A method and system for predicting nitrogen oxide emissions from a thermal power plant

Abstract

This invention relates to the field of emission prediction technology, specifically to a method and system for predicting nitrogen oxide emissions from thermal power plants. The method includes the following steps: collecting oxygen volume fraction, temperature, carbon monoxide concentration, and hydroxyl radical concentration inside the furnace and performing three-dimensional mesh mapping; identifying fault coordinates based on continuous cross-sectional changes in oxygen volume fraction and temperature; determining reaction stages by combining the temporal sequence of carbon monoxide and hydroxyl radical changes; further analyzing the relationship between nitrogen and nitric oxide changes; superimposing downstream generation amounts by combining airflow velocity; and outputting nitrogen oxide emission prediction results. In this invention, by synchronously collecting key components and temperature inside the furnace and constructing a spatial mapping, the reaction faults corresponding to changes in oxygen volume fraction and temperature are identified; the temporal relationship between carbon monoxide, hydroxyl radicals, nitrogen, and nitric oxide is analyzed; and the downstream generation amounts are superimposed by combining airflow velocity, thereby improving the emission trend identification capability and result stability under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of emission prediction technology, and in particular to a method and system for predicting nitrogen oxide emissions from thermal power plants. Background Technology

[0002] The field of emission prediction technology mainly involves modeling, analyzing and predicting trends of pollutant emission processes. Its core aspects include collecting emission source operating status data, organizing historical emission data, analyzing the correlation of operating parameters, and calculating and expressing the emission change patterns. Typically, it combines operating parameters such as boiler load, fuel composition, combustion air ratio, flue gas oxygen content and temperature to systematically describe the generation and emission processes of pollutants such as nitrogen oxides under different operating conditions, thereby forming a cognitive and data support system for emission change patterns in the power production process.

[0003] One traditional method and system for predicting nitrogen oxide emissions from thermal power plants involves installing nitrogen oxide concentration detectors, flue gas flow meters, temperature measuring points, and oxygen analyzers at the boiler and flue locations to acquire operating data of the unit at different load stages. Based on manually compiled historical operating records, an emission data comparison table is established. Furthermore, a fixed functional relationship is used to substitute boiler load, coal consumption, air volume, and the measured nitrogen oxide concentration into the calculation, or linear extrapolation is performed using historical emission records arranged in chronological order to estimate emission values ​​at subsequent times, thereby completing the prediction of nitrogen oxide emissions during thermal power generation.

[0004] Existing technologies rely on collecting macroscopic operating parameters from external measuring points in the flue and boiler. The data sources are mostly discrete and have a coarse time scale, making it difficult to reflect the detailed changes in combustion reactions inside the furnace. Furthermore, calculations based on historical records and fixed functional relationships result in static mapping of parameters, which fails to reflect the differences in reactions in different spatial regions and transient coupling characteristics. Prediction accuracy is easily affected by load fluctuations or changes in fuel properties. Linear extrapolation methods lack adaptability to nonlinear emission evolution, leading to lags in emission trend judgments. This makes it difficult to support the needs of refined regulation and affects the timeliness and reliability of emission control decisions. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for predicting nitrogen oxide emissions from thermal power plants, comprising the following steps: S1: Oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration are acquired simultaneously by sensors inside the furnace of a thermal power plant and then mapped in three dimensions to generate combustion state data. S2: Calculate the fault location based on the oxygen volume fraction and temperature at the continuous cross section of the combustion state data, and filter the coordinates according to the negative sign of the oxygen volume fraction difference and the positive sign of the temperature difference to generate a fault coordinate set; S3: Based on the fault coordinate set and the combustion state data, analyze the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compare the order of time to generate a reaction stage table; S4: Based on the reaction stage table, obtain the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration. When the oxygen volume fraction difference is negative and the temperature increases, determine the change time of nitrogen concentration and nitric oxide concentration, and generate a hysteresis instruction set. S5: Obtain the airflow velocity and nitric oxide concentration according to the reaction stage table and the hysteresis instruction set, obtain the downstream spatial cross-section data and calculate the total nitrogen oxide generation, and combine the nitric oxide concentration and airflow velocity to generate the nitrogen oxide emission prediction result.

[0006] As a further aspect of the present invention, the combustion state data includes a spatial coordinate index, an oxygen volume fraction field, a temperature distribution field, and a carbon monoxide concentration field; the fault coordinate set includes oxygen volume fraction abrupt change points, temperature transition points, and symbol-filtered coordinate identifiers; the reaction stage table includes the starting time of carbon monoxide decrease, the starting time of hydroxyl radical increase, and stage sequence identifiers; the hysteresis instruction set includes oxygen volume fraction difference symbol identifiers, temperature increase identifiers, and nitrogen concentration change delays; and the nitrogen oxide emission prediction results include the total amount of nitrogen oxides in the downstream section and the inverse superimposed concentration sequence.

[0007] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Obtain oxygen volume fraction, temperature, carbon monoxide concentration, hydroxyl radical concentration, nitrogen concentration, nitric oxide concentration and airflow velocity at the same time through sensors inside the furnace of a thermal power plant and align them in time. Match the multi-parameter sequences according to time consistency constraints to generate multi-parameter synchronous distribution data. S102: Based on the multi-parameter synchronous distribution data, call the spatial coordinate index of multi-coordinate points, rearrange the three-dimensional coordinates of the multi-parameter values ​​according to the spatial coordinate order, map the discrete coordinates to the regular grid point set, and perform neighborhood consistency interpolation calculation on the parameters of adjacent nodes to obtain the three-dimensional grid parameter field. S103: The oxygen volume fraction, temperature, carbon monoxide concentration, nitrogen concentration, airflow velocity and hydroxyl radical concentration in each spatial region are fused according to the three-dimensional grid parameter field, and combustion state data are generated based on the weighted superposition and normalization of parameter vectors.

[0008] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Obtain the oxygen volume fraction and temperature at a continuous cross section of the combustion state data and align them with spatial coordinates. Perform adjacent spatial point difference based on the oxygen volume fraction sequence corresponding to multiple coordinate points within the same cross section, and synchronously perform corresponding spatial point difference in combination with the temperature value sequence to generate an oxygen-temperature difference correlation sequence. S202: Based on the oxygen-temperature difference correlation sequence, perform sign discrimination on the oxygen volume fraction difference corresponding to the multi-coordinate index and extract negative values, and at the same time perform sign discrimination on the temperature difference at the same index position and extract positive values, mark the coordinates where the oxygen volume fraction difference is negative and the temperature difference is positive, and generate a sign-matching coordinate index set; S203: Based on the symbol matching coordinate index set, extract the original coordinate point sequence within the corresponding spatial section by index mapping, aggregate all the marked indexed coordinate points and reconstruct the spatial position sequence to generate a fault coordinate set.

[0009] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Based on the fault coordinate set and the combustion state data, obtain the carbon monoxide concentration and hydroxyl radical concentration at the corresponding spatial coordinate points, extract the concentration values ​​of multiple spatial coordinate points according to the time series, and align the two types of concentration values ​​corresponding to the same coordinate point on the time axis to obtain a time series concentration pairing set. S302: Based on the time-series concentration pairing set, perform time-by-time difference operation on the carbon monoxide concentration sequence, extract the concentration decrease rate of change and compare it with a preset decrease trigger threshold, and at the same time perform the same type of difference operation on the hydroxyl radical concentration sequence and match it with a preset increase trigger threshold to generate start time index pairs. S303: Based on the aforementioned start time index pair, calculate the difference between the carbon monoxide falling time index and the hydroxyl radical rising time index at the same spatial coordinate point, and classify and encode them according to the time difference value range. Organize and sequence map the corresponding encoding results of multiple coordinate points to obtain the reaction stage table.

