Intelligent control method and system for flue gas denitration and ammonia supply of coal power plant

Abstract

The invention relates to a coal power plant flue gas denitration ammonia supply intelligent control method which comprises the following steps: acquiring historical data of coal power plant flue gas denitration and real-time detection data of a plurality of point locations on denitration inlet and outlet flue gas cross sections, and recording adjustment data of ammonia spraying branch pipe valves corresponding to ammonia spraying leveling of each point location, the historical data and the adjustment data are input into the LSTM neural network model to obtain nitrogen oxide prediction data at the next moment, the nitrogen oxide prediction data, the real-time detection data and the adjustment data are input into the random forest model to obtain a prediction value of the nitrogen oxide concentration of each point location, and the opening degree of an ammonia supply main adjustment valve is adjusted according to the average nitrogen oxide concentration to obtain the nitrogen oxide concentration of each point location. And adjusting the valve opening of each ammonia spraying branch pipe based on the predicted value of the nitrogen oxide concentration of each point location. Through the method and the device, the problem of low nitrogen supply amount control accuracy is solved, historical monitoring data and real-time detected multi-point detection data are integrated, and the accuracy of a prediction result is improved; leveling data are optimized through ammonia spraying, and a branch pipe control valve opening degree control strategy is provided.

Description

Technical Field

[0001] This application relates to the field of denitrification technology in coal-fired power plants, and in particular to intelligent control methods and systems for flue gas denitrification and ammonia supply in coal-fired power plants. Background Technology

[0002] With increasingly stringent environmental protection requirements and the need for generating units to participate in deep peak shaving, SCR systems are facing challenges related to inlet NO. X Large concentration fluctuations, flue gas temperature deviations from the optimal window, increased ammonia escape, and catalyst blockage / poisoning pose multiple challenges, placing higher demands on the precision of ammonia injection control.

[0003] SCR denitrification technology for thermal power units is currently the most advanced technology for controlling nitrogen oxides (NOx) in coal-fired power plants. X The main method for emission control is PID-based control of ammonia supply valves, referencing historical NO monitoring data at the denitrification outlet. X The concentration and control setpoint are used to adjust the ammonia supply in real time to control the NO concentration at the denitrification outlet. X Although the concentration is below the environmental control limit, the control effect is not precise enough when the nitrogen supply is adjusted based on the single nitrate outlet nitrogen oxide concentration prediction. Summary of the Invention

[0004] This application provides an intelligent control method, system, electronic device, and storage medium for ammonia supply in flue gas denitrification of coal-fired power plants, which at least solves the problem of low accuracy in nitrogen supply control in related technologies.

[0005] In a first aspect, embodiments of this application provide an intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants, the method comprising: Historical data on flue gas denitrification in coal-fired power plants and real-time detection data at multiple points on the flue gas inlet and outlet cross sections of denitrification are obtained. Based on the real-time detection data, the ammonia injection at each point is leveled, and the adjustment data of the corresponding ammonia injection branch valve is recorded. The historical data and the adjustment data are input into a pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet; The predicted nitrogen oxide concentrations, the real-time detection data, and the adjustment data are input into a pre-built random forest model to obtain the predicted values ​​of nitrogen oxide concentrations at each point on the flue gas cross section at the denitrification outlet. Based on the predicted values ​​of nitrogen oxide concentration at each location, the average nitrogen oxide concentration is obtained. The opening of the main ammonia supply valve is adjusted based on the average nitrogen oxide concentration, and the opening of each ammonia injection branch valve is adjusted based on the predicted values ​​of nitrogen oxide concentration at each location.

[0006] In some embodiments, adjusting the opening of each of the ammonia injection branch valves based on the predicted nitrogen oxide concentrations at each location includes: The standard deviation of nitrogen oxide concentration at each location is obtained based on the predicted value of nitrogen oxide concentration at each location and the real-time concentration of nitrogen oxide at each location in the real-time detection data. Based on the partial dependency analysis method, the sensitivity of the ammonia injection branch valve opening adjustment to the standard deviation is analyzed, and the opening of each ammonia injection branch valve is adjusted according to the sensitivity analysis results.

[0007] In some embodiments, the method further includes: The historical data, the real-time detection data, and the adjustment data are converted and expanded into moving averages at the second and minute levels; Feature filtering is performed on the transformed and augmented historical data, real-time detection data, and adjustment data based on SHAP values.

