An evaluation method for the synergistic emission reduction effect of air pollutants and greenhouse gases

Abstract

The present invention discloses a method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases, which relates to the technical field of air pollution emission reduction and includes the following steps: At the initial stage of the entire process, according to the historical average empirical value, a starting reference flow rate is set for the flue gas and introduced into the flue; when the flue gas enters the flue at the starting reference flow rate, multi-type sensors are deployed at key positions in the flue to monitor the flow state of the flue gas in real time and perform high-frequency sampling; the collected flue gas flow state data is comprehensively analyzed by using a pre-trained deep learning model to identify potential turbulent risks in the flue gas. The present invention intelligently identifies and controls turbulent risks through a deep learning model, dynamically optimizes the flue gas flow rate, effectively improves the SCR reaction efficiency and flow field uniformity, solves the problems of NOx exceeding the standard and ammonia escape, and realizes the synergistic emission reduction of pollutants and greenhouse gases and the multi-objective optimization of system operation.

Description

Technical Field

[0001] The present invention relates to the technical field of air pollution reduction, and particularly to a method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases. Background Art

[0002] The evaluation of the synergistic emission reduction effect of air pollutants and greenhouse gases refers to the process of systematically analyzing and quantifying the combined emission reduction effect of such measures on traditional air pollutants (such as sulfur dioxide, nitrogen oxides, particulate matter, etc.) and greenhouse gases (such as carbon dioxide, methane, etc.) when formulating and implementing environmental protection or emission reduction measures. By evaluating the comprehensive impacts of these synergistic emission reduction measures in aspects such as economy, environment, and society, the best strategies for simultaneously reducing the emissions of conventional pollutants and greenhouse gases can be identified, helping decision-makers achieve an overall balance between environmental quality improvement and climate change response.

[0003] In industrial parks, to reduce nitrogen oxide (NOx) emissions, some enterprises have installed selective catalytic reduction (SCR) systems. This system injects ammonia into the flue gas at a suitable temperature of 250°C - 400°C, and under the action of a catalyst, reduces NOx to harmless nitrogen and water, achieving pollutant reduction.

[0004] The existing technologies have the following deficiencies:

[0005] Existing technologies usually introduce flue gas into the flue at a constant flow rate. However, when there is local turbulence in the flue gas flow field, continuing to maintain a constant flow rate will cause uneven flow velocity distribution of the flue gas in the SCR reactor, resulting in significant differences in NOx concentration in different regions. At this time, the ammonia injection system still adds ammonia according to the "average NOx concentration", resulting in insufficient ammonia injection in the high-concentration region, where NOx cannot fully react and the direct emissions exceed the standard. At the same time, the turbulent disturbance also affects the uniform mixing of ammonia and NOx, and the local reaction conditions are limited, causing some ammonia to be discharged without participating in the reaction, resulting in ammonia escape and secondary pollution.

[0006] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide a method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases. Through the identification and quantitative analysis of potential turbulence risks by a deep learning model, it can accurately predict the trend of turbulence formation in advance and dynamically and intelligently adjust the flue gas inlet flow rate to adapt to complex and changeable working conditions. This data-driven active optimization strategy not only reduces the phenomenon of local NOx emission exceeding the standard and ammonia slip caused by turbulence, effectively improves the flow field uniformity and reaction efficiency in the SCR reactor, but also reduces the secondary pollution and catalyst loss caused by ammonia slip, achieving multi-objective collaborative optimization in terms of economy, environmental benefits and equipment stability, and comprehensively reflecting the comprehensive effect of the synergistic emission reduction of air pollutants and greenhouse gases, so as to solve the problems in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: A method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases, comprising the following steps:

[0009] At the initial stage of the whole process, according to the historical average empirical value, a starting reference flow rate is set for the flue gas and introduced into the flue.

[0010] When the flue gas enters the flue at the starting reference flow rate, various types of sensors are deployed at key positions in the flue to monitor the flow state of the flue gas in real time and perform high-frequency sampling.

[0011] The collected flue gas flow state data is comprehensively analyzed using a pre-trained deep learning model to identify potential turbulence risks in the flue gas.

[0012] After identifying the turbulence risk, further quantitative analysis of the severity of the turbulence is carried out, and the preset flow rate of the flue gas is intelligently regulated according to the turbulence risk level. By dynamically adjusting the flue gas flow rate, the flow field uniformity in the flue is improved and the flue gas turbulence risk is eliminated.

[0013] Preferably, the historical average empirical value refers to the average value of the flue gas flow rate statistically obtained based on actual working condition data during the long-term operation of the SCR system.

[0014] Preferably, the specific steps of comprehensively analyzing the collected flue gas flow state data using a pre-trained deep learning model to identify potential turbulence risks in the flue gas are as follows:

[0015] Preprocess the real-time obtained flue gas flow state data, and extract the features from the preprocessed data that reflect the potential turbulent flow hazards during the flue gas flow process. Among them, the extracted features include the irregularity degree of the flue gas flow trajectory and the ratio of the fluctuation amplitude of the flue gas flow velocity to the average velocity per unit time. During the monitoring period, after deeply analyzing the extracted features, a reference value of flue gas chaos degree and a reference value of flow velocity pulsation intensity are respectively generated, and the irregular and disordered degree of the flue gas flow trajectory in the flue and the fluctuation strength of the flue gas flow velocity are quantified through the reference value of flue gas chaos degree and the reference value of flow velocity pulsation intensity.

