Dynamic replacement method of catalyst in denitration device

Abstract

The present invention discloses a method for dynamically replacing catalysts in a denitration device, and relates to the technical field of flue gas denitration. The method comprises: collecting poisoning levels and dynamic replacement strategies of multiple catalyst replacement nodes of a denitration device; setting correlation scales, and obtaining pollutant concentration time series data in combination with pollutant monitoring logs; constructing a catalyst dynamic state transfer model based on concentration data and poisoning levels; collecting real-time pollutant data through a flue gas outlet multimodal sensor array, and loading it into a model to predict poisoning states; associating the predicted poisoning states with replacement strategies, generating real-time replacement strategies, and controlling the dynamic replacement of catalysts. Thus, the technical effects of improving the accuracy of replacement nodes, reducing operating costs, and stabilizing denitration efficiency are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of flue gas denitration, and in particular to a method for dynamically replacing a catalyst in a denitration device. Background Art

[0002] Denitrification equipment is an environmental protection equipment widely used in coal-fired power plants, industrial boilers and other equipment. Its main function is to convert nitrogen oxides (NOx) in flue gas into harmless nitrogen and water vapor through the action of catalysts, thereby achieving the purpose of reducing pollutant emissions. The catalyst is the core part of the denitrification device, and its performance directly affects the denitrification efficiency. However, the catalyst will gradually become poisoned during use, that is, its activity will gradually decrease due to the adsorption of toxic substances, which requires regular replacement of the catalyst.

[0003] At present, catalyst replacement is mainly carried out according to the equipment operation time or the degradation of catalyst performance. Generally speaking, after the equipment has been running for a period of time, the catalyst will be checked, and if the catalyst performance is found to have deteriorated, it will be replaced. Although this method is simple, it cannot accurately reflect the actual use status of the catalyst, and may cause the catalyst to be replaced too early or too late. There are technical problems such as high operating costs, lack of predictability, and poor denitrification effect. Summary of the invention

[0004] The present invention provides a method for dynamically replacing the catalyst in a denitration device to solve the technical problems in the prior art of high operating cost, lack of predictability and influence on the denitration effect, and achieves the technical effects of improving the accuracy of replacement nodes, reducing operating costs and stabilizing the denitration efficiency.

[0005] The method for dynamically replacing a catalyst in a denitration device provided by the present invention comprises:

[0006] By collecting local data on the denitrification device, multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes are obtained.

[0007] A data association scale is preset, and the data association scale and a plurality of catalyst replacement nodes are used as constraints to obtain a plurality of pollutant concentration time series data by calling a pollutant monitoring log.

[0008] Catalyst state transition analysis is performed based on the multiple pollutant concentration time series data and the multiple poisoning levels to complete the construction of a dynamic state transition model.

[0009] Real-time pollutant data is acquired through the multi-modal sensor array pre-configured at the flue gas outlet of the denitrification device.

[0010] The real-time pollutant data is loaded into the dynamic state transfer model to predict the poisoning state, and the predicted poisoning state is output.

[0011] Perform state correlation analysis on the predicted poisoning state and multiple dynamic replacement strategies, and output a real-time replacement strategy.

[0012] The real-time replacement strategy is adopted to control the denitration device to dynamically replace the catalyst.

[0013] In a feasible implementation, the real-time replacement strategy is used to control the denitration device to dynamically replace the catalyst, and the method includes:

[0014] Perform replacement rule analysis on the multiple dynamic replacement strategies to obtain a catalyst maintenance process; extract the catalyst unloading amount, high-temperature flue gas introduction time and catalyst loading amount in sequence from the real-time replacement strategy according to the catalyst maintenance process; unload the catalyst in the catalyst bed by opening the bottom unloading port of the denitrification device with the catalyst unloading amount as a constraint; after completing the catalyst unloading, control the flow rate and temperature of the gas covering the catalyst bed from the gas flow conduit in the denitrification device with the high-temperature flue gas introduction time as a constraint; after completing the thermal decomposition of the remaining catalyst in the catalyst bed with the high-temperature flue gas introduction time as a constraint, add the catalyst to the catalyst bed by opening the top feeding port of the denitrification device with the catalyst loading amount as a constraint.

