Effects of Operating Parameters on Combustion Characteristics of Hydrogen-Doped Natural Gas

Abstract

The operational optimization of industrial boilers utilizing hydrogen-enriched natural gas is constrained by two critical gaps: insufficient understanding of the coupled effects of hydrogen blending ratio, equivalence ratio, and boiler load on combustion performance—compounded by unresolved challenges of combustion instability, flashback, and elevated NO_x emissions—and a lack of systematic investigations combining these parameters in industrial-scale systems (prior studies often focus on single variables like hydrogen fraction). To address this, a comprehensive computational fluid dynamics (CFD) analysis was conducted on a 2.1 MW industrial boiler, employing the Steady Laminar Flamelet Model (SLFM) with a modified k-ε turbulence model and the GRI-Mech 3.0 mechanism. Simulations covered hydrogen fractions (f(H₂) = 0–25%), equivalence ratios (Φ = 0.8–1.2), and load conditions (15–100%). All NO_x emissions reported herein are normalized to 3.5% O₂ (mg/Nm³) for regulatory comparison. Results show that increasing the hydrogen content raises the flame temperature and NO_x emissions while reducing CO and unburned hydrocarbons; a higher equivalence ratio elevates temperature and NO_x, with Φ = 0.8 balancing efficiency and emission control; and reducing load significantly lowers furnace temperature and NO emissions. Notably, the boiler’s unique staged-combustion configuration (81% fuel supply to the central rich-combustion nozzle, 19% to the concentric lean-combustion nozzle) was found to mitigate NO_x formation by 15–20% compared to single-inlet burner designs, and its integrated cyclone blades (generating maximum swirling velocity of 14.2 m/s at full load) enhanced fuel–air mixing, which became particularly critical for maintaining combustion stability at low loads (≤20%) and high hydrogen blending ratios (≥20%). This study provides quantitative trade-off insights between combustion efficiency and pollutant formation, offering actionable guidance for the safe, efficient operation of hydrogen-enriched natural gas in industrial boilers.

1. Introduction

The decarbonization of industrial energy systems has positioned hydrogen-blended natural gas as a pivotal transitional fuel. However, its deployment in existing boiler infrastructure faces significant challenges related to combustion stability, safety, and elevated nitrogen oxide (NO_x) emissions. Industrial-scale incidents underscore these risks: hydrogen must be manipulated in safe and reliable circumstances that will restrict equipment contact with H₂ and satisfy certain conditions to avoid leaks, ignition, and possible explosion [1]. Meanwhile, the combustion of CH₄ and H₂ mixtures demonstrated an increase in H₂ proportion correlated with an increase in NO emissions [2]. These findings demonstrate the urgent need for optimized combustion strategies, as well as a critical disconnect between laboratory-scale research and industrial operational demands.

Current research on hydrogen–methane combustion in boilers exhibits limitations that hinder direct industrial translation. While numerous studies have investigated the hydrogen-blending ratio, they often neglect its interaction with other critical operational parameters and the boiler-specific design characteristics that dictate real-world performance. For instance, Huang et al. [3] investigated the effect of different equivalence ratio (Φ) variations, from 0.6 to 0.8, on pure methane combustion through calculations and experiments without considering hydrogen addition. Cheng et al. [4] analyzed the combustion characteristics of hydrogen-doped natural gas in industrial boilers by conducting single-variable experiments under varying hydrogen doping ratios and excess air coefficients; however, they neglected the synergistic effects between these parameters and load. Zhang et al. [5] analyzed the effect of hydrogen blending ratios on the chemical kinetics of CH₄-air mixtures using CHEMKIN software, with hydrogen blending ratios ranging from 0 to 25%. This oversight of hydrogen fraction-Φ-load coupling effects and boiler-specific design influences represents a critical knowledge gap, as industrial boilers frequently operate under dynamic conditions with 15–100% load fluctuations and variable fuel compositions.

