Combustion method and combustor based on non-equilibrium plasma and gan model

Abstract

This application discloses a combustion method and burner based on non-equilibrium plasma and a GAN model, relating to the field of aero-engine combustion. The method includes constructing a non-equilibrium plasma DBD combustion-assisted system with dielectric barrier discharge; adjusting the flow rate of hydrogen / ammonia / methane ternary fuel to change the hydrogen blending ratio and equivalence ratio to set experimental conditions; measuring the intensity and temperature of OH, CH, and NO radicals using a flame radical measurement system; constructing a GAN prediction model and optimizing the prediction results through adversarial training with a generator and discriminator; and using the trained model to predict untrained operating conditions and verifying the results by comparing with experimental data. This achieves improved combustion performance and effective reduction of NOx emissions (up to 10% or more), while the correlation coefficient R of the GAN prediction values ​​is also high. 2 Up to 0.99. This result is significant for aviation NO. X The assessment, control, and prediction of emissions have important guiding significance.

Description

Technical Field

[0001] This application relates to the field of aero-engine combustion technology, and in particular to a combustion method and burner based on non-equilibrium plasma and a GAN model. Background Technology

[0002] According to the International Civil Aviation Organization (ICAO) CAEP standards, NOx emission limits for aircraft engines have been continuously tightened from CAEP / 2 to CAEP / 10. The CAEP / 12 standard is expected to further reduce NOx emission limits by more than 15%, imposing higher requirements on combustion performance and pollutant emissions. The new version of the "Ambient Air Quality Standard" (GB3095-2026), which will be implemented on March 1, 2026, further tightens the concentration limits for nitrogen dioxide (NO2) and nitrogen oxides (NOx), imposing stricter compliance requirements on NOx emission control for the aviation industry and ground power plants. In today's era, aviation will remain the fastest, most accessible, and most efficient mode of transportation in human economic activities, playing an irreplaceable role in enabling domestic and international economic and trade development.

[0003] In the field of aero-engines, hydrogen / ammonia / methane ternary fuels are an important choice for green and sustainable fuels, and their application has attracted widespread attention. Hydrogen combustion produces no carbon emissions, ammonia has low storage and transportation costs and is easy to integrate into existing fuel systems, and methane has high utilization value as a transitional fuel. However, ammonia fuel has an inherently very low flame propagation speed (approximately 0.07 m / s), far lower than that of hydrogen (approximately 2.5 m / s) and aviation kerosene, resulting in difficulties in ignition and poor combustion stability. Furthermore, the nitrogen element in ammonia molecules is easily oxidized to form fuel-type NOx. Existing research lacks systematic studies on plasma transport effects (ion wind-enhanced mixing, increased diffusion coefficient) and ionization effects (free radical generation pathways, electron energy distribution) under wide operating conditions in aero-engines. In particular, the competitive relationship between OH and NH2 free radicals under plasma and their impact mechanism on NOx generation pathways are still unclear, which restricts the development of high-efficiency, low-pollution burners. Furthermore, existing traditional thermal equilibrium plasmas rely on high-temperature pyrolysis to activate fuel, resulting in low energy utilization efficiency and severely exacerbating the formation of thermal NOx in the high-temperature environment. At the same time, the equipment is bulky and has stringent thermal protection requirements. In contrast, the non-equilibrium plasma used in this invention has an electron temperature of 1~10 eV while the gas temperature is maintained below 500 K. High-energy electrons selectively decompose fuel molecules instead of heating the gas, resulting in more efficient energy utilization. At the same time, it effectively suppresses the formation pathway of thermal NOx, achieving a better balance between improving combustion reactivity and controlling pollutant emissions. It is more suitable for the wide operating conditions and high compactness requirements of aero-engine applications.

[0004] In the optimization design of combustion systems, traditional methods rely heavily on extensive experimental trial and error or high-fidelity numerical simulations (such as direct numerical simulation and large eddy simulation), which suffer from drawbacks such as long processing time, high computational cost, and difficulty in covering a wide range of operating conditions. In recent years, Generative Adversarial Networks (GANs), as a deep generative model, have shown great potential in flow field reconstruction and combustion parameter prediction. However, the application of existing GANs in the field of combustion is still in its early stages: on the one hand, they are mostly focused on image super-resolution or flame morphology generation, with limited research on the quantitative prediction of key chemical kinetic parameters (such as OH radical intensity, NH2 radical intensity, and NOx concentration); on the other hand, there is a lack of attempts to combine GANs with non-equilibrium plasma combustion technology, making it impossible to quickly predict the complex mapping relationship between plasma parameters and combustion performance.

[0005] This application addresses key technical challenges in the combustion of hydrogen / ammonia / methane ternary fuels in aviation, including ignition difficulties, flame instability, and high NOx emissions. It proposes a comprehensive technical solution integrating non-equilibrium plasma catalysis and a GAN prediction model. By systematically revealing the regulatory mechanisms of plasma transport and ionization effects on combustion chemistry and kinetics, and aiming at the synergistic optimization of improved combustion performance and low emissions, this application provides a theoretical basis and engineering tools for efficient and low-emission combustion and prediction in aero-engines during the low-carbon transition phase. Summary of the Invention

[0006] The purpose of this application is to provide a combustion method and burner based on non-equilibrium plasma and a GAN model, thereby solving the aforementioned problems in the prior art.

[0007] To achieve the above objectives, this application provides a combustion method based on non-equilibrium plasma and a GAN model, comprising the following steps: S1: Constructing an experimental system for dielectric barrier discharge plasma combustion; S2: Keep the total flow rate of hydrogen, ammonia, methane and air premixed gas constant, and set the experimental conditions by adjusting the flow rates of hydrogen, ammonia and methane to change the hydrogen doping ratio and equivalence ratio; non-equilibrium plasma acts on the combustion process in a synergistic manner through transport effect, kinetic effect and thermal effect; S3: Experimental parameters during combustion are measured using a flame free radical measurement system, which includes a NO-PLIF system, an OH-PLIF system, and an NH2* chemiluminescence system. The two-dimensional spatial distribution of OH and NO free radicals is measured using the NO-PLIF and OH-PLIF systems; the concentration of NH2 free radicals is obtained through NH2 free radical diagnostic luminescence; and the temperature is measured using thermocouples. S4: Collect experimental data to construct a dataset. After preprocessing the dataset by removing outliers and normalizing it, divide it into a training set and a test set. S5: Construct a GAN prediction model, including a generator and a discriminator. The generator takes the concatenation of a random noise vector and a conditional vector as input and outputs a predicted image of a free radical flame, as well as predicted values ​​for OH radical intensity, NH2 radical intensity, and NOx emission concentration. The discriminator takes the concatenation of a conditional vector and a sample to be discriminated as input and outputs a probability value indicating that the sample is real experimental data. The generator and discriminator are optimized through adversarial training and mutual game. S6: Use the trained GAN prediction model to predict the untrained working conditions and output the prediction results. S7: Compare and verify the prediction results with the experimental data. When both the numerical prediction accuracy and the image prediction quality meet the preset conditions, output the GAN prediction model as the final prediction model.

