A method of boiler oxygen control

Abstract

The present application relates to the technical field of energy production, and provides a boiler oxygen content control method, comprising: collecting operation data including combustion state, air distribution state, fuel supply state, ammonia mixing state and stable combustion state; time delay alignment is performed on the detected quantity with transmission lag, and a credibility is generated according to the deviation of the characterization of the oxygen potential reference value in the furnace; a credible effective oxygen content is constructed based on the oxygen content detection value, the credibility and the reduction risk penalty term; wherein the stable combustion state information participates in the calculation of the penalty term and the correction of the lower oxygen content boundary in the form of the flame stability evaluation quantity; the upper and lower operation boundaries of the oxygen content are determined according to the current load and the ammonia mixing condition, and the reduction debt is accumulated synchronously; when the predicted credible effective oxygen content will fall below the lower oxygen content boundary or the reduction debt exceeds the threshold value, local air distribution repair is performed first, total air volume correction is performed only when the local air distribution repair is insufficient, and the reverse order is performed during the retreat.

Description

Technical Field

[0001] This invention relates to the field of energy production technology, specifically a method for controlling oxygen levels in a boiler. Background Technology

[0002] In the oxygen control scenario of a green ammonia-blended coal-fired boiler operating under deep peak-shaving conditions, the boiler operation simultaneously exhibits two characteristics: rapid load increases and decreases and dynamic changes in the ammonia blending ratio. This leads to variations in tail-end oxygen levels, localized furnace oxygen levels, nitrogen oxides, and CO. The coupling relationship between escape and stable combustion states is enhanced; especially during deep peak shaving and green ammonia co-firing operations, local combustion organization is more prone to change, and the oxygen supply demand of the main combustion zone and the burnout zone is no longer synchronized. This causes the boiler to face combustion instability caused by local oxygen deficiency, CO increase, and There is a risk of escape, and there are also problems such as increased nitrogen oxide emissions and decreased boiler efficiency caused by excessive overall air supply.

[0003] Current technologies typically employ a combination of cross-limited combustion control and tail-end flue gas oxygen feedback correction to control the total boiler air volume. Empirical oxygen curves or excess air coefficients are used as control targets, and the total air volume is biased and corrected based on the tail-end flue gas oxygen content. Under green ammonia co-firing conditions, to balance burnout efficiency and pollutant emissions, further measures such as staged air supply, burner modification, optimized ammonia injection location, or stratified co-firing are usually combined to coordinately regulate the combustion process, achieving oxygen control and nitrogen oxide suppression during boiler operation.

[0004] However, in operating scenarios where deep peak shaving and green ammonia co-firing coexist, existing technologies rely solely on tail-end oxygen feedback as the primary control basis, which easily leads to the problem of false oxygen excess. Tail-end oxygen only reflects the overall oxygen content at the flue gas discharge point and cannot directly characterize the local oxygen supply status in the main combustion zone and burnout zone. Therefore, when local areas within the furnace are already oxygen-deficient, the tail-end oxygen may still be within the normal range or even slightly elevated. This makes it impossible to promptly identify localized oxygen deficiency and make targeted adjustments, resulting in unstable combustion, increased CO levels, and other issues. Problems such as escape are difficult to contain in a timely manner. Summary of the Invention

[0005] In view of the existing problems mentioned above, a method for controlling boiler oxygen content is proposed.

[0006] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a boiler oxygen control method, comprising:

[0007] Collect operational data including combustion status, air distribution status, fuel supply status, ammonia blending status, and stable combustion status;

[0008] The detection quantities with transmission lag are time-delay aligned, and the confidence level is generated based on their characterization deviation of the furnace oxygen potential reference value; the furnace oxygen potential reference value is estimated by the soft measurement model based on the current air volume, coal volume, ammonia volume and furnace temperature distribution.

[0009] A reliable effective oxygen content is constructed based on the oxygen content detection value, the aforementioned confidence level, and the reduction risk penalty term; wherein the stable combustion state information participates in the calculation of the penalty term and the correction of the lower boundary of the oxygen content in the form of a flame stability evaluation quantity;

[0010] Determine the upper and lower operating boundaries of oxygen content based on the current load and ammonia blending conditions, and simultaneously accumulate and reduce debt.

[0011] When it is predicted that the credible effective oxygen level will fall below the lower limit of oxygen level or the debt reduction exceeds the preset threshold, local ventilation repair is performed first. Total air volume correction is performed only when local ventilation repair is insufficient, and the process is reversed during pullback.

[0012] Beneficial effects

[0013] Compared with the prior art, the present invention has the following advantages:

[0014] 1. By aligning and compensating for time delays in various detection parameters, the tail oxygen, CO, nitrogen oxides, and... The escape concentrations all correspond to the same furnace state section, eliminating the time misalignment caused by the delay in flue gas flow and avoiding false oxygen excess misjudgment caused by using lag signals as the basis for real-time control from the source.

[0015] 2. A two-dimensional calibration table is established with load and ammonia blending ratio as two dimensions. Dynamic upper and lower boundaries of oxygen are generated in real time through bilinear interpolation, so that the operating boundary can be adjusted in real time according to the change of operating conditions. This solves the problem of inaccurate oxygen supply margin caused by using the rated operating condition boundary under low load or high ammonia blending ratio conditions.

[0016] 3. The CO concentration, Escape concentration and flame stability evaluation values ​​are uniformly converted into equivalent oxygen penalty values, which participate in the construction of credible effective oxygen in a structured manner. This allows various risk signals to suppress credible effective oxygen in advance before the oxygen level falls below the lower boundary, achieving an advanced response to reduction risks and avoiding the problem of delayed intervention when relying on a single tail oxygen feedback. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation on the scope of this application.

[0018] Figure 1This is a schematic flowchart of the boiler oxygen control method provided in an embodiment of the present invention;

[0019] Figure 2 This is a comparison diagram of the effects of the present invention and the prior art, where gray bars represent the prior art and black bars represent the present invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives, and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention.

[0021] Example 1:

[0022] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a boiler oxygen control method, which includes the following steps:

[0023] Multi-source operation data acquisition: Acquires operation data including combustion status, air distribution status, fuel supply status, ammonia blending status, and stable combustion status;

[0024] Delay alignment and credibility assessment: Delay alignment is performed on the detection quantities with transmission lag, and credibility is generated based on their characterization deviation of the reference value of oxygen potential in the furnace.