Accurate prediction of turbulent combustion in H₂/CH₄ mixtures requires careful model selection that accounts for hydrogen’s distinctive properties and the boiler-specific flow and mixing characteristics. The probability density function (PDF) transport model, while theoretically advantageous for capturing turbulence-chemistry interactions, exhibits significant limitations when applied to hydrogen-enriched fuels. This shortcoming stems primarily from the model’s inherent equal-diffusivity assumption, which can lead to substantial inaccuracies—such as those observed in NO_x predictions [6]. In contrast, the modified k-ε turbulence model has demonstrated superior performance for industrial-scale applications and temperature fields-outperforming RNG k-ε (9–11% error) and Reynolds stress models (10–13% error) while maintaining computational efficiency (1200 core-hours versus 2400 core-hours for LES) [7,8,9].

For chemical kinetics and turbulence–chemistry coupling models, selection was guided by industrial relevance, validation rigor, and compatibility with the boiler’s operational parameters. For chemical kinetics and turbulence–chemistry coupling models, selection was guided by industrial relevance and validation rigor and compatibility with the boiler’s operational parameters. The steady laminar flamelet model (SLFM) [10] was adopted based on its proven accuracy in 2.1 MW CH₄/H₂ boiler simulations, demonstrating errors below 7% for temperature and 9% for species concentration. However, its quasi-steady state assumption leads to measurable limitations: in transient conditions with load fluctuations exceeding 20%, SLFM predictions of flame front position deviate by over 15% from experimental data [10]. For chemical kinetics, while modern mechanisms like HyChem [11] and AramcoMech 2.0 [12] offer improved accuracy for hydrogen combustion, this study employs GRI-Mech 3.0 to maintain direct comparability with approximately 80% of prior industrial boiler studies [4,13]. We explicitly recognize this as a limitation and plan subsequent validation with HyChem in future work.

The validated modeling framework enables systematic analysis of NO_x formation mechanisms, which exhibit complex dependencies on operational parameters and boiler-specific design features. Thermal–NO_x, governed by the Zeldovich mechanism via core reactions (1) N₂ + O ↔ NO + N and (2) N + O₂ ↔ NO + O, [14], increases significantly with hydrogen addition due to elevated flame temperatures, showing a 2.3-fold increase in the formation rate as the hydrogen fraction rises from 0 to 25% at Φ = 1.0. In contrast, prompt NO_x production decreases by 25–30% at 25% hydrogen blending, attributed to a 40% reduction in CH radical concentration, which limits fuel-N conversion [15]. The N₂O-intermediate pathway, involving reactions N₂ + OH ↔ N₂O + H and N₂O + O ↔ 2NO, demonstrates particular relevance under lean, low-temperature conditions: it accounts for 35% of total NO_x at Φ = 0.8 and 15% load but less than 10% at Φ = 1.0 and full load. These pathway-specific trends emphasize the necessity of multi-parameter analysis for accurate NO_x prediction.

Global research initiatives highlight the industrial imperative for understanding such parameter-coupling effects. While these large-scale demonstration projects validate the technical feasibility of hydrogen blending (up to 30%, H₂), they lack the systematic parametric analysis needed for operational optimization across diverse conditions [16]. Notably, material–hydrogen interactions, such as hydrogen embrittlement, though critical for long-term equipment durability, require multi-physics simulations integrating solid mechanics and combustion kinetics, which will be addressed in future works [17]. Thermoacoustic instability and flame oscillations, meanwhile, become significant only at hydrogen fractions exceeding 30% [18], placing them beyond the scope of this study’s 0–25% blending range—relevant to near-term industrial implementation.

This study addresses these research gaps through the first comprehensive CFD investigation of coupled hydrogen blending ratio (0–25%), equivalence ratio (0.8–1.2), and load (15–100%) effects in a 2.1 MW industrial boiler. The primary novelty lies in: (1) establishing quantitative relationships between multiple operational parameters and NO_x formation pathways in an industrially relevant geometry; (2) providing validated data for operational optimization under low-load conditions critical for peak shaving; and (3) developing a modeling framework that balances computational accuracy with practical applicability for industrial boiler design. Furthermore, the analysis specifically accounts for the boiler’s staged-combustion burner design, linking its flow and mixing characteristics to the observed global performance and emission trends. This work advances beyond prior CFD studies by capturing parameter-coupling effects that directly impact compliance with stringent emission standards operational safety.

