A Comparative Study of NO_x Emission Characteristics in a Fuel Staging and Air Staging Combustor Fueled with Partially Cracked Ammonia

Abstract

Recently, ammonia is emerging as a potential source of energy in power generation and industrial sectors. One of the main concerns with ammonia combustion is the large amount of NO emission. Air staging is a conventional method of reducing NO emission which is similar to the Rich-Burn, Quick-Mix, Lean-Burn (RQL) concept. In air-staged combustion, a major reduction of NO emission is based on the near zero NO emission at fuel-rich combustion of NH₃/Air mixture. A secondary air stream is injected for the oxidation of unburned hydrogen and NH_x. On the other hand, in fuel-staged combustion, NO emission is reduced by splitting NH₃ injection, which promotes the thermal DeNO_x process. In this study, NO_x emission characteristics of air-staged and fuel-staged combustion of partially cracked ammonia mixture are numerically investigated. First, the combustion system is modeled by a chemical reactor network of a perfectly stirred reactor and plug flow reactor with a detailed chemistry mechanism. Then, the effects of ammonia cracking, residence time, and staging scheme on NO_x emission are numerically analyzed. Finally, the limitations and optimal conditions of each staging scheme are discussed.

1. Introduction

In response to global climate change, the combustion society should develop a combustion technology for carbon-neutral fuels. Inherent carbon-free fuels, ammonia, and hydrogen are the potential alternatives to conventional hydrocarbon fuels in the power generation and industry sectors [1,2,3,4,5,6]. Based on the economic analysis using a global cost analysis considering production, storage, transportation, and CO₂ capture cost of 30 €/tonCO₂ [7], it is expected that, in the near future, ammonia will be produced by renewable hydrogen with a comparative price of methane. The major drawbacks of ammonia combustion are low reactivity and a large amount of NO emission compared to conventional hydrocarbon fuels. The maximum laminar flame speed of NH₃/air-premixed flame is about 7 cm/s, and the heating value of ammonia is 18.6 MJ/kg. Therefore, increasing the combustion intensity is essential to realize ammonia combustion for practical applications. Fundamental studies on ammonia combustion focused on this issue in various aspects. These include control of flow dynamics, fuel blending, and oxygen enhancement. Most ammonia gas turbine combustors utilize highly swirling flow [8,9,10,11,12,13], which generates large recirculation zones inside the combustor. These recirculation zones promote the reactions of unburned mixtures by transporting heat and radicals and reducing local flow velocity. Enhancing turbulent mixing can increase turbulent flame speed higher than the laminar flame speed and assist in the stabilization of the flame [14,15]. The mixing of ammonia with more reactive fuel, like hydrogen [16,17,18,19,20,21,22,23,24], methane [25,26,27,28,29], or cracking ammonia [13,23], is another way to improve the reactivity of the fuel. In the oxidizer enhancement approach, the oxygen concentration in the oxidizer increases by mixing air with an oxygen-enriched mixture [14,15,30,31,32]. The two-stage combustion can be used to reduce NO_x emissions from the combustion of ammonia. Rich-lean two-stage combustion controls local equivalence ratio and temperature by injecting air after the primary combustion zone [33]. The chemical reactions that oxidize unburned hydrogen and NH_x are promoted in the secondary combustion zone. It was reported that air staging can achieve about 32 ppmvd (15% O₂) of NO in the model gas turbine combustor [9]. However, the activation of NO_x reduction strongly depends on both the equivalence ratio and temperature, and the effective reduction is only valid in a very narrow temperature and equivalence range. Therefore, further research is needed to maximize the advantages of the two-stage combustion as well as explore the operating condition required for NO_x reduction.