[0025] Construction of Confidential Effective Oxygen Content: A Confidential Effective Oxygen Content is constructed based on the oxygen content detection value, the aforementioned confidence level, and the reduction risk penalty term; wherein the stable combustion state information participates in the calculation of the penalty term and the correction of the lower boundary of the oxygen content in the form of a flame stability evaluation quantity;

[0026] Dynamic oxygen quantity dual boundary and reduction debt establishment: Determine the upper and lower operating boundaries of oxygen quantity based on the current load and ammonia blending conditions, and simultaneously accumulate the reduction debt;

[0027] Layered air distribution priority control: When it is predicted that the reliable effective oxygen level will fall below the lower limit of oxygen level or the debt reduction exceeds the threshold, local air distribution repair is performed first. Total air volume correction is only performed when local air distribution repair is insufficient, and the reverse order is performed during the rollback.

[0028] In this embodiment, the specific content of multi-source operational data acquisition includes:

[0029] Combustion status data includes oxygen content, CO concentration, nitrogen oxide concentration, and furnace temperature distribution in the tail flue. Oxygen content is measured using a zirconia online oxygen analyzer, installed at the economizer outlet section, with at least two measuring points at each section, and the average value is used. CO and nitrogen oxides are measured using an infrared absorption flue gas analyzer, installed near the same section. Furnace temperature distribution is obtained from a thermocouple array arranged on the walls around the furnace, with at least four measuring points in each layer. Specifically, the average temperature and temperature difference within each layer are calculated for each layer's thermocouple measuring points and used as the temperature distribution characteristic quantity. When a single measuring point is abnormal, the abnormal point is removed, and the calculation is based on the remaining valid measuring points. The tail flue detection value is used as a lagging characterization quantity of the combustion status in the furnace and input into subsequent time delay alignment and reliability assessment steps, and is not directly used as the sole criterion for judging the local oxygen supply status in the furnace.

[0030] The air distribution status data includes total air volume, primary air volume, secondary air volume, staged air volume, and the opening degree of each damper; the total air volume and primary air volume are taken from the differential pressure flow meter at the outlet of the blower and the primary blower; the secondary air volume and staged air volume are taken from the air volume measuring device at the inlet of each air box; the opening degree of each damper is taken from the position feedback sensor of the damper actuator.

[0031] Fuel supply status data includes the amount of coal fed to each layer, which is taken from the weighing belt scale signal of the coal feeder and converted into mass flow rate per unit time. If there is no online detection of coal quality, the rolling average of the recent test calorific value of the coal entering the furnace is used instead.

[0032] The ammonia blending status data includes the ammonia injection amount of each burner layer and the corresponding ammonia injection valve opening feedback, which is taken from the flow meter and the valve actuator position sensor. The escape concentration is measured by an online analyzer located in the flue gas duct that characterizes the unreacted ammonia content at the boiler tail. If this instrument is not available on-site, it is estimated by a soft sensor model based on the current ammonia injection rate, SCR inlet and outlet nitrogen oxide concentrations, inlet flue gas temperature, and historical sample data. The estimation result is simultaneously labeled as a soft sensor value for subsequent reliability assessment. The soft sensor model takes ammonia injection rate, flue gas temperature, and nitrogen oxide-related detection quantities as inputs and outputs the current value. The concentration estimate is output along with a model confidence marker characterizing the reliability of the estimate.

[0033] The stable combustion status data includes flame image signals and flame pulsation acoustic signals. The flame images are acquired by industrial-grade CCD cameras installed at the observation holes of each burner layer, with a frame rate of not less than 25 frames / s. The shooting area covers the main combustion zone of the corresponding burner layer. The mean brightness, standard deviation of brightness, and flame centroid offset of each frame image are extracted as image features. The acoustic signals are acquired by piezoelectric acoustic sensors arranged on the side wall of the furnace. The root mean square value of pulsation in the 100-500Hz frequency band is extracted as acoustic features. The image features and acoustic features are normalized and then weighted to synthesize a flame stability evaluation value. In one embodiment, the image feature weight can be set to 0.6 and the acoustic feature weight can be set to 0.4.

[0034] Specifically, the image and acoustic features are calculated by taking the arithmetic mean of the gray values ​​of all pixels in the effective combustion area of ​​the current frame image to obtain the average brightness of that frame. The effective combustion area is defined by a preset region of interest mask, the mask range of which is calibrated according to the geometric position of the burner during the unit commissioning phase and remains fixed during operation. Within the effective combustion area, the root mean square of the difference between all pixel gray values ​​and the average brightness is calculated to obtain the standard deviation of brightness for that frame. The standard deviation of brightness reflects the uniformity of brightness distribution inside the flame; when combustion is unstable, the dispersion of flame brightness distribution increases, and the standard deviation of brightness increases accordingly. Using the gray values ​​of each pixel within the effective combustion area as weights, the weighted average of pixel coordinates in the horizontal and vertical directions is calculated to obtain... The current frame's flame brightness centroid coordinates are obtained; the difference between these coordinates and the reference centroid coordinates of the burner under stable operating conditions is taken, and the modulus is calculated to obtain the flame centroid offset; the reference centroid coordinates are determined by the average centroid coordinates of 100 consecutive frames of images under stable combustion conditions during the unit commissioning phase; within each control step, the acquired sound pressure time series is processed as follows: first, the original signal is bandpass filtered to retain components in the 100 to 500 Hz frequency band, which corresponds to the main energy distribution range of combustion pulsations in the furnace, filtering out low-frequency mechanical vibrations and high-frequency electrical interference; then, the root mean square value is calculated for the filtered signal sequence, that is, the square root of the average of the squares of the amplitudes of each sampling point in the sequence is taken to obtain the acoustic characteristic quantity within the control step;

[0035] All operational data are timestamped according to a unified control cycle. In one embodiment, 5 seconds is used. Data with a sampling frequency higher than the control cycle is statistically downsampled in each control cycle, while data with a sampling frequency lower than the control cycle update frequency is padded with the previous valid value.