[0015] In a feasible implementation, a catalyst state transition analysis is performed based on the multiple pollutant concentration time series data and multiple poisoning levels to complete the construction of a dynamic state transition model. The method includes:

[0016] Performing principal component analysis on pollutants of the denitrification device to obtain K kinds of sample pollutants; using the K kinds of sample pollutants to split and reorganize the multiple pollutant concentration time series data to obtain K concentration time series parameter sets; performing catalyst state transfer analysis based on the multiple poisoning levels and the K concentration time series parameter sets to construct and obtain K poisoning state transfer matrices; performing state transfer confidence inversion based on the K poisoning state transfer matrices and multiple poisoning levels to output K poisoning state confidence weights; and associatively storing the K kinds of sample pollutants, the K poisoning state confidence weights and the K poisoning state transfer matrices based on the knowledge graph to complete the construction of the dynamic state transfer model.

[0017] In a feasible implementation, catalyst state transition analysis is performed according to the multiple poisoning levels and K concentration time series parameter sets to construct and obtain K poisoning state transition matrices. The method includes:

[0018] A first concentration timing parameter set corresponding to a first sample pollutant is acquired from the K concentration timing parameter sets, wherein the first concentration timing parameter set includes a plurality of concentration timing parameters; a plurality of concentration change rates are calculated based on the plurality of concentration timing parameters; M concentration change intervals are obtained according to the plurality of concentration change rates; a catalyst state transfer characteristic analysis is performed according to the M concentration change intervals and a plurality of poisoning levels, and a first poisoning state transfer matrix is ​​constructed; and similarly, a catalyst state transfer analysis is performed according to the plurality of poisoning levels and the K concentration timing parameter sets, and the K poisoning state transfer matrices are constructed.

[0019] In a feasible implementation, a catalyst state transition characteristic analysis is performed according to the M concentration variation intervals and multiple poisoning levels to construct a first poisoning state transfer matrix. The method includes:

[0020] According to the mapping relationship between the multiple concentration time series parameters and the multiple poisoning levels, the multiple poisoning levels are classified into M groups of poisoning level information corresponding to the M concentration change intervals; the poisoning level recurrence frequency of the M groups of poisoning level information is calculated to obtain M groups of sample level probabilities; multiple sample state level spans are interactively obtained, and the first poisoning state transfer matrix is ​​constructed according to the multiple sample state level spans and the M concentration change intervals; the data of the first poisoning state transfer matrix is ​​filled using the M groups of sample level probabilities to complete the localization of the first poisoning state transfer matrix.

[0021] In a feasible implementation, based on the K poisoning state transfer matrices and multiple poisoning levels, state transfer confidence inversion is performed to output K poisoning state confidence weights. The method includes:

[0022] The multiple pollutant concentration time series data are input into the K poisoning state transfer matrices to obtain multiple groups of sample level data; state consistency analysis is performed on the multiple groups of sample level data and multiple poisoning levels to obtain K state consistency frequencies of the K types of sample pollutants; and weighted calculation is performed on the K state consistency frequencies to obtain the K poisoning state confidence weights.

[0023] In a feasible implementation, the real-time pollutant data is loaded into the dynamic state transfer model to predict the poisoning state, and the predicted poisoning state is output. The method includes:

[0024] A multimodal sensor array is configured at the flue gas inlet of the denitrification device according to the K kinds of sample pollutants; the real-time pollutant data is collected based on the multimodal sensor array, wherein the real-time pollutant data includes K pollution time-varying data; the K pollution time-varying data are mapped and analyzed using the K poisoning state transfer matrices in the dynamic state transfer model to obtain K poisoning state probability distributions; the K poisoning state probability distributions are updated using the K poisoning state confidence weights to obtain K poisoning state update distributions; state aggregation of the K poisoning state update distributions is performed with the span of the multiple sample state levels as a constraint to obtain multiple backup level probability coefficients; after serializing the multiple backup level probability coefficients, the predicted poisoning state is located at the multiple sample state levels according to the sequence maximum.

