Coal-fired boiler in-furnace denitration method based on multi-element data fusion

Abstract

The application provides a coal-fired boiler in-furnace denitration method based on multi-element data fusion, and belongs to the technical field of nitrogen oxide control of coal-fired boilers, which can at least partially solve the problem of low denitration efficiency and low reducing agent utilization rate under wide load conditions caused by the fact that the SNCR denitration in the prior art only relies on temperature field data. The application comprises: arranging a plurality of layers of double-fluid atomizing lances and smoke temperature monitoring probes and CO measuring point probes along the height direction of the boiler furnace; monitoring the cross-sectional temperature field data of each layer of lances in real time; constructing a digital twin model of the NOx concentration field and the CO concentration field based on numerical simulation and machine learning; determining the target operating lance layer within the preset temperature window according to the cross-sectional temperature field data; calculating and correcting the input amount of each lance according to the NOx value and the CO value; and performing feedback adjustment according to the tail flue NOx monitoring data. The application fuses the multi-element data of the temperature field, the NOx concentration field and the CO concentration field, and effectively improves the SNCR denitration efficiency and the reducing agent utilization rate under wide load conditions.

Description

Technical Field

[0001] This invention relates to the field of nitrogen oxide control technology in coal-fired boilers, and specifically to a method for denitrification in coal-fired boilers based on multi-source data fusion. Background Technology

[0002] Against the backdrop of a dual-carbon development environment, my country's new energy sector is rapidly developing. In the future, thermal power will continue to play a crucial role in peak shaving, frequency regulation, and renewable energy absorption. The long-term operation of coal-fired boilers under wide loads poses a significant challenge to ultra-low NOx emission control. At low loads, a low-NOx combustion atmosphere cannot be effectively formed, resulting in excessively high initial NOx emissions at the furnace outlet. This leads to a surge in ammonia injection during SCR denitrification of the tail-end flue gas, ammonia escape, and severe blockage and corrosion of the air preheater, threatening the safe, efficient, environmentally friendly, and economical operation of coal-fired units. Currently, most large coal-fired boilers in my country use SCR denitrification technology to achieve ultra-low NOx emissions, while SNCR denitrification technology is used less frequently.

[0003] Selective non-catalytic reduction denitrification technology involves injecting an amino-containing reducing agent into a furnace zone at a temperature of 850℃ to 1050℃, where the reducing agent rapidly decomposes and releases... It reacts with NOx in the flue gas to produce harmless nitrogen and water. The denitrification efficiency of SNCR technology is generally 30% to 50%. Taking urea as a reducing agent as an example, the main reaction is: , , ; The main factors affecting SNCR denitrification efficiency include the reaction temperature window, residence time within the temperature window, the uniformity of mixing of the injected reducing agent with the flue gas, the initial NOx concentration level, the stoichiometric ratio (NSR), and the ammonia slip concentration. In large pulverized coal boilers, the means of adjusting the residence time within the temperature window are limited, and the NSR is reflected by the utilization rate of the reducing agent, i.e., the ratio of denitrification efficiency to NSR. Therefore, to improve the SNCR denitrification efficiency and reducing agent utilization rate in large pulverized coal boilers, research and optimization should be carried out on three aspects: the reaction temperature window, the uniformity of mixing of the injected reducing agent with the flue gas, and the initial NOx concentration level.

[0004] In terms of SNCR denitrification technology, compared with SCR technology using catalysts, SNCR has a comprehensive investment and operating cost that is at least 30% to 40% lower, making it more economically attractive. However, currently, SNCR in large pulverized coal boilers is limited by factors such as temperature window, mixing effect, and variable load, resulting in low urea utilization and a denitrification efficiency of only 30% to 40%. Further improving SNCR denitrification efficiency and reducing agent utilization is of great significance for reducing the comprehensive investment and operating costs of flue gas denitrification devices.

[0005] Currently, the initial NOx emissions from pulverized coal boilers in my country burning medium-to-high volatile coal types are mostly around 200 mg / m³ at the furnace outlet. A few pulverized coal boilers or fluidized bed boilers burning high-quality bituminous coal have initial NOx emissions of 100 to 150 mg / m³. Given the initial NOx emission level of around 200 mg / m³ in my country's pulverized coal power plant boilers, to further achieve furnace outlet NOx emissions of no more than 50 mg / m³ over a wide load range, it is necessary to adopt new ultra-low NOx combustion technologies other than deep air staged combustion and enhanced SNCR in-furnace denitrification technology, while simultaneously coordinating with operation control technologies deeply coupled with ultra-low NOx combustion and SNCR denitrification. By stably increasing the SNCR denitrification efficiency of large pulverized coal boilers to over 50%, and simultaneously improving the utilization rate of the reducing agent in SNCR technology, these pulverized coal power plant boilers in my country can achieve ultra-low NOx emissions without using SCR technology, reducing overall investment and operating costs, and avoiding the environmental pollution caused by SCR catalyst replacement.

[0006] In the prior art, the patent "High-Efficiency SNCR Control Method (CN202510986497.7)" discloses a high-efficiency SNCR control method. This method acquires the boiler load during operation, the moisture content of the flue gas at the boiler outlet, and the flue gas temperature in each zone of the layer where the temperature measuring device is located. It then uses a pre-established three-dimensional temperature field calculation model to calculate the flue gas temperature at the spray gun in real time, and then deploys the spray guns in different layers accordingly. However, this method only provides temperature field data for each spray gun and does not know the NOx concentration or the CO concentration that affects the denitrification reaction in that area, thus preventing more precise adjustments.

[0007] The patent "SNCR Intelligent Denitrification System Based on Sound Sensing Technology (CN202510856433.5)" discloses an SNCR intelligent denitrification system based on sound sensing technology. It calculates the furnace combustion temperature field using a sound wave transceiver and then matches an analytical denitrification strategy. However, this patent only focuses on the influence of the temperature field on the SNCR denitrification effect, without considering the influence of NOx and CO concentration values ​​on the SNCR denitrification effect.

[0008] The patent "A Combined SNCR and SCR Denitrification System and Method for Coal-fired Boilers (CN201510768292.8)" discloses a combined denitrification system, in which the urea injection system includes short spray guns arranged in multiple layers in the upper part of the boiler furnace combustion zone and multi-hole long spray guns arranged in multiple layers at the boiler flame deflector. This method makes the ammonia-nitrogen molar ratio distribution at the inlet section of the SCR catalyst uniform, improving the denitrification efficiency of the SCR reactor, but it does not involve the optimization of SNCR denitrification efficiency based on multivariate data fusion.

[0009] The patent "A Novel Precision Ammonia Injection Control System for SNCR Denitrification System (CN202511438406.2)" discloses a precision ammonia injection control system that constructs a closed-loop control system for zoned real-time monitoring, trend prediction, and graded quantity control. However, on the one hand, this patent is applied to circulating fluidized bed units, and on the other hand, it uses data measured by a flue gas analyzer arranged on the separator outlet flue to adjust the ammonia injection rate, which cannot achieve real-time adjustment of the ammonia injection rate based on the NOx values ​​at each spray nozzle.