[0036] In this embodiment, the specific content of latency alignment and reliability evaluation includes:

[0037] Detectable quantities with transmission lag include oxygen, CO, nitrogen oxides, and... Escape concentration: All of the above signals have a time delay introduced by the flue gas flowing from the main combustion zone to the detection section, and time delay compensation and alignment need to be completed before entering the subsequent calculations;

[0038] Firstly, during the unit commissioning phase, the time delay parameters are determined through step tests. Specifically, under stable load conditions, such as within a continuous observation window, when both the boiler load change amplitude and load change rate are below a preset threshold, the load stability condition is deemed met. A step disturbance with an amplitude of 5% of the rated value is applied to the total air volume, and the difference between the response start time of each detected quantity and the air volume step time is recorded simultaneously. This difference is the estimated time delay value of each detected quantity at that load point. The response start time is defined as the earliest time when the mean value of the corresponding detected quantity deviates from the baseline before the disturbance by more than a preset threshold and remains continuously for no less than two control cycles. The preset threshold is preferably twice the standard deviation of the steady-state noise of the detected quantity. Calibration is performed separately for rated load, 75% load, 50% load, and 40% load, and a load-time delay correspondence table is established. If a shutdown test is not possible, the effective furnace volume divided by the current flue gas volumetric flow rate is used as the time delay, where the flue gas volumetric flow rate is calculated from the air volume and fuel quantity. For each detected quantity, the corresponding time delay value is obtained by interpolation according to the current load. Shift the historical sampling sequence of this detection quantity forward. Each sampling step is used to align the furnace state at the same time on the time axis. The aligned measurements are considered as synchronous measurement results reflecting the same furnace state section.

[0039] Using the reference value of in-furnace oxygen potential estimated by the soft measurement model based on the current air volume, coal volume, ammonia volume, load, and furnace temperature distribution as a benchmark, the time-delay aligned measured quantities are converted to the equivalent oxygen potential space, and the deviation from the benchmark value is calculated, i.e., the characterization deviation. For the oxygen volume in the tail flue, it is directly used as the oxygen potential characterization value. For carbon monoxide concentration, nitrogen oxide concentration, and... The escape concentration is converted into an equivalent oxygen potential value with the same dimensions as oxygen content, based on the pre-calibrated mapping relationship between the load segment and the ammonia doping segment. This mapping relationship is implemented using a lookup table interpolation function. For each detected quantity, a confidence coefficient is generated.

[0040] ,

[0041] Where i represents the index of the detection quantity. This represents the confidence coefficient corresponding to the i-th detection quantity. This represents the characterization bias of the i-th detection after time delay alignment. The dimensionless normalization coefficient is determined by the deviation distribution during the calibration stage. During the calibration stage, a sample of the characterization deviation of each detection quantity under normal working conditions is collected. The 90th percentile of the absolute value of the sample deviation is taken as the reference value. k is set to a value that makes the confidence coefficient corresponding to the reference value equal to 0.5.

[0042] Specifically, the furnace oxygen potential reference value is calculated using the current total air volume, primary air volume, secondary air volume, staged air volume, coal feed rate, and ammonia injection rate as inputs. The theoretical excess air coefficient is calculated based on the stoichiometric relationship of combustion, i.e., the actual oxygen supply calculated from the current total air volume is divided by the theoretical oxygen demand for complete combustion estimated based on the coal feed rate and coal calorific value. The resulting ratio is the theoretical excess air coefficient. This is then combined with the estimated furnace leakage rate to calculate the theoretical oxygen content of the main combustion zone section. The furnace leakage rate is estimated by the ratio between the differential pressure measurement points inside and outside the furnace and the air volume meter at the blower outlet. The oxygen content of each height section of the furnace is then calculated. Differential pressure measurements are converted into the corresponding cross-section leakage air volume, summed, and divided by the total air volume at the blower outlet. A correction factor is added to the theoretical oxygen content. The correction factor is given by a linear regression model with furnace temperature distribution characteristics, load change rate, and ammonia blending ratio as inputs. The linear regression model expresses the correction factor as a weighted linear combination of the inputs, i.e., the correction factor equals the sum of the three inputs (furnace temperature distribution standard deviation, load change rate, and ammonia blending ratio) multiplied by their respective coefficients, plus a constant bias term. Finally, the reference value of oxygen potential in the furnace equals the theoretical oxygen content plus the correction factor.

[0043] The larger the characterization deviation, the lower the confidence coefficient of the corresponding detection quantity, which drops to close to 0 at the lowest. When the deviation is within the normal fluctuation range, the confidence coefficient is close to 1. The confidence coefficient of each detection quantity is calculated independently and used as the weighting basis for constructing a reliable effective oxygen quantity in the future. It is not output separately as a control command.

[0044] By performing time delay compensation alignment on the detected quantities with transmission lag, the problems of oxygen, CO, nitrogen oxides, and other parameters in the tail flue gas were resolved. The issue of escape concentration being out of sync with the real-time state inside the furnace due to flue gas flow delay was addressed by ensuring that all detection quantities used in subsequent calculations corresponded to the same furnace state section, thus avoiding control misjudgments caused by time misalignment. During the unit commissioning phase, time delay parameters were calibrated for multiple load points, and an estimation alternative based on flue gas flow rate was provided when shutdown testing was not possible. This ensured that the acquisition of time delay parameters balanced calibration accuracy with engineering feasibility, maintaining reasonable compensation accuracy under different operating conditions. By introducing a reference value of oxygen potential inside the furnace as a benchmark, all detection quantities were uniformly converted to an equivalent oxygen potential space before calculating the characterization deviation. Based on this, a confidence coefficient was independently generated for each detection quantity, enabling quantitative identification of detection distortions caused by sensor drift, measuring point contamination, and local flue gas stratification.

[0045] In this embodiment, the specific content of the reliably effective oxygen content structure includes:

[0046] The measured oxygen content in the tail flue after time delay alignment and the furnace oxygen potential reference value output by the soft measurement model are used as two inputs. The furnace oxygen potential reference value has been converted into an equivalent oxygen content characterization value with the same dimension as the measured oxygen content in the tail flue according to the combustion stoichiometry relationship. The unit is vol%. The two can be directly used in the weighted fusion calculation. The corresponding confidence coefficients are read respectively. The oxygen content of each path is multiplied by its corresponding confidence coefficient and summed. Then, it is divided by the sum of the two confidence coefficients to obtain the fused oxygen content. The measured oxygen content refers to the measurement value of the zirconia online oxygen analyzer.