[0025] In a feasible implementation, a state correlation analysis is performed on the predicted poisoning state and a plurality of dynamic replacement strategies, and a real-time replacement strategy is outputted. The method includes:

[0026] Aggregate the multiple dynamic replacement strategies according to the multiple poisoning levels to obtain N groups of dynamic replacement strategies corresponding to the N sample poisoning levels; interactively obtain multiple maintenance costs and multiple steady-state durations of the multiple dynamic replacement strategies; evaluate the effectiveness of the N groups of dynamic replacement strategies based on the multiple maintenance costs and multiple steady-state durations, and obtain N backup replacement strategies based on the evaluation results; associate and store the N sample poisoning levels and the N backup replacement strategies to complete the construction of the catalyst replacement analysis model; synchronize the predicted poisoning state to the catalyst replacement analysis model for state association analysis, and call and output the real-time replacement strategy.

[0027] The present invention discloses a method for dynamically replacing a catalyst in a denitration device, including: performing local data collection on the denitration device, obtaining multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes; presetting a data association scale, and using the data association scale and multiple catalyst replacement nodes as constraints, obtaining multiple pollutant concentration time series data in the pollutant monitoring log call; performing catalyst state transfer analysis based on multiple pollutant concentration time series data and multiple poisoning levels, and completing the construction of a dynamic state transfer model; acquiring real-time pollutant data through a multimodal sensor array pre-configured at the flue gas outlet of the denitration device; loading the real-time pollutant data into a dynamic state transfer model for poisoning state prediction, and outputting a predicted poisoning state; performing state association analysis on the predicted poisoning state and multiple dynamic replacement strategies, and outputting a real-time replacement strategy; and using a real-time replacement strategy to control the denitration device to dynamically replace the catalyst. The method for dynamically replacing a catalyst in a denitration device disclosed by the present invention solves the technical problems of high operating cost, lack of predictability, and influence on the denitration effect, and achieves the technical effects of improving the accuracy of replacement nodes, reducing operating costs, and stabilizing denitration efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic flow diagram of a method for dynamically replacing a catalyst in a denitration device of the present invention;

[0029] Figure 2 It is a schematic flow chart of the construction of a dynamic state transfer model in the method for dynamically replacing the catalyst in the denitration device of the present invention. DETAILED DESCRIPTION

[0030] The technical solution provided in the embodiments of the present invention is to solve the technical problems of high operating cost, lack of predictability and poor denitrification effect in the prior art. The overall idea adopted is as follows:

[0031] First, by collecting local data on the denitrification device, multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes are obtained; then, a data association scale is preset, and with the data association scale and multiple catalyst replacement nodes as constraints, multiple pollutant concentration time series data are called in the pollutant monitoring log; then, based on the multiple pollutant concentration time series data and multiple poisoning levels, catalyst state transfer analysis is performed to complete the construction of a dynamic state transfer model; then, real-time pollutant data is collected through a pre-configured multimodal sensor array at the flue gas outlet of the denitrification device; then, the real-time pollutant data is loaded into the dynamic state transfer model for poisoning state prediction, and a predicted poisoning state is output; finally, a state association analysis is performed on the predicted poisoning state and multiple dynamic replacement strategies, a real-time replacement strategy is output, and the real-time replacement strategy is used to control the denitrification device to perform dynamic replacement of the catalyst.

[0032] 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.

[0033] Example

[0034] Figure 1 The present invention is a schematic flow diagram of a method for dynamically replacing a catalyst in a denitration device, wherein the method comprises:

[0035] By collecting local data on the denitrification device, multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes are obtained.

