Flue gas denitration pollution reduction and carbon reduction method and system based on model predictive control

Abstract

The invention relates to the technical field of industrial flue gas purification, and discloses a flue gas denitration pollution reduction and carbon reduction method and system based on model prediction control. The method comprises the following steps: collecting flue gas data through a sensor to obtain a flue gas distribution state; the state is processed through a preset model, and a nitrogen oxide concentration predicted value is determined; judging a regulation and control demand based on the predicted value and generating an adjustment coefficient sequence; calculating an opening value of an ammonia spraying distributor by adopting a particle swarm optimization algorithm, and determining ammonia spraying amount distribution; a control instruction is sent according to ammonia spraying amount distribution, and a temperature uniformity index is evaluated based on temperature feedback data; adjusting optimization algorithm parameters according to the indexes, and generating an optimized ammonia spraying strategy; updating the predicted value and determining the stable range of the denitration efficiency; verifying the ammonia escape concentration in the stable range to obtain a system performance index; and forming a continuous regulation and control sequence according to cyclic feedback of the indexes. According to the invention, accurate dynamic regulation and control of ammonia spraying amount are realized, denitration efficiency and system stability are effectively improved, and the risk of ammonia escape is significantly reduced.

Description

Technical Field

[0001] This application relates to the field of industrial flue gas purification technology, and in particular to a model predictive control method and system for flue gas denitrification, pollution reduction, and carbon reduction. Background Technology

[0002] Industries such as coal-fired power plants, steel smelting, and cement production generate large amounts of nitrogen oxides (NOx) during combustion. As one of the major air pollutants, NOx not only causes environmental problems such as acid rain and photochemical smog but also exacerbates the greenhouse effect, severely impacting human health and ecosystems. To reduce NOx emissions, selective catalytic reduction (SCR) and selective non-catalytic reduction (SNCR) technologies are widely used. These technologies reduce and remove NOx by injecting reducing agents such as ammonia or urea into the flue gas.

[0003] In existing denitrification engineering practices, the control of ammonia injection mostly relies on traditional feedback control methods such as PID control. While this method offers a degree of real-time capability, it often only adjusts the feedback from a single point or a limited number of monitoring points, making it difficult to reflect the overall distribution characteristics of flue gas flow, nitrogen oxide concentration, and temperature field within the catalyst bed. This limitation easily leads to uneven ammonia injection distribution, resulting in problems such as ammonia slip and fluctuations in denitrification efficiency. This not only increases operating costs but may also cause secondary pollution.

[0004] In recent years, with the development of sensor networks and data acquisition technologies, researchers have attempted to introduce big data analysis and artificial intelligence models into denitrification systems to improve prediction accuracy and control performance. However, existing methods often focus on single prediction models or local optimization strategies, lacking a systematic approach to ammonia injection control, temperature uniformity, ammonia escape risk, and the stability of denitrification efficiency, making it difficult to form an efficient closed-loop control mechanism. Furthermore, traditional optimization algorithms are prone to getting trapped in local optima under high-dimensional, multivariate operating conditions, making it difficult for ammonia injection strategies to consistently meet the dual demands of pollution reduction and carbon reduction.

[0005] Therefore, existing technologies urgently need a denitrification control method and system that can combine real-time data acquisition, model predictive control and intelligent optimization algorithms to achieve dynamic optimization of ammonia injection distribution, while taking into account nitrogen oxide emission reduction efficiency, ammonia escape control and energy consumption reduction, thereby achieving the synergistic goal of pollution reduction and carbon reduction. Summary of the Invention

[0006] This application provides a model predictive control method and system for flue gas denitrification, pollution reduction, and carbon reduction. It is used to overcome the problems of unstable denitrification efficiency, inaccurate ammonia injection control, and secondary pollution caused by complex flue gas distribution, large fluctuations in nitrogen oxide concentration, and high ammonia escape risk in the prior art. It achieves dynamic and precise control of ammonia injection in complex flue gas environments, simultaneously improving denitrification efficiency, reducing ammonia escape risk, and synergistically optimizing carbon emissions.

[0007] In a first aspect, this application provides a model predictive control method for flue gas denitrification, pollution reduction, and carbon reduction, the method comprising: Step 1: Collect real-time data including flue gas flow rate and nitrogen oxide concentration using a high-frequency sampling mechanism through a sensor network, and obtain the flue gas distribution status based on the real-time data; Step 2: Process the flue gas distribution using the first preset model to determine the predicted value of nitrogen oxide concentration; Step 3: Determine the ammonia injection rate control demand based on the predicted values ​​and generate an adjustment coefficient sequence; Step 4: Based on the adjustment coefficient sequence, use an optimization algorithm to calculate the opening value of each ammonia injector distributor and determine the ammonia injection quantity distribution; Step 5: Send control commands to the control actuator based on the ammonia injection distribution, and evaluate the temperature uniformity index based on the acquired feedback data; Step 6: Adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm with adjusted parameters; Step 7: Based on the optimized ammonia injection strategy, obtain the control signal, update the predicted value using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency; Step 8: Verify the ammonia slip concentration within the stable range using a sensor network and determine the system performance indicators; Step 9: Based on the system performance indicators, feed the feedback back to the high-frequency sampling mechanism and obtain the continuous control sequence.

[0008] Secondly, this application provides a model predictive control flue gas denitrification, pollution reduction, and carbon reduction system, the system comprising: The high-frequency sampling module is used to collect real-time data, including flue gas flow rate and nitrogen oxide concentration, through a sensor network using a high-frequency sampling mechanism, and to obtain the flue gas distribution status based on the real-time data; The concentration prediction module is used to process the flue gas distribution state through a first preset model to determine the predicted value of nitrogen oxide concentration. The demand adjustment module is used to determine the ammonia injection volume control demand based on the predicted value and generate an adjustment coefficient sequence. The distribution calculation module is used to calculate the opening value of each ammonia injector distributor based on the adjustment coefficient sequence and an optimization algorithm, and to determine the ammonia injection quantity distribution. The index evaluation module is used to send control commands to the control actuator based on the ammonia injection quantity distribution and evaluate the temperature uniformity index based on the acquired feedback data. The strategy optimization module is used to adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm after parameter adjustment. The range determination module is used to obtain control signals based on the optimized ammonia injection strategy, update the predicted values ​​using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency. The verification and evaluation module is used to verify the ammonia slip concentration within a stable range through a sensor network and to determine the system performance indicators. The closed-loop feedback module is used to cyclically feed back to the high-frequency sampling mechanism based on system performance indicators and obtain a continuous control sequence.