[0008] In some embodiments, converting and expanding the adjusted data into moving averages on the order of seconds and minutes includes: Remove the initial operating condition data from the adjustment data; Based on the SMOTE method, the adjusted data after the removal process is transformed and expanded to obtain the moving average of the adjusted data. The moving average includes the moving average of the time window of 5s, 10s, 20s, 30s, 1min and 5min.

[0009] In some embodiments, the historical data includes historical denitrification inlet and outlet parameters, ammonia supply parameters, the opening degree of each ammonia supply regulating valve main pipe, and other relevant parameters of the unit; The denitrification inlet and outlet parameters include the nitrogen oxide concentrations at the inlet and outlet of the denitrification reactor, the static pressures at the inlet and outlet of the denitrification reactor, the flue gas temperatures at the inlet and outlet of the denitrification reactor, and the oxygen content at the inlet and outlet of the denitrification reactor. The ammonia supply parameters include ammonia flow rate and ammonia-air ratio; Other relevant parameters of the unit include the unit's power generation capacity, real-time air volume of primary and secondary air in the furnace, flue gas volume at the furnace outlet, coal type, and coal quality test information of the coal fed into the furnace.

[0010] In some embodiments, the detection points on the inlet and outlet flue gas cross sections of the denitrification system are set up using a grid method, and the real-time detection data includes the nitrogen oxide concentration, static pressure, flue gas temperature, and flue gas composition at each point; the nitrogen oxide concentration is detected based on non-dispersive infrared absorption method.

[0011] In some embodiments, each ammonia injection branch pipe is equipped with a zoned remote regulating valve, and the method further includes: The LSTM neural network model and the random forest model are used to predict the nitrogen oxide concentration in each region; Based on the predicted nitrogen oxide concentrations in each zone, the opening degree of the remote regulating valve for each zone is adjusted.

[0012] Secondly, embodiments of this application provide an intelligent control system for flue gas denitrification and ammonia supply in coal-fired power plants, the system comprising: The data acquisition module is used to acquire historical data of flue gas denitrification in coal-fired power plants and real-time detection data of multiple points on the flue gas cross-section of the denitrification inlet and outlet. Based on the real-time detection data, the ammonia injection at each point is leveled, and the adjustment data of the corresponding ammonia injection branch valve is recorded. The first prediction module is used to input the historical data and the adjustment data into a pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet. The second prediction module is used to input the nitrogen oxide prediction data, the real-time detection data and the adjustment data into a pre-built random forest model to obtain the predicted value of nitrogen oxide concentration at each point on the flue gas cross section at the denitrification outlet. The adjustment module is used to obtain the average nitrogen oxide concentration based on the predicted value of the nitrogen oxide concentration at each point, adjust the opening of the main ammonia supply valve based on the average nitrogen oxide concentration, and adjust the opening of each ammonia injection branch valve based on the predicted value of the nitrogen oxide concentration at each point.

[0013] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants as described in the first aspect above.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants as described in the first aspect above.

[0015] Compared to related technologies, the intelligent control method for ammonia supply in flue gas denitrification of coal-fired power plants provided in this application, on the one hand, effectively overcomes the problems of poor PID control performance caused by the lag and lack of representativeness of historical monitoring data by integrating historical monitoring data with real-time multi-point detection data; on the other hand, by optimizing and balancing ammonia injection data and using a machine learning model coupled with interpretable output, it can provide a fast and effective local branch valve opening control strategy for denitrification operation of coal-fired power plants. This solves the problem of low accuracy in nitrogen supply control and improves the intelligence and accuracy of the SCR denitrification operation control system of coal-fired units. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of an intelligent control method for flue gas denitrification and ammonia supply in a coal-fired power plant according to an embodiment of this application; Figure 2 This is a schematic diagram of the inlet and outlet detection points for denitrification according to an embodiment of this application; Figure 3 This is a schematic diagram of an intelligent control method for denitrification and ammonia supply according to an embodiment of this application; Figure 4 This is a flowchart of a data processing method according to an embodiment of this application; Figure 5 This is a structural block diagram of the intelligent control system for flue gas denitrification and ammonia supply in a coal-fired power plant according to an embodiment of this application; Figure 6 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0018] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0019] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0020] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0021] This embodiment provides an intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants. Figure 1 This is a flowchart of an intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain historical data of flue gas denitrification in coal-fired power plants and real-time detection data of multiple points on the flue gas cross-section of denitrification inlet and outlet. Adjust the ammonia injection at each point according to the real-time detection data and record the adjustment data of the corresponding ammonia injection branch valve.

[0022] In some embodiments, historical data includes historical denitrification inlet and outlet parameters, ammonia supply parameters, the opening degree of each ammonia supply control valve, and other relevant parameters of the unit.