[0047] The CO concentration value after readout alignment is compared with a preset CO operating limit. In one embodiment, the preset CO operating limit is 200 mg / m³. When the CO concentration is below 80% of the operating limit, the CO equivalent penalty is zero. When the CO concentration is between 80% and 100% of the operating limit, the penalty is determined by linear interpolation, corresponding to an equivalent oxygen correction range of 0.2 to 0.5 vol%. When the CO concentration reaches or exceeds the operating limit, the penalty is the upper limit of 0.5 vol%. Escape concentration value, when When the escape concentration is below 80% of the operating limit, The equivalent penalty is zero; when When the escape concentration is between 80% and 100% of the operating limit, the penalty is determined by linear interpolation; when When the escape concentration reaches or exceeds the operating limit, the penalty is the upper limit of 0.5 vol%; CO and The combined penalty amount is capped at 0.8 vol%, with any excess being truncated to prevent excessive penalty from causing a persistently high total airflow; the CO concentration and The escape concentration was converted into an equivalent oxygen penalty value affecting the oxygen supply safety margin based on historical calibration results and stable operating condition samples.

[0048] To calculate the flame instability penalty, first read the flame stability evaluation value obtained after normalization of image and acoustic features. In one embodiment, if the evaluation value does not exceed a preset stable combustion threshold, it is set to 0.2, and the flame instability penalty is zero. If the evaluation value exceeds the stable combustion threshold, the magnitude of the exceedance is linearly mapped to an equivalent oxygen penalty. For example, if the preset stable combustion threshold is 0.2, the maximum value of the flame instability penalty corresponds to an equivalent oxygen correction of 0.2 vol%, and the upper limit of the index is 1.0. When the current measured value of the flame stability evaluation value is 0.6, its exceedance of the stable combustion threshold is 0.4, and the maximum exceedance of the index is 0.8. Therefore, the current flame instability penalty is 0.4 divided by 0.8 and multiplied by 0.2 vol%, resulting in 0.1 vol%. When the flame stability evaluation value reaches the upper limit of 1.0, the penalty is set to the maximum value of 0.2 vol%. When the index does not exceed the stable combustion threshold of 0.2, the penalty is zero, and no additional correction is applied to the credible effective oxygen content.

[0049] The obtained CO, The three items of flame instability penalty are added together to obtain the total reduction risk penalty, and the total upper limit is truncated to not exceed 1.0 vol%. The reduction risk penalty is subtracted from the fused oxygen to obtain the reliable effective oxygen, which represents the equivalent level of the safe margin of oxygen supply in the furnace after comprehensively considering the reliability of detection and the current reduction risk. The flame stability evaluation value is read. When it exceeds the preset stable combustion threshold, an upward correction amount proportional to the amount of the index exceeding is superimposed on the lower boundary benchmark value of oxygen obtained by interpolation under the current operating conditions, so that the minimum allowable oxygen supply margin when combustion is unstable is automatically narrowed.

[0050] Specifically, the oxygen content lower boundary upward shift correction involves reading the current flame stability evaluation value and comparing it with a preset flammability threshold. When the flame stability evaluation value does not exceed the flammability threshold, the oxygen content lower boundary is taken as the benchmark value obtained by interpolation under the current operating conditions, and no correction is applied. When the flame stability evaluation value exceeds the flammability threshold, the upward shift correction amount is calculated as follows: the magnitude by which the flame stability evaluation value exceeds the flammability threshold is divided by the maximum magnitude by which the index can exceed it, yielding the excess ratio. This excess ratio is then multiplied by the preset maximum lower boundary correction amount to obtain the current upward shift correction amount. For example, if the preset flammability threshold is 0.2, the maximum lower boundary correction amount is 0.5 vol%, and the upper limit of the index is 1.0, when the current flame stability evaluation value is 0.6, the excess magnitude is 0.4, the maximum magnitude that can exceed it is 0.8, and the excess ratio is 0.5. The upward correction amount is 0.5 multiplied by 0.5vol%, which equals 0.25vol%. When the flame stability evaluation value reaches the upper limit of 1.0, the upward correction amount takes the maximum value of 0.5vol%. The upward correction amount is added to the lower boundary reference value of oxygen obtained by interpolation under the current operating condition to obtain the corrected lower boundary of oxygen. The corrected lower boundary of oxygen only takes effect within the current control step. The flame stability evaluation value is reread and recalculated in the next control step, and it does not have a cumulative effect on the value in the reference calibration table. When the flame stability evaluation value falls below the stable combustion threshold, the upward correction amount is zero, and the lower boundary of oxygen is restored to the reference value. The flame stability evaluation value is used in the penalty term calculation to reflect the current decrease in the oxygen supply safety margin, and it is used in the lower boundary correction of oxygen to improve the boundary trigger sensitivity. The two have different objects of action.

[0051] By CO and The escape concentration is converted into an equivalent oxygen penalty value and added to the reduction risk penalty item, resulting in higher CO levels or... The risk signal of excessive escape participates in the construction of credible effective oxygen in a structured manner, rather than acting as an independent trim correction. This allows the credible effective oxygen to be reduced in advance before the oxygen level falls below the lower boundary, achieving an advanced response to reduction risks and avoiding the lag problem of relying on a single tail oxygen feedback. A flame stability evaluation quantity is introduced and given a dual action path. On the one hand, it participates in the construction of credible effective oxygen in the form of a penalty quantity, so that the equivalent characterization value of the oxygen supply safety margin under combustion instability is automatically reduced. On the other hand, it applies an upward correction to the lower boundary of oxygen, so that the boundary triggering condition is automatically met in advance when combustion is unstable. The two paths act on different objects and are independent of each other, which together improves the system's response sensitivity to combustion instability conditions. Moreover, the flame stability evaluation quantity is not directly output as a control command, avoiding unintended interference of the stable combustion signal on the air distribution control.

[0052] In this embodiment, the specific details of establishing the dynamic oxygen dual boundary and reduction debt include:

[0053] First, an oxygen content boundary calibration table is established. During the unit commissioning phase, a two-dimensional calibration table is established with load and ammonia blending ratio as two dimensions. The calibration points cover four load levels: rated load, 75% load, 50% load, and 40% load, and four ammonia blending ratio levels: 0%, 10%, 20%, and 30%, forming a total of 16 calibration points. The ammonia blending ratio is the proportion of ammonia fuel input heat to the total fuel input heat of coal and ammonia per unit time. For each calibration point, when the load stability condition is met and the fluctuation of the ammonia blending ratio within the continuous observation window does not exceed a preset threshold (e.g., the fluctuation within 60 seconds does not exceed 1% of the target value), the oxygen content is gradually reduced from the current operating level, with each reduction not exceeding 0.2 vol%, and each reduction step lasting at least 5 minutes. CO concentration is observed simultaneously. Changes in escape concentration, flame stability evaluation parameters, and nitrogen oxide concentration; CO concentration, The lowest oxygen value that continuously maintains the escape concentration and flame stability evaluation within the acceptable range is recorded as the lower boundary benchmark value of oxygen at the calibration point. The criteria for determining the acceptable range are as follows: the CO concentration does not exceed 80% of the preset CO operating limit. Escape concentration does not exceed preset limit The escape operation limit is 80%; the flame stability evaluation value does not exceed the preset stable combustion threshold; the nitrogen oxide concentration does not show a sudden increase, that is, the change in nitrogen oxide concentration between two adjacent downward steps does not exceed 10 mg / m³; during the gradual increase of oxygen content, when there is a difference between two adjacent upward steps, the CO concentration, When the improvement of escape concentration and flame stability evaluation quantity does not exceed 5% of their respective limits, and the increase in nitrogen oxide concentration exceeds the preset threshold or the increase in flue gas temperature exceeds the preset threshold, wherein the oxide concentration increment threshold is preferably 10 mg / m³ and the flue gas temperature increase threshold is preferably 3℃, the current oxygen value is recorded as the upper boundary benchmark value of oxygen at the calibration point, wherein when the increase in flue gas temperature between two adjacent upward adjustment steps exceeds 2℃ and there is no improvement trend when the upward adjustment continues, it is determined that the loss of flue gas temperature has increased significantly.

[0054] During online operation, the current boiler load and ammonia blending ratio are read, and the load and ammonia blending ratio range is located in the calibration table. The lower and upper oxygen boundary reference values ​​corresponding to the current operating condition are calculated using the bilinear interpolation method. When the load change rate exceeds a preset threshold, a dynamic correction of no more than 0.3 vol% is added to the lower oxygen boundary based on the interpolation result to cope with the transient deviation of the oxygen potential in the furnace when the load changes rapidly. The preset threshold for the load change rate is 1% of the rated load per minute. An interpolation calculation is performed once within each control step to update the current operating boundary in real time. The current flame stability evaluation value is read, and the correction value is calculated according to the oxygen lower boundary upward correction method. This correction value is then superimposed on the oxygen lower boundary reference value to obtain the final effective oxygen lower boundary within this control step.

[0055] Within each control step, the difference between the lower boundary of oxygen and the current credible effective oxygen is calculated. If the difference is greater than zero, it can be believed that the effective oxygen is below the lower boundary of oxygen. The difference is then multiplied by the control step time and added to the reduction debt. If the difference is not greater than zero, the reduction debt is multiplied by a risk release coefficient for attenuation. The risk release coefficient ranges from 0.3 to 0.8 and is determined by historical disturbance conditions. The smaller the risk release coefficient, the faster the reduction debt attenuates. The control step is 5 seconds. The physical meaning of the reduction debt is the product of the accumulated oxygen deficiency and its duration within the current control cycle, reflecting the degree of accumulation of local reduction risk, and is used for subsequent stratified control triggering judgment. When the reduction debt increases continuously and there are load acceleration / deceleration commands at the same time, it is determined that it is about to enter a local reduction accumulation state, triggering the subsequent air distribution repair step. If the reduction debt shows an increasing trend for at least 3 consecutive control steps, it is determined to be a continuous increase.

[0056] By establishing a two-dimensional calibration table with load and ammonia blending ratio as two dimensions, and calibrating the upper and lower boundaries of oxygen content at each calibration point, the operating boundary can be adjusted in real time according to changes in load and ammonia blending conditions. This solves the problem that traditional fixed oxygen content curves cannot adapt to scenarios where deep peak shaving and green ammonia blending co-firing coexist. It also avoids inaccurate oxygen supply margin caused by using the rated operating condition boundary under low load or high ammonia blending ratio conditions. The acceptable range judgment conditions for the lower boundary calibration are clearly defined, including CO concentration, The escape concentration, flame stability evaluation quantity, and nitrogen oxide surge rate are four indicators that are simultaneously effective as quantitative constraints. This ensures that the lower boundary benchmark value of oxygen content corresponds to the true safe operating lower limit at each calibration point, rather than relying on the experience judgment of operators, thus improving the objectivity and repeatability of the calibration results.

[0057] In this embodiment, the specific content of the stratified air distribution priority control includes:

[0058] Within each control step, a condition determination is performed. Based on the current reliable effective oxygen content and its changing trend over the last three control steps, the reliable effective oxygen content for the next control step is predicted by linear extrapolation. If the predicted value is lower than the current effective oxygen content lower boundary, then trigger condition one is determined to be met. If the current reduction debt exceeds the preset reduction debt trigger threshold, which in one implementation is 1.0 vol%·s, then trigger condition two is determined to be met. If either of the above two conditions is met, the stratified air distribution repair execution process is entered.

[0059] Specifically, the method of predicting the credible effective oxygen quantity (OOF) for the next control step length by linear extrapolation involves using the control step length sequence as the independent variable and the corresponding credible effective oxygen quantity as the dependent variable. A straight line is fitted using the least squares method, and the slope of this line is used as an estimate of the rate of change of the current credible effective oxygen quantity. The credible effective oxygen quantities at three time points are denoted as the first, second, and third values ​​in chronological order. The slope of the fitted line at the three points is calculated using the least squares method, which is the weighted average of the differences in oxygen quantities at adjacent time points, with the weights evenly distributed by the time interval. The obtained slope is multiplied by the control step length to obtain the predicted increment, which is then added to the credible effective oxygen quantity at the current time to obtain the predicted value for the next control step length.

[0060] Read the temperature distribution of each burner area and the damper opening feedback of each layer. Mark the layers with low temperature and corresponding damper openings that have reached the lower limit of the adjustment margin as oxygen-deficient layers. Low temperature is defined as the temperature of the layer being 5% lower than the historical average. The lower limit of the damper opening adjustment margin is defined as the difference between the current damper opening and the minimum safe opening allowed by the damper. If stratified oxygen detection is configured on site, the layers with low oxygen content are directly used as the basis for determining the oxygen-deficient layers.

[0061] For hypoxic layers, increase the opening of the secondary or staged dampers at the corresponding layer, with each adjustment not exceeding 5% of the rated damper opening. After adjustment, wait for two control steps and observe the response of the corresponding layer temperature and reliable effective oxygen content. If the site has the capability for interlayer ammonia distribution, simultaneously decrease the ammonia ratio at the hypoxic layer, with the decrease not exceeding 10% of the current ammonia ratio, to reduce the oxygen consumption at that layer. Local air distribution repair actions are performed while keeping the total air volume setpoint unchanged, without triggering the total air volume adjustment command.