[0016] Preferably, input the reference value of flue gas chaos degree and the reference value of flow velocity pulsation intensity as feature vectors into the trained deep learning model, and output a hazard degree coefficient through the deep learning model, and conduct an intelligent prediction on the potential turbulent flow risk during the flue gas flow process based on the hazard degree coefficient.

[0017] Preferably, compare and analyze the hazard degree coefficient generated when predicting the potential turbulent flow risk in the flue gas by using the deep learning model with a preset hazard degree reference threshold to identify the turbulent flow risk. The specific steps are as follows:

[0018] If the hazard degree coefficient is greater than the hazard degree reference threshold, a risk signal is generated, indicating that there is a potential turbulent flow risk in the flue gas; if the hazard degree coefficient is less than or equal to the hazard degree reference threshold, a normal signal is generated, indicating that the flue gas is flowing normally.

[0019] Preferably, after identifying the turbulent flow risk, conduct a quantitative analysis on the severity of the turbulent flow, and conduct an intelligent regulation on the preset flow velocity of the flue gas according to the turbulent flow risk level. The specific steps are as follows:

[0020] After identifying that there is a turbulent flow risk in the flue gas, compare the currently calculated hazard degree coefficient with the preset hazard degree reference threshold, and construct a flow velocity regulation factor according to the deviation degree to quantify the turbulent flow risk level. The calculation expression of the flow velocity regulation factor is: , where: is the flow velocity regulation factor, is the hazard degree coefficient output by the current deep learning model, is the hazard degree reference threshold, is a very small positive number, is the response sensitivity coefficient, used to adjust the sensitivity to anomalies, is the hyperbolic tangent function;

[0021] After obtaining the flow velocity regulation factor , based on the current starting reference flow velocity, calculate the target flow velocity after intelligent regulation to achieve precise intervention in the flue gas flow field. The regulation expression of the target flow velocity is: , where: is the starting reference flow rate, is the maximum regulation amplitude coefficient, which is used to limit the flow rate change range and prevent excessive regulation, is the target flow rate after intelligent correction.

[0022] Preferably, within the monitoring period, the specific steps for generating the reference value of flue gas chaos degree after deeply analyzing the irregularity degree of the flue gas flow trajectory are as follows:

[0023] Extract the velocity direction, amplitude and gradient change information with spatial distribution representativeness from the preprocessed flue gas flow rate vector sequence to construct a multi-dimensional state vector group , where:

[0024] represents the instantaneous flue gas flow rate amplitude at the i th spatial sampling point;

[0025] represents the local velocity gradient at the i th spatial sampling point, which is used to describe the local flow rate change intensity;

[0026] represents the angle between the velocity vector direction and the reference axis at the i th spatial sampling point, reflecting the deviation degree of the flow direction;

[0027] In the state lattice, analyze the curvature change trend of the continuous trajectory, and construct the reference value of flue gas chaos degree through the total curvature change of the continuous trajectory, which reflects the geometric distortion degree of the trajectory in the high-dimensional phase space, that is, the nonlinearity, complexity and irregularity of the flue gas flow. The construction expression of the reference value of flue gas chaos degree is:

[0028] , where: is the reference value of flue gas chaos degree, N represents the number of trajectory points of the spatial sampling point, represents the i th trajectory curvature at the spatial sampling point, and the calculation formula is: , where: and are the first-order and second-order guiding vectors of the state vector group respectively, represents the curve arc length position of the i th spatial sampling point, represents the change rate of curvature along the trajectory, reflecting the local trajectory distortion degree.

[0029] Preferably, within the monitoring period, the specific steps for generating the reference value of flow rate pulsation intensity after deeply analyzing the ratio of the flue gas flow rate fluctuation amplitude to the average flow rate per unit time are as follows:

[0030] During the monitoring period, a continuous sequence of flow velocity data points is collected. Based on the sequence of flow velocity data points, the flow velocity variation frequency response is calculated. The calculation expression is as follows: where: is the flow velocity variation frequency response, represents the instantaneous flow velocity value at the k -th sampling moment, n is the number of sampling points within the monitoring period. The numerator part is the approximate discrete second derivative of the flow velocity, representing the acceleration trend change rate of the flow velocity. The denominator part is a normalization term, introducing the sum of adjacent velocities to control the influence of the numerical scale;

[0031] After obtaining the flow velocity variation frequency response, further combining with the fluctuation energy density of the flow velocity data, a reference value of the flow velocity pulsation intensity for enhancing the sensing ability is constructed. The construction expression is as follows: where: represents the reference value of the flow velocity pulsation intensity, represents the instantaneous jump of the flow velocity within the unit sampling interval, representing the first-order gradient of the flow velocity, emphasizes the sensitivity to severe jumps in the fluctuation energy.

