Flue gas denitration self-adaptive control method of granular denitration catalyst

Abstract

The invention discloses a flue gas denitration self-adaptive control method of a granular denitration catalyst, and relates to the technical field of flue gas denitration, and the method comprises the following steps: an interactive target low-temperature SCR processing system obtains real-time control parameters and working condition parameters, generates reference operation data and expands working condition neighborhoods; and extracting historical neighborhood control data based on the historical processing log and the expansion result, wherein the historical neighborhood control data comprises historical control parameters and results. And by taking the data as a sample, constructing a neighborhood control prediction model, and inputting control parameters to predict a control result. And defining a loss function and performing optimization by taking a neighborhood control parameter space as an optimization space to obtain a neighborhood control strategy set. And finally, performing adaptive control based on the strategy set. Therefore, the technical effects of improving the working condition adaptive capacity and improving the responsiveness and the control accuracy are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of flue gas denitration, and in particular to a flue gas denitration adaptive control method of a granular denitration catalyst. Background Art

[0002] As the core component of the SCR reaction, the performance of the granular denitrification catalyst directly affects the denitrification efficiency and system stability. In practical applications, the low-temperature SCR denitrification system based on the granular denitrification catalyst is significantly affected by fluctuations in operating conditions. For example, changes in flue gas flow, temperature, ammonia concentration, and pollutant content will have a significant impact on the catalyst performance and denitrification effect.

[0003] Most of the existing flue gas denitrification control methods rely on preset static control parameters or rule-based control logic. These methods usually use fixed parameter settings, but fail to fully consider the dynamic fluctuations of process conditions in actual working conditions. In addition, some methods optimize denitrification control parameters based on machine learning models, but they have limitations in real-time and adaptability, and there are technical problems such as insufficient adaptability and real-time performance, and low control accuracy. Summary of the invention

[0004] The present invention provides a flue gas denitrification adaptive control method of a granular denitrification catalyst to solve the technical problems of insufficient adaptability and real-time performance to working conditions and low control accuracy in the prior art, and to achieve the technical effects of improving adaptability to working conditions and enhancing responsiveness and control accuracy.

[0005] The present invention provides a flue gas denitration adaptive control method for a granular denitration catalyst, comprising: The interactive target low-temperature SCR processing system obtains real-time control parameters and real-time operating condition parameters, outputs them as benchmark operating data, and expands the operating condition neighborhood based on the benchmark operating data.

[0006] A historical processing log of a target low-temperature SCR processing system is obtained, and data is extracted from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, wherein the historical neighborhood control data includes a plurality of groups of corresponding historical control parameters and historical control results.

[0007] The historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, wherein the neighborhood control prediction model uses control parameters as input to predict corresponding control results.

[0008] The neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is defined as an optimization space for optimizing control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained.

[0009] Adaptive control of the target low-temperature SCR treatment system is performed according to the neighborhood control strategy set.

[0010] In a feasible implementation, the interactive target low-temperature SCR processing system obtains real-time control parameters and real-time operating condition parameters, outputs them as reference operating data, and expands the operating condition neighborhood based on the reference operating data, including: Taking the real-time operating condition parameters in the benchmark operation data as the neighborhood origin, density clustering is performed on the historical processing log of the target low-temperature SCR processing system to generate a neighborhood clustering cluster.

[0011] Based on the neighborhood clustering clusters, neighborhood operating condition clustering clusters and neighborhood control clustering clusters are extracted respectively.

[0012] According to the neighborhood operating condition clustering cluster and the neighborhood control clustering cluster, a dual-objective neighborhood definition is performed to obtain a neighborhood control parameter space and a neighborhood operating condition parameter space.

[0013] In a feasible implementation, a dual-objective neighborhood definition is performed according to the neighborhood operating condition cluster and the neighborhood control cluster, and a neighborhood control parameter space and a neighborhood operating condition parameter space are obtained, including: A control requirement of a target low-temperature SCR treatment system is obtained, and a confidence interval constraint is defined according to the control requirement.