[0010] As a further aspect of the present invention, the decrease trigger threshold is determined by collecting a carbon dioxide concentration sequence with consecutive timestamps within the initial combustion stage, performing differential operations on the concentration values ​​corresponding to adjacent time points within the sequence to obtain a discrete difference set, extracting all negative difference parameters within the difference set, calculating the statistical mean and standard deviation values ​​corresponding to all negative difference parameters, and summing the statistical mean and standard deviation values ​​to determine the threshold.

[0011] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Based on the reaction stage table and the combustion state data, obtain the oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration at the associated spatial coordinate point, perform differential extraction on the oxygen volume fraction and extract all negative terms, extract the increasing interval of temperature and cross-match it with the all negative terms to generate an oxygen-temperature co-condition set. S402: Based on the overlapping time points in the oxygen-temperature synergistic condition set, call the nitrogen concentration and nitric oxide concentration at the associated spatial coordinate points, perform difference calculations on the two types of concentrations according to time sequence, extract the time node where the difference turns from negative to positive and mark it as the start time of the rise, and analyze the difference to obtain the concentration rise time difference record. S403: Call the concentration rise time difference record, extract the coordinate points corresponding to the time difference exceeding the preset reaction hysteresis time reference value of multiple spatial coordinate points, perform numerical encoding mapping on the coordinate points and corresponding differences, arrange and integrate all encoding results in spatial topological order, and generate hysteresis instruction set.

[0012] As a further aspect of the present invention, the reaction lag time reference value is determined by collecting nitrogen concentration and nitric oxide concentration sequences at continuous timestamps, extracting the values ​​of the two types of concentration sequences and converting them into timestamps corresponding to the points of positive growth, subtracting the two types of timestamps to obtain the experimental time difference set, calculating the arithmetic mean and standard deviation of all values ​​in the experimental time difference set and adding them together.

[0013] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Based on the reaction stage table, the hysteresis instruction set and the combustion state data, obtain the airflow velocity and nitric oxide concentration at the associated coordinate point, extract the time stamps of adjacent instructions in the hysteresis instruction set, perform difference calculation to obtain the time difference sequence, and multiply it with the velocity value to obtain the spatial offset distance set. S502: Based on the spatial offset distance set, perform spatial index mapping on the combustion state data, locate the downstream spatial section coordinate set, accumulate the nitrogen oxide generation values ​​corresponding to multiple coordinates, obtain the section aggregation sequence, and perform vectorization and recombination to obtain the downstream section nitrogen oxide total vector. S503: The total nitrogen oxide emission vector of the downstream section is called and the nitric oxide concentration sequence is superimposed at the corresponding positions to obtain the superimposed numerical sequence. Then, the superimposed numerical sequence is accumulated in reverse order along the flow direction to generate the nitrogen oxide emission prediction result.

[0014] A nitrogen oxide emission prediction system for a thermal power plant, comprising: The combustion status module acquires oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration at the same time through sensors inside the furnace of a thermal power plant, performs three-dimensional mesh mapping, generates combustion status data, and transmits it to the fault location module. The fault location module calculates the fault location based on the oxygen volume fraction and temperature at continuous cross sections of the combustion state data, filters coordinates according to the negative signs of the oxygen volume fraction difference and the positive signs of the temperature difference, generates a fault coordinate set, and transmits it to the reaction stage module. The reaction stage module, based on the fault coordinate set and the combustion state data, analyzes the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compares the order of time, generates a reaction stage table and transmits it to the hysteresis instruction module. The hysteresis instruction module obtains the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration based on the reaction stage table. When the oxygen volume fraction difference is negative and the temperature increases, it judges the change time of nitrogen concentration and nitric oxide concentration, generates a hysteresis instruction set and transmits it to the emission prediction module. The emission prediction module obtains the airflow velocity and nitric oxide concentration based on the reaction stage table and the hysteresis instruction set, acquires downstream spatial cross-sectional data and calculates the total nitrogen oxide generation, and combines the nitric oxide concentration and airflow velocity to generate the nitrogen oxide emission prediction result.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by simultaneously collecting data on multiple key components and temperatures inside the furnace and constructing a spatial grid mapping, fine-grained characterization of the combustion process is achieved. Simultaneously, by combining oxygen volume fraction and temperature changes, reaction fault locations are identified. Furthermore, the evolutionary sequence analysis of carbon monoxide and hydroxyl radicals is introduced to characterize the differences in reaction stages. Based on this, the temporal correlation of nitrogen and nitric oxide changes is determined, forming a hysteresis relationship expression with temporal characteristics. Combined with airflow velocity, the downstream generation is continuously superimposed and calculated, enabling emission prediction to shift from static empirical estimation to spatial and temporal coupled analysis, thus improving the ability to identify emission trends and the stability of results under complex operating conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] Please see Figure 1 This invention provides a method for predicting nitrogen oxide emissions from thermal power plants, comprising the following steps: S1: Oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration are acquired simultaneously by sensors inside the furnace of a thermal power plant and then mapped in three dimensions to generate combustion state data. S2: Calculate the fault location based on the oxygen volume fraction and temperature at continuous cross sections of combustion state data, and filter the coordinates according to the negative signs of the oxygen volume fraction difference and the positive signs of the temperature difference to generate a fault coordinate set; S3: Based on the fault coordinate set and combustion state data, analyze the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compare the order of time to generate a reaction stage table; S4: Based on the reaction stage table, obtain the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration. When the oxygen volume fraction difference is negative and the temperature increases, determine the change time of nitrogen concentration and nitric oxide concentration and generate a hysteresis instruction set. S5: Obtain airflow velocity and nitric oxide concentration based on the reaction stage table and hysteresis instruction set, obtain downstream spatial cross-sectional data and calculate the total nitrogen oxide generation, and combine the nitric oxide concentration and airflow velocity to generate nitrogen oxide emission prediction results.

[0021] Combustion state data includes spatial coordinate index, oxygen volume fraction field, temperature distribution field, and carbon monoxide concentration field. The fault coordinate set includes oxygen volume fraction abrupt change points, temperature transition points, and symbol-filtered coordinate identifiers. The reaction stage table includes the starting time of carbon monoxide decrease, the starting time of hydroxyl radical increase, and the stage sequence identifiers. The hysteresis instruction set includes oxygen volume fraction difference symbol identifiers, temperature increase identifiers, and nitrogen concentration change delays. The nitrogen oxide emission prediction results include the total nitrogen oxide amount in the downstream section and the inverse superimposed concentration sequence.