[0032] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0033] Through the identification and quantitative analysis of potential turbulence risks by the deep learning model, the present invention can accurately predict the formation trend of turbulence in advance and dynamically and intelligently adjust the flue gas inlet flow velocity to adapt to complex and changeable working conditions. This data-driven active optimization strategy not only reduces the phenomenon of local NOx emission exceeding the standard and ammonia slip caused by turbulence, effectively improves the flow field uniformity and reaction efficiency in the SCR reactor, but also reduces the secondary pollution and catalyst loss caused by ammonia slip, achieving multi-objective collaborative optimization in terms of economy, environmental benefits and equipment stability, and comprehensively reflecting the comprehensive effect of the collaborative emission reduction of air pollutants and greenhouse gases. Description of the Drawings

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0035] Figure 1 It is a method flow chart of a method for evaluating the collaborative emission reduction effect of air pollutants and greenhouse gases according to the present invention. Detailed Implementation Modes

[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art.

[0037] The present invention provides a method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases as Figure 1 shown, comprising the following steps:

[0038] At the starting stage of the whole process, a starting reference flow rate is set for the flue gas according to the historical average experience value and introduced into the flue.

[0039] The historical average experience value refers to the average value of the flue gas flow rate statistically obtained based on the actual operating condition data during the long-term operation of the SCR system. This value is stable and adaptable under most operating conditions and can effectively ensure the operation effect of the system. When setting the starting reference flow rate of the flue gas, the historical operating data of the equipment under different conditions such as load, temperature, and pressure are usually referred to, and the flow rate range with the best performance or the most commonly used is selected as the experience value. For example, if a boiler system has been operating most stably at a flue gas flow rate of 12 m / s at 75% load in the past year, with a high NOx conversion rate and a low ammonia slip rate, then this 12 m / s flow rate can be used as the "historical average experience value" for future startup or default operation. This value has a practical basis and is also convenient to use as a reference standard when real-time control is lacking.

[0040] When the flue gas enters the flue at the starting reference flow rate, various types of sensors are deployed at key positions in the flue to monitor the flue gas flow state in real time and perform high-frequency sampling.

[0041] The function of this step is to provide comprehensive and dynamic flue gas condition information: on the one hand, it can master the distribution of flue gas flow rate, pressure, and temperature in different times and spaces, and on the other hand, it can understand the mixing and conversion of NOx and ammonia in the SCR reaction area.

[0042] In the SCR flue gas treatment system, the "critical positions" usually refer to the areas in the flue that can best represent the overall flow field characteristics or are prone to abnormal flow fields. For example, at the inlet section, the initial flow velocity and temperature distribution of the flue gas when it first enters the flue can be captured to timely understand the emission stability of the boiler or other equipment; in areas with obvious geometric changes such as bends or expansion / contraction sections, turbulence, eddy currents, or local pressure surges are likely to occur, so key monitoring is required; sensors should also be deployed at key positions before and after the catalyst bed to evaluate the gas composition and temperature difference in the actual reaction zone; in addition, the outlet section is an important observation point for comprehensively evaluating the overall reaction efficiency, NOx removal rate, and ammonia slip. By arranging monitoring points in these areas, it is convenient to detect possible problems such as abnormal flow fields, catalyst failure, or local overheating in the first time, thus providing accurate data support for subsequent intelligent regulation.

[0043] To accurately grasp the flue gas flow state, multiple types of sensors need to be deployed for "multi-dimensional" monitoring. Flow velocity sensors or pitot tubes are used to capture the flue gas velocity distribution and its fluctuations; pressure or differential pressure sensors help to judge local turbulence, blockage, or air leakage in the flue; temperature sensors can evaluate whether the thermal conditions of combustion and catalytic reactions are in the ideal range; NOx concentration sensors are the key to evaluating the denitrification efficiency and NOx distribution; ammonia sensors or gas analyzers can monitor ammonia slip and assist in optimizing the ammonia injection amount; if necessary, gas analyzers for sulfur dioxide, carbon monoxide, oxygen, etc. can also be deployed for comprehensive diagnosis of flue gas composition. These sensors work together, sampling at high frequency and transmitting the data to the control system in real time.

[0044] Use a pre-trained deep learning model (such as those based on network structures like CNN, RNN, Transformer, etc.) to comprehensively analyze the collected flue gas flow state data, identify potential turbulence risks in the flue gas, and evaluate the potential impact of turbulence on ammonia slip;

[0045] Use a pre-trained deep learning model to comprehensively analyze the collected flue gas flow state data, identify potential turbulence risks in the flue gas. The specific steps are as follows:

[0046] Preprocess the real-time obtained flue gas flow state data, and extract the features from the preprocessed data that reflect potential turbulence hazards during the flue gas flow process. Among them, the extracted features include the irregularity degree of the flue gas flow trajectory and the ratio of the fluctuation amplitude of the flue gas flow velocity to the average flow velocity per unit time. During the monitoring period, after deeply analyzing the extracted features, a reference value for flue gas chaos and a reference value for flow velocity pulsation intensity are respectively generated, and the irregular and disordered degree of the flue gas flow trajectory in the flue and the fluctuation strength of the flue gas flow velocity are quantified through the reference value for flue gas chaos and the reference value for flow velocity pulsation intensity;

[0047] Taking the reference value of flue gas chaos degree and the reference value of flow velocity pulsation intensity as feature vectors, input them into the trained deep learning model, output the hidden danger degree coefficient through the deep learning model, and make an intelligent prediction of the potential turbulent risk in the flue gas flow process based on the hidden danger degree coefficient.