[0023] The inlet and outlet parameters for denitrification include the concentration of nitrogen oxides at the inlet and outlet of the denitrification reactor, the static pressure at the inlet and outlet of the denitrification reactor, the flue gas temperature at the inlet and outlet of the denitrification reactor, and the oxygen content at the inlet and outlet of the denitrification reactor; the ammonia supply parameters include the ammonia supply flow rate and the ammonia-air ratio; other relevant parameters of the unit include the unit's power generation, the real-time air volume of the primary and secondary air in the furnace, the flue gas volume at the furnace outlet, the coal quality type, and the coal quality test information of the coal fed into the furnace.

[0024] In some embodiments, the detection points on the flue gas cross-section of the denitrification inlet and outlet are set up using a grid method, and the real-time detection data includes the nitrogen oxide concentration, static pressure, flue gas temperature and flue gas composition at each point; the nitrogen oxide concentration is detected based on non-dispersive infrared absorption method.

[0025] Traditional single-point or few-point measurements cannot reflect the non-uniformity of flue gas parameters across the entire flue cross-section. Due to factors such as uneven boiler combustion, flue design, and the effectiveness of baffles, imported NO... X There are often significant deviations in flow velocity distribution. This leads to situations where "some areas have high ammonia levels and high NO levels." X Low, some places have low ammonia NO X "Too much" means that excessive ammonia was sprayed in order to take care of the worst-performing areas, resulting in a large amount of ammonia escape and economic losses.

[0026] Figure 2 This is a schematic diagram of the inlet and outlet detection points for denitrification according to an embodiment of this application, such as... Figure 2 As shown, this embodiment sets up multiple detection points. During unit operation, a multi-channel flue gas sampling gun is used to acquire real-time detection data at various points along the entire flue gas cross-section at the denitrification inlet and outlet. The detection data at each point is analyzed, and based on the analysis results, the corresponding zone branch valves at each point are adjusted to control the NO... X In areas with high concentrations, spray more ammonia; in areas with low concentrations, spray less ammonia, ensuring the export NO... X While ensuring the maximum value is met, reduce the total ammonia consumption to decrease the risk of local ammonia escape.

[0027] Using NO X The concentration of nitrogen oxides (NOx) is measured by the selective absorption of infrared light at specific wavelengths by molecules (mainly NO). Based on non-dispersive infrared absorption (NDIR) for detecting NOx concentration, this method is sensitive only to the target gas (NOx), with minimal interference from other common components in flue gas (such as CO2, water vapor, and SO2), ensuring the accuracy and reliability of the measurement data.

[0028] The ammonia injection system was optimized and leveled, and the adjustment data of the corresponding ammonia injection branch valves were recorded, i.e., the changes in the opening degree of the ammonia injection branch valves.

[0029] The ultimate goal of intelligent control is to find an optimal combination of valve openings that results in the most uniform NOx distribution at the outlet and the lowest total ammonia consumption. In this embodiment, valve adjustment data is recorded to allow the model to learn the complex mapping relationship between operating conditions and valve openings.

[0030] Without adjustment data, the model can only provide an "initial" or "theoretical" valve opening suggestion based on inlet conditions. With this data, however, the model can verify whether the previous control action achieved the expected results. If the effect is unsatisfactory (e.g., NOx reduction is not significant in a certain area), the model can learn in the next adjustment, attempting larger or smaller adjustments to the valve, gradually approaching the optimal solution. This process achieves online, continuous self-optimization.

[0031] Valve characteristics are often nonlinear; this can be observed by recording changes in valve opening and the resulting system changes (such as outlet NO). X The model can learn the individual characteristics of each valve (due to wear, blockage, etc., the actual flow-opening curve of each valve is unique), and establish a "digital fingerprint" for each valve, thereby enabling more precise control.

[0032] Furthermore, valve adjustments are not independent; opening one valve more fully may affect the flow and concentration fields in adjacent regions, causing changes in NOx levels in those regions as well. Only by recording the combined effects of adjusting all valves simultaneously can the model learn this complex coupling relationship and thus make collaboratively optimal decisions.

[0033] Continue to refer to Figure 1 After acquiring historical data, real-time detection data, and adjustment data, step S102 is executed.

[0034] Step S102: Input historical data and adjustment data into the pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet.

[0035] Step S103: Input the predicted nitrogen oxide data, real-time detection data, and adjustment data into the pre-built random forest model to obtain the predicted values ​​of nitrogen oxide concentrations at each point on the flue gas cross section at the denitrification outlet.