[0008] Preferably, the dielectric barrier discharge plasma-assisted combustion experimental system includes a ternary gas source, an air source, a mass flow controller, a premixing chamber, a DBD reactor, a swirling burner, a high-frequency high-voltage plasma power supply, an electrode system, a PLIF laser detection system, and a NO... X Analyzer, data acquisition and control terminal, ternary mixed gas source including hydrogen, ammonia and methane; The center frequency of the plasma power supply is 5–20 kHz, the maximum output power point is 10 kHz, and the maximum output voltage is 60 kV. DBD reactor electric field strength E / N With discharge voltage V The relationship is: ; in, d The electrode spacing, N The gas number density; A swirl burner consists of a nozzle and a swirler, with the swirler installed at the nozzle's combustion outlet. The swirl number of the swirler is expressed as: ; in, D i The inner diameter of the hydrocyclone blade; D o The outer diameter of the hydrocyclone blade; θ The angle of the hydrocyclone blades.

[0009] Preferably, the hydrogen doping ratio is expressed as: ; The equivalent ratio is expressed as: ; in, This refers to the hydrogen flow rate; This refers to the ammonia flow rate. This represents the methane flow rate; Airflow rate; The state is the chemical equivalence ratio.

[0010] Preferably, the transport effect promotes the reaction by generating plasma wind and enhancing turbulent mixing; The kinetic effect involves high-energy electron collisions that break down H2, NH3, and CH4 molecules to generate NH2, CH3, H, and OH reactive free radicals, as well as ions and excited-state species. Among these, NH2 + NO → N2 + H2O is the key pathway for NOx suppression. The thermal effect causes the temperature to rise, accelerates particle movement, and increases the reaction rate.

[0011] Preferably, the dataset input parameters include discharge voltage, discharge frequency, equivalence ratio, methane volume fraction, hydrogen gas integral, ammonia gas integral, and total fuel flow rate; the output parameters include OH radical intensity, NH2 radical intensity, NOx emission concentration, and radical flame image.

[0012] Preferably, outlier removal is represented as follows: ; Normalized representation is: ; in, To obtain a measurement value from three repeated experiments under the same operating conditions. This is the average of three measurements taken under this operating condition. Standard deviation This is the original data. and These are the minimum and maximum values ​​of the parameter under all operating conditions, respectively.

[0013] Preferably, the generator takes as input a concatenation of a random noise vector sampled from a Gaussian distribution and a conditional vector, the conditional vector including discharge voltage, equivalence ratio, and fuel flow parameters; the random noise is mapped into a decoupled intermediate latent code through a Mapping Network, and then a radical flame image is generated by a Generator Network through deconvolution and upsampling operations, and the predicted concentration values ​​of OH, NH2, and NOx are output. The discriminator takes the concatenation of the conditional vector and the sample to be discriminated as input, extracts image features through a convolutional neural network, and focuses on the key region with high concentration of OH radicals in the flame through a Minibatch optimization layer and an Attention mechanism, and outputs the probability value that the sample is real experimental data.

[0014] Preferably, the generator loss function is expressed as: ; The discriminator loss function is expressed as: ; The GAN prediction model is optimized through a minimax game, and the total loss function is expressed as: ; in, For the expectation value operator, For balance coefficient, It is a random noise vector. For conditional vectors, Generate data for the generator's predictions. For real data, This represents the probability value output by the discriminator for the real data. For the discriminator to generate data The output probability value; This indicates sampling from the actual experimental data distribution. For the true data distribution, For noise distribution, This indicates sampling from a priori noise distribution.

[0015] Preferably, the numerical prediction accuracy in S7 is determined by the coefficient of determination R. 2 The judgment is made based on the root mean square error (RMSE), and is expressed as follows: ; ; in, M The number of samples in the test set. For the first i The actual experimental values ​​of the group working conditions. For the corresponding predicted value, The arithmetic mean of the actual experimental values; Image prediction quality is evaluated by comparing the free radical prediction image output by the generator with the real free radical image collected in the experiment, and considering three aspects: flame morphology, location of high-intensity free radical region, and intensity gradient distribution.

[0016] A burner for implementing a combustion method based on nonequilibrium plasma and a GAN model, comprising: The gas supply system consists of a ternary mixed gas source of hydrogen, ammonia, and methane, and an air source. The mass flow controller is connected to a hydrogen, ammonia, and methane ternary gas source and an air source respectively, and is used to independently regulate the flow rates of hydrogen, ammonia, methane and air; The premixing chamber, connected to the output of the mass flow controller, is used to mix the gases to form a premixed gas. The DBD reactor, connected to the output of the premixing chamber, includes a quartz dielectric layer, an AC high-voltage electrode, and a low-voltage electrode. A swirling burner includes a nozzle and a swirler. The swirler is installed at the combustion outlet of the nozzle to organize combustion. A plasma power supply, connected to an AC high-voltage electrode, is used to output high-frequency high-voltage AC power. An oscilloscope, connected to the plasma power supply, is used to monitor the discharge voltage and current waveforms in real time. Quartz window provides an optical observation channel while isolating high-temperature flames; The PLIF laser detection system includes an Nd:YAG solid-state laser and a dye laser, along with a sheet-like light source assembly, for exciting and acquiring fluorescence signals of OH and NO free radicals; An ICCD camera, with a 310nm bandpass filter positioned perpendicular to the laser sheet, is used to acquire fluorescence signals; A portable integrated flue gas analyzer is installed at the exhaust port at the tail of the flame to measure the NOx concentration in the flue gas; The PC data acquisition and control terminal is connected to the mass flow controller, plasma power supply, Nd:YAG solid-state laser, dye laser, ICCD camera and portable integrated flue gas analyzer via data connection cables. It is used to read and store data in real time and control the working sequence of each component.