[0014] According to the confidence interval constraint, the confidence space ranges of the neighborhood operating condition clustering cluster and the neighborhood control clustering cluster are determined, and the neighborhood control parameter space and the neighborhood operating condition parameter space are correspondingly defined.

[0015] In a feasible implementation, a historical processing log of a target low-temperature SCR processing system is obtained, and data is extracted from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, including: The neighborhood control parameter space is configured as a first constraint.

[0016] The real-time operating condition parameter is configured as a second constraint.

[0017] Based on the first constraint and the second constraint, the historical neighborhood control data is extracted from the historical processing log, wherein the historical neighborhood control data includes a plurality of sets of corresponding historical control parameters and historical control results.

[0018] In a feasible implementation, the historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, including: The historical neighborhood control data is traversed, and the historical control results are compared with the preset emission limits.

[0019] If the historical control result is greater than or equal to the emission limit value, the corresponding historical control parameter and the historical control result are stored in a negative sample set.

[0020] If the historical control result is less than the emission limit value, the corresponding historical control parameter and the historical control result are stored in a positive sample set.

[0021] The neighborhood control prediction model constructed based on a neural network is trained according to the negative sample set and the positive sample set.

[0022] In a feasible implementation, the neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is used as an optimization space to optimize the control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained, and before that, it also includes: Based on the historical processing log, multiple control indicators are extracted to generate a control indicator set.

[0023] A principal component analysis is performed on the control indicator set, and a key control indicator set is determined based on the principal component analysis result, wherein the key control indicator set includes a plurality of serialized key control indicators.

[0024] The principal component contribution rates of the plurality of key control indicators in the key control indicator set are normalized, and the indicator optimization step lengths of the plurality of key control indicators are defined according to the normalization result.

[0025] In a feasible implementation, the method further includes: Density clustering is performed based on the historical processing logs, and a plurality of secondary reference operation data are defined based on the clustering result.

[0026] Traversing the plurality of said secondary benchmark operation data, constructing the corresponding plurality of secondary neighborhood control prediction models, and taking the plurality of said secondary neighborhood control prediction models as loss functions, respectively optimizing the control parameters, and obtaining the plurality of secondary neighborhood control strategy sets.

[0027] The neighborhood control strategy set is combined with a plurality of the secondary neighborhood control strategy sets to perform adaptive control of a target low-temperature SCR treatment system.

[0028] In a feasible implementation, the method further includes: The adaptive control results are monitored in real time, wherein the adaptive control results include outlet concentration information and removal rate information.

[0029] The adaptive control result is compared with a corresponding control prediction result to obtain a control deviation, wherein the control prediction result is extracted from the neighborhood control strategy set or a plurality of the secondary neighborhood control strategy sets.

[0030] If the control deviation is greater than the preset deviation constraint, feedback optimization is performed on the corresponding neighborhood control prediction model or multiple secondary neighborhood control prediction models based on the control deviation.

[0031] The present invention discloses a flue gas denitration adaptive control method for a particle denitration catalyst, comprising: interacting with a target low-temperature SCR processing system, collecting real-time control parameters and real-time operating parameters, generating benchmark operation data, and expanding the operating condition neighborhood based on the benchmark operation data; obtaining the historical processing log of the target low-temperature SCR processing system, and extracting data from the historical processing log in combination with the operating condition neighborhood expansion result to generate historical neighborhood control data, wherein the historical neighborhood control data includes multiple groups of historical control parameters and their corresponding control results; using the historical neighborhood control data as sample data, constructing and training a neighborhood control prediction model, wherein the prediction model takes the control parameters as input and outputs the corresponding control results; defining the neighborhood control prediction model as a loss function, taking the neighborhood control parameter space as the optimization target, performing parameter optimization, and generating a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result; and implementing adaptive control on the target low-temperature SCR processing system according to the neighborhood control strategy set. The present invention discloses a flue gas denitration adaptive control method for a particle denitration catalyst, which solves the technical problems of insufficient operating condition adaptability and real-time performance and low control accuracy, and achieves the technical effects of improving operating condition adaptability, responsiveness and control accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic flow diagram of a flue gas denitration adaptive control method of a granular denitration catalyst of the present invention; Figure 2 The present invention is a schematic diagram of a process for constructing and training a neighborhood control prediction model in a flue gas denitration adaptive control method of a particulate denitration catalyst. DETAILED DESCRIPTION