[0009] Compared with the prior art, the beneficial effects of the technical solution of this application are at least as follows: 1. By introducing neural network-based model prediction into denitrification control, the future trend of nitrogen oxide concentration can be accurately estimated, which significantly improves the foresight and accuracy of ammonia injection regulation compared with traditional single-point feedback regulation.

[0010] 2. By using intelligent algorithms such as particle swarm optimization, the opening degree of the ammonia spray distributor is iteratively calculated, and constrained by spatial distribution uniformity and historical data, dynamic optimization of ammonia spray distribution and global optimal allocation of ammonia spray amount are achieved, effectively avoiding uneven ammonia spray and local over-spraying.

[0011] 3. By introducing temperature uniformity index and ammonia slip risk level, the optimization algorithm parameters are adaptively adjusted to ensure the stability of denitrification efficiency under different operating conditions, reduce the risk of the system getting trapped in local optima under complex operating conditions, and enhance the stability and robustness of the system.

[0012] 4. While determining the stable range of denitrification efficiency, the ammonia slip concentration is verified and feedback is provided to form a closed-loop control mechanism, which effectively reduces secondary pollution caused by ammonia slip and reduces the total amount of ammonia injected while ensuring denitrification efficiency, thereby achieving synergistic benefits of pollution reduction and carbon reduction. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of a model predictive control method for flue gas denitrification, pollution reduction, and carbon reduction according to this application. Figure 2 This is a schematic diagram of a model predictive control flue gas denitrification, pollution reduction, and carbon reduction system according to this application. Detailed Implementation

[0015] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] For ease of understanding, the specific process of the embodiments of this application is described below. Figure 1 The diagram shows a flowchart of a model predictive control method for flue gas denitrification, pollution reduction, and carbon reduction provided by the present invention. The flowchart specifically includes the following steps: Step 1: Collect real-time data including flue gas flow rate and nitrogen oxide concentration using a high-frequency sampling mechanism through a sensor network, and obtain the flue gas distribution status based on the real-time data.

[0017] In one specific embodiment, the process of performing step 1 may specifically include the following steps: Real-time data is collected by sensors deployed at designated locations in the catalyst bed using a high-frequency sampling mechanism, generating a sequence of initial parameters at the second level. The initial parameter sequence is mapped to the spatial coordinates of the ammonia injector, and combined with the real-time ammonia injection rate, the flue gas distribution state is determined.

[0018] Specifically, multiple sensor nodes arranged at the inlet, outlet, and intermediate regions of the catalyst bed are used to synchronously collect flue gas flow rate and nitrogen oxide concentration data at a sampling frequency of no less than 10 Hz, forming a second-level initial parameter sequence arranged with timestamps. The sensor nodes include flue gas flow meters and nitrogen oxide concentration analyzers. This sequence contains timestamped flow rate and concentration values, directly reflecting the dynamic changes of the flue gas within the catalyst bed. For example, the sequence format at any sensor node is [t1: flow rate a1, concentration b1; t2: flow rate a2, concentration b2].

[0019] To construct a spatially meaningful distribution state, the initial parameter sequence is integrated with the physical layout of the ammonia injection distributor. Specifically, the second-level parameter sequence data (flow rate, concentration) is mapped to corresponding coordinates based on the acquisition point location. The spatial position of the ammonia injection distributor within the catalyst bed is defined by three-dimensional coordinates; for example, coordinates (1,2,0) represent a specific injection point. The ammonia injection rate is then integrated to determine the flue gas distribution state, where the ammonia injection rate is the volume of ammonia injected per minute. This flue gas distribution state integrates four types of parameters: flow rate, concentration, distributor location, and ammonia injection rate. Through the interaction of these data, it accurately characterizes the actual distribution features of the flue gas within the catalyst bed.

[0020] In data processing, the parameter sequence is mapped to the coordinates of the distributor location. For example, a weighted average is calculated for the sequence data at each coordinate point to generate a local flue gas distribution map reflecting the distribution of pollutants in each region of the bed cross-section. The distribution map is adjusted in conjunction with the ammonia injection rate. For example, if the injection rate is 10 liters per minute, the concentration values ​​in the sequence are scaled according to the rate ratio to form a complete flue gas distribution. This integration can improve the uniformity of distribution and reduce ammonia waste.

[0021] By employing high-frequency sampling and spatial mapping processing, local peak regions of nitrogen oxide concentration on the bed cross-section can be detected. For example, the concentration at coordinates (1,2,0) is identified as being 20% ​​higher than the average, thus providing comprehensive and accurate input for subsequent prediction models. The hardware deployment for data acquisition and the data processing algorithm are tightly coupled. High-frequency sampling provides the algorithm with the basis for recognizing dynamic changes, while the spatial mapping algorithm transforms discrete sensor data into a structurally meaningful distribution. The two functions support each other, jointly solving the technical challenge of incomplete sensing, laying the data foundation for achieving precise ammonia injection control, and ultimately improving the overall accuracy and response speed of denitrification control.

[0022] Step 2: Process the flue gas distribution using the first preset model to determine the predicted value of nitrogen oxide concentration.

[0023] In one specific embodiment, the process of performing step 2 may specifically include the following steps: The distribution parameter sequence is extracted from the flue gas distribution state. The distribution parameter sequence includes flue gas flow data, nitrogen oxide concentration data, ammonia injection distributor location coordinates and ammonia injection rate. The opening percentage range of the ammonia injection distribution valve, real-time flue gas flow feedback, and distribution parameter sequence are all input into a preset neural network model to obtain the predicted value of nitrogen oxide concentration. The predicted value represents the concentration change trend within a preset time window.