[0022] Please see Figure 2 The specific steps of S1 are as follows: S101: Obtain oxygen volume fraction, temperature, carbon monoxide concentration, hydroxyl radical concentration, nitrogen concentration, nitric oxide concentration and airflow velocity at the same time through sensors inside the furnace of a thermal power plant and align them in time. Match the multi-parameter sequences according to time consistency constraints to generate multi-parameter synchronous distribution data. Fifty sensor nodes deployed inside the furnace of a thermal power plant synchronously and continuously collect physical and chemical state information. Sampling datasets with timestamps of 1600000000000 milliseconds are directly read from the communication interfaces of each sensor. Specific data read includes oxygen volume fraction, temperature, carbon monoxide concentration, and hydroxyl radical concentration. Time alignment is performed on the multi-source heterogeneous data streams. The local timestamps of each sensor are extracted, with a base timestamp set to 1600000000000 milliseconds. Simultaneously, the local timestamp returned by sensor number 1 (16000000000002 milliseconds) is extracted. The absolute time difference between the local timestamp and the base timestamp is calculated, and the absolute value obtained by subtracting the two yields a timestamp offset of 2 milliseconds. A preset time consistency tolerance threshold of 5 milliseconds is set. This 5-millisecond value is determined by statistically analyzing the normal distribution of 10,000 sensor delays over a historical 30-day period, using the upper limit of the 95% confidence interval of the distribution curve. The 2-millisecond offset is compared with the 5-millisecond threshold. Since 2 milliseconds is less than 5 milliseconds, the data from sensor number 1 meets the consistency constraint and is added to the valid data queue. Parameters in the valid queue are rearranged and reorganized according to the reference timestamp. Blank data bits caused by packet loss are filled using linear proportional interpolation between adjacent timestamps. The temperature value of 1200 degrees Celsius is extracted from the previous valid sampling time of 1599999999900 milliseconds, and the temperature value of 1210 degrees Celsius is extracted from the next valid sampling time of 1600000000100 milliseconds. The time difference between the two is calculated to be 200 milliseconds, and the temperature difference is calculated to be 10 degrees Celsius. The current time of 1600000000000 milliseconds is 100 milliseconds from the previous time. Based on the fact that the current time accounts for 50% of the time between the previous and next times, the 10-degree Celsius difference is multiplied by 50% to obtain 5 degrees Celsius. Adding 5 degrees Celsius to 1200 degrees Celsius fills the current temperature value to 1205 degrees Celsius. After the node filling is completed, the oxygen volume fraction of 4.5%, the temperature of 1205 degrees Celsius, the carbon monoxide concentration of 150 parts per million, and the hydroxyl radical concentration of 15 parts per million are bundled and encapsulated to generate multi-parameter synchronous distribution data.

[0023] Table 1. Synchronous Distribution Data Acquisition Table of Multi-Parameters in the Furnace

[0024] S102: Based on the multi-parameter synchronous distribution data, call the spatial coordinate index of multi-coordinate points, rearrange the three-dimensional coordinates of the multi-parameter values ​​according to the spatial coordinate order, map the discrete coordinates to the regular grid point set, and perform neighborhood consistency interpolation calculation on the parameters of adjacent nodes to obtain the three-dimensional grid parameter field. Read the generated multi-parameter synchronous distribution data and extract the 3D spatial coordinate values ​​attached to each valid sensor data packet. Separate the lateral width coordinate, longitudinal depth coordinate, and vertical height coordinate from the multi-coordinate point spatial coordinate index. Sort the entire dataset in ascending order according to the lateral width coordinate values, then sort according to the longitudinal depth coordinate values ​​if the lateral coordinate values ​​are the same, and finally sort according to the vertical height coordinate values ​​to complete the 3D coordinate rearrangement. Construct a regular 3D grid point set covering the entire furnace space, setting the lateral, longitudinal, and vertical grid spacing to 0.5 meters. Map the rearranged discrete sensor coordinates to the nearest regular grid cell center point. For blank grid points without direct sensor data, perform neighborhood consistency interpolation calculation for adjacent node parameters. Set the neighborhood search radius threshold to 2.0 meters, which is based on the spatial distance measured when the thermal radiation intensity decays to 50% over 100 historical combustion cycles. Search for discrete mapping points with valid values ​​within a 2.0-meter radius around the target blank point, and extract the temperature values ​​and straight-line distances of each adjacent node. Calculate the reciprocal of the distance to each adjacent node, and sum all the reciprocals to obtain the total reciprocal sum. Divide the reciprocal of the distance to each node by the total reciprocal sum to calculate the distance weight coefficient for each node. Multiply the node temperature value by the distance weight coefficient and sum them to obtain the interpolated temperature value of the target point. Substituting into a practical example, setting the target point to be 1.0 meter horizontally, 1.0 meter vertically, and 1.0 meter in all directions, we obtain the distance to the first adjacent node as 0.5 meters and the distance to the second adjacent node as 1.0 meter. Calculate the reciprocal of the distance to the first node as 2.0, the reciprocal of the distance to the second node as 1.0, and the total reciprocal sum as 3.0. The weight coefficient of the first node is 2.0 divided by 3.0, rounded to three decimal places, which equals 0.667; the weight coefficient of the second node is 1.0 divided by 3.0, rounded to three decimal places, which equals 0.333. Substituting the temperatures of node 1 (1200 degrees Celsius) and node 2 (1230 degrees Celsius) into the values, we get: 1200 * 0.667 = 800.4, 1230 * 0.333 = 409.59. Adding 409.59 to 800.4 gives us the target grid point temperature value of 1209.99 degrees Celsius. To ensure accuracy in subsequent calculations, we round it to the nearest integer, 1210 degrees Celsius. We then perform the same calculations on the remaining three parameters, saving the complete point set numerical matrix to obtain the three-dimensional mesh parameter field.

[0025] S103: The oxygen volume fraction, temperature, carbon monoxide concentration, nitrogen concentration, airflow velocity and hydroxyl radical concentration in each spatial region are fused according to the three-dimensional grid parameter field. Combustion state data are generated based on the weighted superposition and normalization of parameter vectors. The oxygen volume fraction, temperature, carbon monoxide concentration, and hydroxyl radical concentration are extracted from each discrete spatial volume cell in the generated 3D mesh parameter field. Multi-parameter fusion and normalization operations are performed on the extracted four parameters. Extreme value baseline data for each parameter are extracted from the historical data over a continuous 6-month period, with the historical maximum temperature baseline being 1600 degrees Celsius and the minimum being 800 degrees Celsius. The target mesh point temperature value of 1210 degrees Celsius calculated in the previous steps is retrieved. Subtracting the minimum baseline value of 800 from 1210 yields a difference of 410. Subtracting the minimum baseline value of 800 from the maximum baseline value of 1600 yields an extreme value span of 800. Dividing the difference of 410 by the extreme value span of 800 yields a normalized temperature vector value of 0.5125. Similarly, extracting the corresponding node values ​​and performing the same subtraction and division processes, the normalized vector values ​​for oxygen are calculated to be 0.45, carbon monoxide 0.60, and hydroxyl radicals 0.35. Multi-parameter fusion weight coefficients are then applied, setting the weights for temperature (0.4), oxygen (0.3), carbon monoxide (0.2), and hydroxyl radicals (0.1). These weights are obtained through 5000 iterations using the gradient descent algorithm. The loss function used in training consists of the mean squared error term of the predicted heat loss and the actual physical heat loss, plus an L2 regularization term to prevent overfitting. The final coefficients are determined by continuously updating the weight values ​​along the reverse direction of the loss function gradient, allowing the mean squared error term to approach its minimum. Multiplying the normalized temperature value (0.5125) by a weight of 0.4 yields a product of 0.205; multiplying the normalized oxygen value (0.45) by a weight of 0.3 yields a product of 0.135; multiplying the normalized carbon monoxide value (0.60) by a weight of 0.2 yields a product of 0.120; and multiplying the normalized hydroxyl radical value (0.35) by a weight of 0.1 yields a product of 0.035. Summing these four products, 0.205 + 0.135 + 0.120 + 0.035 = 0.495, yields the comprehensive state vector value for the spatial region as 0.495. The comprehensive state vector value is calculated for all regions. The face connectivity and adjacency relationships of each grid cell are extracted to construct a spatial topology matrix. The comprehensive state vector value is then filled into the corresponding node positions in the spatial topology matrix to generate combustion state data.