[0048] The preprocessing of the real-time acquired flue gas flow state data usually includes steps such as data denoising, outlier detection and removal, missing value filling, data smoothing processing, and data format conversion. Among them, data denoising is used to eliminate the interference signals in the sensor acquisition process to ensure the signal quality in subsequent analysis; outlier detection and removal can eliminate the extreme outliers caused by instrument errors or short-term failures to prevent misjudgment of the model; missing value filling can repair the missing parts of the sampling through interpolation or statistical methods to ensure data integrity; data smoothing processing helps to eliminate local sharp fluctuations or jitters and improve the smoothness and reliability of the data; data format conversion ensures that the preprocessed data can seamlessly connect to the subsequent deep learning model training and analysis processes in a suitable format. This series of preprocessing steps provides high-quality and highly reliable data support for subsequent turbulent intelligent analysis, risk assessment, and flow velocity regulation.

[0049] When the flue gas is introduced into the flue, if its flow trajectory shows obvious irregularities, that is, it shows characteristics such as drastic changes, frequent disturbances, variable directions, or non-linear deviations, it usually indicates that there is a high potential turbulent risk in the flue gas. Because in the stable laminar flow state, the flue gas flow trajectory should be relatively smooth, with a consistent direction and a gradual change in velocity; once phenomena such as trajectory distortion, vortex, shedding, and backflow occur, it means that the velocity gradient, pressure distribution, or thermal disturbance in the local area has reached a certain threshold, greatly increasing the possibility of inducing turbulence. The irregularity of the trajectory reflects the non-uniformity of energy transfer and the complexity of fluid disturbance in the flow field, and is an important precursor feature before the formation of turbulence. Therefore, monitoring and analyzing the irregular degree of the flue gas flow trajectory is one of the key means to identify turbulent hidden dangers, give early warnings, and implement intelligent regulation.

[0050] The specific steps to generate the reference value of flue gas chaos degree after deeply analyzing the irregular degree of the flue gas flow trajectory within the monitoring period are as follows:

[0051] Extract the velocity direction, amplitude, and gradient change information with spatial distribution representativeness from the preprocessed flue gas flow velocity vector sequence to construct a multi-dimensional state vector group , where:

[0052] represents the instantaneous flue gas flow velocity amplitude at the i th spatial sampling point;

[0053] represents at the iThe local velocity gradient of the spatial sampling points is used to characterize the intensity of the local flow velocity change;

[0054] It represents the angle between the velocity vector direction at the i th spatial sampling point and the reference axis, reflecting the degree of deviation of the flow direction;

[0055] All state vector groups are mapped into a high-dimensional phase space to form a dot matrix trajectory group. The distribution pattern of this trajectory group will provide a geometric basis for the next chaos degree calculation. The function of this step is to establish the geometric mapping relationship of the flow trajectory in the state space and provide the original structure for analyzing its geometric irregularity.

[0056] In the state dot matrix, analyze the curvature change trend of the continuous trajectory, and construct a reference value of the flue gas chaos degree through the total sum of the curvature changes of the continuous trajectory, reflecting the geometric distortion degree of the trajectory in the high-dimensional phase space, that is, the nonlinearity, complexity and irregularity of the flue gas flow. The construction expression of the flue gas chaos degree reference value is: , where: is the reference value of the flue gas chaos degree, N represents the number of trajectory points of the spatial sampling point, represents the i th trajectory curvature at the spatial sampling point, and the calculation formula is: , where: and are the first-order and second-order guiding vectors of the state vector group respectively, represents the curve arc length position of the i th spatial sampling point, represents the change rate of the curvature along the trajectory, reflecting the local trajectory distortion degree.

[0057] From the reference value of flue gas chaos, it can be seen that during the monitoring period, the larger the value of the reference value of flue gas chaos generated after in-depth analysis of the irregularity degree of the flue gas flow trajectory, the greater the risk of potential turbulence in the flue gas. Specifically, the reference value of flue gas chaos is an index generated after in-depth analysis of the irregular and non-linear disorder degree of the flue gas flow trajectory, which essentially reflects the distortion and complexity of the flue gas trajectory in the high-dimensional state space. When the value of the reference value of flue gas chaos is large, it means that the movement trajectory of the flue gas is highly irregular, the state is changeable, there is strong disorder, the trajectory change is difficult to predict and control, showing obvious turbulence characteristics, indicating that there is a high risk of potential turbulence in the flue gas; on the contrary, if the value of the reference value of flue gas chaos is small, it means that the movement trajectory of the flue gas is relatively regular, continuous and stable, the operating state of the flue gas in the flow field is relatively stable, the risk of potential turbulence in the flue duct is small, and the flow state is within the normal or stable range. Therefore, through the level of the reference value of flue gas chaos, the potential turbulence risk degree of the flue gas under the actual operating conditions can be intuitively and accurately quantified and evaluated.