2. Modeling and Numerical Methods

2.1. Physical Model and Computational Domain

This study investigates a YY(Q)W-2000Y(Q) organic heat carrier boiler (Hebei Yineng Boiler Co., Ltd., Cangzhou, Hebei, China) with a rated thermal input of 2.1 MW. To balance computational accuracy and efficiency, the computational domain was simplified to its core cylindrical combustion chamber (diameter: 1200 mm, height: 4980 mm). The downstream convective tube bundle was excluded, as its flue gas temperature remains below 600 K, exerting a negligible influence on combustion processes and pollutant formation.

The burner employs a staged combustion design: a central nozzle (Fuel-inlet 1, flow area: 923.628 mm²) for rich combustion and a concentric secondary nozzle (Fuel-inlet 2, flow area: 1017.876 mm²) for lean combustion (

Table 1. Boundary conditions for numerical simulation of hydrogen-enriched combustion.

	Areas/mm2	Hydraulic Diameters/m
Fuel-inlet1	923.628	0.014
Fuel-inlet 2	1017.876	0.036
Air-inlet	48,559.598	0.152
Outlet	890,818.232	1.065

). Fuel is distributed based on the effective flow area, with 81% and 19% of the total fuel supplied to Fuel-inlet 1 and Fuel-inlet 2, respectively. This configuration has been demonstrated to reduce NO_x emissions by 15–20% compared to single-inlet designs [19,20]. Cyclone blades integrated at the air inlet generate a swirling flow (maximum velocity: 14.2 m/s at 100% load) to enhance mixing (Figure 1).

2.2. Numerical Methods, Model Validation, and Reliability

All simulations were conducted using the commercial CFD software ANSYS Fluent 2023 R1. A quantitative comparison of turbulence models was performed to justify the selection of the modified k-ε model, addressing its limitations for capturing detailed turbulent flame dynamics. The comparison, based on data from literature [8,9] (

Table 2. Performance comparison of turbulence models for industrial-scale H₂/CH₄ combustion [8,9].

Model	Velocity Prediction Error	Temperature Prediction	Computational Time (Core-Hours/Case)
Modified k-ε	6–8%	6–8%	1200
RNG k-ε	9–11%	8–10%	1300
RSM	10–13%	9–12%	2000
LES	4–6%	4–6%	2400

), is summarized below:

While the modified k-ε model cannot capture transient phenomena (e.g., flame flickering) due to the Boussinesq isotropic assumption—which may underestimate turbulent mixing in high-swirl regions—it was selected for this study because it provides an optimal balance between accuracy and computational feasibility for steady-state parametric analysis. Given the scope of 90 simulation cases, employing LES would be computationally prohibitive (90 cases × 2400 core-hours = 216,000 core-hours), whereas the modified k-ε model (90 × 1200 = 108,000 core-hours) remains practical while delivering acceptable accuracy for the comparative analysis intended [8,9]. The modified constant C_ε1 = 1.6 in the turbulent k-ε model was determined through sensitivity analysis. A well-established detailed reaction mechanism, GRI-Mech 3.0, was adopted for chemical kinetics modeling, encompassing 325 elementary reactions and 53 chemical species [21].

2.3. Boundary Conditions and Numerical Setup

The fuel and air inlets were defined as velocity inlets, and the outlet was set as a pressure outlet. Walls were modeled using a constant temperature boundary condition. Pressure–velocity coupling was handled using the SIMPLE algorithm, and pressure interpolation using the PRESTO! Scheme. The momentum and energy equations were discretized using a first-order upwind scheme, while all other equations employed a second-order upwind scheme, balancing stability and accuracy. The convergence criteria for all equations were set to residual values below 10⁻⁶ for energy and below 10⁻⁵ for other variables, ensuring numerical accuracy.

