Abstract
There is significant interest in utilizing hydrogen as an energy carrier in a net-zero carbon energy economy. A variety of fundamental and practical issues associated with premixed hydrogen/methane combustion must be addressed, particularly flame flashback, combustion instabilities, and NOx emissions. Significantly less attention has been given to CO emissions from hydrogen blending. However, recent blending engine tests have clearly noted major reductions in CO emissions, reductions which substantially exceed what would be expected based upon C atom mass balances in the fuel. Moreover, NOx, CO, and combustor operability always tradeoff with each other, and so, for example, these CO reductions can also enable operational or design modifications to reduce NOx emissions and/or wider operability windows. This paper addresses blending effects on CO emissions, addressing both the direct influences on CO emissions, as well as how it influences NOx–CO tradeoffs. The paper presents computations that differentiate between the kinetic and equilibrium effects of blending, as well as how blending can lead to reductions in both NOx and CO by moving the Pareto front downward. Finally, this work suggests that blending, in addition to its reductions, can also increase plant operational flexibility and so provide additional value propositions to an evolving electricity market.
1 Introduction
Hydrogen is being evaluated as a zero-carbon energy carrier and energy storage medium for a variety of industrial, transportation, and electric power applications [1]. A variety of demonstrations and pilots are being explored, generally with variable levels of blending with natural gas [2,3]. Natural gas, which is composed primarily of methane, is a dominant source of electrical power, as well as for heat in other commercial and industrial applications. Key emissions of concern from combustion are usually carbon monoxide (CO), oxides of nitrogen (NOx), unburned hydrocarbons, and particulate matter (PM).
A number of recent studies have addressed pollutant emissions from / blends. Because nonpremixed -blended flames are hotter than nonpremixed blends, it is critical to carefully specify what is being held constant when such a comparison is being made. In particular, thermal NO production is a linear function of residence time and concentration, and an exponential function of temperature. For this reason, a number of early studies of / blends reported significant increases in NO emissions [2,4–7] with blending, likely due to elevated flame temperatures leading to higher thermal NO formation rates. However, when comparing / blends at constant temperature, recent work from our group has shown that roughly constant NO levels are produced, with deviations due to product O atom concentrations and fuel-dependent roles of flame NO production mechanisms [8].
NOx and CO tradeoffs play one of the dominant roles in how modern combustors are designed, as well as their operational limits. For example, CO exists at superequilibrium levels in the flame, and the rate at which CO relaxes to equilibrium increases with temperature and pressure. Only if CO levels are at equilibrium does increasing temperature lead to increased CO (due to dissociating back to CO). In general, however, CO emissions are reduced with increasing pressure, temperature, and residence time. NO emissions generally exhibit the opposite tendencies. Longer residence times lead to higher NO formation and higher temperatures/pressures to elevated formation rates. As such, combustor design and operational limits are generally associated with balancing NO at high power/temperature and CO emissions at low power/temperature. Indeed, the fact that CO emissions increase rapidly with decreases in flame temperature has huge ramifications on the global gas turbine fleet operational characteristics, as these engines can only be turned down to a minimum power level (often on the order of 60 max power) without going out of compliance on their CO permit. Typical regulations (such as the U.S. Environmental Protection Agency, EPA) are defined with respect to output-based standards (e.g., the mass of NOx/CO per unit of useful shaft work), which for a given fuel can be converted to the more easily measured concentration-based standard (in units of parts per million by volume at 15 oxygen) [9]. The relationship between normalized mass output standards and concentration is itself fuel dependent, and so output-based standards must be used when comparing emissions across different fuels, which becomes particularly important for -blended fuels [10].
Even while both NOx and CO have such important influences on how combustors are both designed and operated, their actual health and environmental impacts are quite different. For reference, the electric power sector contributed about 1 of total U.S. CO emissions in 2023 [11]. Moreover, CO is not a driver of new regulations because the entire U.S. is currently in attainment of the CO National Ambient Air Quality Standard (NAAQS). is a regulated criteria pollutant because of its potential to cause adverse respiratory health effects and because it contributes to acid deposition. Though all areas in the U.S. are currently meeting the air quality standards for , it is still a concern because it is a precursor to two other criteria pollutants, ozone, and fine particulate matter (PM2.5). As of Mar. 31, 2024, there are 46 areas in the U.S. with 115 million people which fail to meet the 2015 NAAQS for ozone [12]. Additionally, there are five areas in the U.S. with 21 million people in nonattainment of the 2012 PM2.5 NAAQS and 11 areas with 31 million people in nonattainment of the 2006 PM2.5 NAAQS. On Feb. 7, 2024, EPA issued a new PM2.5 NAAQS, and nonattainment area designations are pending for the new annual standard of 9 g .