[0010] In summary, current technology only involves setting different layers of spray guns at different heights in the furnace and using sound waves or other methods to measure and construct the furnace temperature field. The SNCR spray gun operation strategy is then determined based on the temperature field data. However, this approach cannot achieve efficient SNCR denitrification under wide load conditions in large coal-fired boilers. Furthermore, the efficiency of SNCR denitrification is affected not only by temperature but also by parameters such as the CO concentration in the injection zone and the uniformity of the mixing between the injected reducing agent and the flue gas NOx. Moreover, the initial NOx concentration in the injection zone is also instructive for adjusting the amount of reducing agent injected in that zone.

[0011] Therefore, how to integrate multi-dimensional data from temperature field, NOx concentration field, and CO concentration field to improve the SNCR denitrification efficiency and reducing agent utilization rate under wide load conditions of large coal-fired boilers is a technical problem that urgently needs to be solved. Summary of the Invention

[0012] The present invention aims to solve at least one of the technical problems existing in the prior art, and to provide a method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion.

[0013] To achieve the above objectives, the present invention provides a method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion, comprising: Multiple layers of dual-fluid atomizing spray guns are arranged along the height of the boiler furnace. Each dual-fluid atomizing spray gun is equipped with a liquid flow regulating valve and an atomizing air pressure regulating valve. Multiple flue gas temperature monitoring probes are arranged at the arrangement height of each layer of dual-fluid atomizing spray guns, and multiple CO measuring probes are arranged on the boiler furnace wall. The cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer is monitored in real time by the smoke temperature monitoring probe. Digital twin models of the NOx and CO concentration fields in the boiler furnace under different operating conditions were constructed based on numerical simulation and machine learning. The target deployment spray gun layer within the preset temperature window is determined based on the cross-sectional temperature field data. The NOx and CO values ​​of each spray gun in each of the target commissioning spray gun layers are obtained according to the digital twin model. The amount of reducing agent added to each spray gun is calculated based on the NOx value, the target NOx value and the flue gas volume. At the same time, the amount of reducing agent added is corrected according to the CO value. The amount of reducing agent added is adjusted based on the NOx monitoring data from the boiler tail flue to achieve the target denitrification efficiency and reducing agent utilization rate under wide load conditions.

[0014] Furthermore, the boiler furnace is divided along the height direction into... The region, the The number of layers in each area corresponds to the arrangement of the dual-fluid atomizing spray gun and the smoke temperature monitoring probe; The boiler furnace is divided into left and right sections. The region, the The number of areas and the number of dual-fluid atomizing spray guns arranged in each layer are consistent; The boiler furnace is divided along the front-to-back direction into Each region has dimensions that meet the accuracy requirements for constructing the cross-sectional temperature field.

[0015] Furthermore, the real-time monitoring of the cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer via the smoke temperature monitoring probe includes: The radiation image signal of the cross section where the dual-fluid atomizing spray gun is located in each layer is acquired using the smoke temperature monitoring probe; The radiation image signal is analyzed using a radiation image processing method to obtain the cross-sectional temperature field distribution data of each layer of the dual-fluid atomizing spray gun. The number of smoke temperature monitoring probes in each layer is 4 to 6, and the number of layers of the smoke temperature monitoring probes is consistent with the number of layers of the dual-fluid atomizing spray gun.

[0016] Furthermore, the construction of digital twin models of the NOx and CO concentration fields of the boiler furnace under different operating conditions based on numerical simulation and machine learning includes: A numerical calculation model for the combustion of the boiler was established, and the CFD numerical simulation method was used to simulate the NOx distribution generated by the boiler furnace combustion under wide load conditions, different coal mill operation modes, different primary air damper openings and secondary air damper openings, different primary air ratios and secondary air ratios, and different burnout air ratios. The results of the CFD numerical simulation are verified and corrected by combining the measurement data of the CO measuring probe to obtain the CO distribution values ​​of each section of the boiler furnace. The NOx distribution values ​​and CO distribution values ​​are correlated with the historical DCS data of the unit to construct a database of furnace cross-sectional temperature field, NOx concentration field and CO concentration field under different unit operating modes and unit load conditions. Based on the database, a digital twin model is trained using machine learning methods, taking boiler operating parameters as input and the NOx concentration field and the CO concentration field as output.

[0017] Furthermore, the mapping relationship of the digital twin model is expressed as follows: ; ; in, This represents the mapping function obtained by training the machine learning method. For boiler load, The primary wind ratio, The ratio of secondary wind. For the burnout wind ratio, For fuel quantity, This refers to the operation mode of the coal mill. Furnace space coordinates NOx concentration at that location Furnace space coordinates CO concentration value at that location.

[0018] Furthermore, it also includes the step of constructing a database of atomization characteristics for dual-fluid atomizing spray guns: The dual-fluid atomizing spray gun was subjected to cold-state tests under different atomizing compressed air pressures and different liquid flow rates. Obtain the atomized particle size distribution, spray angle, and initial spray velocity data of the dual-fluid atomizing spray gun during the cold test; The atomized particle size distribution, the injection angle, and the initial injection velocity data are used to establish an atomization characteristic database indexed by the atomized compressed air pressure and liquid flow rate.

[0019] Further, determining the target deployment spray gun layer within the preset temperature window based on the cross-sectional temperature field data includes: Determine whether the cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer is within the preset temperature window of 850°C to 1050°C. If the cross-sectional temperature field data is within the preset temperature window, then the dual-fluid atomizing spray gun of that layer is identified as the target commissioning spray gun layer and put into operation; If the temperature of an individual dual-fluid atomizing spray gun in the target deployment spray gun layer is not within the preset temperature window, then the corresponding dual-fluid atomizing spray gun is turned off.

[0020] Furthermore, the step of adjusting the amount of reducing agent added based on the CO value includes: When the CO value is greater than the first preset CO threshold, the dual-fluid atomizing spray gun in the corresponding area is turned off; When the CO value is greater than or equal to the second preset CO threshold and less than or equal to the first preset CO threshold, the calculated amount of reducing agent added will be halved. When the CO value is less than the second preset CO threshold, the calculated amount of reducing agent is used.

[0021] Furthermore, the feedback adjustment is based on an optimization objective function, which is expressed as: ; in, To optimize the target value, , , The preset weighting coefficients, The measured NOx concentration value is from the tail flue of the boiler. For the target NOx value, This represents the amount of ammonia escaped. This represents the amount of reducing agent consumed.

[0022] Furthermore, the step of adjusting the amount of reducing agent added based on NOx monitoring data from the boiler tail flue includes: The measured NOx value at the SCR inlet section is obtained by installing a NOx monitoring instrument in the flue gas duct at the tail of the boiler. Compare the measured NOx value with the target NOx value; When the measured NOx value is greater than the target NOx value, the amount of reducing agent added to the corresponding spray gun is increased; when the measured NOx value is less than the target NOx value, the amount of reducing agent added to the corresponding spray gun is decreased. The flow rate of the reducing agent entering the corresponding spray gun is adjusted in real time by the liquid flow regulating valve installed in front of each of the dual-fluid atomizing spray guns.