[0017] Therefore, the combustion method and burner based on non-equilibrium plasma and GAN model described above have the following advantages: (1) This application improves the intensity of OH radicals by about 30% by applying plasma DBD catalytic technology, making the experimental flame image brighter and improving the combustion performance.

[0018] (2) In this application, in the high voltage range (approximately 9~14 kV), high concentrations of NH2 can reduce NO. X The efficient reduction to harmless N2, combined with the strong reducing power of NH2 and the synergistic suppression in the high voltage range, significantly reduces NOx emissions.

[0019] (3) The prediction results obtained by the GAN prediction model in this application have an error of less than 5% with the experimental results and are highly consistent with the flame images, thus obtaining a set of efficient AI prediction GAN model.

[0020] The technical solution of this application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] Figure 1This is a flowchart of a combustion method based on non-equilibrium plasma and a GAN model in this application; Figure 2 This is a diagram illustrating the non-equilibrium plasma dominance mechanism in the embodiments of this application; Figure 3 This is a schematic diagram of the GAN prediction model in the embodiments of this application; Figure 4 This is a burner diagram from an embodiment of this application; Figure 5 This is a comparison of experimental and GAN model results showing the evolution of NOx emissions with plasma voltage at different methane volume fractions in the embodiments of this application. Figure 6 The diagrams show the evolution of OH radicals with and without plasma under the influence of equivalence ratio in the embodiments of this application; where (a) is the experimental evolution diagram and (b) is the GAN model evolution diagram. Figure 7 The diagrams show the evolution of NH2 radicals with and without plasma under the influence of equivalence ratio in the embodiments of this application; where (a) is the experimental evolution diagram and (b) is the GAN model evolution diagram. Figure label: 1. Gas source system; 2. Mass flow controller; 3. Premixing chamber; 4. DBD reactor; 5. Quartz dielectric layer; 6. AC high-voltage electrode; 7. Low-voltage electrode; 8. Cyclone separator; 9. Plasma power supply; 10. Oscilloscope; 11. Portable integrated flue gas analyzer; 12. Nd:YAG solid-state laser; 13. Dye laser; 14. Sheet light source assembly; 15. 310 nm bandpass filter; 16. ICCD camera; 17. PC data acquisition and control terminal; 18. Data connection cable; 19. Quartz window; 20. Connecting bolts; 21. Cast iron round tube support. Detailed Implementation

[0022] The following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning as understood by a person of ordinary skill in the art to which this application pertains.

[0024] The terms "comprising" or "including," as used in this application, mean that the element preceding the term encompasses the element listed after the term, and do not exclude the possibility of encompassing other elements as well. The terms "inner," "outer," "upper," and "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. When the absolute position of the described object changes, the relative positional relationship may also change accordingly. In this application, unless otherwise expressly specified and limited, the term "attached," etc., should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can refer to a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication of two elements or the interaction relationship between two elements. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0025] Example 1: A combustion method based on non-equilibrium plasma and a GAN model is proposed. First, a non-equilibrium plasma catalytic hydrogen / ammonia / methane ternary fuel combustion experimental platform is constructed, focusing on transport effects (mass mixing, diffusion) and ionization effects (molecular fragmentation, free radical generation). Combustion experiments are conducted by adjusting the hydrogen doping ratio and equivalence ratio by changing the flow rates of the three gases. OH-PLIF and NO-PLIF are used to diagnose the two-dimensional spatial distribution of OH and NO free radicals, and NH2 free radical concentration is obtained through NH free radical diagnosis. Flue gas analyzer data is then collected and transmitted to a PC. The AI ​​prediction part uses the model to predict NOx emissions and establishes a mapping relationship between combustion parameters and NOx emissions. A GAN prediction is performed using a generator and a discriminator: the generator generates predicted data and images from random noise, and the discriminator uses a convolutional neural network combined with a Minibatch optimization layer and an attention mechanism to judge the authenticity and output probability scores. Through adversarial training involving generator and discriminator losses, the prediction continuously approximates the real image. The GAN prediction results were compared and verified with experimental results in four aspects: NO concentration, OH radical intensity, NH radical intensity, and radical flame images. This determined whether the consistency met the requirements, ultimately aiming to improve combustion performance, reduce NOx emissions, and establish a highly efficient AI prediction GAN model. Figures 1-4 As shown, the specific steps include: S1: Constructing an experimental system for dielectric barrier discharge plasma combustion; The dielectric barrier discharge plasma combustion experimental system includes a ternary gas source, an air source, a mass flow controller, a premixing chamber, a DBD reactor, a swirling burner, a high-frequency high-voltage plasma power supply, an electrode system, a PLIF laser detection system, and a NO...X Analyzer, data acquisition and control terminal, ternary mixed gas source including hydrogen, ammonia and methane; The center frequency of the plasma power supply is 5–20 kHz, the maximum output power point is 10 kHz, and the maximum output voltage is 60 kV. DBD reactor electric field strength E / N With discharge voltage V The relationship is: ; in, d This refers to the electrode spacing, which can be taken as 13 mm in practical applications. N This represents the gas number density.

[0026] The swirl combustion system consists of a combustion chamber, a swirl burner, and a gas supply system. The swirl burner comprises a nozzle and a swirler, with the swirler installed at the nozzle's combustion outlet. The swirl number of the swirler is expressed as: ; in, D i The inner diameter of the hydrocyclone blade; D o The outer diameter of the hydrocyclone blade; θ The angle of the hydrocyclone blades is 45°. In practical applications, the outer diameter of the hydrocyclone blades is 35 mm, the inner diameter is 13 mm, the number of blades is 12, the angle of the hydrocyclone blades is 45°, and the number of swirls is 0.73.

[0027] S2: Keep the total flow rate of hydrogen, ammonia, methane and air premixed gas constant, and set the experimental conditions by adjusting the flow rates of hydrogen, ammonia and methane to change the hydrogen doping ratio and equivalence ratio; non-equilibrium plasma acts on the combustion process in a synergistic manner through transport effect, kinetic effect and thermal effect; The hydrogen doping ratio is expressed as: ; The equivalent ratio is expressed as: ; in, This refers to the hydrogen flow rate; This refers to the ammonia flow rate. This represents the methane flow rate; Airflow rate; This refers to the state of stoichiometric ratio. In practical applications, the hydrogen doping ratio is 0.1 ~ 0.3, and the stoichiometric ratio is... φ =0.6 ~ 1.2.