[0058] When the flue gas is introduced into the flue duct, if the ratio of the fluctuation amplitude of the flue gas flow velocity per unit time to its average flow velocity increases significantly, this usually indicates that the flue gas flow state tends to be unstable and there is a potential risk of turbulence. This is because in the laminar flow or stable flow state, the change of the flue gas flow velocity is relatively stable and the fluctuation amplitude is small; while when the flow velocity fluctuates violently, it means that the local gas is affected by factors such as disturbance, shear force or temperature gradient, resulting in the appearance of velocity shear layers, vortex structures or recirculation zones in the flow field, which are exactly the precursors of turbulence occurrence.

[0059] During the monitoring period, the specific steps for generating the reference value of velocity pulsation intensity after in-depth analysis of the ratio of the fluctuation amplitude of the flue gas flow velocity per unit time to the average flow velocity are as follows:

[0060] During the monitoring period, collect a continuous sequence of flow velocity data points , calculate the flow velocity variation frequency response according to the sequence of flow velocity data points, and the calculation expression is: , where: is the flow velocity variation frequency response, represents the instantaneous flow velocity value at the k th sampling moment, n is the number of sampling points within the monitoring period, and the numerator part is the approximate discrete second derivative of the flow velocity, representing the acceleration trend change rate of the flow velocity, and the denominator part is the normalization term, introducing the sum of adjacent velocities to control the influence of the numerical scale;

[0061] This step captures the local non-linear disturbance trend in the flow rate sequence. The more intense the second-order change (such as pulses, mutations), the faster and stronger the fluctuating response of the flue gas flow rate, reflecting potential turbulent inducing factors.

[0062] After obtaining the flow rate variation frequency response, further combine with the fluctuation energy density of the flow rate data to construct a reference value of the flow rate pulsation intensity that enhances the perception ability. The constructed expression is: , where: represents the reference value of the flow rate pulsation intensity, represents the instantaneous jump of the flow rate within a unit sampling interval, representing the first-order gradient of the flow rate, emphasizes the sensitivity to severe jumps in the fluctuation energy (more emphasizing abnormal disturbances than squaring), is used to maintain the stability of the calculation result, prevent the energy density from being too large or over-responding to small jumps. The logarithmic transformation is used to compress the numerical scale, enhance the discrimination in the high-disturbance region, and have fault tolerance for low disturbances;

[0063] This step combines the two dimensions of frequent jumps and strong disturbances to construct a reference value of the flow rate pulsation intensity with "pulsation sensitivity". The higher it is, the more unstable the local air flow is, and it is very likely to develop into turbulence. It is the core decision-making index for quantifying the flue gas pulsation intensity.

[0064] From the reference value of the flow rate pulsation intensity, it can be seen that within the monitoring period, the larger the performance value of the reference value of the flow rate pulsation intensity generated by deeply analyzing the ratio of the fluctuation amplitude of the flue gas flow rate to the average flow rate per unit time, the greater the risk of potential turbulence in the flue gas. The reference value of the flow rate pulsation intensity essentially reflects the fluctuation intensity and severity of the flue gas flow rate within the monitoring period. When the value of the reference value of the flow rate pulsation intensity is large, it means that there are significant and frequent fluctuating changes in the flue gas flow rate during the monitoring period, that is, the flow rate fluctuates violently in a short time or even shows large jumps or irregular movement trends. This phenomenon indicates that the flue gas flow state is unstable, and the airflow structure may show characteristics such as vortices, recirculation or shear disturbances, thereby increasing the potential risk of turbulence occurrence. On the contrary, when the performance value of the reference value of the flow rate pulsation intensity is small, it means that the flue gas flow rate changes relatively smoothly and regularly, and the flue gas shows a relatively stable flow state in the flue duct, and the risk of potential turbulence is significantly reduced. Therefore, the level of the reference value of the flow rate pulsation intensity can more accurately reflect the stability of the flue gas flow field and the strength of the turbulence risk, providing an important reference basis for further flue gas regulation and turbulence risk management.