[0023] The beneficial effects of this invention are as follows: This invention overcomes the limitations of existing SNCR technologies that rely solely on temperature field data for nozzle operation strategy control by integrating multi-dimensional data from temperature field, NOx concentration field, and CO concentration field. It achieves synergistic optimization control based on temperature, NOx concentration, and CO concentration, effectively improving the SNCR denitrification efficiency and reducing agent utilization rate of large coal-fired boilers under wide load conditions.

[0024] This invention employs a digital twin model based on numerical simulation and machine learning, which can acquire NOx and CO values ​​in the spray area of ​​each spray gun in real time. This provides a data basis for the fine adjustment of the reducing agent dosage and enables independent adjustment of the reducing agent dosage of each spray gun.

[0025] This invention introduces a CO concentration correction mechanism, which adjusts the amount of reducing agent added according to the CO concentration value in the spraying area, thus avoiding the problems of decreased denitrification efficiency and ammonia escape caused by spraying too much reducing agent in areas with excessively strong reducing atmosphere.

[0026] This invention establishes a feedback regulation mechanism based on an optimized objective function, which comprehensively considers three indicators: NOx emission deviation, ammonia slip, and reducing agent consumption, thereby achieving a balanced optimization of denitrification efficiency and economy. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the system architecture of the in-furnace denitrification method for coal-fired boilers based on multi-source data fusion, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the regulating valve group and flow meter group of the dual-fluid atomizing spray gun according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the furnace cross-sectional area division according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the in-furnace denitrification method for coal-fired boilers based on multi-source data fusion, according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates this application. It should be understood that the specific embodiments described in this specification are merely for explaining this application and are not intended to limit it.

[0029] The in-furnace denitrification method for coal-fired boilers based on multi-source data fusion of the present invention is implemented based on an SNCR denitrification system. The SNCR denitrification system includes a urea solution supply system, a demineralized water supply system, a compressed air supply system, a metering and dilution module, a distribution module, an injection assembly, a furnace flue gas temperature monitoring system, a furnace wall CO testing system, an industrial control computer, a server, and a monitor.

[0030] The urea solution supply system stores and transports urea solution, delivering it from the storage tank to the metering and dilution module via pipelines. The demineralized water supply system provides dilution water, which is mixed with the urea solution in the metering and dilution module according to a preset ratio to prepare a reducing agent solution of a preset concentration. The compressed air supply system provides compressed air for atomization of the dual-fluid atomizing spray guns; the compressed air pressure can be adjusted within the range of 0.1 MPa to 0.8 MPa. The metering and dilution module precisely meters and mixes the urea solution and demineralized water. The distribution module distributes the prepared reducing agent solution to each dual-fluid atomizing spray gun. The spray assembly includes multi-layer dual-fluid atomizing spray guns arranged along the height of the boiler furnace. The furnace flue gas temperature monitoring system consists of multi-layer CCD probes, monitoring the temperature distribution at various cross-sections of the furnace in real time. The furnace wall CO testing system consists of multiple CO measuring probes installed on the furnace wall to acquire CO concentration data. The industrial control computer is responsible for data acquisition, data processing, and control command output. The server is responsible for running the digital twin model and machine learning algorithms. The monitor displays the system's operating status and parameters.

[0031] Example 1 This embodiment uses a counter-firing boiler in a 660 MW ultra-supercritical coal-fired unit as an application scenario. The boiler has a furnace cross-section of 20 m by 20 m and a furnace height of approximately 60 m. It is equipped with 6 medium-speed coal mills and is designed to burn medium-to-high volatile bituminous coal. The initial NOx emissions across the entire load range are between 180 and 300 mg / m³.

[0032] See Figure 1 The denitrification method in this embodiment includes the following steps: Step S1: Multi-layer dual-fluid atomizing spray gun arrangement and monitoring probe installation.

[0033] Based on the flue gas temperature variation patterns of the 660 MW unit under various loads, three layers of SNCR dual-fluid atomizing spray guns are arranged along the furnace height. The first layer of spray guns is located approximately 38 m below the flame deflector, with 5 spray guns each along the front and rear walls of the boiler furnace, totaling 10 spray guns. The second layer of spray guns is located approximately 42 m above the flame deflector, with 10 spray guns arranged along the front wall of the boiler furnace. The third layer of spray guns is located approximately 46 m above the flame deflector, with 10 spray guns arranged along the front wall of the boiler furnace. A total of 30 spray guns are arranged across the three layers.

[0034] See Figure 2An electric flow regulating valve is installed on the inlet liquid line of each dual-fluid atomizing spray gun, allowing independent adjustment of the reducing agent flow rate entering each spray gun, with a flow rate adjustment range of 0 to 500 L / h. An electric pressure regulating valve is also installed on the inlet compressed air line of each dual-fluid atomizing spray gun, allowing independent adjustment of the atomizing air pressure of each spray gun, with a pressure adjustment range of 0.1 MPa to 0.8 MPa. An electromagnetic flow meter is also installed on the inlet liquid line of each spray gun for real-time monitoring of the actual flow rate.

[0035] CCD flue gas temperature monitoring probes were installed in the three layers of spray gun installation areas. In the first layer, five CCD probes were installed at an elevation of 38 m, positioned on the front wall, rear wall, and both side walls of the furnace. In the second layer, four CCD probes were installed at an elevation of 42 m, two on the front wall and one on each side wall. In the third layer, four CCD probes were installed at an elevation of 46 m, two on the front wall and one on each side wall (or one on the front wall and one on each side wall). The CCD probes in each layer were evenly distributed around the furnace cross-section to ensure complete coverage of the temperature distribution across the entire cross-section.

[0036] CO measuring probes are installed on the walls of the three layers of spray gun areas in the boiler furnace. Four CO measuring probes are installed in the first layer, positioned at the center of each of the four furnace walls. Three CO measuring probes are installed in the second layer. Three CO measuring probes are installed in the third layer. A total of 10 CO measuring probes are installed.

[0037] Step S2: Divide the boiler furnace area.

[0038] See Figure 3 The boiler furnace is divided into zones. Along the height of the furnace, the furnace is divided into... The plan defines three zones, corresponding to the placement of three layers of dual-fluid atomizing spray guns and CCD smoke temperature monitoring probes. Within each zone, based on the cross-sectional temperature field data acquired by the CCD smoke temperature monitoring probes, the cross-section of that layer is divided into a two-dimensional grid. Zone 1 covers an elevation range of 36 m to 40 m, corresponding to the first layer of spray guns. Zone 2 covers an elevation range of 40 m to 44 m, corresponding to the second layer of spray guns. Zone 3 covers an elevation range of 44 m to 48 m, corresponding to the third layer of spray guns.

[0039] Divide the furnace chamber into two sections along the left and right sides. There are 10 zones, corresponding to the number of dual-fluid atomizing spray guns arranged on each layer. Each zone is 2 m wide.

[0040] Divide the furnace chamber into sections along the front-to-back direction. Each region is 1 m in size, which meets the accuracy requirements for measuring the furnace temperature field and constructing the cross-sectional temperature field.

[0041] The CCD flue gas temperature monitoring probe acquires the two-dimensional cross-sectional temperature field of each layer, and combines it with the height direction stratification between each layer to form a quasi-three-dimensional temperature distribution in the furnace.