[0028] like Figure 2As shown, non-equilibrium plasma synergistically influences the combustion process through three effects: the transport effect enhances turbulent mixing and promotes the reaction by generating plasma wind; the kinetic effect breaks down NH3, CH4, and H2 molecules through high-energy electron collisions, generating reactive free radicals such as NH2, CH3, H, and OH, as well as ions and excited-state species such as N2(v) and O2(a). 1 Δg), where NH2 + NO → N2 + H2O is the key pathway for NOx suppression; the thermal effect increases temperature, accelerates particle movement, and enhances the reaction rate. These three synergistic effects promote enhanced combustion, reduce ignition delay time, accelerate flame propagation speed, broaden the combustible limit, improve combustion performance, achieve more complete fuel oxidation, and reduce NOx emissions.

[0029] S3: Experimental parameters during combustion are measured using a flame free radical measurement system, which includes a NO-PLIF system, an OH-PLIF system, and an NH2* chemiluminescence system. The two-dimensional spatial distribution of OH and NO free radicals is measured using the NO-PLIF and OH-PLIF systems; the concentration of NH2 free radicals is obtained through NH2 free radical diagnostic luminescence; and the temperature is measured using thermocouples. Planar laser-induced fluorescence (PLIF) enables two-dimensional planar measurement of low-concentration intermediate free radicals (such as OH, CH, and NO radicals). The main principle of PLIF is to excite the target component to an electronically excited state by tuning the wavelength of a sheet laser to a specific absorption wavelength. Due to energy instability, the excited component undergoes energy level transitions and emits fluorescence, which is detected as a fluorescence radiation signal and captured by an enhanced charge-coupled device (ICCD) camera equipped with corresponding filters. The fluorescence intensity and the free radical intensity satisfy the following relationship: ; Fluorescence intensity For system constants, To excite laser power, C For free radical strength, Φ(λ,T) The fluorescence quantum yield is related to the excitation wavelength and ambient temperature. Chemiluminescence technology, on the other hand, measures the fluorescence based on the autofluorescence of the target component in a combustion flame. Due to the high temperature during combustion, some excited states exist within the target component, emitting fluorescence signals that are then detected and captured. Temperature is relevant for NO. X The generation path and rate of NO have a significant impact, so for NO XFor temperature measurement, this application connects the thermocouple output of the flame free radical system to the junction box of the data acquisition unit. The temperature signal is transmitted and stored in the data acquisition unit via the thermocouple, and then measured on the PC. The dynamic response of the thermocouple is as follows: ; In the formula, T g To correct the flue gas temperature, T To measure temperature, h The convective heat transfer coefficient between the thermocouple and the surrounding gas is denoted as . ε For the emissivity of the thermocouple probe, σ The blackbody radiation constant is 5.67 × 10⁻⁸ W∙(m²). 2 ∙K 4 ) -1 .

[0030] S4: Collect experimental data to construct a dataset. After preprocessing the dataset by removing outliers and normalizing it, divide it into a training set and a test set. The collected experimental data were compiled into a standard dataset suitable for training GAN prediction models. Input parameters included seven indicators: discharge voltage, discharge frequency, equivalence ratio, methane volume fraction, hydrogen gas integral, ammonia gas integral, and total fuel flow rate. Output parameters included three indicators: OH radical intensity, NH2 radical intensity, and NOx emission concentration. Data preprocessing consisted of two steps: first, outliers were removed using the following formula: ; in, To obtain a measurement value from three repeated experiments under the same operating conditions. This is the average of three measurements taken under this operating condition. The standard deviation is used. If a measurement satisfies the above inequality, that data point is marked as an outlier and removed. The average of the remaining data is taken as the true value for that operating condition. Then, all input and output parameters are normalized. ; in, This is the original data. and These are the minimum and maximum values ​​of the parameter under all operating conditions, respectively.

[0031] Normalization maps all input and output parameters to the [0,1] interval, eliminating the order-of-magnitude differences between different units and accelerating the training convergence of the GAN model. After preprocessing, all working condition data are randomly divided into training and test sets in an 8:2 ratio. The training set is used for model training, and the test set is used for final performance evaluation. The test set does not participate in the training process.

[0032] S5: Construct a GAN prediction model, including a generator and a discriminator. The generator takes the concatenation of a random noise vector and a conditional vector as input and outputs a PLIF flame prediction image as well as predicted values ​​for OH radical intensity, NH2 radical intensity, and NOx emission concentration. The discriminator takes the concatenation of a conditional vector and a sample to be discriminated as input and outputs a probability value indicating that the sample is real experimental data. The generator and discriminator are optimized through adversarial training and mutual game. The generator takes as input a concatenation of a random noise vector sampled from a Gaussian distribution and a conditional vector, where the conditional vector contains parameters such as discharge voltage, equivalence ratio, and fuel flow rate. The random noise is mapped into a decoupled intermediate latent code through a Mapping Network, and then the Generator Network generates a free radical flame image and outputs the predicted concentrations of OH, NH2, and NOx through deconvolution and upsampling operations. The discriminator takes the concatenation of the conditional vector and the sample to be discriminated as input, extracts image features through a convolutional neural network, and focuses on key areas with high concentrations of OH and NO free radicals in the flame through a Minibatch optimization layer and an Attention mechanism, and outputs the probability value that the sample is real experimental data.

[0033] The generator loss function is expressed as: ; in, For the expectation value operator, For balance coefficient, It is a random noise vector. For conditional vectors, Generate data for the generator's predictions. The first element, which uses real data, is designed to combat loss and encourage the generator to produce data. It can deceive the output of the discriminator, making The first term approaches 1; the second term is the reconstruction loss, which forces the generated output to approximate the true experimental value through the L2 norm. The discriminator loss function is expressed as: ; This represents the probability value output by the discriminator for the real data. For the discriminator to generate data The output probability value; the first term makes the discriminator more accurate for the true data y. real The output probability of the first term approaches 1, and the second term makes the output probability of the discriminator for the generated data G(z,c) approach 0.1, thereby maximizing the discriminator's ability to distinguish between real data and generated data.