[0036] Specifically, in a denitration device, the use of the catalyst directly affects the removal efficiency of the device. Therefore, it is necessary to first collect local data, determine multiple catalyst replacement nodes, and obtain the poisoning level and dynamic replacement strategy; among them, the catalyst poisoning node refers to the moment when the catalyst performance is significantly reduced or loses catalytic activity, which is usually calibrated by the catalyst's performance attenuation, surface poisoning or reduced reaction efficiency.

[0037] Specifically, the poisoning level is determined based on the pollutant concentration data. For example, in mild poisoning, the change in pollutant concentration is relatively gentle, the concentration is relatively stable, but there is a small upward trend; in moderate poisoning, the pollutant concentration changes faster, showing a more obvious upward trend; in severe poisoning, the pollutant concentration rises sharply, the concentration value changes rapidly, and the catalyst reaction rate decreases rapidly.

[0038] Specifically, after the catalyst poisoning node is determined, the severity of the catalyst poisoning is analyzed by further acquiring the pollutant concentration and calculating the change in the pollutant concentration. In other words, the rate of change of the pollutant concentration within a certain time window is compared with the preset rate threshold. If the rate exceeds a certain critical value, it is determined to be a mild, moderate or severe poisoning level.

[0039] Specifically, different poisoning levels correspond to different dynamic replacement strategies for the catalyst. If the pollutant concentration changes at a faster rate, it means that the catalyst is declining more seriously, and it may be necessary to shorten the catalyst replacement cycle (i.e., a higher replacement amount); conversely, if the concentration changes slowly, the replacement cycle can be appropriately extended.

[0040] A data association scale is preset, and the data association scale and a plurality of catalyst replacement nodes are used as constraints to obtain a plurality of pollutant concentration time series data by calling a pollutant monitoring log.

[0041] Specifically, the preset data association scale and catalyst replacement node are constraints for data collection. The data association scale refers to the setting of the time window length or time step of the pollutant concentration monitoring data, which determines how to obtain data at different time points or time periods during the data analysis process; the catalyst replacement node defines the specific timestamp corresponding to the obtained pollutant monitoring log.

[0042] Specifically, the pollutant concentration time series data is the pollutant concentration values ​​at multiple catalyst replacement nodes within the time window of the preset data association scale, that is, the preset time points before and after the node, which are saved in the log of the pollutant monitoring device; exemplary data include NO X If the catalyst replacement node is at time T and the data association scale is set to 5 minutes, the relevant pollutant concentration time series data will be called from T-5 minutes to T+5 minutes.

[0043] Through the above steps, the pollutant concentration time series data can be accurately called. The obtained time series data reflects the change of pollutant concentration before and after the catalyst replacement node, which helps to understand the change characteristics of pollutant concentration under different poisoning levels.

[0044] Catalyst state transition analysis is performed based on the multiple pollutant concentration time series data and the multiple poisoning levels to complete the construction of a dynamic state transition model.

[0045] Specifically, the dynamic state transition model is constructed based on the time series data of pollutant concentration and the changes in catalyst poisoning level, and is used to describe the transition process of the catalyst from one state to another. In other words, the state transition is not random, but is triggered based on certain specific conditions; illustratively, the change in catalyst poisoning level can be judged by the rate of change of pollutant concentration, periodic changes, etc., that is, the change of catalyst state requires a specific concentration threshold or rate condition.

[0046] This dynamic state transition model can analyze the health of the catalyst in real time and predict the future performance of the catalyst, thus providing a basis for replacement strategies and maintenance decisions.