[0033] The technical solution provided in the embodiments of the present invention is to solve the technical problems of insufficient adaptability and real-time performance of working conditions and low control accuracy in the prior art. The overall idea adopted is as follows: First, the interactive target low-temperature SCR processing system obtains real-time control parameters and real-time operating parameters, outputs them as benchmark operating data, and based on the benchmark operating data, performs operating condition neighborhood expansion; then, the historical processing log of the target low-temperature SCR processing system is obtained, and data is extracted from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, wherein the historical neighborhood control data includes multiple groups of corresponding historical control parameters and historical control results; then, the historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, wherein the neighborhood control prediction model takes control parameters as input and predicts corresponding control results; then, the neighborhood control prediction model is defined as a loss function, and the neighborhood control parameter space is used as an optimization space to optimize control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained; finally, the target low-temperature SCR processing system is adaptively controlled according to the neighborhood control strategy set.

[0034] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0035] Embodiment 1 Figure 1 The present invention is a schematic flow chart of a flue gas denitration adaptive control method for a particulate denitration catalyst, wherein the method comprises: The interactive target low-temperature SCR processing system obtains real-time control parameters and real-time operating condition parameters, outputs them as benchmark operating data, and expands the operating condition neighborhood based on the benchmark operating data.

[0036] Specifically, first interact with the low-temperature SCR system to obtain control parameters and operating parameters in real time, and integrate the collected real-time control parameters and operating parameters into benchmark operating data as the basis for subsequent analysis and expansion. Among them, control parameters refer to the key operating parameters of the low-temperature SCR system, which are used to adjust the operating state of the denitrification process, including ammonia injection flow rate, catalyst temperature, air velocity, etc.; operating parameters refer to the operating environment of the low-temperature SCR system and the characteristics of the inlet exhaust gas, including exhaust gas flow rate, exhaust gas composition (such as NOx concentration, O2 concentration), exhaust gas temperature, etc.

[0037] Specifically, based on the benchmark operating data, the range of adjacent operating conditions is expanded, wherein the neighborhood refers to operating conditions that are close to the parameters of the current benchmark operating point (benchmark data) but slightly changed, and the neighborhood range is represented by a multidimensional space of multiple operating condition parameters.

[0038] In the low-temperature SCR treatment system, by acquiring control parameters and operating parameters in real time, outputting benchmark operating data and further expanding the operating condition neighborhood, it helps to improve the system's adaptability to complex operating conditions, improve overall control efficiency and processing effects, and lay the foundation for performance optimization, fault prediction and intelligent control.

[0039] In some embodiments, the interactive target low temperature SCR processing system obtains real-time control parameters and real-time operating condition parameters, outputs them as reference operating data, and performs operating condition neighborhood expansion based on the reference operating data, including: Taking the real-time operating condition parameters in the benchmark operation data as the neighborhood origin, density clustering is performed based on the historical processing log of the target low-temperature SCR processing system to produce a neighborhood clustering cluster; based on the neighborhood clustering cluster, neighborhood operating condition clustering clusters and neighborhood control clustering clusters are respectively extracted; based on the neighborhood operating condition clustering clusters and the neighborhood control clustering clusters, a dual-target neighborhood definition is performed to obtain a neighborhood control parameter space and a neighborhood operating condition parameter space.