[0034] The GAN prediction model is optimized through a minimax game, and the total loss function is expressed as: ; This indicates sampling from the actual experimental data distribution. For the true data distribution, For noise distribution, This indicates sampling from a priori noise distribution. The discriminator attempts to maximize the loss function to better distinguish between real and fake data, while the generator attempts to minimize the loss function to deceive the discriminator. By alternately optimizing the generator and discriminator, continuously seeking the optimal parameters, the model ultimately achieves high-precision combustion parameter prediction.

[0035] S6: Use the trained GAN prediction model to predict the untrained working conditions and output the prediction results. Generalization ability refers to the model's ability to learn the inherent mapping relationship between input parameters and output results based on limited training data, thereby enabling it to accurately predict operating conditions without prior training. In this application, the range of operating conditions covered by the experimental conditions is limited, for example, the training operating condition with a discharge voltage V = 0~20 kV (the operating condition is denoted as...). X i The corresponding output prediction value is Y i However, in practical applications, voltage values ​​between training points or extreme operating conditions exceeding the training range may be encountered, which are difficult to measure experimentally one by one. In such cases, by utilizing a trained GAN prediction model and the nonlinear mapping relationship between combustion parameters learned from existing operating conditions and OH, NH2, and NOx concentrations, the corresponding output can be directly predicted from the input parameters of unseen operating conditions without the need for actual experiments. The generalization error of this process can be expressed by the following formula: ; in, These are predicted operating condition parameters obtained without experimental training. This indicates that the working condition is determined by the GAN model. X i The GAN prediction model in this application is trained adversarially between the generator and the discriminator, which enables it to capture the latent distribution characteristics of the training data rather than mechanically memorizing input-output pairs, thus exhibiting good generalization ability.

[0036] S7: Compare and verify the prediction results with the experimental data. When both the numerical prediction accuracy and the image prediction quality meet the preset conditions, output the GAN prediction model as the final prediction model.

[0037] After completing the training and generalization evaluation of the GAN prediction model, this application systematically compares the model's prediction output with experimental data and evaluates the model's effectiveness based on preset judgment criteria. For the predicted values ​​of OH intensity, NH2 intensity, and NOx concentration, NOx concentration is plotted with voltage as the x-axis and concentration as the y-axis (X) at different methane volume fractions. CH4 Curves showing the variation of free radical concentrations (0.7, 0.5, and 0.3) with discharge voltage, scatter plots comparing OH and NH2 free radical intensities, were used to verify the accuracy of the model's prediction of free radical concentrations. This can be verified using R... 2 Using RMSE to determine, it can be expressed as: ; ; in, M The number of samples in the test set. For the first i The actual experimental values ​​of the group working conditions. For the corresponding predicted value, R is the arithmetic mean of the true experimental values. 2 The value ranges from [0,1], with values ​​closer to 1 indicating a stronger interpretability of the experimental data. A smaller RMSE value indicates higher prediction accuracy. For flame image prediction, the predicted free radical image output by the generator is compared with the actual free radical image collected in the experiment. The image generation quality is evaluated from three aspects: flame morphology, the location of the high-intensity region of OH free radicals, and the intensity gradient distribution. If the coordinate curves and the image are highly consistent, the predictive ability of this GAN model is considered good.

[0038] Regarding combustion performance, ammonia fuel is difficult to ignite due to its extremely low flame propagation speed. This application utilizes the ionization effect of non-equilibrium plasma to solve this problem. Figure 6 (a) Figure 7 As shown in line graph (a), when the volume fraction of methane is X... CH4 Under the condition of equivalence ratio 0.5, the intensities of OH and NH2 under DBD plasma were higher than those without DBD, indicating that the high-energy electrons and active particles generated by the discharge can promote the activation of fuel molecules and nitrogen-containing species, accelerating the conversion of NH3 to intermediates such as NH2. In terms of average improvement, DBD increased the average OH intensity by approximately 5 (average increase of approximately 12%) and the average NH2 intensity by approximately 6 (average increase of approximately 25%) at equivalence ratios of 0.7, 0.9, and 1.2, and the bright areas in the flame images were also more pronounced, indicating that DBD has a certain promoting effect on the formation of nitrogen-containing intermediates. Vibrationally excited states promote fuel pyrolysis by lowering the reaction activation energy, while electronically excited states generate a large number of active free radicals through collisional energy transfer: O( 1D) Reaction with NH3 produces NH2 and OH. Collisions between N2 and H2O or O2 further transfer energy to generate OH radicals. At the same time, the pathway of direct NH3 cracking by high-energy electrons (e+NH3→e+NH2+H) is also enhanced, thereby accelerating the conversion of NH3 to intermediates such as NH2. This demonstrates the promoting effect of DBD on the formation of nitrogen-containing intermediates and the combustion chain reaction, thereby improving combustion performance.

[0039] In NO X Reduction is achieved through two pathways: high-voltage suppression and enhanced reducing power of NH2. For example... Figure 5 As shown, with an equivalence ratio of 1.2, the NOx variation with voltage under different methane ratios can be summarized as a two-stage characteristic: "first remaining stable, then gradually decreasing." Under different methane ratios, the decrease is concentrated in the optimal voltage range (approximately 9~14 kV), for example, for X... CH4 =0.3, NOx fluctuates relatively little overall, remaining basically stable between 1130–1190 ppm in the low to medium voltage range, and decreasing by 10% to approximately 1060 ppm in the high voltage range; for X CH4 =0.5, the high-voltage region decreases by about 5% from 2100 ppm to 2000 ppm; X CH4 =0.7, the highest NOx level, decreasing from approximately 2300 ppm to 2250 ppm in the high-voltage region with relatively small fluctuations. It can be seen that when the voltage reaches a certain threshold, the NOx levels of all three methane proportions show a decreasing trend in the high-voltage region, indicating that high voltage has a significant effect on NOx... X While NH2 reduction has an inhibitory effect on NOx formation, the inhibitory effect of NH2 reduction on NOx formation gradually strengthens with increasing voltage. At high voltages, the increased electron density enhances the vibrational and electronic excitation of N2. The resulting excited-state N2* undergoes energy transfer through collisions with H2O or O2, further replenishing OH radicals and promoting more complete fuel oxidation, indirectly reducing the contribution of unburned intermediates to NOx formation. Therefore, under fuel-rich conditions and in the high-voltage region, the synergistic effect of NH2 reduction and enhanced combustion leads to a continuous downward trend in NOx emissions.