As might be expected, blending should reduce CO emissions from flames due to simple mass balance considerations, as there are fewer carbon atoms in the fuel. To illustrate, Fig. 1 quantifies the reduction by showing a calculation of the equilibrium CO emissions produced from / blends at constant adiabatic flame temperature. The nonzero CO emissions at 100 come from atmospheric .
However, field data from / blending testing show dramatically larger, even orders of magnitude larger, CO reductions than illustrated by Fig. 1 [13]. At a field test of a NYPA Brentwood GE LM6000 test, there was a 70 reduction in CO as hydrogen increased from 0% to 35 hydrogen by volume blend, which corresponds to an approximately 20 decrease in emissions. This was measured while NOx was held constant through an increase in water to fuel ratio, which tends to increase CO emissions. At the Southern Company Mitsubishi 501G testing, baseload CO emissions remained below the analyzer detection limit at the high baseload firing temperature from 0% to 20 blending volume. At 50 load, controlled to a lower firing temperature, CO emissions were reduced between 85 and 98 for an increase of hydrogen from 0% to 20 by volume, which corresponds to an approximately 7 decrease in emissions [14,15].
These observations are a key motivator behind this paper. These CO reductions could be due to two additional factors, beyond the obvious reduction in available carbon atoms with increased blending. First, blending can accelerate CO oxidation to . For example, the important reaction CO + OH → + H would certainly be impacted by altered levels of both OH and H in flames with blending. Second, blending is well known to increase turbulent flame speeds, thus making the flame shorter and thereby increasing postflame residence times for CO oxidation.
The purpose of this paper is to further explore and quantify this first effect, namely, the kinetic influences of – blending on CO emissions. In addition to their fundamental interest, this relationship will have important impacts on both the operation of existing combustors and the design of future ones. For example, a significant number of gas turbines in the U.S. are operated intermittently, shutting down and turning back on once or even twice a day, as their operational window is limited by low-load CO emissions. blending thereby could expand operational envelopes and increase their utilization factors. Second, new dedicated designs for elevated levels could operate with shorter residence times than current systems, thereby enabling reductions in NO emissions for a given CO emission. This paper presents the results of computations to provide further insight into these questions and to understand the role of on CO relaxation rates, with attention given to fundamental influences of pressure, temperature, residence time, and blending levels.
2 Numerical Methods
2.1 Test Matrix.
A test matrix was developed to individually study the effects of fuel composition, adiabatic flame temperature, and pressure conditions. The varying pressure cases were run with different preheat temperature values to simulate different applications and/or power levels of a system. The test matrix was developed by varying equivalence ratios for three power conditions to achieve constant product temperatures across fuel blends. Adiabatic flame temperatures of 1600–2000 K were analyzed across three initial power conditions: “low” power (1 bar, 300 K), “medium” power (5 bar, 650 K), and “high” power (20 bar, 800 K). The low and high-power conditions were chosen to correspond to “idle” and “full power” of an F-class industrial gas turbine, respectively. Eight hydrogen blends from 0% to 100 by volume were studied. Also, note that adiabatic flame temperatures were held constant to ±6 K of the quoted values. The full test matrix, where the value in the table indicates the equivalence ratio, is shown in Table 1. In order to capture CO emissions at the 100 cases, the simulated air includes 412 ppm of atmospheric [16].