Numerical simulation could provide helpful guidelines for the design and optimization process of a multi-parameter system. Li et al. [34] developed a Chemical Reactor Network (CRN) of NH₃/CH₄ fueled gas turbine combustor to analyze the effects of fuel blending and Air-Staged (AS) combustion. Their numerical results showed that low NO_x and CO emissions can be achieved for NH₃/CH₄/air mixtures at H/J class gas turbine conditions with AS combustion. Mashruk et al. numerically assessed humidified Rich-Burn, Quick-Mix, Lean-Burn (RQL) combustors with a CRN [35]. In their research, each part of the reaction zone inside the combustor is modeled as a single reactor based on the thermodynamic state and flow residence time obtained from the RANS CFD analysis. The CRN connects each reactor based on the spatial distribution of the reactors. The fuel is 70–30 (vol%) NH₃/H₂ mixture, and the oxidizer is an air and steam mixture. This model shows that the humidified RQL system can produce flue gas of 99.97% water with NO_x emissions of about 100 ppmv. Bioche et al. [13] performed a large eddy simulation of fuel-rich NH₃/H₂/air combustion in a model gas turbine combustor. This gas turbine combustor is originally designed for the CH₄/air mixture. To adjust the flame characteristics of the mixture to that of methane, they added hydrogen in the NH₃/air mixture (X_H2 = 0.46). According to their simulation results, there were high emissions of unburned hydrogen and NH_x, and the combustion efficiency was about 66% under operating conditions. Authors argued that by adopting AS combustion, combustion efficiency could be increased up to 99.5%.

On the other hand, a stabilization method of pure ammonia flame has to be developed to utilize ammonia in the industrial and power generation sectors. Ammonia has poor combustion characteristics, and its flammable range is much narrower than methane. The maximum laminar flame speed of ammonia is about 1/5 of that of methane. Fuel blending is one way to overcome poor combustion characteristics. Blending ammonia with hydrogen reduces the ignition delay time and the auto-ignition temperature [22] and increases laminar flame speed [21]. Moreover, ammonia is a hydrogen carrier [36]. Hydrogen can be directly converted from ammonia with the cracking process. Mei et al. [23] studied laminar flame propagation of partially cracked NH₃/air mixtures in a constant-volume combustion vessel. At 40% of cracking, the laminar burning velocity of the partially cracking ammonia/air mixture is 38.1 cm/s which is close to that of the methane/air mixture. Another interesting result is that the NO formation of a partially cracked ammonia/air mixture exhibits non-monotonic dependency on the cracking ratio. The authors concluded that the transition from ammonia chemistry to cracked gas chemistry is the main source of this behavior.

Even though the development of commercial ammonia-fueled gas turbines has been started [37], there is no systematic study on the staging combustion strategy of ammonia and partially cracked ammonia. Moreover, most study on two-stage combustion is biased toward the rich-lean two-stage combustion concept [9,12,33,34,35,38,39]. The objective of this work is a comparison of conventional and novel two-stage combustion strategies, including air-staging (AS) and fuel-staging (FS) for partially cracked ammonia. The numerical simulations are carried out using a CRN specially developed for two-stage combustion. With this model, we evaluate the concentrations of NO, NO₂, and N₂O at the exhaust of the combustion system over a wide range of residence time and equivalence ratio.

2. Chemical Reactor Network Model

We developed a CRN model of a partially cracked ammonia combustor with the staged injection to analyze the effect of the staged combustion on the emission characteristics of the NH₃/H₂/N₂/air mixture. The CRN model consists of one Perfectly Stirred Reactor (PSR) and one Plug Flow Reactor (PFR). The PSR represents the primary combustion zone, and the PFR presents the secondary combustion zone. The outlet stream of PSR and secondary stream are mixed, and the mixture flows into the PFR. Figure 1 illustrates a schematic diagram of the CRN model used in this study. The simulation code is written in python using Cantera [40], open source software for thermodynamics and reaction kinetics.