To comprehensively investigate the coupled effects of hydrogen-blending ratio, equivalence ratio, and load, multiple operating conditions were systematically simulated, as detailed in

Table 3. Combustion boundary conditions.

Boiler Load	Φ	f(H2)	Air Flow (m3/h)	Air-Inlet Flow Rate (m/s)	Flue-Inlet1 Flow Rate (m/s)	Flue-Inlet2 Flow Rate (m/s)
100	0.8	0	2489.00	14.24	51.16	10.89
		5	2482.36	14.20	53.01	11.28
		10	2475.22	14.16	54.99	11.70
		15	2467.52	14.12	57.13	12.16
		20	2459.20	14.07	59.45	12.65
		25	2450.18	14.02	61.96	13.19
	1.0	0	1991.20	11.39	51.16	10.89
		5	1985.89	11.36	53.01	11.28
		10	1980.17	11.33	54.99	11.70
		15	1974.02	11.29	57.13	12.16
		20	1967.36	11.25	59.45	12.65
		25	1960.14	11.21	61.96	13.19
	1.2	0	1659.33	9.49	51.16	10.89
		5	1654.91	9.47	53.01	11.28
		10	1650.15	9.44	54.99	11.70
		15	1645.01	9.41	57.13	12.16
		20	1639.47	9.38	59.45	12.65
		25	1633.45	9.34	61.96	13.19
80	0.8	0	1991.20	11.39	40.93	8.71
		5	1985.89	11.36	42.40	9.03
		10	1980.17	11.33	43.99	9.36
		15	1974.02	11.29	45.71	9.73
		20	1967.36	11.25	47.56	10.12
		25	1960.14	11.21	49.57	10.55
	1.0	0	1592.96	9.11	40.93	8.71
		5	1588.71	9.09	42.40	9.03
		10	1584.14	9.06	43.99	9.36
		15	1579.21	9.03	45.71	9.73
		20	1573.89	9.00	47.56	10.12
		25	1568.11	8.97	49.57	10.55
	1.2	0	1327.47	7.59	40.93	8.71
		5	1323.92	7.57	42.40	9.03
		10	1320.12	7.55	43.99	9.36
		15	1316.01	7.53	45.71	9.73
		20	1311.57	7.50	47.56	10.12
		25	1306.76	7.48	49.57	10.55
50	0.8	0	1244.50	7.12	25.58	5.44
		5	1241.18	7.10	26.50	5.64
		10	1237.61	7.08	27.50	5.85
		15	1233.76	7.06	28.57	6.08
		20	1229.60	7.03	29.72	6.33
		25	1225.09	7.01	30.98	6.59
	1.0	0	995.60	5.70	25.58	5.44
		5	992.94	5.68	26.50	5.64
		10	990.09	5.66	27.50	5.85
		15	987.01	5.65	28.57	6.08
		20	983.68	5.63	29.72	6.33
		25	980.07	5.61	30.98	6.59
	1.2	0	829.67	4.75	25.58	5.44
		5	827.45	4.73	26.50	5.64
		10	825.07	4.72	27.50	5.85
		15	822.51	4.71	28.57	6.08
		20	819.73	4.69	29.72	6.33
		25	816.73	4.67	30.98	6.59
20	0.8	0	497.80	2.85	10.23	6.01
		5	496.47	2.84	10.60	6.23
		10	495.04	2.83	11.00	6.46
		15	493.50	2.82	11.43	6.71
		20	491.84	2.81	11.89	6.98
		25	490.04	2.80	12.39	7.28
	1.0	0	398.24	2.28	10.23	6.01
		5	397.18	2.27	10.60	6.23
		10	396.03	2.27	11.00	6.46
		15	394.80	2.26	11.43	6.71
		20	393.47	2.25	11.89	6.98
		25	392.03	2.24	12.39	7.28
	1.2	0	331.87	1.90	10.23	6.01
		5	330.98	1.89	10.60	6.23
		10	330.03	1.89	11.00	6.46
		15	329.00	1.88	11.43	6.71
		20	327.89	1.88	11.89	6.98
		25	326.69	1.87	12.39	7.28
15	0.8	0	373.35	2.14	7.67	4.51
		5	372.35	2.13	7.95	4.67
		10	371.28	2.12	8.25	4.84
		15	370.13	2.12	8.57	5.03
		20	368.88	2.11	8.92	5.24
		25	367.53	2.10	9.29	5.46
	1.0	0	298.68	1.71	7.67	4.51
		5	297.88	1.70	7.95	4.67
		10	297.03	1.70	8.25	4.84
		15	296.10	1.69	8.57	5.03
		20	295.10	1.69	8.92	5.24
		25	294.02	1.68	9.29	5.46
	1.2	0	248.90	1.42	7.67	4.51
		5	248.24	1.42	7.95	4.67
		10	247.52	1.42	8.25	4.84
		15	246.75	1.41	8.57	5.03
		20	245.92	1.41	8.92	5.24
		25	245.02	1.40	9.29	5.46