fuel blend | |||||||||
---|---|---|---|---|---|---|---|---|---|
0 | 25 | 50 | 75 | 90 | 95 | 99 | 100 | ||
1 bar, 300 K | 1600 | 0.561 | 0.553 | 0.542 | 0.525 | 0.507 | 0.499 | 0.489 | 0.487 |
1700 | 0.616 | 0.608 | 0.596 | 0.576 | 0.559 | 0.549 | 0.538 | 0.535 | |
1800 | 0.672 | 0.665 | 0.653 | 0.632 | 0.61 | 0.601 | 0.59 | 0.586 | |
1900 | 0.732 | 0.725 | 0.712 | 0.69 | 0.668 | 0.656 | 0.644 | 0.641 | |
2000 | 0.793 | 0.788 | 0.774 | 0.751 | 0.727 | 0.715 | 0.702 | 0.698 | |
5 bar, 650 K | 1600 | 0.413 | 0.407 | 0.399 | 0.384 | 0.368 | 0.361 | 0.355 | 0.352 |
1700 | 0.465 | 0.459 | 0.449 | 0.434 | 0.416 | 0.408 | 0.4 | 0.397 | |
1800 | 0.519 | 0.512 | 0.502 | 0.484 | 0.465 | 0.456 | 0.447 | 0.444 | |
1900 | 0.575 | 0.568 | 0.557 | 0.537 | 0.516 | 0.506 | 0.496 | 0.493 | |
2000 | 0.633 | 0.626 | 0.614 | 0.593 | 0.57 | 0.559 | 0.547 | 0.543 | |
20 bar, 800 K | 1600 | 0.346 | 0.342 | 0.333 | 0.321 | 0.308 | 0.301 | 0.295 | 0.292 |
1700 | 0.396 | 0.392 | 0.382 | 0.368 | 0.353 | 0.346 | 0.338 | 0.336 | |
1800 | 0.45 | 0.443 | 0.434 | 0.418 | 0.4 | 0.394 | 0.387 | 0.384 | |
1900 | 0.503 | 0.496 | 0.487 | 0.469 | 0.451 | 0.444 | 0.435 | 0.431 | |
2000 | 0.561 | 0.552 | 0.542 | 0.523 | 0.5072 | 0.494 | 0.484 | 0.485 |
fuel blend | |||||||||
---|---|---|---|---|---|---|---|---|---|
0 | 25 | 50 | 75 | 90 | 95 | 99 | 100 | ||
1 bar, 300 K | 1600 | 0.561 | 0.553 | 0.542 | 0.525 | 0.507 | 0.499 | 0.489 | 0.487 |
1700 | 0.616 | 0.608 | 0.596 | 0.576 | 0.559 | 0.549 | 0.538 | 0.535 | |
1800 | 0.672 | 0.665 | 0.653 | 0.632 | 0.61 | 0.601 | 0.59 | 0.586 | |
1900 | 0.732 | 0.725 | 0.712 | 0.69 | 0.668 | 0.656 | 0.644 | 0.641 | |
2000 | 0.793 | 0.788 | 0.774 | 0.751 | 0.727 | 0.715 | 0.702 | 0.698 | |
5 bar, 650 K | 1600 | 0.413 | 0.407 | 0.399 | 0.384 | 0.368 | 0.361 | 0.355 | 0.352 |
1700 | 0.465 | 0.459 | 0.449 | 0.434 | 0.416 | 0.408 | 0.4 | 0.397 | |
1800 | 0.519 | 0.512 | 0.502 | 0.484 | 0.465 | 0.456 | 0.447 | 0.444 | |
1900 | 0.575 | 0.568 | 0.557 | 0.537 | 0.516 | 0.506 | 0.496 | 0.493 | |
2000 | 0.633 | 0.626 | 0.614 | 0.593 | 0.57 | 0.559 | 0.547 | 0.543 | |
20 bar, 800 K | 1600 | 0.346 | 0.342 | 0.333 | 0.321 | 0.308 | 0.301 | 0.295 | 0.292 |
1700 | 0.396 | 0.392 | 0.382 | 0.368 | 0.353 | 0.346 | 0.338 | 0.336 | |
1800 | 0.45 | 0.443 | 0.434 | 0.418 | 0.4 | 0.394 | 0.387 | 0.384 | |
1900 | 0.503 | 0.496 | 0.487 | 0.469 | 0.451 | 0.444 | 0.435 | 0.431 | |
2000 | 0.561 | 0.552 | 0.542 | 0.523 | 0.5072 | 0.494 | 0.484 | 0.485 |
2.2 Kinetics Calculations.
In this study, a kinetic model of a laminar, one-dimensional premixed flame was set up in ansyschemkin using the premix package [17]. The calculations were performed over sufficient axial distance to capture in-flame and postflame processes. The grid independence study performed by Breer et al. [8] was used to extend the same numerical results across the analysis of CO emissions. The mechanism implemented in this study is a HyChem kinetic model [18]. The HyChem model was chosen following the comparison conducted by Breer et al. [8] between HyChem, GRI 3.0 [19], Glarborg [20], and UCSD [21] models. The test matrix shown in Table 1 was replicated in chemkin using a parametrized sweep across the fifteen cases corresponding to five cases at each power condition. Grid independence studies were performed to confirm appropriate gridding across the cases.