[0036] The model is constructed using a first-layer output and a second-layer correction approach. The first layer employs an LSTM neural network, and the second layer uses a random forest. Firstly, processed historical data and adjusted data are input into the first-layer model to predict the NO₂ output of the denitrification outlet during the next timescale of online monitoring. X Concentration; then the output of the first-layer model is added to the dataset to detect NO online at the current time. X Using concentration data as a feature, the nitrogen oxide concentration at each point on the cross-section of the denitrification outlet flue gas is predicted.

[0037] Preferably, the training model employs multi-fold cross-validation, and the mean squared error (RMSE) metric is used for performance evaluation. The mathematical expression is as follows:

[0038] in, For the observed values, n represents the fitted values ​​for each model. samples This represents the number of data entries included in the model.

[0039] In this embodiment, multiple types of random forest models are trained based on the different fluctuation characteristics of the unit load, and the most suitable model is switched and called according to the real-time load fluctuation situation. The load fluctuation situation includes stable load conditions and frequent load fluctuations.

[0040] Preferably, the classification criteria are set empirically based on the unit capacity type. Optionally, the criterion is that the unit load fluctuation exceeds 20% of the unit capacity within half an hour. If the difference between the unit's maximum load and minimum load (fluctuation range) exceeds 20% of the capacity in the past half hour, it is judged that the current state is "frequent load fluctuation"; if the load fluctuation range in the past half hour is less than or equal to 20% of the capacity, it is judged that the current state is "stable load".

[0041] Each random forest sub-model only needs to handle a relatively single running state, so the patterns it learns are more consistent, it has a stronger ability to generalize to new data of the same type, and it reduces the risk of overfitting.

[0042] LSTM's unique gating mechanism enables it to excel at learning long-term and short-term dependencies in time series. The "ammonia injection-reaction" process in denitrification exhibits a significant physical lag, making LSTM highly suitable for modeling this lag effect. An LSTM layer outputs a master prediction of the outlet NOx concentration at a future time, and also generates a high-dimensional feature vector containing numerous learned temporal patterns, which encapsulates information about the current system state.

[0043] The output of the first model (NO) X The main predicted value of the concentration (feature vector) and real-time detection data are used as inputs to the random forest model. Random forest excels at performing high-precision nonlinear mapping based on static and cross features. The corrected output of the random forest model is the predicted concentration of nitrogen oxides at each point on the flue gas cross section at the denitrification outlet.

[0044] LSTM overcomes the weakness of Random Forest in handling long-term dependencies in pure time series data. Random Forest, in turn, overcomes the limitations of LSTM, which may be inefficient or require large amounts of data when dealing with static feature interactions and tabular data. The two-layer model in this embodiment combines the advantages of deep learning and traditional machine learning, fully considering and utilizing the strong temporal and multimodal (spatiotemporally static) characteristics simultaneously present in denitrification system data, thus improving prediction accuracy and reliability.

[0045] Continue to refer to Figure 1 After obtaining the predicted values ​​of nitrogen oxide concentrations at each location, step S104 is executed.

[0046] Step S104: Based on the predicted value of nitrogen oxide concentration at each point, obtain the average nitrogen oxide concentration, adjust the opening of the main ammonia supply valve based on the average nitrogen oxide concentration, and adjust the opening of each ammonia injection branch valve based on the predicted value of nitrogen oxide concentration at each point.

[0047] Using the currently predicted average nitrogen oxide concentration as the target of PID control, the opening of the ammonia supply main control valve is adjusted to ensure the outlet NO... X The average concentration consistently meets environmental protection requirements. Using predicted values ​​instead of measured values ​​for control overcomes the control challenges of lagging systems.

[0048] In some embodiments, adjusting the opening degree of each ammonia injection branch valve based on the predicted value of the nitrogen oxide concentration at each point in step S104 includes: The standard deviation of nitrogen oxide concentration at each location is obtained based on the predicted value of nitrogen oxide concentration at each location and the real-time concentration of nitrogen oxide at each location in the real-time detection data.

[0049] Based on the partial dependency analysis method, the sensitivity of the ammonia injection branch pipe valve opening adjustment to the standard deviation is analyzed, and the opening of each ammonia injection branch pipe valve is adjusted according to the sensitivity analysis results.