[0026] Please see Figure 3 The specific steps of S2 are as follows: S201: Obtain the oxygen volume fraction and temperature at a continuous cross section of combustion state data and align them with spatial coordinates. Perform differential analysis between adjacent spatial points based on the oxygen volume fraction sequence corresponding to multiple coordinate points within the same cross section. Combine this with the temperature value sequence to synchronously perform differential analysis between corresponding spatial points, generating an oxygen-temperature difference correlation sequence. Read and integrate the generated combustion state data, and extract the oxygen volume fraction and temperature values ​​at a specific horizontal continuous cross-section. The first continuous cross-section is the horizontal plane with a height coordinate of 10.0 meters. Perform spatial coordinate alignment on all discrete data points within this cross-section, extracting the lateral width coordinate and longitudinal depth coordinate of each point. Arrange these coordinates in ascending order of lateral coordinates, and if lateral coordinates are the same, arrange them in ascending order of longitudinal coordinates, constructing a two-dimensional coordinate grid for the cross-section. Perform differential processing on adjacent spatial points based on the oxygen volume fraction sequence corresponding to multiple coordinate points within the same cross-section. Extract the longitudinal depth coordinate of the current target grid point at 2.0 meters, with an oxygen volume fraction of 4.2%. Extract the longitudinal depth coordinate of the next adjacent grid point at 2.5 meters on the same lateral side, with an oxygen volume fraction of 4.5%. Subtract the current point's 4.2% from the next point's 4.5% oxygen content to calculate the first oxygen difference of 0.3%. Perform corresponding spatial point differential processing simultaneously with the temperature value sequence. The temperature value of the current target grid point (1210 degrees Celsius) is extracted, and the temperature value of the next adjacent grid point (1190 degrees Celsius) is extracted. The temperature of the current point (1210 degrees Celsius) is subtracted from the temperature of the next adjacent grid point (1190 degrees Celsius) to calculate the first temperature difference as -20 degrees Celsius. Subsequently, the oxygen content at the horizontal coordinate of 1.5 meters and vertical coordinate of 2.0 meters corresponding to the current target point (4.5%) is extracted, and the oxygen content at the horizontal coordinate of the next adjacent grid point (2.0 meters and 2.0 meters) is extracted (4.1%). The oxygen content at the next adjacent grid point is subtracted from the oxygen content at the current grid point to obtain the second oxygen difference as -0.4%. Simultaneously, the temperature of the current target point (1205 degrees Celsius) and the temperature of the horizontally adjacent grid point (1225 degrees Celsius) are extracted to obtain the second temperature difference as 20 degrees Celsius. The above coordinate positions are associated with the corresponding first oxygen difference, first temperature difference, second oxygen difference, and second temperature difference, and the data is aligned and bundled to generate an oxygen-temperature difference correlation sequence.

[0027] S202: Based on the oxygen-temperature difference correlation sequence, the sign of the oxygen volume fraction difference corresponding to the multi-coordinate index is determined and the negative value is extracted. At the same time, the sign of the temperature difference at the same index position is determined and the positive value is extracted. The coordinates with negative oxygen volume fraction difference and positive temperature difference are marked to generate a sign-matching coordinate index set. The generated oxygen-temperature difference correlation sequence is read, and the signs of the oxygen volume fraction differences corresponding to the multi-coordinate indices in the sequence are determined, with negative values ​​extracted. Zero value is set as the sign determination benchmark. The aforementioned second oxygen difference value of -0.4% is retrieved and compared with the zero benchmark. Since -0.4% is less than the zero benchmark, the value is confirmed to be negative, indicating a decreasing oxygen concentration trend in the corresponding direction. This coordinate is then added to the candidate anomaly queue. Simultaneously, the signs of the temperature differences at the same index are determined, with positive values ​​extracted. The aforementioned second temperature difference value of 20 degrees Celsius, with the same index as the second oxygen difference, is retrieved and compared with the zero benchmark. Since 20 degrees Celsius is greater than the zero benchmark, the value is confirmed to be positive, indicating a rising temperature trend at this point. Special coordinate points with negative oxygen volume fraction differences and positive temperature differences are marked. Double verification is performed on all coordinates in the candidate anomaly queue; a successful match is assigned only when a coordinate point simultaneously satisfies both a less than zero oxygen difference and a greater than zero temperature difference. Substituting into the previous example, the coordinate point with a horizontal distance of 1.5 meters and a vertical distance of 2.0 meters has a horizontal oxygen difference of -0.4% and a horizontal temperature difference of 20 degrees Celsius, meeting both conditions, and is therefore marked as a special state point. However, the coordinate point with a first oxygen difference of 0.3% and a first temperature difference of -20 degrees Celsius does not meet the conditions and is therefore removed. By traversing all associated difference sequences across the entire cross-section, the index set of all marked coordinate points is encapsulated to generate a symbol-matching coordinate index set.

[0028] S203: Based on the symbol matching coordinate index set, the original coordinate point sequence within the corresponding spatial section is extracted by index mapping, all the marked indexed coordinate points are aggregated and the spatial position sequence is reconstructed to generate the fault coordinate set; The generated symbol matching coordinate index set is read, and the original coordinate point sequence within the corresponding spatial section is extracted by index mapping based on the marked data in the set. The index of the successfully marked outlier point (1.5 meters horizontally and 2.0 meters vertically) is retrieved, and its position is located in the original 3D coordinate matrix of the section. The corresponding absolute spatial coordinate value set is extracted, covering 1.5 meters horizontally, 2.0 meters vertically, and 10.0 meters vertically. According to this mapping rule, the discrete absolute spatial coordinate data corresponding to all marked indices are extracted sequentially. All extracted discrete coordinate points are aggregated and the spatial position sequence is reconstructed. The Euclidean distance aggregation benchmark threshold is set to 0.8 meters, which is based on the statistical analysis of the geometric characteristics of 50 historical combustion faults. The first extracted coordinate point (1.5 meters horizontally and 2.0 meters vertically) is selected, and the next extracted marked point (2.0 meters horizontally and 2.3 meters vertically) is retrieved. The coordinate difference between the two points is calculated: the horizontal difference is 2.0 minus 1.5, yielding 0.5 meters; the vertical difference is 2.3 minus 2.0, yielding 0.3 meters. The lateral difference of 0.5 meters is squared to obtain 0.25 square meters, and the longitudinal difference of 0.3 meters is squared to obtain 0.09 square meters. Adding 0.25 and 0.09 gives 0.34 square meters. Performing a square root operation on 0.34 square meters and retaining three decimal places yields a linear Euclidean distance of 0.583 meters. Comparing this distance of 0.583 meters with the aggregation threshold of 0.8 meters, it is determined that 0.583 meters is less than 0.8 meters, confirming that the two marker points belong to the same anomalous combustion connected region. They are then spatially clustered and merged. Based on the lateral extension direction of this connected region, the aggregated coordinate positions are reordered to reconstruct a continuous fault spatial position sequence. Traversing all marker points within the entire cross-section, point cloud aggregation and reconstruction processing that meets the distance threshold condition is completed, generating a fault coordinate set.

[0029] Please see Figure 4 The specific steps of S3 are as follows: S301: Based on the fault coordinate set and combustion state data, obtain the carbon monoxide concentration and hydroxyl radical concentration at the corresponding spatial coordinate points, extract the concentration values ​​of multiple spatial coordinate points according to the time series, and align the two types of concentration values ​​corresponding to the same coordinate point on the time axis to obtain a time series concentration pairing set. Read the generated fault coordinate set and combustion state data. Extract the coordinate points from the fault coordinate set, which are 1.5 meters horizontally, 2.0 meters vertically, and 10.0 meters high vertically. Based on these coordinate points, perform a three-dimensional spatial indexing operation in the combustion state data matrix to locate the carbon monoxide concentration sequence and hydroxyl radical concentration sequence of this specific coordinate point within the historical continuous sampling period. Set the start timestamp of the continuous sampling period to 16000000000000 milliseconds, the end timestamp to 1600000005000 milliseconds, and the time step to 100 milliseconds. Extract the multi-source parameter data stream within the aforementioned time interval and perform data cleaning and preprocessing operations. Scan the carbon monoxide concentration field and the hydroxyl radical concentration field line by line, removing invalid negative values ​​or missing records with concentration values ​​below 0 parts per million. For the valid continuous sequences retained after cleaning, perform time axis alignment processing on the two types of concentration values ​​corresponding to the same coordinate point. The carbon monoxide concentration timestamp is extracted at the current sampling time of 1600000000100 milliseconds. Simultaneously, the corresponding hydroxyl radical concentration timestamp of 1600000000105 milliseconds is retrieved. The absolute time difference between the two timestamps is calculated to be 5 milliseconds. If this 5-millisecond time difference is less than the preset 10-millisecond time tolerance, the pairing is confirmed as successful. The carbon monoxide concentration value of 150 parts per million (ppm) and the hydroxyl radical concentration value of 15 ppm are then combined. The aforementioned 10-millisecond time tolerance was determined by testing the asynchronous maximum delay limit of the sensor network under extreme load for 1000 hours. Similarly, the timeline is iterated through to the next moment, continuously generating aligned data entries.