[0042] After the above three-dimensional region division, the entire furnace is divided into: Each area is approximately 4 m high, 2 m wide, and 1 m deep.

[0043] Step S3: Real-time cross-sectional temperature field monitoring.

[0044] Radiation image signals from various cross-sections of the furnace are acquired using CCD flue gas temperature monitoring probes installed on each layer. The CCD probes are area array CCD sensors, operating in the visible to near-infrared band, with a response range covering 400 nm to 1100 nm. The acquisition frequency of each CCD probe is set to 10 frames per second.

[0045] A radiation image processing method is used to analyze the radiation image signals acquired by a CCD probe. The specific process of the radiation image processing method is as follows: dark current correction and non-uniformity correction are performed on the radiation images acquired by the CCD probe; a mapping relationship between radiation intensity and temperature is established based on Planck's radiation law; and the two-dimensional distribution of the cross-sectional temperature field is reconstructed through a tomographic reconstruction algorithm of multi-view CCD images.

[0046] The mapping relationship between radiation intensity and temperature is based on Planck's radiation law and is expressed as: ; in, wavelength Temperature The corresponding radiation intensity, The first radiation constant, W·m², The second radiation constant, m·K.

[0047] After analysis using radiation image processing methods, the cross-sectional temperature field distribution data of each spray gun section was obtained. Taking the first section as an example, the cross-sectional temperature field at an elevation of 38 m under the 660 MW full-load operation condition exhibits a distribution characteristic of high temperature in the center and low temperature around the perimeter. The temperature in the central region is approximately 1000℃ to 1100℃, while the temperature near the surrounding walls is approximately 800℃ to 900℃.

[0048] Step S4: Construct digital twin models of the NOx concentration field and the CO concentration field.

[0049] The construction of a digital twin model is divided into three stages: CFD numerical simulation, database construction, and machine learning training.

[0050] In the CFD numerical simulation phase, a three-dimensional CFD numerical calculation model of boiler combustion was established using the actual boiler of the 660 MW ultra-supercritical coal-fired unit as a prototype. The CFD model was built using FLUENT software, with a mesh size of approximately 3 million. Turbulence model, P1 radiation model, pulverized coal particle discrete phase model and NOx emission model.

[0051] The furnace combustion process of this boiler under a wide load condition was simulated using CFD numerical simulation. The simulation covered a load range of 300 MW to 660 MW, divided into 13 load levels with a step size of 30 MW. At each load level, different pulverizer operation modes were simulated, including conditions with 5 pulverizers, 4 pulverizers, and 3 pulverizers. Simultaneously, different primary and secondary air damper openings were simulated, with the primary air damper opening varying from 60% to 100% in 10% steps, and the secondary air damper opening varying from 40% to 100% in 10% steps. Different primary air ratios, secondary air ratios, and burnout air ratios were also simulated, with the primary air ratio varying from 20% to 30%, the secondary air ratio from 40% to 55%, and the burnout air ratio from 15% to 35%.

[0052] The NOx distribution values ​​at various cross-sections of the boiler furnace under different operating conditions were obtained through the aforementioned CFD numerical simulations. Simultaneously, the CO distribution results obtained from the CFD simulations were verified and corrected by combining the measurement data from CO measuring probes installed at the boiler wall. The verification and correction method involved comparing the measured CO concentration values ​​from the CO measuring probes with the corresponding CO concentration values ​​from the CFD simulations, and calculating the correction coefficient between the two. The correction factor is applied to the CO distribution values ​​of the entire cross section in the CFD simulation to obtain the corrected CO distribution values ​​for each cross section.

[0053] The formula for calculating the correction factor is: ; in, This is the CO distribution correction factor. This represents the number of CO measurement probes. For the first The measured CO concentration values ​​at each CO measuring point probe. This represents the CO concentration value at the corresponding location in the CFD simulation.

[0054] During the database construction phase, the NOx and CO distribution values ​​obtained from CFD numerical simulations were correlated with the unit's historical DCS data. The historical DCS data included operating parameters such as boiler load, pulverizer operation mode, damper opening, primary air ratio, secondary air ratio, burnout air ratio, total air volume, coal feed rate, and flue gas oxygen content. Through data correlation, databases of furnace cross-sectional temperature, NOx, and CO concentration fields under different unit operating modes and load conditions were constructed. The database contains approximately 2000 sets of data for different operating conditions.

[0055] During the machine learning training phase, a digital twin model is trained using machine learning methods based on the aforementioned database. The digital twin model employs a deep neural network architecture with an input layer dimension of 6, corresponding to the boiler load. Primary wind ratio Secondary wind ratio Burnout wind ratio Fuel quantity Coal mill commissioning method The network contains four hidden layers, with 128, 256, 256, and 128 neurons in each layer, respectively. The hidden layers use the ReLU activation function. The output layer corresponds to the number of regions in the furnace cross-section, outputting the NOx and CO concentration values ​​for each region.

[0056] The mapping relationship of the digital twin model is represented as follows: , ; in, This represents the mapping function obtained from training a deep neural network. For boiler load, The primary wind ratio, The ratio of secondary wind. For the burnout wind ratio, For fuel quantity, This refers to the operation mode of the coal mill. Furnace space coordinates NOx concentration at that location Furnace space coordinates CO concentration value at that location.

[0057] The model was trained using the Adam optimizer with a learning rate of 0.001, 500 training epochs, and a batch size of 32. The ratio of the training dataset to the validation dataset was 8:2. After training, the model's root mean square error (RMSE) for NOx concentration prediction on the validation dataset was less than 15 mg / m³, and its RMS error for CO concentration prediction was less than 50 μL / L.

[0058] Step S5: Construct a database of atomization characteristics for dual-fluid atomizing spray guns.

[0059] A cold-state atomization test was conducted on the dual-fluid atomizing spray gun selected in this embodiment. The cold-state test was carried out on a spray test bench equipped with a phase Doppler analyzer, a high-speed camera, and a flow meter.

[0060] The test conditions were set as follows: the atomizing compressed air pressure varied from 0.1 MPa to 0.8 MPa, with a step size of 0.1 MPa, for a total of 8 pressure levels. The liquid flow rate varied from 50 L / h to 500 L / h, with a step size of 50 L / h, for a total of 10 flow rate levels. (Total...) Group test conditions.

[0061] Under each set of test conditions, the following atomization characteristic parameters were measured and recorded: atomized particle size distribution, including... , and Three characteristic particle size values, among which This indicates the particle size corresponding to 10% of the cumulative volume. This represents the median particle size corresponding to 50% of the cumulative volume. The particle size represents 90% of the cumulative volume; the spray angle is determined by capturing the spray pattern with a high-speed camera and measuring the spray cone angle; the initial spray velocity is determined by measuring the initial velocity of the droplets at the nozzle exit and the atomized particle size distribution using a phase Doppler analyzer.

[0062] Taking the operating conditions of atomized compressed air pressure of 0.3 MPa and liquid flow rate of 200 L / h as an example, the measurements were obtained. μm, μm, μm, the injection angle is 25°, and the initial injection velocity is 100 m / s.