[0050] Based on the predicted nitrogen oxide concentrations at the denitrification outlet from multiple locations, the standard deviation of the nitrogen oxide concentration at the denitrification outlet is provided. Furthermore, a partial dependency analysis method is used to provide a sensitivity analysis of the adjustment of the ammonia supply branch pipe valve opening. Based on the analysis results, the branch pipe valve is adjusted to optimize the NO concentration at the denitrification outlet. X Concentration field. Based on standard deviation assessment and sensitivity analysis, the opening degree of the ammonia injection branch pipe valve is controlled to ensure that the NO concentration at each point of the outlet section is controlled. X The concentration distribution is most uniform, that is, the standard deviation of the concentration field is minimized.

[0051] Sensitivity analysis of the nitrogen oxide concentration in a given zone based on the opening degree of the ammonia injection branch pipe valve using partial dependency analysis can provide numerical references for power plant maintenance personnel to adjust the valves of the local ammonia injection branch pipe. Partial dependency analysis is a sensitivity analysis method based on machine learning, and its principle is as follows:

[0052] Where, x s and X C These are the influencing factors that need to be analyzed and the remaining influencing factors. The selected and constructed model is influenced by each feature, where n refers to the number of data entries in the dataset, and X... C ( i )The remaining influencing factors of the i-th factor are used to estimate the sensitivity curve through Monte Carlo sampling. .

[0053] Figure 3 This is a schematic diagram of an intelligent control method for denitrification and ammonia supply according to an embodiment of this application, as shown below. Figure 3 As shown, the partial dependence response curve of the opening of the ammonia injection branch pipe regulating valve and the nitrogen oxide outlet cross section is used to guide the on-site adjustment of the opening of the ammonia injection branch pipe zone valve.

[0054] By decoupling total quantity control (main control valve - PID) and distributed optimization (branch control valve - data-based optimization), the complexity of a single controller handling multiple conflicting objectives simultaneously is avoided.

[0055] Through the above steps, online monitoring data (historical data) and real-time on-site detection data are integrated, effectively overcoming the limitations of online monitoring NO. X The system addresses the lag issue in concentration; simultaneously, by leveraging the interpretable output of the artificial intelligence model, it provides real-time adjustment instructions for the opening of the ammonia supply valve for operators' reference, thereby improving the intelligence and precision of the SCR denitrification operation control system for coal-fired units.

[0056] In some embodiments, after obtaining historical data, real-time detection data, and adjustment data in step S101, these data also need to be preprocessed. This preprocessing primarily involves removing outliers and abnormal periods such as critical equipment failures to ensure the overall accuracy of the data. The preprocessing methods specifically include: Step S1011: Convert and expand the historical data, real-time detection data, and adjustment data into moving averages at the second and minute levels.

[0057] Second-level moving average (e.g., 10-second window): used to capture very rapid changes and disturbances, such as small adjustments to certain dampers or transient noise in measurement signals, providing high-frequency details.

[0058] Minute-level moving averages (e.g., 5-minute or 10-minute windows): used to capture major process trends, such as the effects of slow load changes and coal quality changes. They can effectively filter out short-term noise and reveal potential trends.

[0059] Industrial sensor data inevitably contains noise. Moving average is a simple and effective smoothing method that can improve data quality and allow models to learn more stable patterns. Furthermore, by calculating the average value, historical information is incorporated to some extent, which helps the model understand the inertia of the process.

[0060] The overall data features are transformed and expanded into moving averages at the second and minute levels, and the detection data is augmented using the SMOTE method. The Synthetic Minority Over-sampling Technique (SMOTE) increases the data for the minority class by "artificially" or "synthetically" creating new samples. Augmenting the detection data using the SMOTE method prevents the model from being biased towards the majority class during training, while allowing the model to better learn complex patterns under rare conditions. This leads to more accurate predictions when facing these conditions, thereby improving the safety and robustness of the control system.

[0061] Step S1012: Feature filtering is performed on the transformed and expanded historical data, real-time detection data, and adjustment data based on the SHAP value.

[0062] Incorporate empirical judgment data and perform feature identification and filtering on the overall feature data based on SHAP values.

[0063] By quantifying and injecting domain expert knowledge into the model, the shortcomings of purely data-driven approaches are compensated. These features provide the model with powerful "prior knowledge," enabling it to quickly identify and adapt to certain special operating conditions and avoid making predictions that violate process principles.

[0064] The importance of each feature is identified based on SHAP values. SHAP (SHapley Additive exPlanations) values ​​are a model interpretability method based on game theory's Shapley values. They can explain complex models while providing physically meaningful contributions of driving factors. They precisely tell us how much each feature (variable) contributes (whether it positively pushes or negatively drags down) when the model makes a particular prediction. The specific calculation method is as follows:

[0065] potential, , representing the expected concentration of nitrogen oxides, is usually taken as the mean of the data included in the model; S is the subset of features excluding feature i; F is the complete set of features; p represents the number of features. The calculation considers a weighted combination of multiple variable values. It can be regarded as the contribution (µg / m3) of the i-th influencing factor to the change in pollutant concentration. The importance analysis is performed by rounding the mean of the absolute values ​​of the contributions in the dataset, and features that are close to 0 are removed.