The main parameters of the CRN model are the residence time of the PSR reactor (${\tau }_{PSR}$), the residence time of the PFR reactor (${\tau }_{PFR}$), global equivalence ratio (${\varphi }_G$), primary zone equivalence ratio (${\varphi }_P$), staging stream equivalence ratio (${\varphi }_S$), partial cracking ratio ($\beta$), and operating pressure (p). ${\varphi }_G$ is the equivalence ratio based on the total injected volumetric flow rate. $\beta$ is the volume fraction of H₂ and N₂ in the fuel stream. The fuel is a mixture of NH₃/H₂/N₂, and the volume ratio of H₂/N₂ is fixed at 3:1 to mimic the partially cracked NH₃ mixture. The cracking process is an endothermic reaction (2NH₃ $\to$ N₂ + 3H₂; ΔH⁰ = 92 kJ/mol⁻¹). However, in order to make the analysis simpler, we did not take into account any heat addition or loss caused by the cracking process. The global reaction of stoichiometric partially cracked ammonia for a given $\beta$ can be written as follows.

$$\left(1-\beta \right){\mathrm{NH}}_3+\beta \left(1.5{\mathrm{H}}_2+0.5{\mathrm{N}}_2\right)+0.75\left({\mathrm{O}}_2+3.76{\mathrm{N}}_2\right)\to 1.5{\mathrm{H}}_2\mathrm{O}+3.32{\mathrm{N}}_2$$

(1)

The overall ranges of these parameters are summarized in

Table 1. The range of the parameters considered in this study.

	Minimum	Maximum
${\tau }_{PSR}$ [s]	10−2	102
${\tau }_{PFR}$ [s]	10−2	102
$\beta$	0.0	0.75
${\varphi }_G$	0.40	0.90
${\varphi }_p$	0.60	0.90
${\varphi }_S$	0.00	1.20

. For the classification of the staged combustion strategies, we use the following notation. If ${\varphi }_S=0$, then it is the AS combustion. If ${\varphi }_S>0$, we define it as the FS combustion.

3. Results and Discussion

3.1. Single PSR Calculation

First, we plot the reactor temperature and NO_x concentrations by changing the equivalence ratio and fuel cracking ratio in a single PSR, as shown in Figure 2. All simulation results are computed with the chemistry mechanism of Okafor [28] unless specified otherwise. The peak temperature of the NH₃/H₂/N₂/air mixture is located at a slightly fuel-rich condition. By increasing the portion of cracked ammonia, the peak temperature increases. The maximum temperature of 100% cracked mixture ($\beta$ = 1) is about 2250 K. On the other hand, the peak of NO is located slightly in fuel-lean condition. Except for $\beta$ = 1, the peak NO concentration increases with increasing H₂/N₂ mole fraction. Since the main source of NO is fuel-bonded nitrogen, the NO formation mechanism of $\beta$ = 1 and that of others are significantly different, and small amounts of NH₃ inside the fuel stream could yield large amounts of NO_x in the NH₃/H₂/N₂/air mixture. Most of the NO in the NH₃/H₂/air mixture is produced by the HNO pathway [23,24,34], and this pathway is limited by the availability of oxygen. Moreover, NO concentration is very sensitive to temperature. For example, at $\varphi$ = 0.6, the temperature range for 0 $\le \beta \le 0.75$ is from 1491 to 1596 K. However, the NO concentration distributes from 474 to 2589 ppmv at that condition. Another important characteristic of the NO concentration profile is the strong non-linearity at the fuel-rich condition. The NO concentration stiffly drops at $\varphi$ = 0.95 and reaches about 50 ppmv at $\varphi$ = 1.55 for $\beta$ = 0. This preliminary computational result suggests that if the primary combustion zone is in a fuel-rich condition, near zero NO, NO₂ and N₂O are achievable [9,12]. Thus, the complete combustion of partially oxidized NH₃ with a secondary air stream is the key to low NO_x combustion. On the other hand, if the primary combustion zone is in fuel-lean condition, then the secondary stream must reduce concentrations of NO, NO₂, and N₂O below acceptable values. To achieve this, the secondary stream is carefully controlled to activate ammonia-based Selective Non-Catalytic Reduction (SNCR) [41], also known as the thermal DeNO_x [42]. It should be noted that the original thermal DeNO_x process is developed for NO_x reduction in a hydrocarbon fuel combustion. Therefore, past studies on thermal DeNO_x were mainly focused on NO reduction. However, the reduction of NO₂ and N₂O is also an important issue in ammonia combustion. We will investigate the effect of staged combustion on the emission characteristics of NO, NO₂, and N₂O in the following section.