[0040] Specifically, first, a multidimensional parameter space with multiple operating condition indicator dimensions as spatial dimensions is constructed. Then, the real-time operating condition parameters (such as exhaust gas flow, NOx concentration and exhaust gas temperature) in the benchmark operation data are used as the neighborhood origin to determine the central reference point of the cluster; then, the operating condition parameters and control parameter records are extracted from the historical processing log, and the radius parameter and the minimum number of samples are set to perform density-based clustering analysis to generate neighborhood clustering clusters. The neighborhood clustering clusters reflect the high-density areas in the historical processing logs that are similar to the current operating conditions. In other words, it can be considered that the neighborhood clustering clusters include multiple high-frequency operating conditions that are close to the current operating conditions.

[0041] Specifically, cluster classification is performed, and the distribution range of operating parameters (such as exhaust gas flow, NOx concentration, and temperature) is extracted from the clustering results as the neighborhood operating condition cluster; the distribution range of control parameters (such as ammonia injection flow and catalyst temperature) is extracted from the clustering results as the neighborhood control cluster, so as to distinguish between operating conditions and control parameters, map complex data into concise parameter ranges, provide a data basis for subsequent multi-objective analysis, and improve the efficiency of neighborhood analysis.

[0042] In some implementations, dual-objective neighborhood definition is performed according to the neighborhood operating condition cluster and the neighborhood control cluster, and a neighborhood control parameter space and a neighborhood operating condition parameter space are obtained, including: The control requirements of the target low-temperature SCR treatment system are obtained, and confidence interval constraints are defined according to the control requirements; according to the confidence interval constraints, the confidence space ranges of the neighborhood operating condition clustering clusters and the neighborhood control clustering clusters are determined, and the neighborhood control parameter space and the neighborhood operating condition parameter space are defined accordingly.

[0043] Specifically, the control requirements are the performance indicators and operating strategies that the target low-temperature SCR treatment system needs to meet according to the denitrification targets and operating conditions, including denitrification efficiency, ammonia escape rate, operating condition adaptability, etc. Based on the control requirements, the upper and lower limits of the operating condition parameters can be calculated through statistical analysis (such as normal distribution, percentiles) as confidence interval constraints, where the confidence interval is a range that may contain the true parameter value in a statistical sense. In this neighborhood definition, the confidence interval constraint is used to limit the parameters within a certain range of variation to ensure that the control requirements are met with a high probability.

[0044] Specifically, the operating parameter space represents the range of changes in the operating environment of the SCR system, which is determined by the confidence space of the neighborhood operating condition clustering clusters, ensuring that the denitrification efficiency target is met with a high probability; the control parameter space represents the range of possible control strategies under different operating conditions, which is determined by the confidence space of the neighborhood control clustering clusters, ensuring that a high-quality control parameter range is obtained.

[0045] By defining a dual-objective neighborhood based on confidence interval constraints and obtaining the neighborhood control parameter space and operating condition parameter space, the control parameter space and operating condition parameter space can be accurately delineated, which helps to adapt to a variety of operating environments and operating condition changes, and provides a reliable optimization basis for optimal control under complex conditions, thereby improving the operating efficiency and adaptability of the low-temperature SCR treatment system: A historical processing log of a target low-temperature SCR processing system is obtained, and data is extracted from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, wherein the historical neighborhood control data includes a plurality of groups of corresponding historical control parameters and historical control results.

[0046] Specifically, according to the operating condition neighborhood expansion result, historical processing log data of the corresponding range is extracted from the historical processing log of the target low-temperature SCR treatment system, thereby generating historical neighborhood control data. The historical neighborhood control data can be considered as a sample control record of the operating condition neighborhood expansion result, which reflects the correspondence between control parameters and control results within the range of the operating condition neighborhood expansion result, and is used to provide high-value historical data support for the intelligent control of the low-temperature SCR treatment system.

[0047] In some embodiments, a historical processing log of a target low-temperature SCR processing system is obtained, and data is extracted from the historical processing log according to an operating condition neighborhood expansion result to obtain historical neighborhood control data, including: Configure the neighborhood control parameter space as a first constraint; configure the real-time operating condition parameter as a second constraint; and extract the historical neighborhood control data from the historical processing log based on the first constraint and the second constraint, wherein the historical neighborhood control data includes multiple sets of corresponding historical control parameters and historical control results.

