flames, the reference flame temperature with respect to NOx emission is therefore chosen to be the (adiabatic) flame temperature at an air factor of The choice of a different reference flame temperature will not significantly alter the results of this study. 3. DIFFUSION TYPE COMBUSTORS The effect of air humidity and ambient air temperature on the NOx emission of diffusion type combustors may be estimated using the "classical" EPA correction equation as recommended by the U.S. Environmental Protection Agency (EPA) in the Code of Federal Regulations (1989). On the other hand, air humidity and ambient air temperature affect the reference flame temperature. The effect of variations in the reference flame temperature on NOx emission may be estimated using relation (5). In this chapter, both methods to estimate the effect of ambient conditions on NOx emission will be compared. Air humidity The humidity of the combustion air affects the flame temperature. Adiabatic flame temperatures at an air factor of 1.05 and at different values of ambient humidity have been calculated assuming a temperature increase of 400 C during compression and usage of a standard Groningen natural gas (LCV 38 MJ/kg). Some results are shown in table 1. The effect of humidity on NOx emission has been estimated using both relation (5) and the correction equation as recommended by the U.S. Environmental Protection Agency (EPA) in the Code of Federal Regulations (1989): NOxco = NOxmeasured * e{19*(hm )} (6) It may be noted that there is excellent agreement between the results obtained with relation (5) and with the humidity correction in the EPA formula (6). TABLE 1: EFFECTS OF AIR HUMIDITY ON FLAME TEMPERATURE AND NOx EMISSION ambient air reference flame temperature temp. humidity air factor = 1.05 NOx emission relative to ISO EPA T-flame (6) (5) C % kg/kg C % difference fuel: Groningen gas LCV : 38 MJ/kg 3

4 Lewis (1981) has demonstrated a similar agreement between the humidity correction in the EPA formula and relation (2). Although, the coefficient in (1) is about 25 % higher than that used in the present study. The difference is that Lewis used the peak flame temperature, while in the present study, a reference flame temperature is defined at an air factor of Water and steam infection Using the same line of thought, the authors have related the EPA correlation between the humidity and the NOx emission to the effect of water and steam injection into the combustion chamber. This resulted in the following model (Visser and Bahlmann, 1993): NOx"t = NOxd'y * e{19*y*hw /(1.05*Lo)) (7) where: NOxa'a = NOx emission including effect of water or steam injection (g/gj) NOx"O' = NOx emission without water or steam injection (g/gj) In this relation, the factor 1.05 refers to the air factor. Hw is a constant with a value of 1 for steam injection and a value of 1.6 for water injection. Relation (7) has been evaluated using a large range of experimental data from various diffusion type gas-fired gas turbine systems, and for the majority of the installations a very good agreement with the measured (EPA-corrected) NOx emission was obtained (Visser and Bahlmann, 1993). In his study in 1981, Lewis has observed a similar relationship, but he mentioned that an additional factor of 0.9 had to be applied to the water/fuel ratio, to obtain the best fit. In the present study, using a 25% smaller effect of flame temperature on NOx emission and a reference flame temperature at an air factor of 1.05, such an additional empirical factor is apparently not necessary. Ambient air temperature Variations in the ambient air temperature cause variations in the air temperature after compression and therefore, variations in the flame temperature. In the EPA correction formula the effect of ambient temperature (T,,b) on NOx emission is described as: NOxCO1 = NOxm` ' * {T mb + 273)/288)} (8) Table 2 presents results of calculations of the effect of ambient air temperature on the NOx emission using relations (5) and the EPA formula (8). The reference flame temperatures have been calculated using a zero ambient humidity to exclude the effect of the water content in the air. The temperature rise during compression is 400 C. The relative NOx emission has been normalised to the NOx emission at 15 C and 0% humidity. There is (again) an excellent agreement between the EPA formula and relation (5). For the derivation of table 2, it has been assumed that the compressor temperature rise is independent of ambient air temperature. For single-shaft industrial engines, connected to the electric grid and running at constant speed, this is approximately valid. But, for double-shaft gas turbines, it is generally not. In these cases the effect of ambient temperature on NOx emission may be different than predicted with the EPA correction formula. TABLE 2: EFFECTS OF AMBIENT AIR TEMPERATURE ON FLAME TEMPERATURE AND NOx EMISSION ambient air reference NOx emission flame temperature relative to ISO difference temp. humidity air factor = 1.05 EPA T-flame (8) (5) C % C % fuel: Groningen gas => LCV : 38 MJ/kg Combined effect of ambient air temperature and humidi Figure 2 presents the combined effect of ambient air temperature and humidity on the NOx emission for three compressor temperature rises of 300, 400 and 500 C. The horizontal axis denotes the calculated reference flame temperature, which is affected by the compressor temperature rise and the ambient conditions. The vertical axis represents the NOx emission. The relative NOx emission has been calculated as follows: for each considered compressor temperature rise, a reference flame temperature (T1so) has been calculated at the ISO-conditions (15 C and 60% humidity). Then, for a given set of ambient conditions (a), the reference flame temperature (T a) is calculated. Using relation (5), the relative NOx emission may be calculated from: NOx t1 O = (1.0065)(Ts TIS& (9) The curves in figure 2 have been derived using relation (9). The data have been calculated using the EPA formula at the conditions: - ambient air temperatures: 0, 10, 20, 30 and 40 C; - humidity: 0%, 25%, 50%, 75% and 100%; - compressor temperature rise: 300, 400 and 500 C. From Figure 2, it may be concluded that for the wide range of conditions considered, the correlation between the EPA formula and relation (5) is excellent. 