3.2. Effect of Staged Combustion

CRN model illustrated in Figure 1 is used to demonstrate the effect of staged combustion. Figure 3 shows the effect of ${\varphi }_G$ in AS with pure ammonia ($\beta =0.0$). It is shown that for the three ${\varphi }_G$ values, the minimum level of NO is located at the upper left corner of the contour maps. The primary combustion zone needs a relatively long time to minimize the concentration of NO. On the other hand, the secondary combustion zone has to provide efficient mixing of secondary air and product of the primary combustion zone in a short time to delay the thermodynamic state of the mixture’s progress to the chemical equilibrium. This large ${\tau }_{PSR}$ and small ${\tau }_{PFR}$ is also the condition for a local minimum of NO₂. On the other hand, N₂O decreases by increasing ${\tau }_{PFR}$. This indicates the trade-off between NO and N₂O at the fuel-lean condition, as shown in Figure 2.

Figure 4 shows the effect of ${\varphi }_G$ in FS CRN. Due to the low reactivity of the mixture, the flame is extinguished when the residence time of the primary combustion zone is short (${\tau }_{PSR}<0.02$). If ${\tau }_{PSR}$ is long enough to stabilize the combustion, the ${\tau }_{PFR}$ determines the overall characteristics of the exhaust emissions. It is identified that the abrupt increase in NO is related to the combustion of the secondary stream NH₃/air mixture. The more NH₃ is injected, the longer residence time is required to oxidize NH₃. For instance, the ${\tau }_{PFR}$ = 100 s is not enough to secondary injected NH₃/air mixture to react in case of ${\varphi }_G$ = 0.9. Relatively low NO and N₂O with a certain level of NO₂ are found at the upper left of the contour. However, a large amount of unburned NH₃ is also found in this region. The trade-off between NO and N₂O is the primary issue with FS combustion at high global equivalence ratios. In Figure 4, the contours of NO and N₂O have the same interface where the stiff variations of species concentration with opposite directions are observed for ${\varphi }_G$ = 0.7 and 0.8. It shows the trade-off between NO and N₂O and the limitations of FS combustion with high-temperature exhaust gas. Figure 5 shows the effect of ${\varphi }_P$ and ${\varphi }_S$ on the emission map of two-stage combustion. Three operating conditions with fixed ${\varphi }_G$ are simulated. On the left side of the figure, the computational results of ${\varphi }_P$ = 0.6/${\varphi }_S$ = 0 are shown. It is AS combustion with a lean primary combustion zone. The figure shows a very high level of NO concentration for all stable combustion, but the NO₂ and N₂O levels are very low if ${\tau }_{PFR}$ is larger than 1 s. There is no effective regime for NO_x reduction. Thus, AS combustion is not an effective way of NO_x reduction when ${\varphi }_P$ is fuel-lean. On the other hand, the other two cases simulate FS combustion. It is clearly shown that a small amount of ammonia addition in the secondary stream yields a remarkable reduction of NO. Moreover, for ${\varphi }_P$ = 0.6/${\varphi }_S$ = 0.1, there exists a clean combustion regime where NO, NO₂, and N₂O levels are below 10 ppmv. It seems that ${\tau }_{PFR}$ should be longer than a certain limit to achieve this condition, and this critical ${\tau }_{PFR}$ depends on the equivalence ratio of the secondary injected stream. The critical residence time for ${\varphi }_P$ = 0.6/${\varphi }_S$ = 0.1 and ${\varphi }_P$ = 0.6/${\varphi }_S$ = 0.2 is about 10 and 100 s, respectively.