[0024] Specifically, a sequence of distribution parameters, including flue gas flow rate, nitrogen oxide concentration, ammonia injector location coordinates, and ammonia injection rate, is extracted from the flue gas distribution. This sequence is high-frequency time-series data at the second level. The distribution parameter sequence, the percentage range of the ammonia injection valve opening reflecting the current control state, and the real-time flue gas flow feedback characterizing instantaneous disturbances are input into a pre-defined neural network model. This model is a multilayer perceptron structure trained on historical flue gas distribution data. Its input layer receives multi-dimensional sequence data, the hidden layer processes the nonlinear coupling relationship between parameters using the ReLU activation function, and the output layer generates a predicted concentration value within a preset time window. This predicted value represents the concentration change trend (increasing, stabilizing, or decreasing) within the preset time window (e.g., the next 10 seconds). This step correlates the ammonia injection system operating parameters (opening range, flow feedback) with flue gas distribution characteristics, enabling the model to comprehensively consider the influence of ammonia injection conditions and flue gas state on concentration changes.

[0025] Neural network models learn the mapping relationship between parameter sequences and concentration changes in historical data, enabling them to integrate spatial distribution information with real-time dynamic signals. For example, when real-time flue gas flow feedback shows that the flow rate jumps from 500 cubic meters per hour to 550 cubic meters per hour, while the opening percentage is within the range of 50% to 80%, the model can predict that the nitrogen oxide concentration may increase by 20% within the next 10 seconds, thereby adjusting the ammonia injection strategy in advance to avoid a decrease in denitrification efficiency or excessive ammonia escape.

[0026] The algorithm features (neural network model structure and training method) and technical features (multi-source data input and load trend calculation) support each other in function. Through the interaction between data, the changes in nitrogen oxide concentration can be accurately predicted, providing a basis for subsequent ammonia injection regulation. Together, they solve the technical problems of low denitrification efficiency and high ammonia escape risk caused by complex flue gas distribution and inaccurate ammonia injection matching, thus improving the control accuracy and system stability of the denitrification process.

[0027] Step 3: Determine the ammonia injection rate control demand based on the predicted values ​​and generate an adjustment coefficient sequence.

[0028] In one specific embodiment, the process of performing step 3 may specifically include the following steps: Determine whether the predicted value exceeds the preset value. If so, obtain the concentration time series data corresponding to the nitrogen oxide concentration, input the concentration time series data into the second preset model, and extract the fluctuation characteristics of the nitrogen oxide concentration. The fluctuation characteristics include the amplitude and frequency of the change in nitrogen oxide concentration. Based on the comparison results between the fluctuation characteristics and the corresponding preset thresholds, the ammonia injection volume control requirements are determined. By combining valve response delay and pressure difference compensation factors, dynamic compensation is performed on the control requirements to generate an adjustment coefficient sequence for controlling the opening of the ammonia injection distributor.

[0029] Specifically, when the predicted value exceeds a threshold set based on historical data (e.g., 150 mg / m³), the control demand analysis process is triggered. At this point, the time-series concentration data corresponding to the nitrogen oxide concentration is obtained from the original data on which the predicted value was generated, and this time-series data is input into a second preset model (i.e., a Long Short-Term Memory network model) to extract fluctuation characteristics. This LSTM model resolves the time dependence of the concentration sequence through its hidden state output and calculates its amplitude and frequency of change. The combined characteristics of amplitude and frequency can characterize the instability of flue gas nitrogen oxide distribution, thereby reflecting the impact of boiler load fluctuations, uneven combustion, and flue gas flow rate changes on denitrification efficiency.

[0030] Based on the extracted fluctuation characteristics, they are compared with corresponding preset characteristic thresholds to determine the need for ammonia injection rate adjustment. For example, if the fluctuation amplitude exceeds a preset amplitude threshold (e.g., 50 mg / m³) or the frequency is higher than a preset frequency threshold (e.g., twice per minute), it indicates a rapid increase in nitrogen oxide concentration, and it is determined that the ammonia injection rate needs to be increased to suppress the concentration rise. Conversely, if the fluctuation amplitude is small and the frequency is lower than the threshold, it indicates that the system is in a relatively stable state, and the ammonia injection rate can be maintained or slightly adjusted. This judgment process relies on the mapping relationship between characteristics and effects in historical control data, which can ensure denitrification efficiency while avoiding ammonia escape due to over-response.

[0031] Before generating specific adjustment commands, dynamic compensation is required to adapt the control demand to the actual industrial environment. The compensation process combines valve response delay and pressure difference compensation factor. Valve response delay refers to the time lag between the ammonia injection valve receiving the opening adjustment command and actually reaching the target opening. Failure to consider this lag will lead to a deviation between the actual ammonia injection volume and the predicted demand. This lag is obtained through historical data measurement; for example, a measured delay of 3 seconds. The pressure difference compensation factor is a correction value based on the pressure difference between the inlet and outlet of the ammonia injection pipeline. This parameter affects the uniformity of ammonia distribution and injection intensity. For example, if the standard pressure difference is 100 Pa and the actual pressure difference is 115 Pa, then the compensation factor is (115-100) / 100=0.15. Combining the valve response delay, pressure difference compensation factor, and control demand value, an adjustment coefficient sequence is formed. For example, the calculation formula can be set as control demand value × (1 + compensation factor) / (1 + delay / standard time). The resulting set of adjustment coefficients accurately reflects the dynamic compensation of the actual execution effect of the ammonia distributor under industrial field conditions.

[0032] For example, in a scenario of sudden boiler load increase, the predicted value of nitrogen oxides exceeds 150 mg / m³, and the extracted fluctuation characteristics show an amplitude of 60 mg / m³ and a frequency of 3 times per minute. It is determined that the ammonia injection rate needs to be increased by 20%. Based on this, considering a valve response delay of 3 seconds and a pressure difference compensation factor of 0.15, the adjustment coefficient sequence is calculated as [1.1, 1.2, 1.15]. The corresponding ammonia injection distributor opening is gradually adjusted from 50% to 55%, 60%, and 57%. This method can suppress the peak fluctuation of nitrogen oxide concentration and maintain the denitrification efficiency above 85%. For example, under a relatively stable load, when only local concentrations are abnormal, the predicted characteristic amplitude is 40 mg / m³ and the frequency is once per minute. It is determined that only a 5% fine adjustment of the ammonia injection amount is needed. The adjustment coefficient sequence generated by combining a 2-second delay and a compensation factor of 0.1 is [1.05, 1.04, 1.06]. The corresponding opening is fine-tuned to 52%, 51%, and 53%, which maintains the uniformity of the temperature field and controls the ammonia escape concentration below 5 mg / m³.