[0066] By using SHAP values ​​to identify and filter features from overall feature data, the risk of overfitting can be reduced, the model training speed can be accelerated, and a more concise and efficient model can be obtained. Figure 4 This is a flowchart of a data processing method according to an embodiment of this application.

[0067] In some embodiments, step S1011, which involves converting and expanding the adjusted data into moving averages at the second and minute levels, includes: The initial operating condition data in the adjustment data is removed. Based on the SMOTE method, the adjustment data after removal is transformed and expanded to obtain the moving average value of the adjustment data. The moving average value includes the moving average value of time windows of 5s, 10s, 20s, 30s, 1min and 5min.

[0068] The Synthetic Minority Oversampling (SMOTE) technique is used to augment the field detection dataset. This method avoids neglecting the features of the detection data during subsequent model training due to the relatively small amount of online data in the field detection dataset, thus balancing the overall dataset to improve the accuracy of subsequent models.

[0069] The addition of experience and supplementary judgments to the data includes, but is not limited to: the number of catalyst layers, the process characteristic indicators of the catalyst, and the commissioning time of each catalyst layer. The data expanded to include second-level and minute-level time-series information, primarily using the moving average values ​​of time windows of 5s, 10s, 20s, 30s, 1min, and 5min.

[0070] The current power generation capacity is classified into three categories based on the ratio of power generation capacity to the maximum rated output: high load (80%-100%), medium load (60%-80%), and low load (<60%).

[0071] Assessing the current power generation load, whether there are significant load changes in a short period of time, and the impact of peak shaving, etc., are crucial, as both the size and changes in power generation load have a significant impact on the operating efficiency of denitrification equipment.

[0072] In some embodiments, each ammonia injection branch pipe is equipped with a zoned remote regulating valve, and the method further includes: The concentration of nitrogen oxides in each region was predicted using an LSTM neural network model and a random forest model. Based on the predicted nitrogen oxide concentrations in each zone, the opening of the remote control valve in each zone is adjusted.

[0073] If the NO outlet of the denitrification facility of a coal-fired power plant X The concentration field has poor uniformity, and there are remote regulating valves in each ammonia injection branch. The model can predict the nitrogen oxide concentration in each zone and feed it back to the zone regulating valve for automatic control.

[0074] It should be noted that the remote control valve for each zone is a technically modified denitrification device. It can remotely adjust the opening of the branch pipe (electric / pneumatic valve) in the central control room to control the branch pipe, and can directly realize remote intelligent control.

[0075] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0076] This embodiment also provides an intelligent control system for flue gas denitrification and ammonia supply in a coal-fired power plant. This system is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the terms "module," "unit," "subunit," etc., can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0077] Figure 5 This is a structural block diagram of the intelligent control system for flue gas denitrification and ammonia supply in a coal-fired power plant according to an embodiment of this application, as shown below. Figure 5 As shown, the system includes: The data acquisition module 51 is used to acquire historical data of flue gas denitrification in coal-fired power plants and real-time detection data of multiple points on the flue gas cross-section of the denitrification inlet and outlet. Based on the real-time detection data, the ammonia injection at each point is leveled, and the adjustment data of the corresponding ammonia injection branch valve is recorded.

[0078] The first prediction module 52 is used to input historical data and adjustment data into a pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet.

[0079] The second prediction module 53 is used to input the predicted nitrogen oxide data, real-time detection data and adjustment data into a pre-built random forest model to obtain the predicted value of nitrogen oxide concentration at each point on the flue gas cross section of the denitrification outlet.

[0080] The adjustment module 54 is used to obtain the average nitrogen oxide concentration based on the predicted value of nitrogen oxide concentration at each point, adjust the opening of the main ammonia supply valve based on the average nitrogen oxide concentration, and adjust the opening of each ammonia injection branch valve based on the predicted value of nitrogen oxide concentration at each point.

[0081] In some embodiments, the adjustment module 54 includes: The analysis module is used to obtain the standard deviation of nitrogen oxide concentration at each location based on the predicted value of nitrogen oxide concentration at each location and the real-time concentration of nitrogen oxide at each location in the real-time detection data.