[0063] Taking the operating conditions of atomized compressed air pressure of 0.6 MPa and liquid flow rate of 200 L / h as an example, the measurements were obtained. μm, μm, μm, the injection angle is 35°, and the initial injection velocity is 120 m / s.

[0064] The atomization characteristic parameters of all 80 test conditions were established into an atomization characteristic database indexed by atomizing compressed air pressure and liquid flow rate. In actual operation, the atomizing air pressure setting value of each spray gun can be determined by querying the atomization characteristic database according to the required reducing agent flow rate and target atomized particle size.

[0065] Step S6: Determine the target spray gun layer.

[0066] During boiler operation, the industrial control computer receives cross-sectional temperature field data in real time from the CCD flue gas temperature monitoring probes at each level after radiation image processing.

[0067] Determine whether the cross-sectional temperature field data of each layer of dual-fluid atomizing spray gun is within a preset temperature window of 850℃ to 1050℃. If the average temperature of the cross-sectional temperature field of a certain layer is within the range of 850℃ to 1050℃, then the dual-fluid atomizing spray gun layer of that layer is identified as the target spray gun layer for operation and put into operation.

[0068] Taking the 660 MW unit operating at a 500 MW load as an example, the cross-sectional temperature field monitoring showed that: the average temperature of the first cross-section was 980℃, within the preset temperature window, and was therefore identified as the target spray gun layer for commissioning. The average temperature of the second cross-section was 910℃, also within the preset temperature window, and was therefore identified as the target spray gun layer for commissioning. The average temperature of the third cross-section was 820℃, lower than the lower limit of the preset temperature window of 850℃, and was therefore not put into operation.

[0069] In the first and second floors, which were designated as the target spray gun operation layers, the temperature of each spray gun was further checked individually. Taking the first floor as an example, among the 10 spray guns, the temperature of the third spray gun on the front wall was 830℃, which is below 850℃, so this spray gun was shut down. The temperatures of the remaining 9 spray guns were all within the range of 850℃ to 1050℃, and they were put into normal operation. The temperatures of all 10 spray guns on the second floor were within the range of 850℃ to 1050℃, and all were put into operation.

[0070] Step S7: Calculate the amount of reducing agent added to each spray gun.

[0071] The NOx and CO values ​​at the location of each spray gun in each target spray gun layer are obtained based on the digital twin model. Taking the 660 MW unit at a 500 MW load as an example, the current operating parameters are input into the digital twin model: boiler load. MW, primary wind ratio Secondary wind ratio Burnout wind ratio Coal feed The coal mill has a capacity of t / h and operates with 5 units in operation.

[0072] The digital twin model outputs the NOx concentration values ​​at each spray gun location in the first layer. The NOx concentration at the first spray gun location on the front wall is 280 mg / m³, the NOx concentration at the second spray gun location on the front wall is 310 mg / m³, and the NOx concentrations at each spray gun location on the rear wall range from 260 to 340 mg / m³. The CO concentration values ​​at each spray gun location in the first layer range from 80 to 350 μL / L.

[0073] The reducing agent dosage for each spray nozzle was calculated based on the NOx reading, target NOx reading, and flue gas volume. The target NOx reading was set at 140 mg / m³. The flue gas volume generated by the boiler at a 500 MW load is approximately... Nm³ / h. A total of 19 spray guns are in operation. The amount of flue gas covered by a single spray gun is: ; in, The amount of smoke covered by a single spray gun. This refers to the total number of spray guns in operation.

[0074] Taking the first spray gun on the front wall of the first floor as an example, the NOx concentration at the location of this spray gun is: mg / m³, the target NOx concentration is mg / m³. The amount of NOx to be removed is: ; Based on the stoichiometric relationship of the SNCR denitrification reaction, and considering a stoichiometric ratio (NSR) of 1.1, the required urea solution flow rate for a single spray gun is: ; in, g / mol is the molar mass of urea. g / mol is the molar mass of NO. kg / L is the density of urea solution. This refers to the mass concentration of the urea solution. It is the chemical equivalence ratio.

[0075] Step S8: Adjust the amount of reducing agent added based on the CO value.

[0076] The calculated amount of reducing agent added is corrected based on the CO concentration at each spray gun location. The correction rule sets the first preset CO threshold at 1000 μL / L and the second preset CO threshold at 250 μL / L.

[0077] When the CO concentration at the location of the spray gun exceeds 1000 μL / L, the spray gun is turned off. In this embodiment, the CO concentration values ​​at each spray gun location in the first and second layers do not exceed 1000 μL / L, so there is no need to turn off the spray gun.

[0078] When the CO concentration at the location of the spray gun is greater than or equal to 250 μL / L and less than or equal to 1000 μL / L, the calculated amount of reducing agent should be halved. Taking the fourth spray gun on the front wall of the first floor as an example, the CO concentration at this location is 350 μL / L, which is within the range of 250 to 1000 μL / L, so the amount of reducing agent for this spray gun should be halved.

[0079] When the CO concentration at the location of the spray gun is less than 250 μL / L, the calculated amount of reducing agent is used. Taking the first spray gun on the front wall of the first floor as an example, the CO concentration at this location is 120 μL / L, which is less than 250 μL / L, so the amount of reducing agent calculated in step S7 is used directly.

[0080] After CO correction is applied to the reducing agent input of the 19 operating spray guns in the first and second layers, the reducing agent flow rate entering each spray gun is automatically adjusted by the electric flow regulating valve installed in front of each spray gun, and the atomizing air pressure is set according to the atomization characteristic database by the electric pressure regulating valve installed in front of each spray gun.

[0081] Step S9: Feedback adjustment of reducing agent input.

[0082] The measured NOx values ​​at the SCR inlet section were obtained using a NOx monitoring instrument installed at the SCR inlet of the boiler tail flue. The NOx monitoring instrument used an online continuous monitoring method with a sampling period of 5 seconds.

[0083] The measured NOx value was compared with the target NOx value of 140 mg / m³. In this embodiment, after adjustments in steps S7 and S8, the measured NOx value at the SCR inlet section was 155 mg / m³, which is higher than the target NOx value of 140 mg / m³.

[0084] Feedback adjustment is based on an optimization objective function. The optimization objective function is expressed as: ; in, To optimize the target value, , , The preset weighting coefficients, The measured NOx concentration value is from the boiler tail flue. mg / m³ is the target NOx value. This represents the amount of ammonia escaped. This represents the amount of reducing agent consumed.

[0085] Since the measured NOx value of 155 mg / m³ was higher than the target NOx value of 140 mg / m³, the system increased the amount of reducing agent added to each operating spray gun according to a preset adjustment step size. The adjustment step size was set to 5% of the current calculated value. After three adjustment cycles, the measured NOx value at the SCR inlet section decreased to 145 mg / m³, close to the target NOx value of 140 mg / m³.

[0086] Simultaneously, ammonia slip is monitored. An NH3 slip monitoring instrument is installed at the SCR outlet of the boiler tail flue to monitor ammonia slip in real time. When the ammonia slip exceeds a preset ammonia slip threshold, the reducing agent input of the corresponding spray gun is reduced. In this embodiment, the ammonia slip threshold is set to 7 mg / m³.