In Figure 6, we compare two staged combustion strategies at ${\varphi }_G$ = 0.45. NO_x mole fraction and temperature versus ${\tau }_{PFR}$ are plotted. For FS combustion, a high level of NH₃ is present in the exhaust gas for ${\tau }_{PFR}$ < 0.6 s. When ${\tau }_{PFR}$ > 0.6 s, NH₃ and N₂O converted to the N₂ with a small formation of NO. However, AS combustion is not an effective way of NO reduction for ${\varphi }_G$ = 0.45. The level of NO and NO₂ is nearly constant, and only N₂O drops from 5 to 0.02 ppmv by increasing ${\tau }_{PFR}$. Figure 7 compares two staged combustion strategies at ${\varphi }_G$ = 0.7. As shown in Figure 4, FS combustion is ineffective for reducing NO in high-temperature exhaust gas. The NO level in the exhaust gas is higher than 1000 ppmv. A long residence time is required to reduce it, and its equilibrium concentration is not at an environmentally acceptable level. On the other hand, the NO level in AS combustion is about 200 ppmv, and levels of NO₂ and N₂O are lesser than 3 ppmv for ${\tau }_{PFR}$ < 1 s. However, the NO, NO₂, and N₂O levels of AS combustion become similar to those values of FS combustion when ${\tau }_{PFR}$ is about 1000 s. It implies that the two staged combustion strategies yield the same equilibrium condition at these operating conditions.

3.3. Effect of Ammonia Cracking Ratio

Next, we investigate the effect of the cracked ammonia ratio. Figure 8 and Figure 9 show the contour maps of NO_x emissions for $\beta$ = 0.25, 0.5, and 0.75 in an AS CRN and FS CRN, respectively. Essentially, the same trends are found for both AS and FS cases with respect to the variation of $\beta$. At fixed ${\tau }_{PSR}$ and ${\tau }_{PFR}$, the level of NO increases with increasing $\beta$. The cracked gas accelerates chain-branching reactions which produce O and OH radicals [23]. This chemical effect leads to the increase of NO. However, the overall landscapes of the contour maps are not varying significantly with $\beta$. This means that the staged combustion strategies can be applied to a wide range of cracked ammonia and ammonia mixture without a major modification. Additionally, increasing $\beta$ decreases ${\tau }_{PFR}$ required for activation of NO reduction in an FS combustion. This result is consistent with the experimental study of rich-lean combustion in NH₃/H₂-fueled combustors [9].