[0033] The calculation scheme of this invention achieves accurate characterization of concentration dynamics at the algorithm level through prediction models and fluctuation feature extraction, and achieves precise regulation of ammonia injection quantity at the control execution level through delay and pressure compensation correction. The two interact functionally to jointly solve the problems of uneven ammonia injection distribution, response lag and high risk of ammonia escape. Thus, it achieves the beneficial effects of stable denitrification efficiency, reduced ammonia escape and optimized energy consumption in flue gas denitrification, pollution reduction and carbon reduction applications.

[0034] Step 4: Based on the adjustment coefficient sequence, use an optimization algorithm to calculate the opening value of each ammonia injector distributor and determine the ammonia injection quantity distribution.

[0035] In one specific embodiment, the process of performing step 4 may specifically include the following steps: With the adjustment coefficient sequence as the optimization objective, the particle swarm optimization algorithm is used for iterative solution. In the solution process, the spatial distribution uniformity index of the ammonia spray distributor and the historical opening sequence are used as constraints and are incorporated into the fitness function for calculation. The global optimal solution is output as the optimal opening value for each ammonia injection distributor, and the final ammonia injection quantity distribution is determined based on each opening value and its maximum injection rate.

[0036] Specifically, the adjustment coefficient sequence provides real-time input data for the optimization algorithm, ensuring that the optimization process is highly real-time and dynamic. For example, the adjustment coefficient sequence includes the amplitude and frequency of concentration fluctuations, reflecting the current state of the flue gas system, and also considers the response delay of valves. These factors serve as important constraints in the particle swarm optimization algorithm, ensuring that the optimized ammonia injection rate can effectively cope with the dynamic changes of the system.

[0037] The particle swarm optimization algorithm iteratively optimizes the opening value of each ammonia injector distributor, aiming to achieve a globally optimal ammonia injection distribution. In each iteration, the algorithm calculates a fitness value based on the current opening value of the ammonia injector distributor and compares it with historical opening sequences and distribution uniformity indices to form a comprehensive evaluation. The fitness function incorporates multiple factors, such as the deviation of the adjustment coefficient sequence, the uniformity of ammonia injection distribution, and the stability of historical opening changes. These factors collectively determine the optimization direction of the ammonia injector distributor opening value. Specifically, the fitness function includes constraints on the uniformity of ammonia injection distribution. By calculating the spatial distribution of each ammonia injector distributor, the uniformity of ammonia injection is evaluated, and weighting coefficients are added to the algorithm to ensure that uniformity is maximized during the optimization process. For example, the spatial distribution uniformity index is obtained by calculating the Euclidean distance variance using the three-dimensional coordinates of the ammonia injector distributor. For instance, if the variance of 16 distributors in a 4×4 grid is 0.2, it is considered high uniformity, and the penalty weight is reduced. Furthermore, the historical opening sequence is used to constrain drastic changes in the opening value of the ammonia injector distributor, preventing excessive fluctuations in the system and thus ensuring the stability of the ammonia injection distribution and the long-term stable operation of the system. The historical opening sequence records the opening values ​​of the previous few time steps and is used to calculate the opening change rate. If the change rate exceeds a threshold (e.g., 10%), a stability penalty term is added to the fitness function to constrain drastic fluctuations. For example, if the opening sequence of the most recent five time steps is [50%, 52%, 55%, 53%, 54%], the change rate is 1.5%, and no stability penalty is applied if it is below the 10% threshold.

[0038] The core of the optimization algorithm lies in the iterative process of particle swarm optimization. In particle swarm optimization, a particle represents the opening value of a set of ammonia injector distributors, with its initial position randomly generated within the opening range. The particle swarm continuously updates the position and velocity of particles by comparing individual optimal solutions with the global optimal solution, searching for the global optimum. In each iteration, the fitness function scores the quality of the combination of opening values, thereby adjusting the position of the particles, and finally outputting the global optimum solution, which is the optimal opening value for each ammonia injector distributor. This process not only optimizes the distribution of ammonia gas but also adapts to the ammonia injection distribution requirements under different loads by adjusting optimization parameters, such as the number of particles and the number of iterations.

[0039] By optimizing the ammonia injection rate distribution, the system can determine the final ammonia injection volume based on the opening value and maximum injection rate of each ammonia distributor. For example, the optimized ammonia injection rate distribution ensures a slightly higher ammonia volume in the central area and a moderately lower volume in the peripheral areas, effectively avoiding local oversaturation, improving denitrification efficiency, and reducing the risk of ammonia escape. This optimization method can also ensure stable adjustment of the ammonia injection rate under low load or sudden load changes, effectively controlling energy consumption and gas emissions during the denitrification process, reducing ammonia waste and valve wear.

[0040] In the process of flue gas denitrification, pollution reduction, and carbon reduction, optimizing the distribution of ammonia injection is a key step in improving denitrification efficiency, reducing ammonia slip, and minimizing energy waste. By adjusting the coefficient sequence and applying optimization algorithms, the opening value of each ammonia injector is accurately calculated, ultimately determining the ammonia injection distribution, thereby optimizing ammonia distribution and improving denitrification performance. In this process, particle swarm optimization (PSO) is used as the core technology, leveraging its global search capability and adaptive adjustment characteristics to effectively solve problems such as poor uniformity of ammonia injection distribution, unstable control, and wasted ammonia injection. This improves denitrification efficiency, reduces ammonia slip, and enhances system reliability and long-term operational stability.

[0041] Step 5: Send control commands to the control actuator based on the ammonia injection distribution, and evaluate the temperature uniformity index based on the acquired feedback data.

[0042] In one specific embodiment, the process of performing step 5 may specifically include the following steps: Control commands are generated based on the ammonia injection volume distribution and sent to the control actuator to adjust the valve opening of the ammonia injection distributor; Temperature feedback data for each region of the catalyst bed after receiving the execution control command; The escape concentration threshold is calculated based on temperature feedback data, and risk classification is performed according to the predefined escape concentration risk level. The risk classification results are adjusted based on the real-time ammonia injection rate to obtain the ammonia escape risk level; The initial temperature uniformity index is calculated based on the standard deviation and mean of the temperature feedback data. The initial temperature uniformity index is corrected using the ammonia escape risk level to generate a temperature uniformity index.