[0087] After feedback adjustment and stabilization, this embodiment achieves an SNCR denitrification efficiency of approximately 50% and a reducing agent utilization rate of approximately 45% under a 500 MW load condition. The SCR inlet NOx concentration is stabilized between 140 and 145 mg / m³, and the ammonia slip is maintained between 6 and 7 mg / m³.

[0088] Example 2 This embodiment uses a W-flame boiler in a 350 MW subcritical coal-fired unit as an example. The W-flame boiler has a furnace cross-section of 18 m by 12 m and a furnace height of approximately 50 m. It is equipped with four medium-speed coal mills, designed to burn anthracite, and its initial NOx emissions are between 550 and 900 mg / m³ across the entire load range.

[0089] Step S1: Multi-layer dual-fluid atomizing spray gun arrangement and monitoring probe installation.

[0090] Based on the temperature variation patterns of the furnace flue gas under various loads in the 350 MW unit's W-flame boiler, four layers of SNCR dual-fluid atomizing spray guns are arranged along the furnace height. The first layer of spray guns is located approximately 32 m below the flame deflector, with four guns along the front wall and four along the rear wall, totaling eight guns. The second layer is located approximately 36 m above the flame deflector, with eight guns along the front wall, totaling eight guns. The third layer is located approximately 40 m above the flame deflector, with eight guns along the front wall, totaling eight guns. The fourth layer uses multi-nozzle long spray guns, located approximately 44 m above the furnace outlet screen superheater and high-temperature superheater, with one multi-nozzle long spray gun on each side wall, totaling two. A total of 26 spray guns are used across the four layers.

[0091] Each dual-fluid atomizing spray gun is equipped with an electric flow regulating valve on its inlet liquid line, with a flow rate adjustment range of 0 to 400 L / h. Each dual-fluid atomizing spray gun is also equipped with an electric pressure regulating valve on its inlet compressed air line, with a pressure adjustment range of 0.1 MPa to 0.8 MPa.

[0092] CCD smoke temperature monitoring probes were installed in the four spray gun installation areas, with four CCD probes installed on each layer, for a total of 16 CCD probes. CO measuring probes were also installed on the walls of the four spray gun areas, with four CO measuring probes installed on each layer, for a total of 16 CO measuring probes.

[0093] Step S2: Divide the boiler furnace area.

[0094] The furnace of the W-flame boiler is divided into three-dimensional regions. Along the height of the furnace, the furnace is divided into... Each area corresponds to the placement of four layers of dual-fluid atomizing spray guns and CCD flue gas temperature monitoring probes. The furnace is divided into sections along the left-right direction. Each zone corresponds to the number of dual-fluid atomizing spray guns arranged in each layer. The furnace is divided into zones along the front-to-back direction. The furnace is divided into several zones, each measuring 1 meter. A small area.

[0095] Step S3: Real-time cross-sectional temperature field monitoring.

[0096] Radiation image signals from various cross-sections of the furnace were acquired using CCD flue gas temperature monitoring probes installed at each level. Radiation image processing methods were then used to analyze the radiation image signals, obtaining the cross-sectional temperature field distribution data for each level. Due to its unique combustion method of ignition above the arch and combustion below the arch, the W-flame boiler exhibits higher temperatures in the lower part of the furnace and lower temperatures in the upper part, resulting in a temperature field distribution that differs from that of an opposed-fired boiler. Under 350 MW full-load operation, the average temperature of the first cross-section is approximately 1020℃, the second cross-section is approximately 950℃, the third cross-section is approximately 880℃, and the fourth cross-section is approximately 790℃.

[0097] Step S4: Construct digital twin models of the NOx concentration field and the CO concentration field.

[0098] A numerical combustion model for the W-flame boiler was established. Because the W-flame boiler burns anthracite, its combustion characteristics differ significantly from those of the offset combustion boiler burning bituminous coal, resulting in higher initial NOx emission levels.

[0099] The CFD numerical simulation covered a load range of 180 MW to 350 MW, divided into 9 load levels with a step size of 20 MW. At each load level, different coal mill operation methods, different damper opening combinations, and different airflow ratios were simulated. Approximately 1500 sets of operating data were simulated in total.

[0100] The CFD simulation results were verified and corrected using measured data from CO measuring probes. A database was constructed, and a digital twin model was trained using machine learning methods. The model's input parameters and network architecture were the same as in Example 1, and the output corresponds to the number of regions in the furnace cross-section of the W-flame boiler.

[0101] Step S5: Construct a database of atomization characteristics for dual-fluid atomizing spray guns.

[0102] The dual-fluid atomizing spray gun used in this embodiment is the same model as in Embodiment 1, and the atomization characteristic database can be directly reused. For the multi-nozzle long spray gun used in the fourth layer, a cold-state atomization test was conducted to establish an atomization characteristic database for the multi-nozzle long spray gun.

[0103] Step S6: Determine the target spray gun layer.

[0104] Taking the 350 MW unit operating at a 250 MW load as an example, the cross-sectional temperature field monitoring results show that: the average temperature of the first cross-section is 960℃, within the preset temperature window; the average temperature of the second cross-section is 890℃, also within the preset temperature window; the average temperature of the third cross-section is 840℃, below 850℃, and therefore not in operation; the average temperature of the fourth cross-section is 750℃, below 850℃, and therefore not in operation.

[0105] In the first and second layers, which were designated as the target spray gun operation areas, the temperature of each spray gun was checked. In the first layer of eight spray guns, the temperature of the fourth spray gun on the back wall was 840℃, below 850℃, so this spray gun was shut down. The remaining seven spray guns were put into normal operation. In the second layer of eight spray guns, the temperatures of all locations were within the range of 850℃ to 1050℃, and all were put into operation. The total number of target spray guns in operation was 15.

[0106] Step S7: Calculate the amount of reducing agent added to each spray gun.

[0107] Input the current operating parameters into the digital twin model: boiler load MW, primary wind ratio Secondary wind ratio Burnout wind ratio Coal feed The coal mill has a capacity of t / h and operates with 3 units in operation.

[0108] The digital twin model outputs NOx and CO concentration values ​​at each spray nozzle location. Because the W-flame boiler burns anthracite, the NOx concentration at each spray nozzle location is generally high, ranging from 320 to 420 mg / m³.

[0109] The target NOx value is set at 210 mg / m³. The amount of flue gas produced by the boiler at a 250 MW load is approximately... Nm³ / h. A total of 15 spray guns were put into operation. The reducing agent input for each spray gun was calculated using the same method as in Example 1.

[0110] Step S8: Adjust the amount of reducing agent added based on the CO value.

[0111] Due to its unique combustion method, the W-flame boiler has a strong reducing atmosphere in the lower part of the furnace, resulting in a generally high CO concentration. The CO concentration at each spray gun location in the first layer ranges from 150 to 600 μL / L, with some spray gun locations exceeding 250 μL / L.

[0112] According to the revised rules, the amount of reducing agent added to spray guns with CO concentrations in the range of 250 to 1000 μL / L should be halved. The CO concentration at the second spray gun location on the front wall of the first layer is 450 μL / L, so the amount of reducing agent added is halved. The CO concentration at the second spray gun location on the rear wall of the first layer is 520 μL / L, so the amount of reducing agent added is halved. For all spray gun locations on the second layer, the CO concentration is between 80 and 200 μL / L, all less than 250 μL / L, so the calculated amount of reducing agent added is used.