[0065] A trained deep learning model refers to an artificial intelligence model that has been trained with a large amount of sample data, with fixed parameters, and can give stable prediction results under given input conditions. Specifically, this model is usually based on neural network structures, such as frameworks like multi-layer perceptron (MLP), convolutional neural network (CNN), or attention mechanism (Transformer). Through learning from a large amount of historical operating condition data in the early stage, it has mastered the characteristic laws and complex mapping relationships closely related to the occurrence of turbulence during the flue gas flow process. In actual operation, historical or simulated flue gas flow data will first be labeled or scored. For example, significant turbulence phenomena that have occurred are defined as high-risk cases, and relatively stable flow processes are defined as low-risk cases. Then, through an iterative training process, the model will continuously adjust its internal weights so that when it sees new inputs (such as feature vectors like "flue gas chaos reference value" and "flow velocity pulsation intensity reference value"), it can also accurately output a quantifiable "hazard degree coefficient". Such a formed deep learning model often has strong generalization ability. Even when facing operating condition changes that have not occurred before, it can give relatively reliable risk prediction results by capturing the correlations between features. It is worth emphasizing that "trained" does not mean it stops evolving. If new measured data or extreme operating conditions are continuously added, online learning or incremental training methods can be used to enable the model to continuously improve itself during operation, so as to still maintain a high level of accuracy and robustness when facing dynamic situations such as multi-source flue gas merging, non-constant load, or seasonal climate impacts.

[0066] In specific use, when taking the "flue gas chaos reference value" and the "flow velocity pulsation intensity reference value" together as input features, the deep learning model will comprehensively consider the different aspects of information contained in these two indicators: the former mainly measures the chaos degree of the flue gas flow trajectory and energy distribution, reflecting the possibility of the existence of irregular structures such as turbulence, vortices, or shear layers inside the flue gas; the latter focuses more on the sharp fluctuations and acceleration changes of the flow velocity within a limited time or space interval, quantifying the fluctuation intensity and whether the pulse jumps are frequent. After these input features undergo multiple non-linear transformations in the neural network and interact with a large amount of implicit knowledge and weights stored in the model, a continuous "hazard degree coefficient" is finally output. The higher the value of this hazard degree coefficient, the greater the probability of turbulence occurring under the current operating conditions, and higher-level early warning or linkage control strategies need to be triggered; if the hazard degree coefficient is low, it indicates that the overall flue gas flow remains stable, and there is no need to perform large-scale flow velocity regulation or emergency operating condition switching for the time being. Such a prediction mechanism can not only enable the production management system to perceive the turbulence risk in advance, but also be combined with automated scheduling algorithms to timely adjust the flue gas flow rate, ammonia injection rate, and temperature control, etc., reducing the occurrence of problems such as NOx over-standard and ammonia escape, enabling the SCR system to achieve stable operation with high efficiency and low emissions within a wider range of operating conditions.

[0067] The deep learning model is not specifically limited herein, and any deep learning model that can achieve comprehensive analysis of the reference value of flue gas chaos and the reference value of flow velocity pulsation intensity to generate a hidden danger degree coefficient is acceptable. To implement the technical solution of the present invention, the present invention provides a specific implementation manner; the expression for generating the hidden danger degree coefficient is: , where in the formula, are respectively the reference value of flue gas chaos and the reference value of flow velocity pulsation intensity preset proportionality coefficients, and are all greater than 0. The preset proportionality coefficient refers to the weight factor assigned to each index (such as the chaos index and the reference value of flow velocity pulsation intensity ) when comprehensively calculating the hidden danger degree coefficient, that is, and . Their numerical values are usually determined by experience, experimental data or model tuning results, and are used to balance the proportion of different indexes in the final turbulence assessment. If an index has a relatively greater impact on the turbulence risk, a higher proportionality coefficient can be assigned; otherwise, it is lower. The purpose of doing this is to enable the final "hidden danger degree coefficient" to more accurately reflect the contribution degree of each index to the overall turbulence risk, so as to achieve a better prediction and evaluation effect in comprehensive judgment.

[0068] It can be seen from the hidden danger degree coefficient that during the monitoring period, the larger the value of the reference value of flue gas chaos generated after in-depth analysis of the irregular degree of the flue gas flow trajectory, and the larger the value of the reference value of flow velocity pulsation intensity generated after in-depth analysis of the ratio of the flow velocity fluctuation amplitude to the average flow velocity per unit time, that is, the larger the value of the hidden danger degree coefficient generated when using the pre-trained deep learning model to predict the potential turbulence risk in the flue gas, it indicates that the risk of potential turbulence in the flue gas is greater; on the contrary, it indicates that the risk of potential turbulence in the flue gas is smaller.

[0069] Compare and analyze the hidden danger degree coefficient generated when using the deep learning model to predict the potential turbulence risk in the flue gas with the pre-set hidden danger degree reference threshold to identify the turbulence risk. The specific steps are as follows:

[0070] If the hidden danger degree coefficient is greater than the hidden danger degree reference threshold, a risk signal is generated, indicating that there is a potential turbulence risk in the flue gas; if the hidden danger degree coefficient is less than or equal to the hidden danger degree reference threshold, a normal signal is generated, indicating that the flue gas is flowing normally.

[0071] After identifying the risk of turbulence, further quantitatively analyze the severity of the turbulence, and intelligently adjust the preset flow rate of the flue gas according to the turbulence risk level. By dynamically adjusting the flue gas flow rate, improve the flow field uniformity in the flue and eliminate the risk of flue gas turbulence;

[0072] Specifically, the rate of flue gas entering the flue and the flow field distribution can be dynamically changed by adjusting the speed of the induced draft fan inverter, opening and closing the pipeline valve or the angle of the guide vane, etc.