. Table 3 summarizes key boundary conditions, including air flow rates and inlet velocities, for different loads (100%, 80%, 50%, 20%, 15%), equivalence ratios (Φ = 0.8, 1.0, 1.2), and hydrogen-blending ratios (f(H₂) = 0–25%). The hydrogen-blending ratio f(H₂) at 300 K and 1 atm is defined as follows:

$$f\left(H_2\right)={\displaystyle \frac{V\left(H_2\right)}{V\left(H_2\right)+V\left({CH}_4\right)}}$$

(1)

where V(H₂) is the volume flow rate of hydrogen (m³/h); V(CH₄) is the volume flow rate of methane (m³/h); and f(H₂) is the hydrogen-doping ratio of natural gas (0~0.25). The fuel fractions under different f(H₂) conditions are summarized in

Table 4. Mixed fuel components.

Hydrogen-Enriched Natural Gas Components	f(H2)
	0.00	0.05	0.10	0.15	0.20	0.25
CH4 (%)	96.00	91.20	86.40	81.60	76.80	72.00
N2 (%)	2.00	1.90	1.80	1.70	1.60	1.50
C2H6 (%)	1.92	1.82	1.73	1.63	1.54	1.44
C3H8 (%)	0.08	0.08	0.07	0.07	0.06	0.06
H2 (%)	0.00	5.00	10.00	15.00	20.00	25.00

.

2.4. Meshing and Grid Independence Study

Unstructured polyhedral meshes were generated using Fluent Meshing. Local refinement was applied to critical regions like the burner to accurately resolve flow and combustion details (Figure 2). The minimum orthogonal quality of the mesh was better than 0.15, and the maximum skewness was below 0.7.

A grid independence study was conducted using three mesh densities (0.69 million, 0.85 million, and 1.13 million cells) to quantify and minimize numerical uncertainties associated with grid resolution. Grid dependency was assessed by comparing radial velocity profiles at an axial position of z = 500 mm (Figure 3) and key parameters like peak temperature.

Quantitative analysis revealed excellent agreement between the medium (0.85 million cells) and fine (1.13 million cells) grids. The maximum relative velocity error between them was less than 2% (

Table 5. Quantitative error comparison of key parameters.

Grid Density (Number of Cells)	Peak Temperature Error (%) (Relative to Fine Grid)	Radial Velocity Error (%) (Relative to Fine Grid)	NO Mole Fraction Error (%) (Relative to Fine Grid)	CO Mole Fraction Error (%) (Relative to Fine Grid)	Grid Density (Number of Cells)	Peak Temperature Error (%) (Relative to Fine Grid)
0.69 × 106	3.8	4.5	5.2	4.9	0.69 × 106	3.8
0.85 × 106	0.9	1.8	1.5	1.2	0.85 × 106	0.9
1.13 × 106 (Fine Grid)	0 (Reference)	0 (Reference)	0 (Reference)	0 (Reference)	1.13 × 106 (Fine Grid)	0 (Reference)

), and the peak temperature difference was below 1.0%. These errors fall within the acceptable tolerance for this study. Consequently, the grid configuration with 0.85 million cells was selected for all subsequent simulations, ensuring computational accuracy while maintaining efficiency.