[0048] Specifically, the neighborhood control parameter space is used as the first constraint to ensure the consistency of the extracted historical control parameters with the current target control parameter space, supporting the optimization strategy under the current working conditions. The real-time working condition parameters are used as the second constraint to ensure that the acquired historical neighborhood control data are all control records under the same working conditions, which will help to accurately construct the corresponding relationship between control parameters and control results in the subsequent process.

[0049] Specifically, the historical operating parameters corresponding to the multiple groups of historical control parameters are consistent with or similar to the real-time operating parameters.

[0050] The historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, wherein the neighborhood control prediction model uses control parameters as input to predict corresponding control results.

[0051] Specifically, based on the extracted historical neighborhood control data, a neighborhood control prediction model is constructed and trained to achieve efficient mapping of control parameters to control results in the low-temperature SCR treatment system, thereby providing support for subsequent optimization of control strategies and improvement of denitrification efficiency.

[0052] In some embodiments, Figure 2 As shown, the historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, including: The historical neighborhood control data is traversed, and the historical control results are compared with the preset emission limit; if the historical control result is greater than or equal to the emission limit, the corresponding historical control parameters and the historical control results are stored in a negative sample set; if the historical control result is less than the emission limit, the corresponding historical control parameters and the historical control results are stored in a positive sample set; according to the negative sample set and the positive sample set, the neighborhood control prediction model constructed based on the neural network is trained.

[0053] Specifically, the neighborhood control prediction model is a machine learning model established with historical neighborhood control data as sample data. The input of the neighborhood control prediction model is control parameters (such as ammonia injection flow rate, catalyst temperature), and the corresponding output is control results (such as denitrification efficiency, ammonia escape rate).

[0054] Specifically, traverse the historical neighborhood control data, compare the historical control results with the preset emission limits, if the historical control results are greater than or equal to the emission limits, then store the corresponding historical control parameters and historical control results in the negative sample set; if the historical control results are less than the emission limits, then store the corresponding historical control parameters and historical control results in the positive sample set. By dividing the samples of the historical neighborhood control data, it helps to clarify the expected input-output relationship, thereby ensuring the accuracy of the prediction.

[0055] Specifically, select a suitable neural network architecture (such as multilayer perceptron, convolutional neural network, etc.), define the input layer, hidden layer and output layer, select the appropriate number of layers and neurons according to the data characteristics, and build a neighborhood control prediction model; then train the neighborhood control prediction model based on the negative sample set and the positive sample set, and use indicators such as accuracy, precision, recall rate and F1 score to evaluate the model performance.

[0056] Optionally, after completing the model training based on the negative sample set and the positive sample set, further obtain historical processing log data within the neighborhood operating condition parameter space range, obtain expanded sample data, and then perform enhanced training on the neighborhood control prediction model based on the expanded sample data, thereby improving the robustness and adaptability of the neighborhood control prediction model within the neighborhood operating condition parameter space range, thereby ensuring the prediction accuracy.

[0057] The neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is defined as an optimization space for optimizing control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained.

[0058] Specifically, the neighborhood control prediction model is used as the loss function, that is, the control effect corresponding to a set of control parameters is predicted by the neighborhood control prediction model, so as to evaluate the pros and cons of the set of control parameters; further, the neighborhood control parameter space is the optimized parameter space, which is used to determine the range of control parameters to be optimized.

[0059] Specifically, through optimization algorithms, such as genetic algorithms, particle swarm optimization, Bayesian optimization, etc., the control parameters are optimized in the neighborhood control parameter space, so as to determine the best control parameter combination under multiple working conditions, and the output is a neighborhood control strategy set, wherein the neighborhood control strategy set includes multiple control strategies, and each control strategy is associated with the predicted control effect and the corresponding calling working condition, so as to ensure that the neighborhood control strategy set can be correctly called.