[0047] In some embodiments, Figure 2 As shown, based on the multiple pollutant concentration time series data and multiple poisoning levels, catalyst state transition analysis is performed to complete the construction of a dynamic state transition model. The method includes:

[0048] Performing principal component analysis on pollutants of the denitrification device to obtain K kinds of sample pollutants; using the K kinds of sample pollutants to split and reorganize the multiple pollutant concentration time series data to obtain K concentration time series parameter sets; performing catalyst state transfer analysis based on the multiple poisoning levels and the K concentration time series parameter sets to construct and obtain K poisoning state transfer matrices; performing state transfer confidence inversion based on the K poisoning state transfer matrices and multiple poisoning levels to output K poisoning state confidence weights; and associatively storing the K kinds of sample pollutants, the K poisoning state confidence weights and the K poisoning state transfer matrices based on the knowledge graph to complete the construction of the dynamic state transfer model.

[0049] Specifically, first, a principal component analysis of pollutants is performed on the denitrification device to obtain K types of sample pollutants. Principal component analysis is used to extract the key pollutants that mainly affect catalyst poisoning from complex pollutant data, that is, the top K pollutants that contribute most to catalyst poisoning, thereby reducing data redundancy, focusing on key pollutants and improving the efficiency and accuracy of subsequent analysis.

[0050] Specifically, the original pollutant concentration time series data are split into K independent concentration time series data sets according to the results of principal component analysis, so as to establish an independent time series analysis framework for each pollutant, which helps to enhance the separation and understanding of the impact of complex pollutants; among which, each concentration time series parameter set corresponds to the concentration change of one of the above-mentioned sample pollutants (i.e., the key pollutant).

[0051] Specifically, based on multiple poisoning levels and K concentration time series parameter sets, K poisoning state transfer matrices are constructed respectively, where the current pollutant concentration determines the current catalyst state (i.e., poisoning level), and the current state and the changing trend of the pollutant concentration jointly determine the next state; then, according to the state transfer matrix, the confidence of the impact of each pollutant on the catalyst poisoning state is calculated, and the obtained multiple impact confidences are normalized to obtain the weight of the quantitative pollutant's impact on the catalyst poisoning state, and the value range of the weight is 0,1; finally, through the knowledge graph technology, the pollutants, poisoning state transfer matrix and confidence weights are stored in a structured form for query and analysis, where the nodes of the knowledge graph represent entities such as pollutant types, poisoning levels, and state transfer matrices, the edges represent the relationship between pollutants and catalyst poisoning states, and the attributes are used to store information such as node and edge weights and state transition probabilities.

[0052] In some implementations, performing catalyst state transition analysis according to the multiple poisoning levels and K concentration time series parameter sets to construct and obtain K poisoning state transition matrices, the method includes:

[0053] A first concentration timing parameter set corresponding to a first sample pollutant is acquired from the K concentration timing parameter sets, wherein the first concentration timing parameter set includes a plurality of concentration timing parameters; a plurality of concentration change rates are calculated based on the plurality of concentration timing parameters; M concentration change intervals are obtained according to the plurality of concentration change rates; a catalyst state transfer characteristic analysis is performed according to the M concentration change intervals and a plurality of poisoning levels, and a first poisoning state transfer matrix is ​​constructed; and similarly, a catalyst state transfer analysis is performed according to the plurality of poisoning levels and the K concentration timing parameter sets, and the K poisoning state transfer matrices are constructed.

[0054] Specifically, first, the concentration time series parameters of the first sample pollutant are collected from K concentration time series parameter sets, wherein the first sample pollutant is any one of the K sample pollutants. Exemplarily, the data format of the first concentration time series parameter set is {C 1， C 2， ..., C t}, where C t is the pollutant concentration at time t.

[0055] Preferably, the first sample pollutant is a sample pollutant having the highest principal component contribution among the K sample pollutants.

[0056] Specifically, based on the acquired first concentration time series parameter set, the average concentration change rate in the corresponding multiple time intervals is calculated and output as multiple concentration change rates. Then, according to the distribution of the concentration change rates, they are divided into M concentration change intervals, wherein the M concentration change intervals can be based on equal interval division, quantile division (such as 25%, 50%, 75% quantiles) or cluster division.