[0062] After the local air distribution repair action is executed, wait for 3 control steps, reread the predicted value of credible effective oxygen and the reduction debt. If the predicted value has risen above the lower boundary of oxygen and the reduction debt no longer increases, the local repair is deemed sufficient, the current air distribution status is maintained, and no total air volume correction is executed. If the predicted value is still below the lower boundary of oxygen or the reduction debt continues to rise, the local repair is deemed insufficient. When the local air distribution repair is insufficient, a correction increment is added to the current total air volume. The correction increment is determined by the difference between the credible effective oxygen and the lower boundary of oxygen and the current value of the reduction debt. Specifically, the base increment is obtained by multiplying the absolute value of the difference between the credible effective oxygen and the lower boundary of oxygen by a preset proportional coefficient. Then, the base increment is amplified by the proportion of the reduction debt exceeding the trigger threshold. This is the total air volume correction increment for this time. Each correction increment does not exceed 3% of the rated value of the total air volume. After correction, wait for 3 control steps to evaluate whether further correction is needed until the trigger condition is removed.

[0063] When the credible effective oxygen (OO) prediction value remains consistently above the lower limit of OO and the reduction debt is less than 50% of the trigger threshold, and this condition is maintained for 5 consecutive control steps, a pullback is initiated. The pullback is executed in the reverse order of the triggering sequence: first, the total air volume setpoint is gradually pulled back to the pre-correction level in increments not exceeding 2% of the rated total air volume per step. After each pullback step, 3 control steps are waited to confirm that the credible effective OO has not approached the lower limit again. After the total air volume pullback is completed, the distribution of secondary and staged air at each level is gradually restored to the pre-correction state. Finally, the inter-layer ammonia blending ratio is restored to the pre-correction level. During the pullback process, the credible effective OO prediction value and the reduction debt are continuously monitored at each control step. If any indicator meets the triggering condition again, the pullback is immediately stopped and the condition determination is re-entered.

[0064] By simultaneously setting two independent triggering conditions—the predicted value falling below the lower boundary and the debt recovery exceeding the trigger threshold—and stipulating that either condition triggers the subsequent process, the system can simultaneously cover two risk scenarios: rapid oxygen decline and slow accumulation of hypoxia. This prevents the underestimation of persistent low-intensity hypoxia when relying solely on instantaneous predicted values. Before implementing total air volume correction, local air distribution repair is prioritized, and the inter-layer distribution adjustment of secondary and tiered air is used as the first response measure. This allows hypoxic layers to receive targeted oxygen supplementation without changing the total air volume, avoiding problems such as overoxygenation in non-hypoxic layers, increased nitrogen oxide emissions, and increased smoke exhaust heat loss caused by a global increase in total air volume. This improves accuracy and economy.

[0065] like Figure 2A comparative diagram of the effects of a boiler oxygen control method is shown. Gray bars represent the baseline scheme of existing technology, and black bars represent the scheme of the present invention. Regarding the frequency of CO exceedances, the existing technology has 18.4 times / 100 hours, while the present invention has 4.2 times / 100 hours, a reduction of approximately 77.2%. This improvement is attributed to the introduction of a CO equivalent penalty term in the credible effective oxygen construction step, which allows the risk of high CO levels to be reflected in the control decision before the oxygen level falls below the lower boundary, thereby suppressing incomplete combustion caused by local oxygen deficiency. Regarding the escape and over-limit rate, the existing technology has a rate of 23.6%, while this invention has a rate of 5.8%, a reduction of approximately 75.4%. This improvement in the indicator stems from... The equivalent penalty term's structured correction of the credible effective oxygen content, and the simultaneous reduction of the ammonia blending ratio in high-risk layers during the stratified air distribution repair step, together constrain the escape risk under ammonia blending conditions. Regarding the local hypoxia false alarm rate, the existing technology has a rate of 31.2%, while this invention has a rate of 6.1%, a reduction of approximately 80.4%. This indicator is a core issue addressed by this invention. The time delay alignment and reliability assessment steps eliminate the time misalignment between the tail-end detection quantity and the real-time state inside the furnace. The debt restoration criterion further identifies scenarios where the instantaneous value of the credible effective oxygen content is normal, but local hypoxia has been accumulating. These two aspects work together to effectively identify local hypoxia. Regarding the frequency of total airflow over-adjustment, the existing technology has a rate of 27. The efficiency of air volume was 0.5 times / 100 hours, while that of this invention was 7.3 times / 100 hours, a reduction of approximately 73.5%. This improvement stems from the hierarchical air distribution priority control strategy. By prioritizing the adjustment of local air distribution rather than directly increasing the total air volume, unnecessary disturbances to the total air volume caused by global air increase are reduced. At the same time, the reverse orderly execution mechanism during the withdrawal further prevents frequent repetition of control actions. The comprehensive improvement of the above four indicators shows that this invention, through a complete causal control chain of multi-source data fusion, credibility weighting, restoration of accumulated debt, and hierarchical control priority switching, effectively solves the problem of delayed response and missed detection of local hypoxia caused by single tail oxygen feedback in scenarios where deep peak shaving and green ammonia co-firing coexist.

[0066] Example 2:

[0067] Based on Example 1, another boiler oxygen control method is provided to address the need to control nitrogen oxide emissions, carbon monoxide emissions, ammonia slip, and stable combustion in deep peak-shaving green ammonia co-fired coal-fired boilers while further reducing operating oxygen content;

[0068] Existing control methods mostly rely on fixed empirical oxygen curves or single-layer PID to correct the total air volume. In order to ensure operational safety, a large oxygen margin is usually reserved, which makes it difficult to simultaneously and stably balance low oxygen operation, nitrogen oxide control, carbon monoxide control, ammonia slip suppression and stable combustion requirements.

[0069] In this embodiment, a multi-objective collaborative optimization layer is added between the reliable effective oxygen content construction and the stratified air distribution priority control, to satisfy nitrogen oxide emissions, CO emissions, and Under the premise of four constraints including escape and flammability stability, the lowest oxygen setpoint that can be safely executed under the current operating conditions is found in real time to replace the fixed empirical oxygen curve, so as to achieve low oxygen operation and multiple emission and flammability stability constraints at the same time.