[0043] Specifically, the ammonia injection rate distribution parameters are converted into control signals, which are then sent to the control actuator in the form of digital pulses. The control actuator then drives the dynamic adjustment of the valve opening of each ammonia injection distributor, thereby forming an ammonia injection distribution pattern in the catalyst bed space that is consistent with the ammonia injection rate distribution. In this process, the opening value of the ammonia injection distributor corresponds one-to-one with the actual ammonia injection rate. The pulse width and frequency of the control signal directly determine the amplitude and rate of change of the valve opening, thus realizing the mapping relationship between the ammonia injection rate distribution data and the physical execution action.

[0044] After valve adjustment, feedback data is collected by temperature sensors deployed in different areas of the catalyst bed. This feedback data reflects the spatial temperature distribution of the bed in the form of a multi-point sequence. For example, temperature sequences were collected for three areas of the bed: [350℃, 352℃, 348℃], [345℃, 347℃, 346℃], and [340℃, 342℃, 341℃]. The average value represents the overall thermal state of that area, while the temperature gradient within the bed is revealed compared to the midstream and downstream temperature distributions. Through this data collection, a causal relationship is established between the ammonia injection distribution command and the catalyst temperature state. The temperature data not only reflects the results after the ammonia injection control is executed but also provides an input basis for subsequent risk and uniformity calculations.

[0045] After obtaining temperature feedback data, the ammonia escape concentration threshold is further calculated. This threshold is estimated based on the difference in temperature distribution. The formula uses the difference between the maximum and minimum temperature values ​​as a quantitative indicator of the degree of imbalance, and combines this with a preset concentration baseline and proportional coefficient. For example, the calculation formula is: Threshold = Base Value + (T_max - T_min) * Coefficient, where T_max and T_min are the highest and lowest temperatures in the distribution, respectively. For example, with a base value of 3 ppm, a coefficient of 0.5, T_max of 352℃, and T_min of 340℃, the calculated escape concentration threshold is 9 ppm. The escape concentration threshold is then mapped to three risk levels: low, medium, and high, according to risk grading standards. For example, the predefined risk levels are: below 4 ppm is low risk, 4–6 ppm is medium risk, and above 6 ppm is high risk. In this case, the escape concentration threshold is 9 ppm, which falls under the high risk category.

[0046] Further risk level adjustments are made by incorporating real-time ammonia injection rate. If the ammonia injection rate exceeds the baseline value by 10%, the medium-risk scenario is directly upgraded to a high-risk scenario. This correction rule ensures that the risk assessment is sensitive to the ammonia injection volume, avoiding underestimation of escape risk under high flow conditions. Temperature feedback data and ammonia injection rate form dual inputs in the algorithm. The calculation results not only characterize the potential ammonia escape risk caused by temperature imbalances but also enhance the dynamic adaptability of risk assessment through rate correction.

[0047] The initial temperature uniformity index is calculated based on temperature feedback data. This index measures the degree of uniformity by the ratio of the standard deviation to the mean of the temperature distribution; the smaller the standard deviation, the more uniform the temperature distribution. For example, the formula for calculating the uniformity index is: Uniformity = 1 - (T_std / T_avg), where T_std is the standard deviation of the temperature distribution and T_avg is the average temperature. For instance, with T_avg = 345.67℃ and T_std = 4.04℃, the calculated uniformity value is approximately 0.988, indicating a highly uniform overall temperature distribution in the bed.

[0048] The initial temperature uniformity index is corrected based on the ammonia slip risk level. If the risk level is high, the uniformity value is multiplied by a correction factor (e.g., 0.8). After correction, the index decreases to 0.790, indicating that although the temperature distribution appears uniform, the high-risk conditions still require vigilance regarding the overall operating conditions. Through this correction, the temperature uniformity index not only reflects the statistical characteristics of the temperature distribution but also incorporates the impact of ammonia slip risk on the safety and stability of denitrification, thus expanding its function from a single statistical quantity to a comprehensive operational evaluation.

[0049] The technical solution of this invention features mutually supportive functionalities in its methodological and technical characteristics. Temperature monitoring technology provides a data source for risk assessment, while the risk assessment algorithm quantifies the non-uniformity of temperature distribution into operable risk indicators, together forming the feedback verification link in closed-loop control. Through this process, local temperature anomalies and escalating escape risks caused by uneven ammonia injection distribution can be detected in a timely manner. For example, when the corrected uniformity index drops to 0.79, it indicates that the ammonia injection strategy needs to be adjusted to avoid a decrease in catalyst efficiency or excessive ammonia escape, thereby improving the system's adaptability and operational reliability, enhancing the accuracy of ammonia injection control, solving the problem of accumulated control deviations caused by the lack of feedback in open-loop control, and achieving the unity of stable denitrification efficiency and pollution reduction and carbon reduction goals.

[0050] Step 6: Adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm with adjusted parameters.

[0051] In one specific embodiment, the process of performing step 6 may specifically include the following steps: Determine if the temperature uniformity index is lower than the preset threshold; if so, initiate the optimization algorithm parameter adjustment process. The efficiency fluctuation range is calculated based on historical data of denitrification efficiency, and a predefined temperature correlation coefficient is obtained; The efficiency fluctuation range and temperature correlation coefficient are input into the risk assessment function to obtain the current ammonia escape risk level; The search parameters of the particle swarm optimization algorithm are dynamically adjusted based on the ammonia escape risk level. An optimized ammonia injection strategy is generated using a particle swarm optimization algorithm with adjusted parameters to reduce the risk of ammonia escape.

[0052] Specifically, when the temperature uniformity index is lower than the preset threshold, it indicates that there is a significant difference in the temperature distribution inside the catalyst bed. This imbalance will lead to a decrease in the reaction efficiency of ammonia and flue gas and amplify the risk of ammonia escape. Therefore, it is necessary to start the parameter adjustment process to correct the search path of the optimization algorithm.

[0053] In this process, the system calculates the efficiency fluctuation range based on historically collected denitrification efficiency data. This range is obtained by calculating the variance or standard deviation of the efficiency sequence over a time window, reflecting the stability of efficiency over time. Large fluctuations indicate that the system is in an unstable operating state, increasing the uncertainty weight in risk assessment. Simultaneously, the system calls a predefined temperature correlation coefficient, derived from historical correlation analysis between temperature uniformity and ammonia slip concentration. The closer the value is to 1, the more significant the impact of temperature on slip concentration.