[0040] This application has clear engineering application value in improving combustion performance and reducing NOx emissions. Regarding improved combustion performance, it enables low-temperature rapid ignition and wide-range stable combustion, suitable for high-altitude re-ignition of aero-engines, ground start-up, and low-load operation of industrial gas turbines, thus improving fuel adaptability and operational reliability. In terms of reducing NOx emissions, it lowers NOx emissions through a high-voltage NH2 reduction pathway, applicable to scenarios with stringent emission requirements such as urban gas turbines, marine power, distributed energy stations, and industrial boilers. It is particularly suitable for low-carbon retrofit projects involving hydrogen / ammonia blending of natural gas, achieving pollutant reduction without replacing the main equipment, reducing environmental compliance costs, shortening the technical retrofit cycle, and providing a feasible engineering solution for the energy and power industry's smooth transition from fossil fuels to zero-carbon fuels.

[0041] like Figure 5 As shown, the scatter plots of GAN-predicted NOx and experimental NOx are basically distributed near the diagonal, and the calculated correlation coefficients both reach R0. 2 =0.99, indicating that the dashed line obtained by the GAN prediction model matches the experimental scatter plot well overall. It can simultaneously capture the stable changes in the low voltage region, the rising or falling trends in the high voltage region, and the NOx stratification relationship between different methane ratios. Moreover, the deviation between most prediction points and experimental values ​​is within a reasonable range of about 3%–8%. Figure 6 (b) and Figure 7 As shown in (b), the line graph generated by the GAN prediction model and Figure 6 (a) Figure 7Compared to (a) in the previous example, the differences between the upward trend changes and the two cases with and without an equilibrium body catalyzing DBD are basically consistent, and the difference between the two cases is basically controlled at around 5%. From the combustion images, the predicted images are basically consistent with the experimental images, and there are also cases where the high-intensity regions in the predicted images are enhanced. These results are directly related to the internal mechanism of the GAN prediction model. The generator in this application starts from random noise and combustion parameters, and generates the predicted free radical intensity through deconvolution and upsampling, which is completely consistent with the experimental measurements, indicating that the generator G has successfully learned the recognition role of DBD on fuel activation. By improving the training stability through the Minibatch optimization layer and the Attention mechanism, the high concentration region of free radicals in the flame is focused, and the key reaction zone features are extracted. The experimental images and data focus on the flame center of this region, and the prediction results also show the same spatial distribution. The generator and the discriminator are constantly playing a game in the adversarial process. The generator G tries to make the discriminator D believe that the generated image is real, and the discriminator D tries to distinguish between the real and generated images. After iteration, the generator has achieved a reconstruction loss of less than 0.02 and the accuracy of the discriminator is stable at around 85%. The trained model does not directly solve the chemical reaction equations, but instead learns the nonlinear relationship between flame image brightness, average intensity, and operating parameters from the data. Through the GAN model's ability to understand and penetrate microscopic reaction mechanisms, it predicts similar results.

[0042] Traditional burner development relies heavily on extensive experimental trial and error or high-fidelity numerical simulations, which are time-consuming and costly. This application utilizes a GAN prediction model to achieve efficient prediction of plasma-assisted combustion effects. In practical applications, it can provide efficient and low-cost intelligent solutions for scenarios such as aero-engine ground testing, low-NOx retrofitting of industrial boilers, and low-carbon fuel switching in gas turbines. This reduces traditional, lengthy debugging methods and ineffective test schemes, automatically optimizing parameters based on the model to obtain the best results, significantly shortening prediction time. The model can also predict data that is difficult to obtain directly through experiments. The prediction results provide a reference for the ultimate performance assessment of combustion and the definition of safety boundaries, avoiding equipment damage and safety accidents caused by high-risk experiments. It also expands the data coverage, providing more comprehensive data support for combustion performance optimization and emission control strategy formulation.

[0043] This application employs non-equilibrium plasma-catalyzed hydrogen / ammonia / methane combustion. Through the transport effect (enhanced mixing and diffusion) and ionization effect of plasma (generating high-energy electrons to crack H2 / NH3 / CH4, producing free radicals such as H, OH, NH2, and CH3), a high-efficiency, low-pollution non-equilibrium plasma-catalyzed hydrogen / ammonia / methane combustor system is developed, achieving low-temperature ignition, flame stability, and NO reduction. XPollutant suppression. Simultaneously, this application constructs a GAN prediction model, which generates flame images and free radical intensities from random noise and combustion condition parameters using a generator G. A discriminator D, employing a convolutional neural network combined with a Minibatch optimization layer and an Attention mechanism, judges the authenticity of the images. The two continuously optimize through adversarial training, ultimately yielding prediction results. These results are then compared and verified with experimental results, achieving high consistency between experimental and AI predictions.

[0044] Example 2: A burner for implementing a combustion method based on nonequilibrium plasma and a GAN model, comprising: Gas source system 1: a ternary mixed gas source of hydrogen, ammonia, and methane, and an air source; Mass flow controller 2 is connected to a ternary mixed gas source of hydrogen, ammonia, and methane and an air source respectively, and is used to independently regulate the flow rates of hydrogen, ammonia, methane and air; The premixing chamber 3 is connected to the output of the mass flow controller 2 and is used to mix the gases to form a premixed gas. DBD reactor 4 is connected to the output end of premixing chamber 3 and includes quartz medium layer 5, AC high voltage electrode 6 and low voltage electrode 7. A swirling burner includes a nozzle and a swirling element 8, which is installed at the combustion outlet of the nozzle to organize combustion. The plasma power supply 9 is connected to the AC high-voltage electrode 6 and is used to output high-frequency high-voltage AC power. Oscilloscope 10, connected to plasma power supply 9, is used to monitor discharge voltage and current waveforms in real time; Quartz window provides an optical observation channel while isolating high-temperature flames; The PLIF laser detection system includes an Nd:YAG solid-state laser 12, a dye laser 13, and a sheet-like light source assembly 14, used to excite and collect OH free radical fluorescence signals; An ICCD camera 16, in conjunction with a 310nm bandpass filter 15, is positioned perpendicular to the laser sheet to collect fluorescence signals. A portable integrated flue gas analyzer 11 is installed at the exhaust port at the tail of the flame to measure the NOx concentration in the flue gas; The PC data acquisition and control terminal 17 is connected to the mass flow controller 2, plasma power supply 9, Nd:YAG solid-state laser 12, dye laser 13, ICCD camera 16 and portable integrated flue gas analyzer 11 via data connection cable 18, respectively, for real-time reading and storage of data and control of the working sequence of each component.