[0113] Step S9: Feedback adjustment of reducing agent input.

[0114] The measured NOx value is obtained by the NOx monitoring instrument at the SCR inlet. The measured NOx value is compared with the target NOx value of 210 mg / m³, and feedback adjustment is performed according to the optimized objective function.

[0115] Due to the high initial NOx emissions of W-flame boilers, a large amount of reducing agent is required, necessitating close monitoring of ammonia slip during feedback regulation. In this embodiment, the weighting coefficient is set to... , , Compared with Example 1, the weight of ammonia slip is increased.

[0116] After feedback adjustment and stabilization, this embodiment achieved an SNCR denitrification efficiency of approximately 48% and a reducing agent utilization rate of approximately 40% under a 250 MW load condition. The SCR inlet NOx concentration was stabilized between 210 and 220 mg / m³, and the ammonia slip was maintained between 6 and 7 mg / m³.

[0117] Example 3 This embodiment uses a tangential combustion boiler in a 300 MW subcritical coal-fired unit as an application scenario. The furnace cross-section of this tangential combustion boiler is 14 m by 14 m, the furnace height is about 55 m, it is equipped with 5 medium-speed coal mills, and it is designed to burn medium-to-high volatile bituminous coal. The original NOx emissions are between 150 and 220 mg / m³ across the entire load range.

[0118] Step S1: Multi-layer dual-fluid atomizing spray gun arrangement and monitoring probe installation.

[0119] Based on the temperature variation patterns of the furnace flue gas under various loads in the 300 MW unit's tangential combustion boiler, three layers of SNCR dual-fluid atomizing spray guns are arranged along the furnace height. The first layer of spray guns is located approximately 35 m below the flame deflector, with 6 spray guns along the front wall of the boiler furnace and 2 spray guns on each side wall, totaling 10 spray guns. The second layer of spray guns is located approximately 39 m above the flame deflector, with 7 spray guns along the front wall of the boiler furnace. The third layer of spray guns is located approximately 43 m above the flame deflector, with 7 spray guns along the front wall of the boiler furnace. A total of 24 spray guns are arranged across the three layers.

[0120] In a tangential combustion boiler, the airflow rotates, and the temperature and concentration fields exhibit rotational characteristics. The installation angle of each dual-fluid atomizing nozzle must take into account the direction of airflow rotation to ensure that the injected reducing agent is fully mixed with the rotating airflow.

[0121] Electric flow control valves and electric pressure control valves are installed on the inlet liquid pipes and compressed air pipes of each spray gun. CCD flue gas temperature monitoring probes are installed in the three spray gun installation areas, with six CCD probes per layer, totaling 18 CCD probes across the three layers. The temperature field of the tangential combustion boiler exhibits rotational characteristics, requiring more CCD probes per layer to ensure the accuracy of temperature field reconstruction. CO measuring probes are installed on the walls of the three spray gun areas, with four CO measuring probes per layer, totaling 12 CO measuring probes across the three layers.

[0122] Step S2: Divide the boiler furnace area.

[0123] The furnace of a tangentially circular combustion boiler is divided into three-dimensional regions. Along the height of the furnace, the furnace is divided into... The furnace is divided into several areas along the left-right direction. The furnace is divided into several areas along its front-to-back direction. The furnace is divided into several zones, each measuring 1 meter. A small area.

[0124] Step S3: Real-time cross-sectional temperature field monitoring.

[0125] Radiation image signals from various cross-sections of the furnace were acquired using CCD flue gas temperature monitoring probes installed on each layer, and analyzed using radiation image processing methods. The temperature field of the tangentially circular combustion boiler exhibits a rotating distribution characteristic, and the temperature field reconstruction algorithm must consider the asymmetry in temperature distribution caused by airflow rotation. Under 300 MW full-load operation, the average temperature of the first cross-section is approximately 1030℃, the average temperature of the second cross-section is approximately 940℃, and the average temperature of the third cross-section is approximately 860℃.

[0126] Step S4: Construct digital twin models of the NOx concentration field and the CO concentration field.

[0127] A numerical combustion model for the tangential combustion boiler was established. The CFD model of the tangential combustion boiler needs to accurately simulate the airflow rotation characteristics of tangential combustion at the four corners. The simulation conditions covered a load range of 150 MW to 300 MW, divided into 11 load levels with a step size of 15 MW. Approximately 1800 sets of operating condition data were simulated.

[0128] The data from the CO measurement probes were used for verification and correction, a database was constructed, and a digital twin model was trained using machine learning methods.

[0129] Step S5: Construct a database of atomization characteristics for dual-fluid atomizing spray guns.

[0130] The method for constructing the atomization characteristic database is the same as in Example 1, and the atomization characteristic database established in Example 1 can be directly reused.

[0131] Step S6: Determine the target spray gun layer.

[0132] Taking the 300 MW unit operating at a 200 MW load as an example, the cross-sectional temperature field monitoring results show that: The average temperature of the first cross-section is 960℃, which is within the preset temperature window. The average temperature of the second cross-section is 880℃, which is also within the preset temperature window. The average temperature of the third cross-section is 810℃, which is below 850℃, and therefore it will not be put into operation.

[0133] In the first and second layers, designated as the target spray gun deployment areas, the temperature of each spray gun's location was checked. The temperatures of all eight spray guns on the first layer were within the range of 870℃ to 1030℃, and all were put into operation. On the second layer, the temperature of the second spray gun on the left wall was 845℃, below 850℃, so this spray gun was shut down. The remaining seven spray guns were put into normal operation. The total number of target deployment spray guns was 15.

[0134] Step S7: Calculate the amount of reducing agent added to each spray gun.

[0135] The current operating parameters are input into the digital twin model to obtain the NOx and CO concentration values ​​at each spray nozzle location. The tangential combustion boiler uses medium-to-high volatile bituminous coal, resulting in relatively low initial NOx emissions, with NOx concentrations at each spray nozzle location ranging from 180 to 260 mg / m³.

[0136] The target NOx value is set at 110 mg / m³. The amount of flue gas produced by the boiler at a 200 MW load is approximately... Nm³ / h. The reducing agent input for each spray gun was calculated using the same method as in Example 1.

[0137] Step S8: Adjust the amount of reducing agent added based on the CO value.

[0138] The CO concentration distribution in the tangential combustion boiler exhibits a non-uniform distribution due to the influence of airflow rotation. The CO concentration at each spray nozzle location in the first layer ranges from 50 to 280 μL / L, with higher CO concentrations observed at some nozzles located in the convergence zone of the rotating airflow. The amount of reducing agent added was adjusted according to the correction rules.

[0139] Step S9: Feedback adjustment of reducing agent input.

[0140] The measured NOx value is obtained through the NOx monitoring instrument at the SCR inlet, and feedback adjustment is performed. The weighting coefficient is set to... , , .