[0054] Efficiency fluctuation amplitude and temperature correlation coefficient are simultaneously input into the risk assessment function. The model uses a weighted summation or a nonlinear function to map the two parameters to a risk level score. For example, the risk assessment function is: Risk Level = 0.6 × Efficiency Fluctuation Amplitude + 0.4 × Temperature Correlation Coefficient. For instance, when the efficiency fluctuation amplitude is 0.2 and the temperature correlation coefficient is 0.75, the model outputs a risk level of 0.42, close to 0.5, indicating a medium-to-high risk threshold.

[0055] The risk level directly drives the parameter adjustment logic of the optimization algorithm. In particle swarm optimization, search parameters such as inertia weight, learning factor, and particle size determine the search range and convergence speed. If the risk level is high, the system increases the inertia weight to enhance the global search capability and avoid the algorithm getting trapped in local optima. If the risk level is in the medium range, the number of particles is moderately increased to improve the diversity of ammonia injection distribution solutions by expanding the search dimension. If the risk level is low, the parameter convergence remains unchanged, and only the random factor in the velocity update formula is fine-tuned to reduce the energy consumption caused by over-adjustment.

[0056] After parameter adjustment, the modified particle swarm optimization algorithm receives flue gas operation data, ammonia injection valve opening, and temperature feedback data as input conditions. It calculates the ammonia slip risk cost function under different ammonia injection distribution schemes and iteratively updates to find the ammonia injection distribution scheme that minimizes the cost. In this process, the output of the risk assessment function interacts with the fitness function of the algorithm iteration, ensuring that the algorithm not only pursues maximizing denitrification efficiency but also considers the goal of suppressing ammonia slip risk, forming a dual-objective coupled optimization logic. The generated ammonia injection strategy is output to the control actuator in the form of a numerical matrix, allocating the flow rate ratio of each ammonia injection nozzle, thus realizing the correspondence between the physical execution level and the optimization result. Preferably, the optimized ammonia injection strategy includes the adjusted ammonia injection distribution value.

[0057] For example, under high-load operating conditions, the temperature uniformity index was 0.78, lower than the threshold of 0.8, the historical efficiency fluctuation range was 0.18, the temperature correlation coefficient was 0.82, and the risk assessment function output a risk value of 0.56, indicating a high risk. Therefore, the system adjusted the inertia weight of the particle swarm optimization algorithm from 0.6 to 0.9 and the number of particles from 50 to 100 to enhance global search capabilities and improve solution accuracy. The optimized ammonia injection strategy reduced the ammonia injection rate in the central region by 10% and increased it by 5% in the peripheral region, effectively suppressing ammonia escape caused by localized high temperatures. The ammonia concentration decreased from 3 ppm to 1.5 ppm while maintaining a denitrification efficiency above 90%.

[0058] The technical solution of this invention realizes the linkage between adaptive parameter adjustment and dynamic optimization of ammonia injection distribution. This enables the optimization algorithm to not only rely on static model prediction, but also to make dynamic corrections based on actual operating data. It solves the technical problem that traditional ammonia injection control strategies cannot take into account both efficiency stability and escape risk. It achieves the beneficial effect of significantly reducing ammonia escape concentration and improving system safety and stability while ensuring denitrification efficiency.

[0059] Step 7: Based on the optimized ammonia injection strategy, obtain the control signal, update the predicted value using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency.

[0060] Specifically, in the flue gas denitrification system, a control signal is sent to the ammonia injection distributor based on the optimized ammonia injection strategy. This signal includes the distributor's opening command set and ammonia injection rate data. The control signal acts on the actual ammonia injection valve through the execution unit, realizing real-time adjustment of the ammonia injection quantity and distribution. These control signals are fed back as new input conditions to the first preset model, namely the neural network model. This model combines real-time temperature, flue gas flow data, and inlet nitrogen oxide concentration monitoring data to update the predicted value of nitrogen oxide concentration within a future preset time window. At the same time, the system integrates the set range of the ammonia-nitrogen molar ratio (NSR) and the span of the prediction time window, and simulates the response curve of denitrification efficiency under different NSR values ​​through the model. For example, the NSR is gradually increased from 0.8 to 1.2, and the trend of predicted efficiency changes is observed.

[0061] Based on the updated predicted values ​​and the NSR-efficiency response relationship, the system determines the stable range of denitrification efficiency. For example, if the prediction shows that the efficiency can be stably maintained between 88% and 93% within the next 60 seconds, and the predicted ammonia slip concentration is below 5 ppm, this range is determined to be an effective stable operating interval. This process uses the multivariate processing capability of the neural network model to dynamically correlate the control strategy, reactant ratio, and time factor, ultimately outputting an operating boundary that balances efficiency and safety.

[0062] The technical solution of this invention can dynamically define a safe and efficient operating range, avoid conservative or risky control behavior caused by fixed threshold settings, reduce the fluctuation range of denitrification efficiency to within 5%, and ensure that the risk of ammonia escape is controlled, thereby improving the economy and reliability of system operation.

[0063] Step 8: Verify the ammonia slip concentration within the stable range using a sensor network and determine the system performance indicators.

[0064] In one specific embodiment, the process of performing step 8 may specifically include the following steps: When the denitrification efficiency is within a stable range, ammonia escape concentration data is collected by sensors deployed at preset positions in the catalyst bed at a preset verification sampling frequency. The ammonia escape concentration data is compared with the preset escape concentration threshold to generate a verification result that includes the deviation value. The system performance index used to evaluate the stability of denitrification control is calculated based on the deviation value.

[0065] Specifically, ammonia slip concentration data is collected via a sensor network within the stable range of denitrification efficiency. To ensure data accuracy and representativeness, sensors are deployed downstream of the catalyst bed, in an area where ammonia slip concentration changes significantly. The sensors monitor ammonia concentration in real time using electrochemical principles and collect data according to a set validation sampling frequency (e.g., once per second to once per minute). Each collection of ammonia slip concentration data forms a time series, covering multiple moments within the stable range of denitrification efficiency, ensuring a comprehensive reflection of system behavior. For example, when the denitrification efficiency is between 90% and 95%, the sensor network may collect data 10 times per minute, forming a dense data point sequence, ensuring both timeliness and coverage of the data.