2.5. Data Processing and Analysis

2.5.1. Data Acquisition and Pre-Processing

Raw simulation data (temperature, species mole fractions, velocity) were extracted from 100 uniformly distributed sampling points within the combustion chamber (5 axial positions, 20 radial positions). To improve the data quality, a Gaussian filter (σ = 0.05 m) was applied to remove high-frequency numerical fluctuations, and outliers were excluded using the Grubbs test (α = 0.05) [22].

2.5.2. Calculation of Key Results

Key quantitative results were derived from the simulation data using specific calculation methods. The peak temperature was determined as the maximum value recorded within the entire combustion chamber, while the average temperature at the outlet was obtained by computing the area-weighted average across the furnace outlet cross-section. For pollutant emissions, NO_x and CO levels were quantified as area-weighted average mole fractions at the outlet section. To eliminate the effect of excess air, the NO_x mole fractions were converted to mass concentration under a reference oxygen content of 3.5% (mg/Nm³, dry basis) [23].

2.5.3. Repeatability and Computational Conditions

To ensure repeatability, three independent simulations were performed for the benchmark condition (100% load, Φ = 0.8, f(H₂) = 20%). The coefficient of variation (CV) for all key parameters (temperature, NO_x, and CO) was less than 2%, meeting repeatability requirements. All simulations were performed on a workstation equipped with 120 Intel Xeon Gold 6348 cores and 512 GB RAM. The average computational time per case was approximately 20 h (until convergence criteria were met).

3. Simulation Results and Analysis

3.1. Effect of the Hydrogen-Doping Ratio on the Combustion Process of the Burner

The load levels selected for this study (15–100%) are based on actual operational data from industrial boilers, covering peak production, normal operation, partial load, and the minimum stable load for this boiler type. This study limits the range to 0–25% to reflect current natural gas infrastructure compatibility constraints and near-term industrial implementation targets.

Temperature is a critical parameter characterizing combustion reactions and directly affects pollutant concentrations. Consistent with the conclusions in the literature [19,24], as the hydrogen-blending ratio increases from 0 to 25%, the maximum temperature and the average outlet temperature increase approximately linearly, with the maximum combustion temperature rising by approximately 20 K (Figure 4 and Figure 5,

Table A1. Temperature at outlet, NO/CO concentrations.