[0073] After identifying the risk of turbulence, quantitatively analyze the severity of the turbulence, and intelligently adjust the preset flow rate of the flue gas according to the turbulence risk level. The specific steps are as follows:

[0074] After identifying the risk of turbulence in the flue gas, compare the currently calculated hidden danger degree coefficient with the preset hidden danger degree reference threshold, and construct a flow rate regulation factor according to the deviation degree to quantitatively analyze the turbulence risk level. The calculation expression of the flow rate regulation factor is: , where: is the flow rate regulation factor, is the hidden danger degree coefficient output by the current deep learning model, is the hidden danger degree reference threshold, is a very small positive number to prevent the denominator from being zero and enhance numerical stability, usually taking 0.0001, is the response sensitivity coefficient, used to adjust the sensitivity to anomalies, is the hyperbolic tangent function, used to smoothly map the result to (-1, 1) to prevent extreme jumps;

[0075] This step constructs a continuous, non-linear and anomaly-sensitive flow rate regulation factor. When the hidden danger degree coefficient is slightly higher than the hidden danger degree reference threshold, a mild response is output; when the hidden danger degree coefficient is much higher than the hidden danger degree reference threshold, a rapidly amplified output is generated, driving the system into a strong regulation state, ensuring that the regulation behavior has the progressive characteristics of early warning, inhibition and intervention.

[0076] After obtaining the flow rate regulation factor , based on the current starting reference flow rate, calculate the target flow rate after intelligent regulation to achieve precise intervention in the flue flow field. The regulation expression of the target flow rate is: , where: is the starting reference flow rate, which comes from the optimal value under the historical stable working condition, is the maximum regulation amplitude coefficient , used to limit the flow rate change range to prevent over-regulation, is the target flow rate after intelligent correction;

[0077] This step realizes the dynamic non-linear regulation of the flue gas flow rate. When the turbulence risk is relatively high the target flow rate will be lower than the starting reference flow rate, thus prolonging the residence time in the reactor and alleviating the turbulence trend. When the system tends to be stable the flue gas flow rate returns to the reference state to maintain the SCR efficiency and energy balance. If it is necessary to cope with extremely weak turbulence in the future this structure also allows for forward speed regulation, with good scalability and adaptability.

[0078] Through the above solution, real-time intelligent monitoring and precise regulation of the flue gas flow state can be achieved, effectively overcoming the limitations and negative impacts brought by traditional constant flow rate and average ammonia injection strategies, and significantly improving the refined control level and response timeliness of NOx removal during the flue gas treatment process in the SCR system. Specifically, through the identification and quantitative analysis of potential turbulence risks by the deep learning model, the formation trend of turbulence can be accurately predicted in advance, and the flue gas inlet flow rate can be dynamically and intelligently adjusted to adapt to complex and changeable working conditions. This data-driven active optimization strategy not only reduces the phenomenon of local NOx emission exceeding the standard and ammonia slip caused by turbulence, effectively improves the flow field uniformity and reaction efficiency in the SCR reactor, but also reduces the secondary pollution and catalyst loss caused by ammonia slip, thus achieving multi-objective collaborative optimization in terms of economy, environmental benefits and equipment stability, and comprehensively reflecting the comprehensive effect of the collaborative emission reduction of air pollutants and greenhouse gases.

[0079] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0080] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

[0081] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0082] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0083] In several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0084] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0085] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0086] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0087] As described above, it is only the specific implementation manner of the present application. However, the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claimed rights.