[0045] Specifically, the hydrogen / ammonia / methane ternary gas source and the air source enter the premixing chamber 3 after their flow rates are regulated by the mass flow controller 2, forming a premixed gas with an adjustable equivalence ratio. At the same time, the PC data acquisition and control terminal 17 reads the flow data of the mass flow controller 2 in real time through the data connection cable 18, and automatically adjusts the flow ratio of each gas according to the preset operating conditions to achieve precise fuel ratio control.

[0046] After the premixed gas enters the DBD reactor 4, the plasma power supply 9 outputs high-frequency high-voltage alternating current, generating a strong electric field between the high-voltage alternating current electrode 6 and the low-voltage alternating current electrode 7. Through the blocking effect of the quartz dielectric layer 5, a dielectric barrier discharge is formed, generating non-equilibrium plasma.

[0047] The oscilloscope 10 monitors the discharge waveform in real time through voltage and current probes, and transmits the data synchronously to the PC terminal for calculating discharge power and judging discharge stability. The PLIF laser detection system mainly consists of an Nd:YAG solid-state laser 12 that emits fundamental frequency light, which is then frequency-doubled and used to pump a dye laser 13 to generate 283 nm ultraviolet laser light.

[0048] The PC terminal controls the emission timing and energy of the Nd:YAG solid-state laser 12 and the dye laser 13 via data connection cable 18, ensuring that the laser pulses are synchronized with the plasma discharge cycle. The ultraviolet laser is shaped by the sheet-like light source assembly 14 and passes vertically through the PLIF detection zone at the center of the flame, exciting OH and NO free radicals to generate fluorescence signals. At the same time, the laser energy is monitored in real time by the energy meter and fed back to the PC terminal.

[0049] An ICCD camera 16, in conjunction with a 310 nm bandpass filter 15, acquires fluorescence signals perpendicular to the laser plate direction. The bandpass filter is used to filter out stray light and background radiation. A PC terminal controls the gating timing and gain parameters of the ICCD camera 16 via a data connection cable 18, and achieves nanosecond-level synchronization with the laser and plasma power supply 9 to ensure that the fluorescence signal is acquired within the optimal time window. A portable integrated flue gas analyzer 11 is installed at the exhaust port at the flame tail. The sampling probe extends into the high-temperature region at the flame tail, and the flue gas is transported to the analyzer via a heat tracing pipe.

[0050] The portable integrated flue gas analyzer 11 transmits the real-time measured NO and NOx concentration data to a PC terminal via a data connection cable 18. The data is then associated with and stored with the discharge parameters and PLIF images under the current operating conditions for subsequent analysis of the plasma's effect on NOx emission suppression.

[0051] The PC data acquisition and control terminal 17 connects to and controls the mass flow controller 2, plasma power supply 9, Nd:YAG solid-state laser 12, dye laser 13, and ICCD camera 16 via data connection cable 18. The timing controller is integrated into the PC terminal or used as a separate module to synchronize laser pulses, plasma discharge, and camera exposure, ensuring that OH-PLIF / NO-PLIF signals are acquired within the discharge stabilization period and avoiding the influence of plasma discharge electromagnetic interference on the ICCD camera 16 signal.

[0052] During the experiment, the quartz window 19 provides an optical observation channel while isolating the high-temperature flame. The cast iron round tube support 21 and connecting bolts 20 fix the entire burner structure to prevent high-voltage discharge and high-temperature flame from harming the operator.

[0053] Therefore, this application adopts the aforementioned combustion method and burner based on non-equilibrium plasma and GAN model, through a complete technical solution integrating non-equilibrium plasma catalysis and GAN prediction model. By systematically revealing the regulatory mechanism of plasma transport and ionization effects on combustion chemical kinetics, and aiming at the synergistic optimization of combustion performance improvement and low-pollution emissions, this not only provides a theoretical basis and engineering tools for efficient and low-pollution combustion of aero-engines in the low-carbon transition phase, but also provides technical support for meeting increasingly stringent international emission regulations such as CAEP / 12 and domestic ambient air quality standards. It has significant application value and guiding significance for promoting the green and sustainable development of the aviation industry.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and not to limit them. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of this application, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of this application.

Claims

1. A combustion method based on non-equilibrium plasma and a GAN model, characterized in that, Includes the following steps: S1: Constructing an experimental system for dielectric barrier discharge plasma combustion; S2: Keep the total flow rate of hydrogen, ammonia, methane and air premixed gas constant, and set the experimental conditions by adjusting the flow rates of hydrogen, ammonia and methane to change the hydrogen doping ratio and equivalence ratio; non-equilibrium plasma acts on the combustion process in a synergistic manner through transport effect, kinetic effect and thermal effect; S3: Experimental parameters during combustion are measured using a flame free radical measurement system, which includes a NO-PLIF system, an OH-PLIF system, and an NH2* chemiluminescence system. The two-dimensional spatial distribution of OH and NO free radicals is measured using the NO-PLIF and OH-PLIF systems; the concentration of NH2 free radicals is obtained through NH2 free radical diagnostic luminescence; and the temperature is measured using thermocouples. S4: Collect experimental data to construct a dataset. After preprocessing the dataset by removing outliers and normalizing it, divide it into a training set and a test set. S5: Construct a GAN prediction model, including a generator and a discriminator. The generator takes the concatenation of a random noise vector and a conditional vector as input and outputs predicted values ​​of OH radical intensity, NH2 radical intensity, NOx emission concentration, and a PLIF flame prediction image. The discriminator takes the concatenation of a conditional vector and a sample to be discriminated as input and outputs a probability value indicating that the sample is real experimental data. The generator and discriminator are optimized through adversarial training and mutual game. S6: Use the trained GAN prediction model to predict the untrained working conditions and output the prediction results. S7: Compare and verify the prediction results with the experimental data. When both the numerical prediction accuracy and the image prediction quality meet the preset conditions, output the GAN prediction model as the final prediction model.