[0060] Optionally, control parameter optimization is performed. First, a parameter set is initialized in the neighborhood control parameter space to generate an initial strategy set. Then, the control effects of multiple initial strategies in the initial strategy set are evaluated through a neighborhood control prediction model, and the initial strategy parameter set is iteratively adjusted until a predetermined number of iterations are completed or the optimal control effect obtained becomes stable. The corresponding initial strategy is then output as the neighborhood control strategy under the corresponding working condition.

[0061] Through the above steps, the neighborhood control prediction model can be used as the loss function to optimize the control strategy in the parameter space, thereby obtaining the optimal control strategy set that adapts to different working conditions.

[0062] In some embodiments, the neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is used as an optimization space to optimize control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained, before that, it also includes: Based on the historical processing log, multiple control indicators are extracted to generate a control indicator set; principal component analysis is performed on the control indicator set, and a key control indicator set is determined based on the principal component analysis result, wherein the key control indicator set includes multiple serialized key control indicators; the principal component contribution rates of the multiple key control indicators in the key control indicator set are normalized, and the indicator optimization step sizes of the multiple key control indicators are defined according to the normalization result.

[0063] Specifically, multiple control indicators are first extracted from the historical processing log to form an initial control indicator set. For example, the initial control indicator set includes specific operating parameters such as temperature, ammonia flow, gas velocity, oxygen content, and water vapor content. Then, principal component analysis is performed on the control indicator set to determine the contribution of each indicator to the total variance, and the control indicators corresponding to the principal components with higher contribution rates are selected to form a key control indicator set to reduce the control dimension.

[0064] Specifically, the proportional scaling method is used to normalize the principal component contribution rates of the key control indicators so that the total is 1, and then the optimization step size of each key control indicator is defined based on the normalized contribution rate. For example, the higher the contribution rate, the smaller the step size of the indicator, thereby ensuring fine-grained optimization of indicators with high contribution rates and improving optimization accuracy.

[0065] Specifically, the key control indicator set includes multiple serialized key control indicators, and the serialization is used to define the calling order of multiple key control indicators to ensure that the control indicators with a greater impact on the control effect are optimized first, which helps to further improve the parameter optimization efficiency.

[0066] Adaptive control of the target low-temperature SCR treatment system is performed according to the neighborhood control strategy set.

[0067] Specifically, the operating status of the system is monitored in real time, and the current operating data, including key parameters such as temperature, pressure, and flow, are collected. Then, according to the current operating conditions, a similarity algorithm or rule matching is used to select the most matching strategy from the neighborhood control strategy set, and the selected control strategy is applied to the target low-temperature SCR treatment system, thereby achieving adaptive control of the target low-temperature SCR treatment system and ensuring efficient operation of flue gas denitrification under different operating conditions.

[0068] In some embodiments, the method further comprises: Density clustering is performed based on the historical processing logs, and based on the clustering results, a plurality of secondary benchmark operation data are defined; the plurality of the secondary benchmark operation data are traversed to construct a corresponding plurality of secondary neighborhood control prediction models, and the plurality of the secondary neighborhood control prediction models are used as loss functions to optimize control parameters respectively to obtain a plurality of secondary neighborhood control strategy sets; the neighborhood control strategy set is combined with the plurality of the secondary neighborhood control strategy sets to perform adaptive control of the target low-temperature SCR processing system.

[0069] Specifically, the plurality of secondary reference operation data are a plurality of common reference operation data other than the reference operation data, corresponding to a plurality of common flue gas input states of the target low-temperature SCR treatment system.

[0070] Specifically, based on the same method principle as above for constructing the neighborhood control prediction model, multiple secondary neighborhood control prediction models are constructed for different operating conditions of the target low-temperature SCR treatment system. The multiple neighborhood control prediction models are used to adaptively reflect the mapping relationship between the control parameters and the control effects of the target low-temperature SCR treatment system in different neighborhood operating condition parameter spaces, thereby ensuring the control accuracy under different operating conditions, and at the same time helping to reduce the difficulty of constructing neighborhood control prediction models and neighborhood control prediction models.