[0082] The branch valve adjustment module is used to analyze the sensitivity of the ammonia injection branch valve opening adjustment based on the standard deviation using a partial dependency analysis method, and adjust the opening of each ammonia injection branch valve according to the sensitivity analysis results.

[0083] In some embodiments, the system further includes: The preprocessing module is used to convert and expand historical data, real-time detection data, and adjustment data into moving averages at the second and minute levels.

[0084] The filtering module is used to perform feature filtering on transformed and expanded historical data, real-time detection data, and adjustment data based on SHAP values.

[0085] In some embodiments, the preprocessing module includes: The data removal module is used to remove initial operating condition data from the adjustment data.

[0086] The transformation module is used to transform and expand the adjusted data after the removal process based on the SMOTE method to obtain the moving average of the adjusted data. The moving average includes the moving average of the time window of 5s, 10s, 20s, 30s, 1min and 5min.

[0087] In some embodiments, historical data includes historical denitrification inlet and outlet parameters, ammonia supply parameters, opening degree of each ammonia supply regulating valve main pipe, and other relevant parameters of the unit; The inlet and outlet parameters for denitrification include the nitrogen oxide concentrations at the inlet and outlet of the denitrification reactor, the static pressure at the inlet and outlet of the denitrification reactor, the flue gas temperature at the inlet and outlet of the denitrification reactor, and the oxygen content at the inlet and outlet of the denitrification reactor. Ammonia supply parameters include ammonia flow rate and ammonia-to-air ratio; Other relevant parameters of the unit include the unit's power generation capacity, real-time air volume of primary and secondary air in the furnace, flue gas volume at the furnace outlet, coal type, and coal quality test information of the coal fed into the furnace.

[0088] In some embodiments, the detection points on the flue gas cross-section of the denitrification inlet and outlet are set up using a grid method, and the real-time detection data includes the nitrogen oxide concentration, static pressure, flue gas temperature and flue gas composition at each point; the nitrogen oxide concentration is detected based on non-dispersive infrared absorption method.

[0089] In some embodiments, each ammonia injection branch pipe is equipped with a zoned remote regulating valve, and the system further includes: The remote control module is used to predict the nitrogen oxide concentration in each zone using an LSTM neural network model and a random forest model; and to adjust the opening of the remote control valve in each zone based on the predicted nitrogen oxide concentration.

[0090] By integrating historical monitoring data with real-time multi-point detection data, the system effectively overcomes the problems of poor PID control performance caused by the lag and lack of representativeness of historical monitoring data. By optimizing and balancing ammonia injection data and using machine learning models coupled with interpretable output, a fast and effective local branch valve opening control strategy can be provided for the denitrification operation of coal-fired power plants. This solves the problem of low accuracy in nitrogen supply control and improves the intelligence and precision of the SCR denitrification operation control system of coal-fired units.

[0091] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0092] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0093] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0094] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program: S1: Acquire historical data of flue gas denitrification in coal-fired power plants and real-time detection data of multiple points on the flue gas cross-section of denitrification inlet and outlet. Adjust the ammonia injection at each point according to the real-time detection data and record the adjustment data of the corresponding ammonia injection branch valve.

[0095] S2 inputs historical data and adjustment data into a pre-built LSTM neural network model to obtain the predicted nitrogen oxide data at the denitrification outlet at the next moment.

[0096] S3. The predicted nitrogen oxide concentrations at each point on the flue gas cross section at the denitrification outlet are obtained by inputting the predicted nitrogen oxide concentrations, real-time detection data, and adjustment data into a pre-built random forest model.

[0097] S4. Based on the predicted value of nitrogen oxide concentration at each point, the average nitrogen oxide concentration is obtained. The opening of the main ammonia supply valve is adjusted based on the average nitrogen oxide concentration, and the opening of each ammonia injection branch valve is adjusted based on the predicted value of nitrogen oxide concentration at each point.

[0098] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0099] In one embodiment, Figure 6 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application, such as... Figure 6 As shown, an electronic device is provided, which can be a server, and its internal structure diagram can be as follows. Figure 6 As shown, this electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements an intelligent control method for flue gas denitrification and ammonia supply in a coal-fired power plant.