[0066] After data acquisition, the system compares the real-time ammonia slip concentration data with a preset threshold to generate verification results. The preset ammonia slip concentration threshold is typically set based on industry standards or empirical values; for example, the upper limit for ammonia slip concentration might be set at 10 ppm. By analyzing the difference between the real-time value and the threshold, the deviation value at each time point can be obtained, reflecting whether the ammonia slip concentration is controlled within the predetermined range. If the ammonia slip concentration exceeds the preset threshold, the system will issue an alarm signal to prompt operators to take measures to adjust the ammonia injection rate or optimize the denitrification efficiency; if the deviation value is small, it indicates that the system is operating stably and the denitrification effect is good.

[0067] Based on the deviation values, the overall system performance indicators are further calculated. The core of these performance indicators lies in evaluating the stability of the denitrification control, primarily quantified by comparing the degree of deviation in the results. For example, the system calculates the difference between the real-time ammonia slip concentration at each time point and a preset threshold, then calculates the average absolute value of the difference, and uses the reciprocal of the average absolute value as the stability score. The stability score is a key indicator for measuring the performance of the denitrification control system, helping maintenance personnel understand whether the system is operating optimally and whether problems such as excessive ammonia slip are possible. Changes in performance indicators under different operating conditions can help schedule and optimize operational strategies, thereby improving the overall stability and response speed of the denitrification system.

[0068] For example, when the boiler load is 80%, the ammonia slip concentration collected by the sensor network is 3 ppm, and the system calculates a deviation of -1 ppm based on the set threshold. At this point, by calculating the average deviation of multiple data points, a stability index (e.g., 4.5) can be obtained, indicating that the denitrification system is operating well under the current load. If the load increases to 90%, the same calculation method will yield a new performance index, and the index may show an increase (e.g., reaching 4.8), indicating that the system's stability has improved under higher loads, thus providing a basis for subsequent optimized control.

[0069] The ammonia slip concentration within the stable range of denitrification efficiency was verified by using a sensor network, and system performance indicators were determined. This aimed to evaluate the stability of the denitrification system, improve the safety and controllability of the denitrification process, and ensure that the system maintains efficient and stable operation under complex working conditions.

[0070] Step 9: Based on the system performance indicators, feed the feedback back to the high-frequency sampling mechanism and obtain the continuous control sequence.

[0071] Specifically, based on the overall system performance indicators, the feedback is cyclically fed back to the high-frequency sampling mechanism to determine whether further optimization is needed. At the same time, the distributor location coordinates and risk classification standards are linked to obtain a continuous control sequence.

[0072] Based on the system performance indicators obtained from sensor network verification (such as a stability index reaching 4.5), a feedback signal is generated after comparing this indicator with a preset target threshold. This signal dynamically adjusts the operating parameters of the high-frequency sampling mechanism through a control loop. For example, when the performance indicators show increased fluctuations in ammonia escape concentration, the feedback signal triggers an increase in the sampling frequency from 10Hz to 15Hz to enhance the ability to capture dynamic changes in flue gas; simultaneously, the weights of the sensor network's acquisition points are adjusted, prioritizing increased monitoring density in historically abnormal areas. The updated sampling mechanism collects flue gas data in real time and inputs it into the flue gas distribution state construction process in step 1, initiating a new round of closed-loop control from state perception, predictive regulation, to verification and optimization, thereby forming a continuous regulation sequence that continuously adapts to the system state. The specific form of the regulation sequence is a time-stamped control instruction set sequence, which is essentially a series of control commands arranged in chronological order that can be directly issued to the actuators to achieve continuous dynamic adjustment of the ammonia injection quantity.

[0073] Taking high-load operation as an example, if high-frequency sampling finds that the ammonia escape concentration has increased from 2 ppm to 3 ppm, and the system performance indicators are judged to have a risk of deviation, the feedback logic returns this information to the high-frequency sampling mechanism, and generates an updated continuous control sequence in the next sampling cycle, which increases the ammonia injection rate of the edge nozzles by 5% and decreases the ammonia injection rate of the center nozzles by 8%, thereby achieving a decrease in ammonia escape concentration and maintaining it within a stable range.

[0074] This technology features a real-time closed-loop relationship between performance indicators and a high-frequency sampling mechanism. The algorithm provides dynamic correction capabilities in the feedback loop, while the sampling mechanism ensures the timeliness and integrity of the input data. The two support each other functionally, solving the problem of inaccurate control caused by feedback lag, and achieving the beneficial effect of improving the stability of denitrification efficiency while reducing ammonia slip and energy consumption.

[0075] The above describes a model predictive control method for flue gas denitrification, pollution reduction, and carbon reduction in embodiments of this application. The following describes a model predictive control system for flue gas denitrification, pollution reduction, and carbon reduction in embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 This application provides a schematic diagram of an embodiment of a model predictive control flue gas denitrification, pollution reduction, and carbon reduction system. The system includes: The high-frequency sampling module 10 is used to collect real-time data including flue gas flow rate and nitrogen oxide concentration through a sensor network and a high-frequency sampling mechanism, and to obtain the flue gas distribution status based on the real-time data.

[0076] The concentration prediction module 20 is used to process the flue gas distribution state through a first preset model to determine the predicted value of nitrogen oxide concentration.

[0077] The demand adjustment module 30 is used to determine the ammonia injection volume control demand based on the predicted value and generate an adjustment coefficient sequence. The distribution calculation module 40 is used to calculate the opening value of each ammonia injector distributor based on the adjustment coefficient sequence and an optimization algorithm, and to determine the ammonia injection quantity distribution.

[0078] The index evaluation module 50 is used to send control commands to the control actuator based on the ammonia injection quantity distribution and to evaluate the temperature uniformity index based on the acquired feedback data.

[0079] The strategy optimization module 60 is used to adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm after parameter adjustment.

[0080] The range determination module 70 is used to obtain control signals based on the optimized ammonia injection strategy, update the predicted values ​​using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency.

[0081] The verification and evaluation module 80 is used to verify the ammonia slip concentration within a stable range through a sensor network and to determine the system performance indicators.

[0082] The closed-loop feedback module 90 is used to cyclically feed back to the high-frequency sampling mechanism based on the system performance indicators and obtain a continuous control sequence.