[0070] Read the nitrogen oxide concentration and CO concentration after latency alignment. Escape concentration and flame stability evaluation parameters are calculated by determining the distance between each indicator and its operating limit, and normalizing it to a margin coefficient in the range of 0 to 1. The larger the margin coefficient, the farther away from the limit and the more sufficient the current safety margin. Specifically, the nitrogen oxide margin coefficient is equal to the nitrogen oxide operating limit minus the current nitrogen oxide concentration, and then divided by the nitrogen oxide operating limit; the CO margin coefficient is equal to the CO operating limit minus the current CO concentration, and then divided by the CO operating limit. Escape margin coefficient equal to escape operation limit minus current After escaping concentration, divide by Escape operation limit; the stable combustion margin coefficient is equal to the preset stable combustion threshold minus the current flame stability evaluation value and then divided by the preset stable combustion threshold. When the flame stability evaluation value exceeds the stable combustion threshold, the stable combustion margin coefficient is zero; when the concentration of any index exceeds the corresponding operation limit, the corresponding margin coefficient is zero, not negative, and an instruction to increase the oxygen setpoint is immediately triggered without waiting for the comprehensive constraint tension determination.

[0071] The minimum value among the four margin coefficients is taken as the comprehensive constraint tension index. This index reflects the constraint direction closest to the limit in the current operating state. When the comprehensive constraint tension is higher than the preset easing threshold, it is set to 0.4 in one implementation, indicating that there is sufficient margin for all current constraints, allowing for a further reduction in the oxygen setpoint in the next control step. When the comprehensive constraint tension is lower than the preset tightening threshold, it is set to 0.15 in one implementation, indicating that at least one constraint is close to the limit, requiring an increase in the oxygen setpoint or triggering the stratified air distribution repair process. When the comprehensive constraint tension is between the two thresholds, the current oxygen setpoint is maintained unchanged.

[0072] When it is determined that a reduction in oxygen is permissible, the oxygen setpoint is lowered by a fixed step size of 0.05 vol%. The lowered oxygen setpoint must not be lower than the currently effective lower oxygen limit. When it is determined that an increase in oxygen is required, the oxygen setpoint is increased by a fixed step size of 0.1 vol%. The increased oxygen setpoint must not be higher than the currently effective upper oxygen limit. The upward adjustment step size is greater than the downward adjustment step size to ensure that the response speed is faster than the bottoming speed when the constraint tightens, and to prevent the adjustment lag from causing the constraint to exceed the limit. The above adjustment command takes effect before entering the stratified air distribution priority control step, and serves as the benchmark input for the oxygen lower limit comparison used to trigger the determination of this step.

[0073] The system introduces constraint direction identification and directional air distribution pre-adjustment. When the overall constraint tension is below the tightening threshold, it identifies the constraint direction with the lowest current tension and performs directional pre-adjustment before triggering stratified air distribution repair: if the most tense constraint is CO, the burnout air damper opening is preferentially increased; if the most tense constraint is... If escape occurs, the ammonia blending ratio of the corresponding layer should be reduced first; if the most critical constraint is nitrogen oxides, the ratio of staged air and secondary air should be adjusted first to enhance the staged combustion effect; if the most critical constraint is stable combustion, the primary air volume should be increased first to maintain the ignition stability of pulverized coal; the directional pre-adjustment range should not exceed 3% of the rated opening of the corresponding damper or 5% of the ammonia blending ratio each time. After pre-adjustment, wait for 2 control steps and then reassess the overall constraint intensity. If the intensity has risen back to above the tightening threshold, the pre-adjustment is terminated; otherwise, the stratified air distribution repair process is entered.

[0074] To prevent the failure to promptly identify constraint exceedances during continuous reduction of the oxygen setpoint due to lag in the response of various detection quantities, a bottom-out protection mechanism is set up as follows: Record the moment of the most recent reduction of the oxygen setpoint. Within six consecutive control steps from that moment, the oxygen setpoint is prohibited from being reduced again. After the observation window ends, the overall constraint tension is reassessed before deciding whether to continue reducing it. If the overall constraint tension shows a downward trend within the observation window, the observation is immediately terminated and an upward adjustment command is executed. Without waiting for the observation window to end, if the overall constraint tension shows a decreasing trend for two consecutive control steps within the observation window, a downward trend is determined, the observation is immediately terminated, and an upward adjustment command is executed.

[0075] Based on the complete control chain of Example 1, this embodiment uses a multi-objective collaborative optimization layer to find the minimum safe oxygen setpoint that satisfies the four constraints in real time. Compared with a fixed empirical oxygen curve, it can adaptively tighten the oxygen margin under different loads and ammonia blending ratios, avoiding the excessive nitrogen oxide levels and operational economic losses caused by the fixed curve reserving too much margin to ensure safety. The constraint direction identification and directional air distribution pre-adjustment mechanism ensures that the air volume adjustment action directly corresponds to the current most stringent constraint direction, reducing non-targeted global air volume disturbances. The bottom-out protection mechanism ensures that the oxygen reduction speed does not exceed the observable response speed of the system under the condition that there is a transmission lag in the detection quantity, preventing the problem of continuous constraint overruns caused by response lag.

[0076] The embodiments of the present invention described above are available for reference only. Without departing from the embodiments of the present invention and its broader aspects, those skilled in the art can make data modifications and method changes based on these descriptions in specific operations.