4. LEAN-PREMIXED COMBUSTORS In the case of premixed combustors, the situation is different from conventional diffusion type combustors. This will be illustrated using the following example of a fully premixed combustor with conditions: - turbine inlet temperature of 1200 C; - 40% of the engine air being used as dilution air; - compressor temperature rise 400 C; The conclusions, derived in this chapter are insensitive to the choice of these conditions. 4

5 NOx relative to I80 conditions 140% O Calculated with EPA. Tcomp - 300: C 1 20%... : Tcomp : C 100% % Tcomp - 500: C O. 0-: CQ:... V.... Q G'^... ::I.....y... a^ p,/':. 40% Reference flame temperature ( C) FIGURE 2: COMPARISON OF RELATIVE NOx EMISSION CALCULATED USING EPA AND USING REFERENCE FLAME TEMPERATURE The flame temperature in this example is calculated as follows: it is assumed that the engine control of the gas turbine will maintain a constant turbine inlet temperature of 1200 C. Thus, the total air factor of the gas turbine may be calculated from the ambient conditions, the compressor temperature rise, the gas composition, and the turbine inlet temperature. The " flame air factor" is a fraction of the total air factor. In this example, 60% enters the combustion zone and thus, the flame air factor is 60% of the total air factor in the engine. The flame temperature is calculated at that flame air factor. Air humidity Table 3 shows the calculated total air factor, the flame air factor and the flame temperature, and the NOx emission relative to ISO conditions (15 C and 60% relative humidity). The relative NOx emission has been calculated using both relation (5) and the EPA correction formula. From table 3, it may be noted that the combined effect of air humidity and engine control, maintaining a constant turbine inlet temperature, results in a fairly constant flame temperature. Therefore, the NOx emission, calculated from the flame temperature, is approximately independent of air humidity. The same conclusion has been deduced by Lewis (1981). This implies that for gas turbines with lean-premixed combustors, the correction factor for humidity in the EPA formula should be omitted. Ambient air temperature The turbine inlet temperature is determined by the mixture of relative hot combustion gases of the premixed flame and relative cold dilution air. With increasing ambient air temperature, the temperature of the dilution air increases. Maintaining a constant turbine inlet temperature, the control system will then decrease the fuel flow to generate cooler combustion gases. Thus, the ambient air temperature will affect the NOx emission for premixed systems. The effects of the ambient temperature on the flame temperature and, using relation (5), on the NOx emission are illustrated in table 4. The calculated NOx emission in this example decreases by 12% when ambient temperature increases from 0 C to 40 C. The application of the EPA formula would lead to an estimated increase in the NOx emission by 20%. This implies that, for premixed combustors, an increase in the ambient air temperature results in a decrease in the flame temperature and thus in a decrease of the NOx emission. This effect is opposite to the effect encountered with conventional diffusion combustors. 5. FUEL GAS COMPOSITION The NOx emission of a gas turbine is dependent on the fuel composition. Experience at Gasunie has shown that the NOx emission of natural gas fired gas turbines varies approximately linearly with the LCV value (in MJ/kg) of the natural gas used. In the Netherlands, the NOx limit for gas turbines varies according to this rule of thumb:

6 TABLE 3: EFFECT OF AMBIENT HUMIDITY ON NOx NOXC0Ir = NOxGr- * LCV/ 38 (10) EMISSION FROM A LEAN-PREMIXED COMBUSTOR where: NOx O1r = corrected NOx emission limit (g/gj ISO) NOxG1O = NOx emission limit for Groningen gas ambient relative total flame flame NOx emission (LCV = 38 MJ/kg) air humidity air air temp. relative to ISO difference temp. factor factor T-flame EPA LCV = Lower Calorific Value of the fuel gas applied (5) (6) (in MJ/kg) oc % oc % fuel: Groningen gas LCV : 38 MJ/kg TABLE 4: EFFECT OF AMBIENT TEMPERATURE ON NOx EMISSION FROM A LEAN-PREMIXED COMBUSTOR ambient relative total flame flame NOx emission air humidity air air temp. relative to ISO difference temp. factor factor T-flame EPA (5) (8) oc o^ oc % fuel: Groningen gas LCV : 38 MJ/kg TABLE 5: EFFECT OF FUEL GAS TYPE ON FLAME TEMPERATURE AND NOx EMISSION fuel gas type LCV reference NOx emission flame temperature relative to difference at air factor 1.05 Groningen gas (5) LCV MJ/kg C % Groningen gas G-gas H-gas Ekofisk gas De Lier gas Steen gas K15 gas Refinery gas ISO ambient conditions : 15 C and 60% relative