[0083] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A model-based predictive control method for flue gas denitrification, pollution reduction, and carbon reduction, characterized in that, The method includes: Step 1: Collect real-time data including flue gas flow rate and nitrogen oxide concentration using a high-frequency sampling mechanism through a sensor network, and obtain the flue gas distribution status based on the real-time data; Step 2: Process the flue gas distribution state using the first preset model to determine the predicted value of nitrogen oxide concentration; Step 3: Determine the ammonia injection rate control requirement based on the predicted values ​​and generate an adjustment coefficient sequence; Step 4: Based on the adjustment coefficient sequence, use an optimization algorithm to calculate the opening value of each ammonia injector distributor and determine the ammonia injection quantity distribution; Step 5: Send control commands to the control actuator based on the ammonia injection distribution, and evaluate the temperature uniformity index based on the acquired feedback data; Step 6: Adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm with adjusted parameters; Step 7: Based on the optimized ammonia injection strategy, obtain the control signal, update the predicted value using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency; Step 8: Verify the ammonia slip concentration within the stable range using a sensor network and determine the system performance indicators; Step 9: Based on the system performance indicators, feed the feedback back to the high-frequency sampling mechanism and obtain a continuous control sequence.

2. The method according to claim 1, characterized in that, Step 1 includes: The real-time data is collected by sensors deployed at designated locations in the catalyst bed using a high-frequency sampling mechanism, generating an initial parameter sequence at the second level. The initial parameter sequence is mapped to the spatial coordinates of the ammonia injection distributor, and combined with the real-time ammonia injection rate, the flue gas distribution state is determined.

3. The method according to claim 2, characterized in that, Step 2 includes: Extract a sequence of distribution parameters from the flue gas distribution state. The sequence of distribution parameters includes flue gas flow data, nitrogen oxide concentration data, ammonia injector location coordinates, and ammonia injection rate. The opening percentage range of the ammonia injection distribution valve, the real-time flue gas flow feedback, and the distribution parameter sequence are input together into a preset neural network model to obtain the predicted value of nitrogen oxide concentration, which represents the concentration change trend within a preset time window.

4. The method according to claim 1, characterized in that, Step 3 includes: Determine whether the predicted value exceeds the preset value. If so, obtain the concentration time series data corresponding to the nitrogen oxide concentration, input the concentration time series data into the second preset model, and extract the fluctuation characteristics of the nitrogen oxide concentration, wherein the fluctuation characteristics include the change amplitude and frequency of the nitrogen oxide concentration. Based on the comparison results between the fluctuation characteristics and the corresponding preset threshold, the ammonia injection volume control requirements are determined. By combining valve response delay and pressure difference compensation factors, the control requirements are dynamically compensated to generate the adjustment coefficient sequence used to control the opening of the ammonia injection distributor.

5. The method according to claim 1, characterized in that, Step 4 includes: Using the aforementioned adjustment coefficient sequence as the optimization objective, the particle swarm optimization algorithm is employed for iterative solution. In the solution process, the spatial distribution uniformity index of the ammonia spray distributor and the historical opening sequence are used as constraints and are incorporated into the fitness function for calculation. The global optimal solution is output as the optimal opening value for each ammonia injection distributor, and the final ammonia injection quantity distribution is determined based on each opening value and its maximum injection rate.

6. The method according to claim 1, characterized in that, Step 5 includes: Control commands are generated based on the ammonia injection quantity distribution and sent to the control actuator to adjust the valve opening of the ammonia injection distributor; After the control command is executed, the temperature feedback data of each region of the catalyst bed is obtained; The escape concentration threshold is calculated based on the temperature feedback data, and the risk is classified according to the predefined escape concentration risk level. The risk classification results are adjusted based on the real-time ammonia injection rate to obtain the ammonia escape risk level; Based on the standard deviation and mean of the temperature feedback data, the initial temperature uniformity index is calculated; The initial temperature uniformity index is corrected using the ammonia escape risk level to generate the temperature uniformity index.

7. The method according to claim 1, characterized in that, Step 6 includes: Determine whether the temperature uniformity index is lower than the preset index threshold. If so, start the optimization algorithm parameter adjustment process. The efficiency fluctuation range is calculated based on historical data of denitrification efficiency, and a predefined temperature correlation coefficient is obtained; The efficiency fluctuation range and the temperature correlation coefficient are input into the risk assessment function to obtain the current ammonia escape risk level. The search parameters of the particle swarm optimization algorithm are dynamically adjusted based on the ammonia escape risk level. The optimized ammonia injection strategy for reducing the risk of ammonia escape is generated using a particle swarm optimization algorithm with adjusted parameters.

8. The method according to claim 1, characterized in that, Step 8 includes: When the denitrification efficiency is within the aforementioned stable range, ammonia escape concentration data is collected at a preset verification sampling frequency using sensors deployed at preset positions in the catalyst bed. The ammonia escape concentration data is compared with a preset escape concentration threshold to generate a verification result that includes the deviation value. Based on the deviation value, a system performance index for evaluating the stability of denitrification control is calculated.

9. A model predictive control flue gas denitrification, pollution reduction, and carbon reduction system, used to implement the method as described in any one of claims 1 to 8, characterized in that, The system includes: The high-frequency sampling module is used to collect real-time data including flue gas flow rate and nitrogen oxide concentration through a sensor network and a high-frequency sampling mechanism, and to obtain the flue gas distribution state based on the real-time data; The concentration prediction module is used to process the flue gas distribution state through a first preset model to determine the predicted value of nitrogen oxide concentration. The demand adjustment module is used to determine the ammonia injection volume control demand based on the predicted value and generate an adjustment coefficient sequence. The distribution calculation module is used to calculate the opening value of each ammonia injector distributor according to the adjustment coefficient sequence and to determine the ammonia injection quantity distribution. The index evaluation module is used to send control commands to the control actuator based on the ammonia injection quantity distribution and evaluate the temperature uniformity index based on the acquired feedback data. The strategy optimization module is used to adjust the optimization algorithm parameters according to the temperature uniformity index, and generate an optimized ammonia injection strategy based on the optimized algorithm with adjusted parameters. The range determination module is used to obtain control signals according to the optimized ammonia injection strategy, update the predicted values ​​using the first preset model, integrate the ammonia-nitrogen molar ratio and the prediction time window, and determine the stable range of denitrification efficiency. The verification and evaluation module is used to verify the ammonia slip concentration within the stability range through a sensor network and to determine the system performance indicators. The closed-loop feedback module is used to cyclically feed back to the high-frequency sampling mechanism based on the system performance indicators and obtain a continuous control sequence.

Images