[0030] Table 2. Time-series concentration pairing data for target coordinate points

[0031] Table 2 lists some of the time-series correlation data after alignment processing. The aforementioned time matching process is performed on all remaining marked coordinate points in the fault coordinate set, integrating the paired records generated from all coordinate points to generate a time-series concentration pairing set.

[0032] S302: Based on the time-series concentration pairing set, perform time-by-time difference operation on the carbon monoxide concentration sequence, extract the concentration decrease rate of change and compare it with the preset decrease trigger threshold, and at the same time perform the same difference operation on the hydroxyl radical concentration sequence and match it with the preset increase trigger threshold to generate start time index pairs. Read the extracted time-series concentration pairing data and lock onto a specific spatial coordinate point data sequence with a horizontal distance of 1.5 meters and a vertical distance of 2.0 meters. Extract the continuous carbon monoxide concentration sequence corresponding to this coordinate point and perform a time-by-time difference operation. Retrieve the carbon monoxide concentration value of 150 parts per million (ppm) at the previous time of 1600000000100 ms and the carbon monoxide concentration value of 145 ppm at the current time of 1600000000200 ms. Subtract the concentration value of 145 ppm from the previous time of 150 ppm to obtain a concentration difference of 5 ppm. Extract the time span of 100 ms between the current time and the previous time. Divide the concentration difference of 5 ppm by the time span of 100 ms to calculate the rate of decrease in carbon monoxide concentration as 0.05 ppm per ms. A preset carbon monoxide (CO) decrease trigger threshold of 0.03 parts per million (ppm) per millisecond is extracted. This threshold is calculated by retrieving a sample set of the maximum concentration change rate from 50,000 slight fluctuations under normal and stable combustion conditions over 90 days, performing an arithmetic mean calculation, and adding a 20% safety margin. The calculated CO decrease threshold of 0.05 ppm per millisecond is compared with the trigger threshold of 0.03 ppm per millisecond. Since 0.05 is greater than 0.03, the trigger condition is met, and the current time is extracted as the CO decrease time index, recorded as 1600000000200 ms. Simultaneously, a similar difference operation is performed on the hydroxyl radical concentration sequences in the paired set. The hydroxyl radical concentration value of 15 ppm per millisecond at the previous time of 1600000000150 ms is retrieved, and the concentration value of 18 ppm per millisecond at the current time of 1600000000250 ms is retrieved. The difference of 3 ppm per millisecond is obtained by subtracting the previous time value of 15 from the current time value of 18. Dividing the concentration difference of 3 parts per million by the time span of 100 milliseconds yields a hydroxyl radical rise rate of 0.03 parts per million concentration per millisecond. A hydroxyl radical rise trigger threshold of 0.02 parts per million concentration per millisecond is extracted; this threshold is calibrated based on the minimum generation rate of the shock wave at the forefront of a basic chemical chain reaction. Since the calculated 0.03 is greater than the threshold 0.02, the trigger condition is met, and the hydroxyl radical rise time index at this moment is recorded as 1600000000250 milliseconds. The two confirmed time index data are packaged and combined to generate a start time index pair.

[0033] S303: Based on the start time index pair, the difference between the carbon monoxide falling time index and the hydroxyl radical rising time index at the same spatial coordinate point is calculated, and the time difference values ​​are classified and coded according to the range. The coding results of multiple coordinate points are sorted and sequence mapped to obtain the reaction stage table. Read the generated start time index pairs and locate the time index values ​​at spatial coordinates of 1.5 meters horizontally and 2.0 meters vertically. Extract the carbon monoxide fall time index (1600000000200 milliseconds) calculated and recorded in the previous steps, and simultaneously extract the hydroxyl radical rise time index (1600000000250 milliseconds) corresponding to the same reaction chain. Perform a time difference calculation operation on the extracted time indices of the two dimensions. Subtract the carbon monoxide fall time index (1600000000200 milliseconds) from the hydroxyl radical rise time index (1600000000250 milliseconds) to calculate the absolute time difference reflecting the response delay of the two substances, which is 50 milliseconds. Call the preset time difference classification coding dictionary set, which is derived by classifying the evolution time intervals of 10,000 independent spark excitation images collected in a standard combustion test chamber under controlled conditions. The time difference value within the closed interval of 0 to 30 milliseconds is defined as the rapid free radical excitation stage, assigned category code value 1; the time difference value greater than 30 milliseconds and less than or equal to 80 milliseconds is defined as the transition fuel control stage, assigned category code value 2; the time difference value greater than 80 milliseconds and less than or equal to 150 milliseconds is defined as the hysteresis tail oxidation stage, assigned category code value 3. The calculated time difference value of 50 milliseconds is substituted into the judgment logic to perform interval matching comparison. It is determined that the 50 milliseconds value is greater than 30 milliseconds and less than 80 milliseconds, which meets the set range of the transition fuel control stage, and the corresponding coordinate grid point is assigned the output code value 2. All other trigger points in the global 3D mesh are traversed, and the time span difference calculation and conditional interval discrimination coding actions are repeatedly performed. The acquired absolute position data of all multi-coordinate points and their corresponding code values ​​2 are mapped using a one-dimensional sequence association. The entire spatial interface is arranged according to the increasing horizontal size rule, and the data are merged to construct a global view table, resulting in the reaction stage table.

[0034] Please see Figure 5 The specific steps of S4 are as follows: S401: Based on the reaction stage table and combustion state data, obtain the oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration at the associated spatial coordinate points, perform differential extraction on the oxygen volume fraction and extract all negative terms, extract the increasing interval of temperature and cross-match it with the all negative terms to generate the oxygen-temperature co-condition set. Read the specific spatial coordinates (1.5 meters horizontally, 2.0 meters vertically, and 10.0 meters vertically) corresponding to the coded value 2 in the reaction stage table. Extract the oxygen volume fraction, temperature, nitrogen concentration, and nitric oxide concentration sequences for this coordinate point within the time interval from 1600000000200 ms to 1600000000300 ms from the combustion state data matrix. Before extracting the data sequences, perform missing value imputation on the oxygen volume fraction and temperature sequences using linear interpolation to numerically repair the missing values. Perform a difference operation on the time-series adjacent points of the repaired oxygen volume fraction sequence, retrieving the oxygen volume fraction of 4.5% at 1600000000200 ms and 4.2% at 1600000000210 ms. Subtract the previous time-series 4.5% from the later 4.2%, calculating the oxygen volume fraction time-series difference as -0.3%. Setting zero as the baseline, the calculated -0.3% is compared to zero. If it is less than zero, the difference is confirmed as a negative number. The entire time interval is traversed to extract the set of all time points where the calculated result is less than zero, generating a set of all negative numbers. Temperature sequences within the same time interval are extracted simultaneously, and adjacent values ​​with the same time step are subtracted. The temperatures of 1210 degrees Celsius at 1600000000200 ms and 1230 degrees Celsius at 1600000000210 ms are retrieved. Subtracting 1210 degrees Celsius from 1230 degrees Celsius yields a temperature time difference of 20 degrees Celsius. This 20 degrees Celsius is compared to zero; if it is greater than zero, the corresponding continuous interval is marked as a temperature increasing interval. The time point and temperature increase interval in the all-negative number terms are retrieved and cross-matched. It is determined that 1600000000210 milliseconds exist in both the all-negative number terms and the temperature increase interval. The time point and its corresponding coordinate position information are encapsulated and bound to generate the oxygen-temperature synergistic condition set.