2.3 Data Processing.
Here, the emissions are quantified using two different metrics—mass of emissions produced per energy released by fuel using lower heating value of the fuel, and in terms of parts-per-million by volume, dry (ppmvd) corrected to 15. Because the conversion between these two quantities is concentration dependent [10], the ppm values were only calculated for the 0 case (if not, two lines would be needed for each hydrogen blending value). This is done in order to plot the data in both units simultaneously and provide a benchmark comparison at 0. The purpose of including these concentration values on many of the secondary y-axes is that many emissions regulations are written in terms of concentrations, and many analyzers report measured values using these concentration units. However, it is emphasized that only the mass normalized metric should be used when comparing CO production values across different blends as the ppm values presented are only relevant to the 0 case.
This factor weighs the CO concentration by the ratio of concentrations for the / blended mixture and pure . As such, variations in levels with blending can be interpreted as a kinetic effect of blending. If this ratio does not change with blending, then variations in CO emissions with blending are purely due to a variation in C atom mass in the reactants.
3 Results and Discussion
3.1 Baseline CO Trends.
Figure 2 plots typical CO trends for . It shows the initial rise in CO, its peak to superequilibrium levels on the order of ppm, and its relaxation toward equilibrium. While the peak superequilibrium CO levels are not strongly temperature dependent, increases in flame temperature make the flame thinner (as manifested by the width of the CO peak), increase the relaxation rate of CO toward equilibrium, and increase equilibrium concentration. While results are not shown for multiple pressure/power levels, increasing pressure decreases flame thickness, increases relaxation rates toward equilibrium, and decreases equilibrium CO levels.
Consider next Fig. 3, which illustrates blending effects. Increasing levels decreases flame thickness and accelerates relaxation rates toward equilibrium, causing decreases in CO emissions at all residence time values.
Having considered baseline trends, next are the more detailed characterizations. There are several useful ways to quantify and analyze this problem. These include characterizing blending effects on (1) CO and emissions at a fixed residence time, and (2) residence time required to achieve a given CO values. Furthermore, given the intimate tradeoff between CO and NO emissions in combustor design, (3) examines NO emissions at fixed CO values. The rest of this paper summarizes these different representations in turn.
3.2 CO Emissions at Fixed Residence Time.
Start by considering CO emissions as a function of blending at different residence times values. Figure 4 plots the low-power case, including the equilibrium CO levels, at residence times of 5, 15, and 25 ms. The figure shows the monotonically decreasing equilibrium CO levels with increased blending noted earlier. This reduction is due to both kinetic effects and equilibrium effects. For example, the 15 ms case has essentially reached equilibrium at this condition, and so the reduction is entirely an equilibrium effect due to reduced C atoms in the fuel. In contrast, the 5 ms case is a mix of both effects.
Similar results are observed at the other power conditions, with the key difference being that the medium power case reaches equilibrium in 3 ms and the high-power case in 1.9 ms. Figure 5 replots this data, normalizing CO emissions by their equilibrium value at that blending composition. This normalization allows us to differentiate the kinetic effects of blending from the fact that the overall C atom balance is also varying. If this ratio does not vary with concentration, then the variations shown above in absolute CO concentrations are purely due to C atom balance considerations. If this ratio decreases with concentration, then adding also causes CO reductions by accelerating the oxidation of CO to . The figure shows that this effect is particularly prominent at the 5 ms condition, where CO emissions are about six times their equilibrium value at 0% H2. In contrast, at 90 blending, the CO emissions are 1.3 times above their equilibrium value. This CO/ ratio decreases with residence time, as the products approach equilibrium. Similar behavior is observed at the higher power conditions at lower residence times. Figure 6 illustrates a similar point by plotting both CO and emissions as a function of blending. Note that both curves have the same value at 0 by definition. The two curves are almost on top of each other at the lowest residence time case, 5 ms, indicating that the observed CO reductions are almost entirely due to the kinetic effect of addition. In contrast, the curve is almost flat at 15 ms, even while CO emissions fall, indicating that the observed CO reduction is simply manifesting as a result of the drop in C atom mass in the reactants.