[0141] After feedback adjustment and stabilization, this embodiment achieved an SNCR denitrification efficiency of approximately 58% and a reducing agent utilization rate of approximately 48% under a 200 MW load condition. The SCR inlet NOx concentration remained stable between 110 and 115 mg / m³, and the ammonia slip was maintained between 6 and 7 mg / m³. Due to its lower initial NOx emissions, the tangential combustion boiler requires less reducing agent, resulting in better denitrification efficiency and reducing agent utilization than both Embodiment 1 and Embodiment 2.

[0142] In summary, the embodiments of the present invention have at least the following technical effects: This invention overcomes the limitations of existing SNCR technologies that rely solely on temperature field data for nozzle operation strategy control by integrating multi-dimensional data from temperature field, NOx concentration field, and CO concentration field. It has achieved a significant improvement in SNCR denitrification efficiency in different types of large coal-fired boilers.

[0143] This invention employs a digital twin model based on numerical simulation and machine learning, which is applicable to different boiler types such as offset combustion boilers, W-flame boilers, and tangential combustion boilers, and has good versatility and adaptability.

[0144] The CO concentration correction mechanism and feedback regulation mechanism based on the optimized objective function established in this invention can effectively control ammonia slip and reducing agent consumption while ensuring denitrification efficiency.

[0145] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion, characterized in that, include: Multiple layers of dual-fluid atomizing spray guns are arranged along the height of the boiler furnace. Each dual-fluid atomizing spray gun is equipped with a liquid flow regulating valve and an atomizing air pressure regulating valve. Multiple flue gas temperature monitoring probes are arranged at the arrangement height of each layer of dual-fluid atomizing spray guns, and multiple CO measuring probes are arranged on the boiler furnace wall. The cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer is monitored in real time by the smoke temperature monitoring probe. Digital twin models of the NOx and CO concentration fields in the boiler furnace under different operating conditions were constructed based on numerical simulation and machine learning. The target deployment spray gun layer within the preset temperature window is determined based on the cross-sectional temperature field data. The NOx and CO values ​​of each spray gun in each of the target commissioning spray gun layers are obtained according to the digital twin model. The amount of reducing agent added to each spray gun is calculated based on the NOx value, the target NOx value and the flue gas volume. At the same time, the amount of reducing agent added is corrected according to the CO value. The amount of reducing agent added is adjusted based on the NOx monitoring data from the boiler tail flue to achieve the target denitrification efficiency and reducing agent utilization rate under wide load conditions.

2. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The boiler furnace is divided along the height direction into The region, the The number of layers in each area corresponds to the arrangement of the dual-fluid atomizing spray gun and the smoke temperature monitoring probe; The boiler furnace is divided into left and right sections. The region, the The number of areas and the number of dual-fluid atomizing spray guns arranged in each layer are consistent; The boiler furnace is divided along the front-to-back direction into Each region has dimensions that meet the accuracy requirements for constructing the cross-sectional temperature field.

3. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The real-time monitoring of the cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer via the smoke temperature monitoring probe includes: The radiation image signal of the cross section where the dual-fluid atomizing spray gun is located in each layer is acquired using the smoke temperature monitoring probe; The radiation image signal is analyzed using a radiation image processing method to obtain the cross-sectional temperature field distribution data of each layer of the dual-fluid atomizing spray gun. The number of smoke temperature monitoring probes in each layer is 4 to 6, and the number of layers of the smoke temperature monitoring probes is consistent with the number of layers of the dual-fluid atomizing spray gun.

4. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The construction of digital twin models of the NOx and CO concentration fields in the boiler furnace under different operating conditions based on numerical simulation and machine learning includes: A numerical calculation model for the combustion of the boiler was established, and the CFD numerical simulation method was used to simulate the NOx distribution generated by the boiler furnace combustion under wide load conditions, different coal mill operation modes, different primary air damper openings and secondary air damper openings, different primary air ratios and secondary air ratios, and different burnout air ratios. The results of the CFD numerical simulation are verified and corrected by combining the measurement data of the CO measuring probe to obtain the CO distribution values ​​of each section of the boiler furnace. The NOx distribution values ​​and CO distribution values ​​are correlated with the historical DCS data of the unit to construct a database of furnace cross-sectional temperature field, NOx concentration field and CO concentration field under different unit operating modes and unit load conditions. Based on the database, a digital twin model is trained using machine learning methods, taking boiler operating parameters as input and the NOx concentration field and the CO concentration field as output.

5. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 4, characterized in that, The mapping relationship of the digital twin model is expressed as follows: ； ； in, This represents the mapping function obtained by training the machine learning method. For boiler load, The primary wind ratio, The ratio of secondary wind. For the burnout wind ratio, For fuel quantity, This refers to the operation mode of the coal mill. Furnace space coordinates NOx concentration at that location Furnace space coordinates CO concentration value at that location.

6. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, It also includes the step of building a database of atomization characteristics for dual-fluid atomizing spray guns: The dual-fluid atomizing spray gun was subjected to cold-state tests under different atomizing compressed air pressures and different liquid flow rates. Obtain the atomized particle size distribution, spray angle, and initial spray velocity data of the dual-fluid atomizing spray gun during the cold test; The atomized particle size distribution, the injection angle, and the initial injection velocity data are used to establish an atomization characteristic database indexed by the atomized compressed air pressure and liquid flow rate.

7. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The step of determining the target deployment spray gun layer within the preset temperature window based on the cross-sectional temperature field data includes: Determine whether the cross-sectional temperature field data of the section where the dual-fluid atomizing spray gun is located in each layer is within the preset temperature window of 850°C to 1050°C. If the cross-sectional temperature field data is within the preset temperature window, then the dual-fluid atomizing spray gun of that layer is identified as the target commissioning spray gun layer and put into operation; If the temperature of an individual dual-fluid atomizing spray gun in the target deployment spray gun layer is not within the preset temperature window, then the corresponding dual-fluid atomizing spray gun is turned off.

8. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The step of adjusting the amount of reducing agent added based on the CO value includes: When the CO value is greater than the first preset CO threshold, the dual-fluid atomizing spray gun in the corresponding area is turned off. When the CO value is greater than or equal to the second preset CO threshold and less than or equal to the first preset CO threshold, the calculated amount of reducing agent added will be halved. When the CO value is less than the second preset CO threshold, the calculated amount of reducing agent is used.

9. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The feedback adjustment is based on an optimization objective function, which is expressed as follows: ； in, To optimize the target value, , , The preset weighting coefficients, The measured NOx concentration value is from the tail flue of the boiler. For the target NOx value, This represents the amount of ammonia escaped. This represents the amount of reducing agent consumed.

10. The method for in-furnace denitrification of coal-fired boilers based on multi-source data fusion according to claim 1, characterized in that, The step of adjusting the amount of reducing agent added based on NOx monitoring data from the boiler tail flue includes: The measured NOx value at the SCR inlet section is obtained by installing a NOx monitoring instrument in the flue gas duct at the tail end of the boiler. Compare the measured NOx value with the target NOx value; When the measured NOx value is greater than the target NOx value, the amount of reducing agent added to the corresponding spray gun is increased; when the measured NOx value is less than the target NOx value, the amount of reducing agent added to the corresponding spray gun is decreased. The flow rate of the reducing agent entering the corresponding spray gun is adjusted in real time by the liquid flow regulating valve installed in front of each of the dual-fluid atomizing spray guns.

Images