2. The combustion method based on non-equilibrium plasma and GAN model as described in claim 1, characterized in that, The dielectric barrier discharge plasma combustion experimental system includes a ternary gas source, an air source, a mass flow controller, a premixing chamber, a DBD reactor, a swirling burner, a high-frequency high-voltage plasma power supply, an electrode system, a PLIF laser detection system, and a NO... X Analyzer, data acquisition and control terminal, ternary mixed gas source including hydrogen, ammonia and methane; The center frequency of the plasma power supply is 5–20 kHz, the maximum output power point is 10 kHz, and the maximum output voltage is 60 kV. DBD reactor electric field strength E / N With discharge voltage V The relationship is: ； in, d The distance between the electrodes. N The gas number density; A swirl burner consists of a nozzle and a swirler, with the swirler installed at the nozzle's combustion outlet. The swirl number of the swirler is expressed as: ； in, D i The inner diameter of the hydrocyclone blade; D o The outer diameter of the hydrocyclone blade; θ The angle of the hydrocyclone blades.

3. The combustion method based on non-equilibrium plasma and GAN model according to claim 2, characterized in that, The hydrogen doping ratio is expressed as: ； The equivalent ratio is expressed as: ； in, This refers to the hydrogen flow rate; This refers to the ammonia flow rate; This represents the methane flow rate. Airflow rate; The state is the chemical equivalence ratio.

4. The combustion method and burner based on non-equilibrium plasma and GAN model as described in claim 3, characterized in that, The transport effect promotes the reaction by generating plasma wind and enhancing turbulent mixing; The kinetic effect involves high-energy electron collisions that break down H2, NH3, and CH4 molecules to generate NH2, CH3, H, and OH reactive free radicals, as well as ions and excited-state species. Among these, NH2 + NO → N2 + H2O is the key pathway for NOx suppression. The thermal effect causes the temperature to rise, accelerates particle movement, and increases the reaction rate.

5. The combustion method based on non-equilibrium plasma and GAN model according to claim 1, characterized in that, The dataset input parameters include discharge voltage, discharge frequency, equivalence ratio, methane volume fraction, hydrogen gas integral, ammonia gas integral, and total fuel flow rate; the output parameters include OH radical intensity, NH2 radical intensity, NOx emission concentration, and radical flame image.

6. The combustion method based on non-equilibrium plasma and GAN model according to claim 1, characterized in that, Outlier removal is represented as: ； Normalized representation is: ； in, To obtain a measurement value from three repeated experiments under the same operating conditions. This is the average of three measurements taken under this operating condition. Standard deviation This is the original data. and These are the minimum and maximum values ​​of the parameter under all operating conditions, respectively.

7. The combustion method based on non-equilibrium plasma and GAN model according to claim 1, characterized in that, The generator takes as input a concatenation of a random noise vector sampled from a Gaussian distribution and a conditional vector, the conditional vector containing discharge voltage, equivalence ratio, and fuel flow parameters; the random noise is mapped into a decoupled intermediate latent code through a Mapping Network, and then a PLIF flame prediction image is generated through deconvolution and upsampling operations by a Generator Network, and the predicted concentrations of OH, NH2, and NOx are output. The discriminator takes the concatenation of the conditional vector and the sample to be discriminated as input, extracts image features through a convolutional neural network, and focuses on the key region of high intensity of free radicals in the flame through a Minibatch optimization layer and an Attention mechanism, and outputs the probability value that the sample is real experimental data.

8. The combustion method based on non-equilibrium plasma and GAN model according to claim 1, characterized in that, The generator loss function is expressed as: ； The discriminator loss function is expressed as: ； The GAN prediction model is optimized through a minimax game, and the total loss function is expressed as: ； in, For the expectation value operator, For balance coefficient, It is a random noise vector. For conditional vectors, Generate data for the generator's predictions. For real data, This represents the probability value output by the discriminator for the real data. For the discriminator to generate data The output probability value; This indicates sampling from the actual experimental data distribution. For the true data distribution, For noise distribution, This indicates sampling from a priori noise distribution.

9. The combustion method based on non-equilibrium plasma and GAN model according to claim 1, characterized in that, In S7, the accuracy of numerical prediction is determined by the coefficient of determination R. 2 The judgment is made based on the root mean square error (RMSE), and is expressed as follows: ； ； in, M The number of samples in the test set. For the first i The actual experimental values ​​of the group working conditions. For the corresponding predicted value, The arithmetic mean of the actual experimental values; Image prediction quality is evaluated by comparing the PLIF predicted image output by the generator with the real PLIF image acquired in the experiment, and considering three aspects: flame morphology, location of high-intensity free radical regions, and intensity gradient distribution.

10. A burner for implementing the combustion method based on non-equilibrium plasma and GAN model as described in any one of claims 1-9, characterized in that, include: The gas supply system consists of a ternary mixed gas source of hydrogen, ammonia, and methane, and an air source. The mass flow controller is connected to a hydrogen, ammonia, and methane ternary gas source and an air source respectively, and is used to independently regulate the flow rates of hydrogen, ammonia, methane and air; The premixing chamber, connected to the output of the mass flow controller, is used to mix the gases to form a premixed gas. The DBD reactor, connected to the output of the premixing chamber, includes a quartz dielectric layer, an AC high-voltage electrode, and a low-voltage electrode. A swirling burner includes a nozzle and a swirler. The swirler is installed at the combustion outlet of the nozzle to organize combustion. A plasma power supply, connected to an AC high-voltage electrode, is used to output high-frequency high-voltage AC power. An oscilloscope, connected to the plasma power supply, is used to monitor the discharge voltage and current waveforms in real time. Quartz window provides an optical observation channel while isolating high-temperature flames; The PLIF laser system, including an Nd:YAG solid-state laser and a dye laser, along with a sheet-like light source assembly, is used to excite and collect fluorescence signals of OH and NO free radicals; An ICCD camera, with a 310nm bandpass filter positioned perpendicular to the laser sheet, is used to acquire fluorescence signals; A portable integrated flue gas analyzer is installed at the exhaust port at the tail of the flame to measure the NOx concentration in the flue gas; The PC data acquisition and control terminal is connected to the mass flow controller, plasma power supply, Nd:YAG solid-state laser, dye laser, ICCD camera and portable integrated flue gas analyzer via data connection cables. It is used to read and store data in real time and control the working sequence of each component.

Images