3.3 Residence Time Required to Achieve a Given CO Level.
Consider next the effect of blending on the residence time required to achieve a given CO value. Figure 7 plots the pressure scaled time () required to achieve CO levels of 1.1 of the pure case as a function of blending. The value of 1.1 here is arbitrary and only influences the results for pure , as the time to reach this equilibrium value is infinite. Picking some representative values, this required residence time is about 15 ms for the 1800 K low-power case with 100, that drops to about 9 ms with 50 and to about 2 ms with 90. This ratio drops by about a factor of 8 for low hydrogen content at the high-power condition, scaling almost linearly with pressure (for 90 it is approximately four times lower) as seen in the pressure scaled representation of the . In other words, the above values become about 1.9, 1.1, and 0.5 ms with 0%, 50%, and 90. These shorter residence times become particularly interesting when one begins to consider combustor design options for minimizing NOx and CO, as discussed in Sec. 3.4. We do not show results for the 95 and higher cases as their highest CO peaks are lower than the equilibrium CO in the 0 case.
3.4 Interrelationship Between NO and CO Emissions.
Prior studies have examined blending effects on NO emissions at a fixed residence time [8]. However, combustion systems optimized for elevated levels can actually reduce NO emissions if residence times are reduced, which could potentially be done since CO levels drop with increased blending. Such an option is an interesting regulatory option, whereby regulators could decrease NO emissions of the fleet by allowing designers to develop systems optimized for a certain fraction. This section provides several visualizations of the relationship between NO and CO.
Figure 8 plots the relationship between NO and CO at the low-power condition, where is a parameter. Iso- values of 1.5 to 20 ms are also plotted in the figure. The different lines denote increased blending. For a given fraction, the figure shows the well-known tradeoff between NO and CO. In other words, increasing residence times decreases CO emissions but leads to increased NO emissions. Each of the curves shown in Fig. 8 represents a “Pareto front” or set of optimal solutions in a multi-objective problem for a fixed fuel blend. The figure also shows the monotonic shifting of the Pareto front with increased blending, i.e., whereas the classical design tradeoff is one of balancing NO and CO, addition allows one to decrease both. For example, at a fixed CO value, increasing blending leads to reductions in NO. Similarly, at a fixed NO value, increasing blending decreases CO. Similar points can be seen at the high-power condition plotted in Fig. 9, where the calculations have been run out to sufficient that CO has reached equilibrium by about 2 ms and does not change further, even while NO values monotonically rise with increased residence time. Figures 8 and 9 compare NO and CO under the same operating conditions. However, combustor design typically requires finding a residence time that optimizes CO at low power with NO at high-power conditions. To understand how blending influences this tradeoff, Fig. 10 plots NO values at the high-power condition (800 K, 20 bar) as a function of CO emissions at the low-power condition (300 K, 1 bar) at the same residence times for = 1800 K. This plot better illustrates the tradeoff in balancing NO at high temperature conditions, with CO emissions at low temperature conditions for a system designed with a given residence time. The figure again illustrates that addition can provide benefits in NO, CO, or both by shifting the Pareto front down. This figure also demonstrates how addition can enable increased turndown capabilities for gas turbines, by maintaining a fixed NO value at full power but being able to run to lower power conditions while maintaining the same CO levels.
4 Conclusion
One of the most important tradeoffs in combustor design is balancing CO and NOx emissions, which generally have opposite tendencies. There has been significant attention to the question of how blending will influence NOx emissions. This study has shown the important implications of blending on CO emissions as well as how simultaneous consideration of NOx and CO can lead to a reconceptualization of both design and operational spaces. At a fixed residence time, addition monotonically decreases CO emissions, due to both reduction in C atoms as well as kinetic effects. Similarly, addition increases the rate at which superequilibrium CO formed in the flame relaxes down to postflame equilibrium levels. Finally, considering the NOx–CO tradeoff, addition can lead to simultaneous reductions in both. These results have interesting implications for the utilization of not only as a means for reducing the carbon intensity of gas turbine power but also expanding its operational flexibility.
Acknowledgment
This study has been partially supported by the Electric Power Research Institute (EPRI) and LCRI project titled “Investigation for Hydrogen GT DLN NOx Entitlement Curve,” and the Department of Energy (Grant No. DE-FE0032079).
Data Availability Statement
The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.