Boiler Load	Φ	f(H2)	Outlet Concentration (mg/Nm3, 3.5% O2)
			Temperature	NO	CO
100	0.8	0	1073.81	42.7	34.4
		5	1074.37	45.1	34.2
		10	1074.57	46.8	34.1
		15	1074.21	47.4	33.8
		20	1075.05	49.1	33.7
		25	1075.96	52.8	642
	1.0	0	1083.60	199	638
		5	1085.45	204	633
		10	1087.30	217	629
		15	1089.07	249	624
		20	1091.05	298	619
		25	1094.94	463	3600
	1.2	0	1107.13	490	3570
		5	1108.15	509	3520
		10	1108.27	527	3480
		15	1108.59	572	3430
		20	1108.72	646	3380
		25	1108.59	45.3	36.4
80	0.8	0	1013.88	45.5	36.2
		5	1014.07	45.9	36
		10	1014.31	50.6	35.8
		15	1014.55	55.7	35.7
		20	1014.81	56.1	34.8
		25	1015.11	191	642
	1.0	0	1010.06	199	638
		5	1011.75	204	633
		10	1013.45	205	629
		15	1015.04	205	624
		20	1016.81	231	619
		25	1018.47	487	3600
	1.2	0	1032.57	490	3570
		5	1033.50	496	3520
		10	1033.55	499	3480
		15	1033.80	508	3430
		20	1033.86	520	3380
		25	1033.81	43	36.4
50	0.8	0	889.91	43.2	36.1
		5	890.32	44	36
		10	890.75	45.1	35.8
		15	891.16	45.8	35.7
		20	891.57	46	34.8
		25	891.97	174	641
	1.0	0	901.67	178	637
		5	901.65	180	633
		10	902.53	181	629
		15	903.50	183	624
		20	904.55	187	619
		25	906.11	516	3600
	1.2	0	896.61	529	3570
		5	897.56	534	3520
		10	898.05	535	3480
		15	898.96	546	3430
		20	899.89	549	3380
		25	900.82	15.9	1460
20	0.8	0	923.75	15.9	1430
		5	926.36	16.3	1410
		10	927.31	16.8	1400
		15	928.34	18.5	1390
		20	929.45	19.2	1370
		25	930.66	21.3	5740
	1.0	0	784.30	22.1	5680
		5	785.39	22.9	5610
		10	787.76	23.8	5540
		15	790.24	25	5460
		20	793.03	26.1	5370
		25	795.98	29.8	7440
	1.2	0	636.76	30.8	7380
		5	636.98	31.8	7320
		10	637.50	32.7	7240
		15	638.20	35	7160
		20	639.16	39.9	7070
		25	640.41	13.3	1460
15	0.8	0	923.75	14.9	1430
		5	926.36	15.7	1410
		10	927.31	19.2	1400
		15	928.34	23.5	1390
		20	929.45	24	1370
		25	930.66	21.8	5740
	1.0	0	784.27	24.3	5680
		5	785.37	25.5	5610
		10	787.74	27	5540
		15	790.22	28.9	5460
		20	793.01	30.4	5370
		25	795.95	33.8	7440
	1.2	0	613.68	35.3	7380
		5	615.35	36.4	7320
		10	618.42	37.4	7240
		15	621.67	38.3	7160
		20	625.31	39.7	7070
		25	629.20	46.1	34.4

). This trend is attributed to hydrogen’s higher adiabatic flame temperature and laminar flame speed compared to methane, leading to increased local flame temperature and a more compact flame structure [25,26].

From a combustion chemistry perspective, hydrogen addition not only increases temperature but also influences pollutant formation by altering the radical pool. The CO concentration initially increases, then decreases, and finally stabilizes along the furnace axis (Figure 6), with the peak concentration decreasing by over 30% as the hydrogen-blending ratio increases. This phenomenon can be explained by key reaction pathways: HCO + H ↔ H₂ + CO and OH + CO ↔ H + CO₂. Hydrogen addition increases the OH radical concentration, promoting CO consumption by enhancing the CO oxidation pathway while simultaneously reducing the conversion rate of CH₄-derived intermediates (e.g., HCO) to CO.

The axial variation trend in NO concentration is consistent with the literature [24], showing an increase with higher hydrogen blending ratios (Figure 7), with peak concentration rising by approximately 15%. This is primarily due to enhanced thermal-NO formation (Zeldovich mechanism: N₂ + O ↔ N + NO) promoted by the increased flame temperature. Although hydrogen addition suppresses the prompt-NO route (CH + N₂ ↔ HCN + N), the thermal-NO pathway prevails within the temperature range investigated in this study. When normalized to 3.5% O₂, the NO_x emissions at the furnace outlet increase from approximately 43 to 53 mg/Nm³ as f(H₂) increases from 0 to 25% at Φ = 0.8 and full load.

3.2. Effect of Equivalence Ratio on Burner Combustion Processes

Under full load conditions with f(H₂) = 0, increasing the equivalence ratio from 0.8 to 1.2 raises temperatures at multiple locations within the furnace and increases the average outlet temperature from 1073 K to 1107 K (Figure 8 and Figure 9, Table A1). Aligned with industrial operational data, this temperature rise is attributed to enhanced chemical energy release due to increased fuel concentration.