Claims

1. A method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases, characterized in that, It includes the following steps: At the initial stage of the whole process, set a starting reference flow rate for the flue gas according to the historical average experience value and introduce it into the flue. When the flue gas enters the flue at the starting reference flow rate, deploy multiple types of sensors at key positions in the flue to monitor the flow state of the flue gas in real time and perform high-frequency sampling. Use a pre-trained deep learning model to comprehensively analyze the collected flue gas flow state data and identify potential turbulence risks in the flue gas. When the turbulence risk is identified, further quantitatively analyze the severity of the turbulence, and intelligently adjust the preset flow rate of the flue gas according to the turbulence risk level. By dynamically adjusting the flue gas flow rate, improve the flow field uniformity in the flue and eliminate the flue gas turbulence risk. Use a pre-trained deep learning model to comprehensively analyze the collected flue gas flow state data and identify potential turbulence risks in the flue gas. The specific steps are as follows: Preprocess the real-time obtained flue gas flow state data, and extract features from the preprocessed data that reflect potential turbulence hazards during the flue gas flow process. Among them, the extracted features include the irregularity degree of the flue gas flow trajectory and the ratio of the fluctuation amplitude of the flue gas flow rate to the average flow rate per unit time. During the monitoring period, after deeply analyzing the extracted features, generate a flue gas chaos degree reference value and a flow rate pulsation intensity reference value respectively, and quantify the irregular disorder degree of the flue gas flow trajectory in the flue and the fluctuation strength of the flue gas flow rate through the flue gas chaos degree reference value and the flow rate pulsation intensity reference value. The specific steps for generating the flue gas chaos degree reference value after deeply analyzing the irregularity degree of the flue gas flow trajectory during the monitoring period are as follows: Extract the velocity direction, amplitude, and gradient change information with spatial distribution representativeness from the preprocessed flue gas flow velocity vector sequence, and construct a multi-dimensional state vector group , where: Indicates the instantaneous smoke gas flow velocity amplitude at the i th spatial sampling point; Indicates the local velocity gradient at the i th spatial sampling point, which is used to characterize the intensity of local flow velocity change; Indicates the angle between the velocity vector direction at the i th spatial sampling point and the reference axis, reflecting the degree of deviation of the flow direction; In the state lattice, analyze the curvature change trend of the continuous trajectory, and construct the flue gas chaos degree reference value through the total curvature change of the continuous trajectory, which reflects the geometric distortion degree of the trajectory in the high-dimensional phase space, that is, the nonlinearity, complexity and irregularity of the flue gas flow. The construction expression of the flue gas chaos degree reference value is: ， Wherein: is the reference value of flue gas chaos degree, N represents the number of trajectory points of the spatial sampling point, represents the i trajectory curvature at the th spatial sampling point, and the calculation formula is: ， Wherein: and are the first-order guiding vector and the second-order guiding vector of the state vector group respectively, represents the i curvilinear arc length position of the i -th spatial sampling point, represents the rate of change of curvature along the trajectory and reflects the degree of local trajectory distortion; The specific steps for generating the flow rate pulsation intensity reference value after deeply analyzing the ratio of the fluctuation amplitude of the flue gas flow rate to the average flow rate per unit time during the monitoring period are as follows: During the monitoring period, a continuous sequence of flow velocity data points is collected , and the flow velocity variation frequency response is calculated based on the sequence of flow velocity data points. The calculation expression is as follows: ； Wherein: is the flow velocity variation frequency response, represents the instantaneous flow velocity value at the k th sampling moment, n is the number of sampling points within the monitoring period, and the numerator part is the discrete second derivative approximation of the flow velocity, representing the acceleration trend change rate of the flow velocity, and the denominator part is the normalization term, introducing the sum of adjacent velocities to control the influence of the numerical scale; After obtaining the flow rate variation frequency response, further combine the fluctuation energy density of the flow rate data to construct a flow rate pulsation intensity reference value with enhanced perception ability. The constructed expression is: ， Wherein: represents the reference value of the flow velocity pulsation intensity, represents the instantaneous jump of the flow velocity within a unit sampling interval, representing the first-order gradient of the flow velocity, emphasizes the sensitivity to severe jumps in the fluctuation energy.

2. The method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases according to claim 1, characterized in that, The historical average experience value refers to the average value of the flue gas flow rate statistically obtained based on the actual working condition data during the long-term operation of the SCR system.

3. The method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases according to claim 1, wherein Take the flue gas chaos degree reference value and the flow rate pulsation intensity reference value as feature vectors and input them into the trained deep learning model. Output the hazard degree coefficient through the deep learning model, and intelligently predict the potential turbulence risk during the flue gas flow process based on the hazard degree coefficient.

4. An evaluation method for the synergistic emission reduction effect of air pollutants and greenhouse gases according to claim 3, characterized in that, Compare and analyze the hazard degree coefficient generated when predicting the potential turbulence risk in the flue gas using the deep learning model with the preset hazard degree reference threshold to identify the turbulence risk. The specific steps are as follows: If the hidden danger degree coefficient is greater than the hidden danger degree reference threshold, a risk signal is generated, indicating the existence of potential turbulent flow risk in the flue gas; if the hidden danger degree coefficient is less than or equal to the hidden danger degree reference threshold, a normal signal is generated, indicating that the flue gas is flowing normally.

5. The method for evaluating the synergistic emission reduction effect of air pollutants and greenhouse gases according to claim 4, wherein After identifying the turbulent flow risk, quantitatively analyze the severity of the turbulent flow, and intelligently adjust the preset flow rate of the flue gas according to the turbulent flow risk level. The specific steps are as follows: After identifying the existence of turbulent flow risk in the flue gas, compare the currently calculated hidden danger degree coefficient with the preset hidden danger degree reference threshold, and construct a flow rate adjustment factor according to the deviation degree to quantify the turbulent flow risk level. The calculation expression of the flow rate adjustment factor is: ， in: is the flow rate control factor, is the hidden danger degree coefficient output by the current deep learning model, is the reference threshold of hidden danger level, is a very small positive number, is the response sensitivity coefficient, which is used to adjust the sensitivity to abnormalities. is the hyperbolic tangent function; After obtaining the flow velocity regulation factor Based on the current starting reference flow velocity, calculate the target flow velocity after intelligent regulation to achieve precise intervention in the flue gas flow field. The regulation expression of the target flow velocity is as follows: ， Wherein: is the starting reference flow rate, is the maximum regulation amplitude coefficient, which is used to limit the flow rate change range and prevent over-regulation, is the target flow rate after intelligent correction.

Images