[0057] Furthermore, the concentration variation range of the pollutant is associated with the poisoning level of the catalyst, and the transition probability from one poisoning level to another is determined by analyzing the state transition characteristics, thereby determining the first poisoning state transfer matrix; then, the same process as the above-mentioned construction of the first poisoning state transfer matrix is ​​adopted to traverse the K types of sample pollutants, and construct corresponding K poisoning state transfer matrices according to the corresponding K concentration time series parameter sets and multiple poisoning levels.

[0058] The K poisoning state transfer matrices constructed through the above process obtain the dynamic impact relationship of various concentration states of various pollutants on the catalyst under different operating conditions, providing a basis for subsequent analysis of the synergistic impact of multiple pollutants.

[0059] In some implementations, performing catalyst state transition characteristic analysis according to the M concentration variation intervals and the multiple poisoning levels to construct a first poisoning state transition matrix, the method includes:

[0060] According to the mapping relationship between the multiple concentration time series parameters and the multiple poisoning levels, the multiple poisoning levels are classified into M groups of poisoning level information corresponding to the M concentration change intervals; the poisoning level recurrence frequency of the M groups of poisoning level information is calculated to obtain M groups of sample level probabilities; multiple sample state level spans are interactively obtained, and the first poisoning state transfer matrix is ​​constructed according to the multiple sample state level spans and the M concentration change intervals; the data of the first poisoning state transfer matrix is ​​filled using the M groups of sample level probabilities to complete the localization of the first poisoning state transfer matrix.

[0061] Specifically, first, based on the mapping relationship between the multiple concentration time series parameters and the multiple poisoning levels determined above, the multiple poisoning levels are respectively mapped to M concentration change intervals to obtain M groups of poisoning level information, that is, the poisoning level classifications corresponding to the multiple concentration time series parameters are added to the data group corresponding to the concentration change interval to which the pollutant concentration belongs (i.e., M groups of poisoning level information), wherein the poisoning levels in each group of poisoning level information may appear repeatedly.

[0062] Then, based on the statistical analysis method, the occurrence frequency of different poisoning levels in each group of poisoning level information is obtained, and the occurrence frequency is calculated accordingly to obtain the M group of sample level probabilities. Specifically, each group of sample level probabilities includes the occurrence frequencies of multiple poisoning levels, that is, the concentration change range corresponding to the group of sample level probabilities corresponds to the confidence level of the poisoning level.

[0063] Specifically, the multiple sample state level spans correspond to the state transitions between the multiple poisoning levels. For example, if the poisoning levels include 1-3 levels, the corresponding multiple sample state level spans include L0→L0, L0→L1, L0→L2, and L0→L3.

[0064] Specifically, according to multiple sample state level spans and M concentration change intervals, for each concentration change interval, fill the elements of the transfer matrix according to the sample level probability and the multiple sample state level spans to construct a first poisoning state transfer matrix, wherein the element P in the first poisoning state transfer matrix P is ij Indicates that from state S i Transfer to state S j For example, if the first sample pollutant is NO x , the corresponding relationship between the sample level probability and the span of multiple sample state levels is as follows:

[0065] NOx concentration change rate (ppm / h) L0→L0 L0→L1 L0→L2 L0→L3 0.0-0.2 0.8 0.2 0 0 0.2-0.5 0.6 0.3 0.1 0 0.5-1.0 0.4 0.4 0.2 0 ＞1.0 0.2 0.3 0.3 0.2

[0066] Then the first poisoning state transfer matrix P can be expressed as:

[0067] ;

[0068] Among them, the row corresponds to NO x The intervals of concentration change rates (0.0−0.2, 0.2−0.5, 0.5−1.0, >1.0); the columns correspond to multiple sample state level spans (L0→L0, L0→L1, L0→L2, L0→L3).