[0035] S402: Based on the coincident time points in the oxygen-temperature synergy condition set, call the nitrogen concentration and nitric oxide concentration at the associated spatial coordinate points, perform difference calculations on the two types of concentrations according to time sequence, extract the time node where the difference turns from negative to positive and mark it as the start time of the rise, and analyze the difference to obtain the concentration rise time difference record. Read the coincident time point 1600000000210 ms within the generated oxygen-temperature coordinating condition set. Based on this time point, retrieve the nitrogen and nitric oxide concentration sequences within the subsequent time window at the associated spatial coordinates (1.5m horizontal, 2.0m vertical, 10.0m vertical). Extract the nitric oxide concentration sequence and perform a difference operation on adjacent data points with a time step of 10 ms. Retrieve the nitric oxide concentration value of 15.0 parts per million (ppm) at 1600000000250 ms and the concentration value of 14.5 ppm at 1600000000260 ms. Subtract the 15.0 ppm concentration from the 14.5 ppm concentration to calculate the first concentration difference as -0.5 ppm. The concentration value of 16.2 parts per million (ppm) at the next adjacent time interval of 1600000000270 ms is retrieved. Subtracting 14.5 ppm from 16.2 ppm yields a second concentration difference of 1.7 ppm. The first concentration difference of -0.5 ppm and the second concentration difference of 1.7 ppm are then compared sequentially with zero. Since -0.5 is less than zero and 1.7 is greater than zero, it is confirmed that a sign change from negative to positive occurs in the nitric oxide concentration difference at 1600000000270 ms. This 1600000000270 ms is designated as the starting time of the nitric oxide concentration increase. The baseline coincidence time point of 1600000000210 milliseconds was retrieved from the centralized record of oxygen-temperature coordination conditions. The calibrated concentration rise start time of 1600000000270 milliseconds was subtracted from the coincidence time point of 1600000000210 milliseconds, yielding a concentration rise time difference of 60 milliseconds. Simultaneously, the corresponding time difference was calculated on the nitrogen concentration sequence. The 78.0% and 77.8% nitrogen concentrations at the corresponding times were retrieved, and the difference was calculated to obtain a -0.2% concentration. This time difference value was then correlated with coordinate information and entered into the data queue to obtain the concentration rise time difference record.

[0036] S403: Call the concentration rise time difference record, extract the coordinate points corresponding to the time difference exceeding the preset reaction hysteresis time reference value, perform numerical encoding mapping on the coordinate points and corresponding differences, arrange and integrate all encoding results in spatial topological order, and generate hysteresis instructions. The output concentration rise time difference record is retrieved. The time difference values ​​corresponding to multiple spatial coordinate points in the record are extracted and screened. The time difference value of 60 milliseconds corresponding to the previously calculated horizontal (1.5m), vertical (2.0m), and vertical height (10.0m) coordinate points is retrieved. A preset reaction lag time baseline value of 40 milliseconds is extracted. This baseline value is calculated by recording the average time span from the exothermic peak to the first detection of nitrogen oxides in 200 sets of interference-free pure combustion tests at different equivalence ratios in a standard laminar flow combustion furnace. The extracted time difference value of 60 milliseconds is compared with the preset reaction lag time baseline value of 40 milliseconds. If 60 milliseconds is greater than 40 milliseconds, a significant reaction lag phenomenon is confirmed at this spatial coordinate point. This exceeding spatial coordinate point and its corresponding difference value of 60 milliseconds are extracted and added to the encoding queue. The coordinate points and their corresponding differences in the encoding queue are numerically encoded and mapped. The horizontal value of 1.5, the vertical value of 2.0, and the height value of 10.0 are extracted and multiplied by a position magnification factor of 10 to convert them into integer position identifiers of 15, 20, and 100. Extracting the time difference of 60 milliseconds exceeding the limit and subtracting the baseline value of 40 milliseconds, a hysteresis over-limit margin of 20 milliseconds is calculated. The horizontal identifier 15, the vertical identifier 20, the height identifier 100, and the over-limit margin of 20 are concatenated to generate the specific encoding string 152010020. After traversing the entire network to complete the encoding mapping of all over-limit points, all encoding results are arranged and integrated according to spatial topology, sorted in ascending order of horizontal position identifier values ​​(smallest to largest), and in ascending order of vertical values ​​when horizontal positions are the same. These results are then merged and entered into the underlying register address stack to generate the hysteresis instruction.

[0037] Please see Figure 6 The specific steps of S5 are as follows: S501: Based on the reaction stage table, hysteresis command set and combustion state data, obtain the airflow velocity and nitric oxide concentration at the associated coordinate points, extract the time stamps of adjacent commands in the hysteresis command set, perform difference calculation to obtain the time difference sequence, and multiply it with the velocity value to obtain the spatial offset distance set. Read the specific spatial coordinates from the generated reaction stage table and hysteresis instruction set. Extract the associated spatial coordinates at a distance of 1.5 meters horizontally, 2.0 meters vertically, and 10.0 meters high. Perform a continuous time-series read operation from the combustion state dataset to extract the continuous change sequence of airflow velocity and the nitric oxide concentration sequence corresponding to this coordinate point. Retrieve the current velocity acquisition timestamp of 1600000000300 milliseconds from the airflow velocity sequence, and simultaneously retrieve the instruction issuance timestamp of 1600000000270 milliseconds corresponding to the aforementioned hysteresis over-limit margin from the hysteresis instruction set. Perform a point-by-point difference calculation operation between the velocity sequence timestamp and the instruction set timestamp. Subtract the instruction issuance timestamp of 1600000000300 milliseconds from the current velocity acquisition timestamp of 1600000000300 milliseconds to calculate the instruction issuance time difference value as 30 milliseconds. Retrieve the airflow velocity value at this coordinate point at the current acquisition time, which is 15.0 milliseconds per millisecond. The calculated command issuance time difference of 30 milliseconds is multiplied by the retrieved airflow velocity value of 15.0 mm per millisecond, yielding an absolute displacement of 450.0 mm within that time period. This absolute displacement of 450.0 mm is established as the spatial offset distance at the corresponding coordinate point. Substituting these parameters into the computational logic, all spatial coordinate points recorded in the hysteresis command set are traversed, and the results of each multiplication operation are integrated to generate a set of spatial offset distances. The calculated spatial offset distance of 450.0 mm falls within the standard scale range of single-step fluid transport within a conventional furnace. This value indicates that the initial nitric oxide cloud generated by the hysteresis reaction has undergone substantial physical displacement downstream of the furnace. Its correlation with the step results lies in the fact that this displacement calculation provides a dynamic spatial index scale for subsequent cross-regional data tracing.