Claims

1. A method for controlling oxygen levels in a boiler, characterized in that, include: Collect operational data including combustion status, air distribution status, fuel supply status, ammonia blending status, and stable combustion status; The stable combustion state data acquisition process includes: the stable combustion state data includes flame image signals and flame pulsation acoustic signals; the mean brightness, standard deviation of brightness, and flame centroid offset are extracted from the flame image signals as image features; the root mean square value of pulsation is extracted from the acoustic signals as acoustic features; the image features and acoustic features are normalized and then weighted to synthesize a flame stability evaluation value; the flame stability evaluation value participates only in the construction of credible effective oxygen content in the form of a penalty amount and in the correction of the lower boundary of oxygen content in the form of an upper shift correction amount. For detection quantities with transmission lag, time delay alignment is performed, and a confidence level is generated based on their characterization deviation of the furnace oxygen potential reference value. The generation of the confidence level includes: independently generating the confidence level based on the characterization deviation of each detection quantity of the furnace oxygen potential reference value after time delay alignment. The confidence level coefficient is calculated using the following formula: ， Where i represents the index of the detection quantity. This represents the confidence coefficient corresponding to the i-th detection quantity. This represents the characterization bias of the i-th detection after time delay alignment. Represents the dimensionless normalization coefficient; The confidence coefficients of each detection quantity are calculated independently and are used only as the basis for the weighted fusion of reliable effective oxygen content. The reference value of oxygen potential in the furnace is estimated by a soft measurement model based on the current air volume, coal volume, ammonia volume and furnace temperature distribution; A reliable effective oxygen level is constructed based on the oxygen level detection value, the aforementioned confidence level, and a reduction risk penalty term; the reduction risk penalty term includes: CO and... after read delay alignment. Escape concentration value, compared with preset operating limit, CO and The escape penalty is set to zero when the corresponding concentration is below 80% of the operating limit, linearly interpolated when it is between 80% and 100% of the operating limit, and the upper limit is taken when it reaches or exceeds the operating limit. When the flame stability evaluation value does not exceed the preset flammability threshold, the flame instability penalty is set to zero. When the flame stability evaluation value exceeds the flammability threshold, it is linearly mapped to an equivalent oxygen penalty value based on the magnitude of the exceedance. The obtained CO, The total amount of the reduction risk penalty is obtained by adding the three items: flame instability penalty, and the total upper limit is truncated. Among them, the stable combustion state information is used in the calculation of the penalty item and the lower limit correction of oxygen in the form of flame stability evaluation quantity. Determine the upper and lower operating boundaries of oxygen content based on the current load and ammonia blending conditions, and simultaneously accumulate and reduce debt. The process of obtaining the upper and lower operating boundaries of oxygen content includes: establishing a two-dimensional calibration table with load and ammonia blending ratio as two dimensions; and calibrating the lower boundary of oxygen content with CO concentration not exceeding 80% of the operating limit. The judgment is based on the simultaneous fulfillment of four conditions: escape concentration not exceeding 80% of the operating limit, flame stability evaluation quantity not exceeding the stable combustion threshold, and nitrogen oxide concentration change within adjacent detection intervals not exceeding the preset sudden increase judgment threshold; the upper boundary of oxygen content is calibrated based on CO, The termination condition is when the improvement of escape and flame stability evaluation values ​​does not exceed their respective preset limits, and the increase in nitrogen oxide concentration exceeds the preset threshold or the increase in flue gas temperature exceeds the preset threshold. Layered air distribution priority control: When it is predicted that the reliable effective oxygen level will fall below the lower limit of oxygen level or the debt reduction exceeds the preset threshold, local air distribution repair is performed first. Total air volume correction is only performed when local air distribution repair is insufficient, and the reverse order is executed during the rollback.

2. The boiler oxygen control method according to claim 1, characterized in that: The time-delay alignment of the detected quantity with transmission lag includes: During the unit commissioning phase, the time delay parameters are calibrated by step test. Under stable load conditions, a step disturbance is applied to the total air volume, and the difference between the response start time of each detection quantity and the air volume step time is recorded synchronously. This difference is recorded as the estimated time delay value of the corresponding detection quantity at that load point. The load and time delay correspondence table is established for rated load, 75% load, 50% load and 40% load respectively. When the unit cannot be shut down for testing, the effective volume of the furnace is divided by the current flue gas volume flow rate as the initial value of the time delay.

3. The boiler oxygen control method according to claim 1, characterized in that: The process of obtaining the reference value of oxygen potential in the furnace includes: Using the current total air volume, primary air volume, secondary air volume, staged air volume, coal feed rate, and ammonia injection rate as inputs, the actual oxygen supply calculated from the current total air volume is divided by the theoretical oxygen demand for complete combustion estimated based on the coal feed rate and coal calorific value. The resulting ratio is the theoretical excess air coefficient. Combined with the estimated furnace air leakage rate, the theoretical oxygen content of the main combustion zone section is calculated. A correction factor is then added to the theoretical oxygen content. The correction factor is given by a linear regression model using furnace temperature distribution characteristics, load change rate, and ammonia blending ratio as inputs. The reference value of the oxygen potential in the furnace is equal to the theoretical oxygen content plus the correction factor.

4. The boiler oxygen control method according to claim 1, characterized in that: The debt recovery process includes: Within each control step, the difference between the lower boundary of oxygen content and the current reliable effective oxygen content is calculated. If the difference is greater than zero, the difference is multiplied by the control step time and accumulated to the reduction debt. If the difference is not greater than zero, the reduction debt is multiplied by the risk release coefficient to reduce it. When the reduction debt increases continuously and there are load acceleration / deceleration commands at the same time, it is determined that it is about to enter a local reduction accumulation state and trigger the air distribution repair step.

5. The boiler oxygen control method according to claim 1, characterized in that: The stratified air distribution priority control includes: Within each control step, a condition determination is performed. Based on the current credible effective oxygen content and its changing trend within the control step, the credible effective oxygen content for the next control step is predicted by linear extrapolation. If the predicted value is lower than the current effective oxygen content lower boundary, trigger condition one is determined to be met. If the current debt to be restored exceeds the preset debt to be restored trigger threshold, trigger condition two is determined to be met. If either condition is met, the stratified air distribution repair execution process is entered. Increase the opening of the secondary or staged air dampers in the anoxic layer and simultaneously decrease the ammonia blending ratio in that layer; after local repair, wait for multiple control steps and reassess, and only execute the total air volume correction if it is still insufficient; when withdrawing, first withdraw the total air volume, then restore the air distribution, and finally restore the inter-layer ammonia blending ratio. After each withdrawal step, wait for multiple control steps to confirm stability, and immediately stop the withdrawal if any trigger condition is met again.

6. The boiler oxygen control method according to claim 1, characterized in that: Add a multi-objective collaborative optimization layer between the construction of credible effective oxygen content and the priority control of stratified air distribution: Nitrogen oxide concentration, CO concentration, The distances between the escape concentration and flame stability evaluation values ​​and their corresponding operating limits are normalized to margin coefficients. When any index exceeds the operating limit, the corresponding margin coefficient is set to zero and an upward adjustment command is immediately triggered. The minimum value of the four margin coefficients is taken as the comprehensive constraint tension. When it is higher than the lenient threshold, the oxygen setpoint is lowered. When it is lower than the tightening threshold, it is raised or the constraint direction identification is triggered. Directional air distribution pre-adjustment is performed according to the most stringent constraint direction. After each downward adjustment, a bottoming-out protection window is set. When the overall constraint tension continuously decreases within the window, an upward adjustment is immediately executed.

Images