[0069] In some implementations, performing state transition confidence inversion based on the K poisoning state transfer matrices and multiple poisoning levels to output K poisoning state confidence weights, the method includes:

[0070] The multiple pollutant concentration time series data are input into the K poisoning state transfer matrices to obtain multiple groups of sample level data; state consistency analysis is performed on the multiple groups of sample level data and multiple poisoning levels to obtain K state consistency frequencies of the K types of sample pollutants; and weighted calculation is performed on the K state consistency frequencies to obtain the K poisoning state confidence weights.

[0071] Specifically, first, the pollutant concentration time series data is input into K poisoning state transfer matrices to gradually simulate the changes in the poisoning state of the catalyst. Each transfer matrix generates a set of sample level data corresponding to each transfer matrix, indicating the changes in the poisoning level of the catalyst over time under the transfer model.

[0072] Specifically, state consistency refers to the degree of match between sample level data and actual poisoning level. By comparing and counting the number of times that the sample level data corresponding to each poisoning state transfer matrix is ​​consistent with the actual poisoning level, the state consistency frequency is taken as the state consistency frequency. The higher the state consistency frequency, the more accurately the transfer matrix can reflect the actual poisoning state change of the catalyst.

[0073] Furthermore, the state consistency frequency of each state transfer matrix is ​​weighted and calculated to obtain the corresponding poisoning state confidence weight, which reflects the credibility of each transfer matrix in describing the catalyst poisoning state.

[0074] Through the above method, based on K poisoning state transfer matrices and multiple poisoning levels, the state transfer confidence inversion can be performed using the pollutant concentration time series data, the consistency between the sample level data and the actual poisoning level can be comprehensively considered, and K poisoning state confidence weights can be output through weighted calculation, which helps to improve the reliability of the analysis results.

[0075] Real-time pollutant data is acquired through the multi-modal sensor array pre-configured at the flue gas outlet of the denitrification device.

[0076] Specifically, the multimodal sensor array is a sensor network composed of multiple types of sensors that can simultaneously monitor multiple pollutant parameters at the flue gas outlet and other related environmental variables, such as nitrogen oxides (NO x ), sulfur dioxide (SO2), ammonia (NH3), oxygen (O2), etc., thereby providing comprehensive information for the prediction of catalyst poisoning status.

[0077] The real-time pollutant data is loaded into the dynamic state transfer model to predict the poisoning state, and the predicted poisoning state is output.

[0078] In some embodiments, the real-time pollutant data is loaded into the dynamic state transfer model to predict the poisoning state, and the predicted poisoning state is outputted. The method includes:

[0079] A multimodal sensor array is configured at the flue gas inlet of the denitrification device according to the K kinds of sample pollutants; the real-time pollutant data is collected based on the multimodal sensor array, wherein the real-time pollutant data includes K pollution time-varying data; the K pollution time-varying data are mapped and analyzed using the K poisoning state transfer matrices in the dynamic state transfer model to obtain K poisoning state probability distributions; the K poisoning state probability distributions are updated using the K poisoning state confidence weights to obtain K poisoning state update distributions; state aggregation of the K poisoning state update distributions is performed with the span of the multiple sample state levels as a constraint to obtain multiple backup level probability coefficients; after serializing the multiple backup level probability coefficients, the predicted poisoning state is located at the multiple sample state levels according to the sequence maximum.

[0080] Specifically, first, the K collected time-varying pollutant data are input into the dynamic state transfer model, and K poisoning state transfer matrices are used for mapping analysis to generate the probability distribution of each poisoning state, wherein the probability distribution of each poisoning state includes multiple probability parameters of multiple poisoning levels, indicating the possibility of the catalyst being in different poisoning levels; then, based on the K poisoning state confidence weights defined above, the probability parameters of each level of each poisoning state probability distribution are weighted and adjusted to make it more consistent with the actual state of the catalyst, and the output result is K poisoning state update distributions (closer to the actual poisoning state distribution after confidence weight correction); then, the K poisoning state update distributions are state aggregated to obtain multiple backup level probability coefficients, indicating the comprehensive probability of the catalyst at different poisoning levels; finally, the multiple backup level probability coefficients are arranged in the order of poisoning levels to form a state sequence, and the level with the largest probability value is found in the state sequence as the predicted poisoning state of the catalyst, that is, the predicted poisoning level of the catalyst under the current pollutant concentration conditions.