[0100] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0101] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0102] Those skilled in the art should understand that the technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for intelligent control of flue gas denitrification and ammonia supply in coal-fired power plants, characterized in that, The method includes: Historical data on flue gas denitrification in coal-fired power plants and real-time detection data at multiple points on the flue gas inlet and outlet cross sections of denitrification are obtained. Based on the real-time detection data, the ammonia injection at each point is leveled, and the adjustment data of the corresponding ammonia injection branch valve is recorded. The historical data and the adjustment data are input into a pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet; The predicted nitrogen oxide concentrations, the real-time detection data, and the adjustment data are input into a pre-built random forest model to obtain the predicted values ​​of nitrogen oxide concentrations at each point on the flue gas cross section at the denitrification outlet. Based on the predicted values ​​of nitrogen oxide concentration at each location, the average nitrogen oxide concentration is obtained. The opening of the main ammonia supply valve is adjusted based on the average nitrogen oxide concentration, and the opening of each ammonia injection branch valve is adjusted based on the predicted values ​​of nitrogen oxide concentration at each location.

2. The method according to claim 1, characterized in that, The adjustment of the opening degree of each ammonia injection branch valve based on the predicted value of nitrogen oxide concentration at each location includes: The standard deviation of nitrogen oxide concentration at each location is obtained based on the predicted value of nitrogen oxide concentration at each location and the real-time concentration of nitrogen oxide at each location in the real-time detection data. Based on the partial dependency analysis method, the sensitivity of the ammonia injection branch valve opening adjustment to the standard deviation is analyzed, and the opening of each ammonia injection branch valve is adjusted according to the sensitivity analysis results.

3. The method according to claim 1, characterized in that, The method further includes: The historical data, the real-time detection data, and the adjustment data are converted and expanded into moving averages at the second and minute levels; Feature filtering is performed on the transformed and augmented historical data, real-time detection data, and adjustment data based on SHAP values.

4. The method according to claim 3, characterized in that, Converting and expanding the adjusted data into moving averages at the second and minute levels includes: Remove the initial operating condition data from the adjustment data; Based on the SMOTE method, the adjusted data after the removal process is transformed and expanded to obtain the moving average of the adjusted data. The moving average includes the moving average of the time window of 5s, 10s, 20s, 30s, 1min and 5min.

5. The method according to claim 1, characterized in that, The historical data includes historical denitrification inlet and outlet parameters, ammonia supply parameters, the opening degree of each ammonia supply regulating valve, and other relevant parameters of the unit. The denitrification inlet and outlet parameters include the nitrogen oxide concentrations at the inlet and outlet of the denitrification reactor, the static pressures at the inlet and outlet of the denitrification reactor, the flue gas temperatures at the inlet and outlet of the denitrification reactor, and the oxygen content at the inlet and outlet of the denitrification reactor. The ammonia supply parameters include ammonia flow rate and ammonia-air ratio; Other relevant parameters of the unit include the unit's power generation capacity, real-time air volume of primary and secondary air in the furnace, flue gas volume at the furnace outlet, coal type, and coal quality test information of the coal fed into the furnace.

6. The method according to claim 1, characterized in that, The detection points on the inlet and outlet flue gas cross sections of the denitrification system are set up using a grid method. The real-time detection data includes the nitrogen oxide concentration, static pressure, flue gas temperature, and flue gas composition at each point. The nitrogen oxide concentration is detected based on the non-dispersive infrared absorption method.

7. The method according to claim 1, characterized in that, Each ammonia injection branch pipe is equipped with a zoned remote regulating valve, and the method further includes: The LSTM neural network model and the random forest model are used to predict the nitrogen oxide concentration in each region; Based on the predicted nitrogen oxide concentrations in each zone, the opening degree of the remote regulating valve for each zone is adjusted.

8. An intelligent control system for flue gas denitrification and ammonia supply in a coal-fired power plant, characterized in that, The system includes: The data acquisition module is used to acquire historical data of flue gas denitrification in coal-fired power plants and real-time detection data of multiple points on the flue gas cross-section of the denitrification inlet and outlet. Based on the real-time detection data, the ammonia injection at each point is leveled, and the adjustment data of the corresponding ammonia injection branch valve is recorded. The first prediction module is used to input the historical data and the adjustment data into a pre-built LSTM neural network model to obtain the nitrogen oxide prediction data at the next moment of the denitrification outlet. The second prediction module is used to input the nitrogen oxide prediction data, the real-time detection data and the adjustment data into a pre-built random forest model to obtain the predicted value of nitrogen oxide concentration at each point on the flue gas cross section at the denitrification outlet. The adjustment module is used to obtain the average nitrogen oxide concentration based on the predicted value of the nitrogen oxide concentration at each point, adjust the opening of the main ammonia supply valve based on the average nitrogen oxide concentration, and adjust the opening of each ammonia injection branch valve based on the predicted value of the nitrogen oxide concentration at each point.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the intelligent control method for flue gas denitrification and ammonia supply in coal-fired power plants as described in any one of claims 1 to 7.

Images