[0038] S502: Based on the spatial offset distance set, the combustion state data is spatially indexed and mapped to locate the downstream spatial section coordinate set. The nitrogen oxide generation values ​​corresponding to multiple coordinates are accumulated and calculated to obtain the section aggregation sequence and then vectorized and recombined to obtain the total nitrogen oxide vector of the downstream section. Read all records from the generated spatial offset distance set. Retrieve the specific spatial offset distance value of 450.0 mm corresponding to the initial three-dimensional coordinate point with a horizontal distance of 1.5 meters, a vertical distance of 2.0 meters, and a height of 10.0 meters. Extract the single-direction vector vertically upward from the main flow field of the furnace as the reference, convert the original height coordinate value of 10.0 meters to an equivalent of 10000.0 mm, and perform a linear summation operation between the height value of 10000.0 mm and the calculated spatial offset distance value of 450.0 mm to calculate the absolute value of the downstream height coordinate after offset mapping as 10450.0 mm, or 10.45 meters. Keep the horizontal coordinate of 1.5 meters and the vertical coordinate of 2.0 meters unchanged, and combine them to generate the downstream center spatial coordinate point after completing the spatial index mapping. Extract all 120 discrete grid coordinates within the horizontal section with a radius span of 0.5 meters surrounding the downstream center coordinate point to generate the downstream spatial section coordinate set. Traverse all coordinate point data in the downstream spatial section coordinate set and extract the real-time nitrogen oxide generation values ​​corresponding to each section coordinate point from the combustion state data matrix. The generated concentration values ​​of 20.5 parts per million (ppm) for point A, 18.2 ppm for point B, and 22.1 ppm for point C are retrieved. These values ​​(20.5, 18.2, and 22.1) are sequentially added to registers, resulting in a total generated concentration of 60.8 ppm for the local cross-section. The accumulated values ​​from all points within the cross-section are arranged chronologically according to the original time series and integrated into a cross-sectional aggregate sequence. A single-column, multi-row matrix reset and transformation operation is performed on the aggregate sequence to reconstruct a structured downstream cross-sectional nitrogen oxide total concentration vector.

[0039] Table 3 Calculation data of total nitrogen oxides in downstream sections

[0040] Table 3 lists the specific nitrogen oxide generation distribution values ​​for some coordinate points within the downstream section after offset mapping. The calculated 10.45-meter height coordinate and 60.8 parts per million concentration cross-section total value are within a reasonable fluctuation range of the furnace redox boundary zone. This result indicates that the mapped section has become the main area where bottom nitrogen oxides accumulate and penetrate with the flow field. Its correlation with the step results is that the aggregated calculation results provide the total amount base conditions for cross-scale analysis in spatial dimension. Finally, the total nitrogen oxide vector of the downstream section is derived.

[0041] S503: Call the downstream section nitrogen oxide total vector and nitric oxide concentration sequence to perform corresponding position superposition operation to obtain superimposed numerical sequence and perform reverse accumulation operation along the flow direction to generate nitrogen oxide emission prediction results; The reconstructed downstream section nitrogen oxide total vector data is retrieved, and the nitric oxide concentration sequence matching the target section at the time of the historical combustion state data is extracted synchronously. The numerical term from the downstream section nitrogen oxide total vector, namely the previously calculated concentration of 60.8 parts per million, is retrieved, and the corresponding background baseline concentration of 35.2 parts per million at the same spatial dimension position in the nitric oxide concentration sequence is simultaneously located and retrieved. The extracted numerical terms are summed at their corresponding positions. The concentrations of 60.8 and 35.2 are added together to calculate the superimposed concentration at the section location as 96.0 parts per million. The independent calculation results from multiple sections across the entire furnace height are integrated to construct a superimposed numerical sequence. The final reference spatial position preset at the top outlet of the furnace is extracted, and a recursive cumulative operation is performed on each value in the superimposed numerical sequence according to a specific direction in reverse order from the top outlet to the bottom burner area along the airflow direction. The superimposed concentration value of 96.0 parts per million (ppm) from the nearest adjacent section near the exit is retrieved and multiplied by the pre-set reverse diffusion attenuation coefficient of 0.95 to obtain the median value of 91.2 ppm for the reverse-order retention concentration. The superimposed concentration value of 85.0 ppm corresponding to the next forward section is extracted and summed with the 91.2 ppm concentration to obtain the reverse-order cumulative local total concentration of 176.2 ppm. The final cumulative concentration of 176.2 ppm is divided by the total number of section samples (parameter 2) to calculate the averaged predicted scalar concentration as 88.1 ppm. The calculated predicted concentration of 88.1 parts per million is below the safe threshold of the standard emission limit. This result indicates that the pollution emissions reaching the outlet after the current furnace combustion state has undergone delayed generation and superposition still meet the basic setting conditions. Its correlation with the step results is that, through the coefficient weighted accumulation calculation operation of each section of the entire watershed, a direct judgment basis that can be used for feedforward closed-loop control is output, and the nitrogen oxide emission prediction result is derived.

[0042] Please see Figure 7 A nitrogen oxide emission prediction system for thermal power plants, comprising: The combustion status module acquires oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration at the same time through sensors inside the furnace of a thermal power plant, performs three-dimensional mesh mapping, generates combustion status data, and transmits it to the fault location module. The fault location module calculates the fault location based on the oxygen volume fraction and temperature at continuous cross sections of the combustion state data. It filters the coordinates according to the negative signs of the oxygen volume fraction difference and the positive signs of the temperature difference, generates a fault coordinate set, and transmits it to the reaction stage module. The reaction phase module, based on the fault coordinate set and combustion state data, analyzes the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compares the order of time, generates a reaction phase table and transmits it to the hysteresis instruction module. The hysteresis instruction module obtains the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration based on the reaction stage table. When the oxygen volume fraction difference is negative and the temperature increases, it judges the change time of nitrogen concentration and nitric oxide concentration, generates a hysteresis instruction set and transmits it to the emission prediction module. The emission prediction module obtains airflow velocity and nitric oxide concentration based on the reaction stage table and hysteresis instruction set, acquires downstream spatial cross-sectional data and calculates the total nitrogen oxide generation, and combines the nitric oxide concentration and airflow velocity to generate the nitrogen oxide emission prediction result.

[0043] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method of predicting nitrogen oxide emissions from a thermal power plant, characterized in that, Includes the following steps: S1: Oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration are acquired simultaneously by sensors inside the furnace of a thermal power plant and then mapped in three dimensions to generate combustion state data. S2: Calculate the fault location based on the oxygen volume fraction and temperature at the continuous cross section of the combustion state data, and filter the coordinates according to the negative sign of the oxygen volume fraction difference and the positive sign of the temperature difference to generate a fault coordinate set; S3: Based on the fault coordinate set and the combustion state data, analyze the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compare the order of time to generate a reaction stage table; S4: Based on the reaction stage table, obtain the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration. When the oxygen volume fraction difference is negative and the temperature increases, determine the change time of nitrogen concentration and nitric oxide concentration, and generate a hysteresis instruction set. S5: Obtain the airflow velocity and nitric oxide concentration according to the reaction stage table and the hysteresis instruction set, obtain the downstream spatial cross-section data and calculate the total nitrogen oxide generation, and combine the nitric oxide concentration and airflow velocity to generate the nitrogen oxide emission prediction result.

2. The method of claim 1, wherein the method is characterized by: The combustion state data includes a spatial coordinate index, an oxygen volume fraction field, a temperature distribution field, and a carbon monoxide concentration field. The fault coordinate set includes oxygen volume fraction abrupt change points, temperature transition points, and symbol-filtered coordinate identifiers. The reaction stage table includes the starting time of carbon monoxide decrease, the starting time of hydroxyl radical increase, and the stage sequence identifiers. The hysteresis instruction set includes oxygen volume fraction difference symbol identifiers, temperature increase identifiers, and nitrogen concentration change delays. The nitrogen oxide emission prediction results include the total amount of nitrogen oxides in the downstream section and the inverse superimposed concentration sequence.