From a pollutant control perspective, the NO concentration gradually increases with the equivalence ratio (Figure 10). The corresponding NO_x emissions normalized to 3.5% O₂ are presented. Under lean conditions (Φ ≤ 1), increasing the equivalence ratio brings the fuel–air ratio closer to stoichiometric, improving combustion completeness and raising temperature, which promotes thermal NO formation. Under rich conditions (Φ > 1), although the O₂ concentration decreases, localized high-temperature zones and radicals (CH_i and H) produced by the decomposition of excess fuel enhance prompt NO formation via HCN intermediates [15].

Variations in CO and O₂ concentrations further confirm the influence of the equivalence ratio (Figure 11). Under Φ > 1, significant O₂ deficiency leads to incomplete combustion and increased CO production. At Φ = 0.8, moderate excess air ensures consistently low CO emissions while avoiding the NO peak near Φ = 1.0. This compromise condition provides practical guidance for industrial operation, suggesting the use of Φ = 0.8 at medium loads to achieve emission balance.

3.3. Influence of Load on the Combustion Process of the Burner

Using Φ = 0.8 as a representative condition, load reduction leads to contraction of the high-temperature region and shortening of the flame length (Figure 12 and Figure 13). When the load decreases to 50%, the high-temperature zone in the central furnace cross-section contracts significantly; at 20% and 15% loads, the temperature distribution becomes more uniform, and the outlet temperature decreases by 150 °C, compared to full load (Table A1).

The velocity field analysis reveals a direct link between vortex structures and pollutant formation (Figure 14). A counterclockwise vortex forms near the burner outlet across all conditions, induced by the velocity gradient between the high-speed flow at the outlet and the lower-speed flow near the wall. As the load decreases, reduced fuel and inlet flow rates lead to weakened vortex intensity. At 50% load, a significant velocity decrease occurs at the burner outlet; at 20% load, velocity gradients disappear, resulting in a more homogeneous flow field. The weakening of these vortex structures directly impacts mixing intensity and flame stability, particularly at very low loads like 15%, where combustion oscillations may occur, requiring mitigation through burner design adjustments in practical applications.

NO emissions show a strong negative correlation with load (Figure 15), with load reduction from 100% to 15% leading to an over 40% NO reduction. This decrease is directly linked to the furnace temperature drop shown in Figure 13, confirming the dominance of the thermal–NO pathway.

Notable simulation limitations include the steady-state assumption, neglect of soot formation, and limited applicability of GRI-Mech 3.0 at very high hydrogen-blending ratios (>30%). Furthermore, practical low-load operation (especially at 15%) requires consideration of thermal shock risks and combustion stability, recommending full-scale validation before implementation.

4. Conclusions

This study analyzed the combustion characteristics and pollutant emissions of hydrogen-enriched natural gas under different operating conditions through CFD simulations. The main conclusions are as follows:

Hydrogen blending linearly increases the flame temperature within the 25% range while reducing CO emissions but increasing NO_x formation. Based on emission balance and current infrastructure constraints, 20% is recommended as the practical upper limit for operation. Specifically, a 20% blending ratio at Φ = 0.8 can achieve a 25% reduction in CO while limiting the NO_x increase to below 18%.

The equivalence ratio exhibits competing effects on pollutant formation. Φ = 0.8 is recommended as the optimal compromise, achieving CO mole fractions <62.5 mg/Nm³ while maintaining NO_x emissions below 100 mg/Nm³ (3.5% O₂).

Load reduction effectively suppresses NO_x formation, achieving 30% NO reduction (in mg/Nm³ (3.5% O₂) at 50% load, providing clear environmental benefits for part-load operation. However, the weakened vortex structures at very low loads (e.g., 15%) may affect combustion stability, suggesting the need for combined swirler adjustments or staged combustion strategies in practical applications.

Through the multi-parameter-coupling analysis, this study extends the understanding beyond existing single-parameter studies, providing a quantitative design basis for the hydrogen retrofit of industrial boilers. Subsequent work will integrate material compatibility analysis and transient dynamics validation to support the safe implementation of higher blending ratios (>30%).