humidity Although this rule of thumb has been derived using data of natural gas fired systems, it is also used for other gaseous fuels. The maximum correction of the NOx limit allowed for gas composition is 10%. Diffusion-tune combustors Table 5 presents calculated effects of gas composition on the NOx emission for the commercial Groningen and H-gas, as well as a Norwegian Ekofisk gas and gases from De Lier, Sleen and K15 (North sea). The latter three gases are source gases with extreme LCV values and/or Wobbe indices. For the wide range of natural gases considered, there is a good correlation between the NOx emission based on the LCV value and the NOx emission, based on the flame temperature. Table 5 contains also results for a refinery gas, containing large fractions of ethene and hydrogen (but no fuel-bound nitrogen). For this gas, there is a significant discrepancy between the NOx emission, calculated using the LCV correction and the NOx emission calculated using the reference flame temperature. A limited set of emission data from a gas turbine, firing this refinery gas, shows approximately 75% higher NOx emission values compared to the NOx emission from an identical gas turbine fired with G-gas. This suggests that the NOx / flame temperature correlation is more generally applicable than the LCV correlation. Nevertheless, for natural gases, which do not contain large fractions of higher hydrocarbons and/or hydrogen, the LCV correlation provides a good guideline. Lewis (1981) has also shown that the flame temperature could be used effectively for the prediction of the effect of gas composition on the NOx emission. Although, an extra empirical constant (2/3) had to be introduced. With the NOx/temperature correlation used in this paper this additional constant is not necessary. Lean-premixed combustors In the case of premixed combustors, the flame temperature will be balanced by the engine control to obtain a constant turbine inlet temperature. The effect of gas composition on the NOx emission is similar to that of humidity: a change in gas composition does not lead to a significant change in the flame temperature and consequently, it might be expected that the effect of changes in the gas composition on the NOx emission is small. Practical data from gas turbines equipped with lean-premixed combustors, which are necessary to prove this conclusion, are not yet available. 6. CONCLUSIONS The present study focuses on the use of natural gas, which does not contain any fuel-bound nitrogen. The effects of air temperature, air humidity and fuel gas composition on the NOx 6

7 emission of both diffusion-type combustors and lean-premixed combustors have been discussed. From the study, the following conclusions may be derived: Diffusion type combustors The relation between flame temperature and NOx emission, presented in this paper, is in very good agreement with existing engineering correlations between NOx emission and humidity and ambient air temperature (the EPA correction equation) and gas composition (the LCV correction equation). A study, presented by Lewis (1981) and showing similar relations, has been evaluated. It is suggested that Lewis' relation between NOx emission and flame temperature is structurally affected by the effect of pressure on the data which were used to derive it. This resulted in an overestimation of the effect of flame temperature on NOx emission. This may explain why Lewis had to introduce some additional empirical constants in his study. In the present study these factors are no longer necessary. Nevertheless, the ideas of Lewis regarding the presence of a simple engineering relation between NOx emission and flame temperature were fully confirmed. Lean-premixed combustors It is concluded that there will be hardly any effect of humidity and/or fuel gas composition on the NOx emission, because of the engine control, which regulates the turbine inlet temperature. The effect of ambient temperature on the NOx emission is opposite to that for conventional diffusion combustors: with increasing air temperature, NOx emission decreases. This implies that the EPA correction formula should not be used for gas turbines equipped with lean-premixed combustors. REFERENCES Code of Federal Regulations, "Standards of Performance for Stationary Gas Turbines," Altemark, D., and Knauber, R., 1987, "Ergebnisse von Untersuchungen an einem Vormischbrenner unter Druck mit extrem niedriger NOx-Emission". VDI Berichte 645 (German). Anderson, D.N., 1975, "Effects of equivalence ratio and dwell time on exhaust emissions from an experimental premixing, prevaporised burner," NASA TP X Correa, S.M., 1991, "Lean-Premixed Combustion for Gas Turbines: Review and Required Research," ASME Fossil Fuel Combustion, PD-Vol. 33, pp Lefebvre, A.H., 1984, "Fuel effects on Gas Turbine Combustion Ignition, Stability and Combustion Efficiency," Trans ASME Journal of Aircraft, Vol. 21, No. 11, pp Leonard, G. and Stegmaier, J., 1993, "Development of an Aeroderivative Gas Turbine Dry Low Emissions Combustion System", Diesel & Gas Turbine Worldwide, May Lewis, G.D., 1981, "Prediction of NOx emissions" ASME Paper 81-GT-119. Rizk, N. K., and Mongia, H.C., 1992, "Semi-analytical Correlations for NOx, CO, and UHC Emissions", ASME Paper 92-GT-130. Visser, B.M. and Bahlmann, F.C., 1993, "NOx abatement in gas turbines", ErdOl & Kohle and Erdgas - Petrochemie Vol 46, September Visser, B.M. and Levinsky, H.B., 1993, "Premixed combustion in gas fired equipment", VDI Berichte 1090.