[0071] Specifically, the original neighborhood control strategy set is combined with the secondary neighborhood control strategy set to form a more comprehensive control strategy library, thereby achieving multi-condition coverage of the target low-temperature SCR treatment system and improving the response capability and control accuracy of the target low-temperature SCR treatment system under different conditions.

[0072] In some embodiments, the method further comprises: Monitor the adaptive control results in real time, wherein the adaptive control results include outlet concentration information and removal rate information; compare the adaptive control results with corresponding control prediction results to obtain control deviations, wherein the control prediction results are extracted from the neighborhood control strategy set or multiple secondary neighborhood control strategy sets; if the control deviation is greater than a preset deviation constraint, perform feedback optimization on the corresponding neighborhood control prediction model or multiple secondary neighborhood control prediction models based on the control deviation.

[0073] Specifically, the adaptive control results are continuously monitored, including the outlet concentration information and the removal rate information; then, the actual control results are compared with the predicted results, and the control deviation is calculated to determine the difference between the actual effect and the expected effect, wherein the predicted results are extracted from the neighborhood control strategy set or the secondary neighborhood control strategy set.

[0074] Specifically, a preset deviation constraint is set to determine whether the deviation is within an acceptable range. If the control deviation is greater than the preset deviation constraint, it means that the corresponding neighborhood control prediction model or secondary neighborhood control prediction model has poor prediction accuracy and needs to be adjusted. Then, based on the control deviation, the corresponding neighborhood control prediction model or secondary neighborhood control prediction model is feedback optimized, the model parameters are adjusted, and the adaptability of the model to the corresponding working conditions is continuously improved.

[0075] Through the above-mentioned real-time monitoring and feedback optimization mechanism, the control performance of the target low-temperature SCR treatment system can be effectively improved to ensure that the system always maintains efficient operation under different operating conditions.

[0076] In summary, the flue gas denitration adaptive control method of a granular denitration catalyst provided by the present invention has the following technical effects: Through the interactive target low-temperature SCR processing system, real-time control parameters and real-time operating parameters are collected to generate benchmark operation data, and the operating condition neighborhood is expanded based on the benchmark operation data; the historical processing log of the target low-temperature SCR processing system is obtained, and the historical processing log is extracted in combination with the operating condition neighborhood expansion result to generate historical neighborhood control data, the historical neighborhood control data includes multiple groups of historical control parameters and their corresponding control results; the historical neighborhood control data is used as sample data to construct and train a neighborhood control prediction model, the prediction model takes the control parameters as input and outputs the corresponding control results; the neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is used as the optimization target, the parameters are optimized, and the neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is generated; the target low-temperature SCR processing system is adaptively controlled according to the neighborhood control strategy set. Thus, the technical effect of improving the adaptability to the operating conditions, improving the responsiveness and control accuracy is achieved.

[0077] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A flue gas denitration adaptive control method for a granular denitration catalyst, characterized in that: The method comprises: The interactive target low-temperature SCR treatment system obtains real-time control parameters and real-time operating condition parameters, outputs them as benchmark operating data, and expands the operating condition neighborhood based on the benchmark operating data; Obtaining a historical processing log of a target low-temperature SCR processing system, and extracting data from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, wherein the historical neighborhood control data includes a plurality of groups of corresponding historical control parameters and historical control results; Using the historical neighborhood control data as sample data, constructing and training a neighborhood control prediction model, wherein the neighborhood control prediction model uses control parameters as input to predict corresponding control results; The neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is defined as an optimization space for optimizing control parameters, and a neighborhood control strategy set corresponding to the operating condition neighborhood expansion result is obtained; Adaptive control of the target low-temperature SCR processing system is performed according to the neighborhood control strategy set.

2. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 1, characterized in that: The interactive target low temperature SCR treatment system obtains real-time control parameters and real-time operating condition parameters, outputs them as benchmark operating data, and performs operating condition neighborhood expansion based on the benchmark operating data, including: Taking the real-time operating condition parameters in the benchmark operation data as the neighborhood origin, density clustering is performed on the historical processing log of the target low-temperature SCR processing system to generate a neighborhood clustering cluster; Based on the neighborhood clustering clusters, extracting neighborhood operating condition clustering clusters and neighborhood control clustering clusters respectively; According to the neighborhood operating condition clustering cluster and the neighborhood control clustering cluster, a dual-objective neighborhood definition is performed to obtain a neighborhood control parameter space and a neighborhood operating condition parameter space.

3. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 2, characterized in that: According to the neighborhood operating condition clustering cluster and the neighborhood control clustering cluster, a dual-objective neighborhood definition is performed to obtain a neighborhood control parameter space and a neighborhood operating condition parameter space, including: Obtaining a control requirement of a target low-temperature SCR treatment system, and defining a confidence interval constraint according to the control requirement; According to the confidence interval constraint, the confidence space ranges of the neighborhood operating condition clustering cluster and the neighborhood control clustering cluster are determined, and the neighborhood control parameter space and the neighborhood operating condition parameter space are correspondingly defined.

4. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 3, characterized in that: Obtain the historical processing log of the target low-temperature SCR processing system, and extract data from the historical processing log according to the operating condition neighborhood expansion result to obtain historical neighborhood control data, including: Configuring the neighborhood control parameter space as a first constraint; Configuring the real-time operating condition parameter as a second constraint; Based on the first constraint and the second constraint, the historical neighborhood control data is extracted from the historical processing log, wherein the historical neighborhood control data includes a plurality of sets of corresponding historical control parameters and historical control results.

5. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 4, characterized in that: Using the historical neighborhood control data as sample data, a neighborhood control prediction model is constructed and trained, including: Traversing the historical neighborhood control data, and comparing the historical control results with the preset emission limits; If the historical control result is greater than or equal to the emission limit, the corresponding historical control parameter and the historical control result are stored in a negative sample set; If the historical control result is less than the emission limit, the corresponding historical control parameter and the historical control result are stored in a positive sample set; The neighborhood control prediction model constructed based on a neural network is trained according to the negative sample set and the positive sample set.

6. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 1, characterized in that: The neighborhood control prediction model is defined as a loss function, the neighborhood control parameter space is used as an optimization space to optimize the control parameters, and a neighborhood control strategy set corresponding to the working condition neighborhood expansion result is obtained. Before that, it also includes: Based on the historical processing log, extract multiple control indicators and generate a control indicator set; Performing principal component analysis on the control indicator set, and determining a key control indicator set based on the principal component analysis result, wherein the key control indicator set includes a plurality of serialized key control indicators; The principal component contribution rates of the plurality of key control indicators in the key control indicator set are normalized, and the indicator optimization step lengths of the plurality of key control indicators are defined according to the normalization result.

7. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 1, characterized in that: The method further comprises: Performing density clustering based on the historical processing logs, and defining a plurality of secondary benchmark operation data based on the clustering results; Traversing the plurality of secondary benchmark operation data, constructing a plurality of corresponding secondary neighborhood control prediction models, and using the plurality of secondary neighborhood control prediction models as loss functions, respectively optimizing control parameters to obtain a plurality of secondary neighborhood control strategy sets; The neighborhood control strategy set is combined with a plurality of the secondary neighborhood control strategy sets to perform adaptive control of a target low-temperature SCR treatment system.

8. The flue gas denitration adaptive control method of a particulate denitration catalyst according to claim 7, characterized in that: The method further comprises: Real-time monitoring of adaptive control results, wherein the adaptive control results include outlet concentration information and removal rate information; Comparing the adaptive control result with a corresponding control prediction result to obtain a control deviation, wherein the control prediction result is extracted from the neighborhood control strategy set or a plurality of the secondary neighborhood control strategy sets; If the control deviation is greater than the preset deviation constraint, feedback optimization is performed on the corresponding neighborhood control prediction model or multiple secondary neighborhood control prediction models based on the control deviation.

Images