3.4. Effects of Pressure and Chemistry Mechanism

Finally, we investigated the effects of pressure and chemistry mechanisms on the prediction of temperature and NO mole fraction. Figure 10 and Figure 11 show 15% O₂ NO concentration and the corresponding temperature of CRN for two-stage combustion computed with various chemistry mechanisms at 1, 10, and 30 atm. We checked not only NH₃ concentration but also NH₃, NO₂, and N₂O for various chemistry mechanisms. However, NO is the only chemical species that show noticeable deviations among the chemistry mechanisms considered here. The differences between mole fractions are lesser than 5 ppm for other species. The calculation condition for AS CRN is ${\varphi }_G$ = 0.7, ${\varphi }_P$ = 1.2, ${\varphi }_S$ = 0, ${\tau }_{PSR}$ = 100 s, and ${\tau }_{PFR}$ = 1 s. In terms of 15% O₂ NO concentration, each chemistry mechanism predicted different trends and quantities. Okafor’s [28] mechanism predicts the maximum NO level of 16.9 ppm at 1 atm, and it decreases to 7.1 ppm at 30 atm. Otomo’s mechanism [24] shows a nearly constant level of NO. The values of NO concentration at 1, 10, and 30 atm are 19.6, 18.8, and 18.7 ppm. Tian’s mechanism [27] shows an opposite trend with respect to Okafor’s mechanism. The NO concentration increases from 25.8 to 38.1 ppm with increasing pressure. The calculation condition for FS CRN is ${\varphi }_G$ = 0.4, ${\varphi }_P$ = 0.7, ${\varphi }_S$ = 0.1, ${\tau }_{PSR}$ = 1 s, and ${\tau }_{PFR}$ = 10 s. The discrepancies between the chemistry mechanisms in AS combustion are mentioned in [35]. In that study, the simulations are conducted with the chemistry mechanisms of Okafor [28], Glarborg [43], Xiao [19], Tian [29], and Otomo [24]. The major source of NO production at the primary combustion is from HNO, but the calculated production rate of NO from the Glaborg mechanism, Otomo mechanism, and Xiao mechanism are all different. This is partly attributed to uncertainties in important species such as HONO, HNO, NH₂, OH, N, and H [35]. Compared to the AS CRN results, much more consistent predictions are found for FS CRN. All three mechanisms predict the same non-monotonic behavior with pressure increasing. The differences between predicted NO concentrations are lesser than 10 ppm for this condition. For the temperature prediction, differences between chemistry mechanisms are lesser than 2 K. These results show that there exist uncertainties in high-pressure reaction rate, especially for the reactions involved in AS combustion. However, the computational results also confirm that these two NO_x reduction strategies are applicable to a high-pressure environment. At 20 bar, ${\tau }_{PFR}$ for the activation of NO reduction is 0.001 and 0.3 s for AS combustion and FS combustion, respectively. These are values relevant to the operating conditions of practical gas turbine applications.

4. Conclusions

In this study, we systematically investigate the combustion of the NH₃/H₂/N₂ mixture using a staged combustion CRN. A wide range of CRN parameters is considered to find the effective regions of NO_x reduction for air-staged and fuel-staged combustion. Key findings can be summarized as follows.

• Both air- and fuel-staged combustion strategies can be viable solutions for NO reduction in an ammonia combustion system. However, the optimal operation points of these two-stage combustion strategies are distinctively different. The air-staged combustion can be utilized with a fuel-rich primary combustion zone of long residence time. The residence time of the secondary combustion zone for unburned H₂ and NH_x should be short enough to prevent the mixture reaches chemical equilibrium. On the contrary, fuel-staged combustion is an effective way of NO reduction for a fuel-lean primary combustion zone. For this strategy, the secondary combustion zone with a long residence time is required to activate the thermal DeNO_x process. Moreover, the temperature of the secondary combustion zone should be low enough to prevent an abrupt NO formation.
• Considering the predicted concentrations of NO, N₂O, NO₂, and NH₃ from fuel-staged CRN, fuel-staged combustion could be a promising way of low NO_x ammonia combustion. Unrealistically long residence time required for low NO_x staging combustion indicates that low reactivity of NH₃ could be problematic for some real applications. However, at elevated temperature and pressure, the CRN model shows the reduction of NO_x with the residence time of O(1) ms.
• The cracked ammonia ratio essentially does not change the emission characteristics when the cracked ammonia ratio is not close to 1.0. However, the increment of the reactivity with ammonia cracking slightly increases the level of NO and decreases residence time for NH₃ oxidation.
• Non-linearity of NO formation in terms of reactor residence time, mixture composition, and staged injection, together with a narrow region of effective NO_x reduction, implies that accurate control of flow dynamic is necessary for a low NO_x ammonia combustion system.
• The three chemistry mechanisms of Okafor, Otomo, and Tian are used to estimate the uncertainties in simulation results. The differences in temperature, NH₃, NO₂, and N₂O predictions are negligible. However, they show qualitatively and quantitatively different predictions in terms of NO concentration, especially at high-pressure conditions. Although reaction rates in high-pressure simulations lead to large discrepancies in NO predictions, all three mechanisms demonstrate that both air- and fuel-staged combustion strategies are still applicable to high-pressure environments.