3. The method of claim 1, wherein the method is characterized by: The specific steps of S1 are as follows: S101: Obtain oxygen volume fraction, temperature, carbon monoxide concentration, hydroxyl radical concentration, nitrogen concentration, nitric oxide concentration and airflow velocity at the same time through sensors inside the furnace of a thermal power plant and align them in time. Match the multi-parameter sequences according to time consistency constraints to generate multi-parameter synchronous distribution data. S102: Based on the multi-parameter synchronous distribution data, call the spatial coordinate index of multi-coordinate points, rearrange the three-dimensional coordinates of the multi-parameter values ​​according to the spatial coordinate order, map the discrete coordinates to the regular grid point set, and perform neighborhood consistency interpolation calculation on the parameters of adjacent nodes to obtain the three-dimensional grid parameter field. S103: The oxygen volume fraction, temperature, carbon monoxide concentration, nitrogen concentration, airflow velocity and hydroxyl radical concentration in each spatial region are fused according to the three-dimensional grid parameter field, and combustion state data are generated based on the weighted superposition and normalization of parameter vectors.

4. The method of claim 1, wherein the method is a method of predicting nitrogen oxide emissions from a thermal power plant. The specific steps of S2 are as follows: S201: Obtain the oxygen volume fraction and temperature at a continuous cross section of the combustion state data and align them with spatial coordinates. Perform adjacent spatial point difference based on the oxygen volume fraction sequence corresponding to multiple coordinate points within the same cross section, and synchronously perform corresponding spatial point difference in combination with the temperature value sequence to generate an oxygen-temperature difference correlation sequence. S202: Based on the oxygen-temperature difference correlation sequence, perform sign discrimination on the oxygen volume fraction difference corresponding to the multi-coordinate index and extract negative values, and at the same time perform sign discrimination on the temperature difference at the same index position and extract positive values, mark the coordinates where the oxygen volume fraction difference is negative and the temperature difference is positive, and generate a sign-matching coordinate index set; S203: Based on the symbol matching coordinate index set, extract the original coordinate point sequence within the corresponding spatial section by index mapping, aggregate all the marked indexed coordinate points and reconstruct the spatial position sequence to generate a fault coordinate set.

5. The method of claim 1, wherein the method is a method of predicting nitrogen oxide emissions from a thermal power plant. The specific steps for S3 are as follows: S301: Based on the fault coordinate set and the combustion state data, obtain the carbon monoxide concentration and hydroxyl radical concentration at the corresponding spatial coordinate points, extract the concentration values ​​of multiple spatial coordinate points according to the time series, and align the two types of concentration values ​​corresponding to the same coordinate point on the time axis to obtain a time series concentration pairing set. S302: Based on the time-series concentration pairing set, perform time-by-time difference operation on the carbon monoxide concentration sequence, extract the concentration decrease rate of change and compare it with a preset decrease trigger threshold, and at the same time perform the same type of difference operation on the hydroxyl radical concentration sequence and match it with a preset increase trigger threshold to generate start time index pairs. S303: Based on the aforementioned start time index pair, calculate the difference between the carbon monoxide falling time index and the hydroxyl radical rising time index at the same spatial coordinate point, and classify and encode them according to the time difference value range. Organize and sequence map the corresponding encoding results of multiple coordinate points to obtain the reaction stage table.

6. The method of claim 5, wherein the method further comprises: The decrease trigger threshold is determined by collecting a carbon dioxide concentration sequence with consecutive timestamps during the initial combustion phase, performing differential operations on the concentration values ​​corresponding to adjacent time points within the sequence to obtain a discrete difference set, extracting all negative difference parameters within the difference set, calculating the statistical mean and standard deviation values ​​corresponding to all negative difference parameters, and summing the statistical mean and standard deviation values ​​to determine the threshold.

7. The method of claim 1, wherein the method is a method of predicting nitrogen oxide emissions from a thermal power plant. The specific steps of S4 are as follows: S401: Based on the reaction stage table and the combustion state data, obtain the oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration at the associated spatial coordinate point, perform differential extraction on the oxygen volume fraction and extract all negative terms, extract the increasing interval of temperature and cross-match it with the all negative terms to generate an oxygen-temperature co-condition set. S402: Based on the overlapping time points in the oxygen-temperature synergistic condition set, call the nitrogen concentration and nitric oxide concentration at the associated spatial coordinate points, perform difference calculations on the two types of concentrations according to time sequence, extract the time node where the difference turns from negative to positive and mark it as the start time of the rise, and analyze the difference to obtain the concentration rise time difference record. S403: Call the concentration rise time difference record, extract the coordinate points corresponding to the time difference exceeding the preset reaction hysteresis time reference value of multiple spatial coordinate points, perform numerical encoding mapping on the coordinate points and corresponding differences, arrange and integrate all encoding results in spatial topological order, and generate hysteresis instruction set.

8. The method of claim 7, wherein the method further comprises: The baseline value of the reaction lag time is determined by collecting nitrogen and nitric oxide concentration sequences at continuous timestamps, extracting the values ​​of the two types of concentration sequences and converting them into timestamps corresponding to the points of positive growth, subtracting the two types of timestamps to obtain the experimental time difference set, calculating the arithmetic mean and standard deviation of all values ​​in the experimental time difference set and adding them together.

9. The method of claim 1, wherein the method is a method of predicting nitrogen oxide emissions from a thermal power plant. The specific steps of S5 are as follows: S501: Based on the reaction stage table, the hysteresis instruction set and the combustion state data, obtain the airflow velocity and nitric oxide concentration at the associated coordinate point, extract the time stamps of adjacent instructions in the hysteresis instruction set, perform difference calculation to obtain the time difference sequence, and multiply it with the velocity value to obtain the spatial offset distance set. S502: Based on the spatial offset distance set, perform spatial index mapping on the combustion state data, locate the downstream spatial section coordinate set, accumulate the nitrogen oxide generation values ​​corresponding to multiple coordinates, obtain the section aggregation sequence, and perform vectorization and recombination to obtain the downstream section nitrogen oxide total vector. S503: The total nitrogen oxide emission vector of the downstream section is called and the nitric oxide concentration sequence is superimposed at the corresponding positions to obtain the superimposed numerical sequence. Then, the superimposed numerical sequence is accumulated in reverse order along the flow direction to generate the nitrogen oxide emission prediction result.

10. A nitrogen oxide emission prediction system for thermal power plants, characterized in that, The system is used to implement the nitrogen oxide emission prediction method for thermal power plants according to any one of claims 1-9, the system comprising: The combustion status module acquires oxygen volume fraction, temperature, nitrogen concentration, airflow velocity, carbon monoxide concentration and hydroxyl radical concentration at the same time through sensors inside the furnace of a thermal power plant, performs three-dimensional mesh mapping, generates combustion status data, and transmits it to the fault location module. The fault location module calculates the fault location based on the oxygen volume fraction and temperature at continuous cross sections of the combustion state data, filters coordinates according to the negative signs of the oxygen volume fraction difference and the positive signs of the temperature difference, generates a fault coordinate set, and transmits it to the reaction stage module. The reaction stage module, based on the fault coordinate set and the combustion state data, analyzes the time difference between the decrease in carbon monoxide concentration and the increase in hydroxyl radical concentration and compares the order of time, generates a reaction stage table and transmits it to the hysteresis instruction module. The hysteresis instruction module obtains the corresponding spatial coordinates of oxygen volume fraction, temperature, nitrogen concentration and nitric oxide concentration based on the reaction stage table. When the oxygen volume fraction difference is negative and the temperature increases, it judges the change time of nitrogen concentration and nitric oxide concentration, generates a hysteresis instruction set and transmits it to the emission prediction module. The emission prediction module obtains the airflow velocity and nitric oxide concentration based on the reaction stage table and the hysteresis instruction set, acquires downstream spatial cross-sectional data and calculates the total nitrogen oxide generation, and combines the nitric oxide concentration and airflow velocity to generate the nitrogen oxide emission prediction result.

Images