[0081] Through the above steps, the real-time pollutant data collected by the multimodal sensor array is combined with the dynamic state transition model and the poisoning state confidence weight to achieve accurate prediction of the catalyst poisoning state.

[0082] Perform state correlation analysis on the predicted poisoning state and multiple dynamic replacement strategies, and output a real-time replacement strategy.

[0083] Specifically, based on the predicted poisoning state, correlation analysis is performed to match the corresponding most suitable dynamic replacement strategy, and the output is a real-time replacement strategy.

[0084] In some embodiments, performing state correlation analysis on the predicted poisoning state and multiple dynamic replacement strategies and outputting a real-time replacement strategy, the method includes:

[0085] Aggregate the multiple dynamic replacement strategies according to the multiple poisoning levels to obtain N groups of dynamic replacement strategies corresponding to the N sample poisoning levels; interactively obtain multiple maintenance costs and multiple steady-state durations of the multiple dynamic replacement strategies; evaluate the effectiveness of the N groups of dynamic replacement strategies based on the multiple maintenance costs and multiple steady-state durations, and obtain N backup replacement strategies based on the evaluation results; associate and store the N sample poisoning levels and the N backup replacement strategies to complete the construction of the catalyst replacement analysis model; synchronize the predicted poisoning state to the catalyst replacement analysis model for state association analysis, and call and output the real-time replacement strategy.

[0086] Specifically, first, according to the multiple poisoning levels of the catalyst, the corresponding dynamic replacement strategies are aggregated to establish a preliminary association between the poisoning level and the replacement strategy; then, for each group of dynamic replacement strategies, through interactive or historical data analysis, the parameters involving maintenance costs and steady-state duration are obtained, such as catalyst replacement costs, indirect costs of reduced operating efficiency, and the length of time the catalyst can maintain normal operation after implementing the replacement strategy.

[0087] Furthermore, for each group of dynamic replacement strategies, a weighted calculation method is used to comprehensively evaluate the relative importance of maintenance cost and steady-state duration, and multiple groups of dynamic replacement strategies are sorted according to the effectiveness scores. The strategy with the highest score is selected as the best strategy for the current poisoning level, and then N sample poisoning levels and N backup replacement strategies are obtained and stored in association.

[0088] Specifically, the predicted poisoning state is synchronized to the catalyst replacement analysis model, and state correlation analysis is performed, and the corresponding backup replacement strategy is extracted and output as a real-time replacement strategy. Through the above steps, both economy and operational stability can be taken into account, and the correlation analysis between the predicted poisoning state of the catalyst and the dynamic replacement strategy can be realized, and the optimal replacement strategy can be output in real time.

[0089] The real-time replacement strategy is adopted to control the denitration device to dynamically replace the catalyst.

[0090] Specifically, the catalyst of the denitration device is dynamically replaced according to the catalyst replacement cycle and replacement ratio (quantity) defined in the real-time replacement strategy.

[0091] In some embodiments, the real-time replacement strategy is used to control the denitration device to dynamically replace the catalyst, and the method includes:

[0092] Perform replacement rule analysis on the multiple dynamic replacement strategies to obtain a catalyst maintenance process; extract the catalyst unloading amount, high-temperature flue gas introduction time and catalyst loading amount in sequence from the real-time replacement strategy according to the catalyst maintenance process; unload the catalyst in the catalyst bed by opening the bottom unloading port of the denitrification device with the catalyst unloading amount as a constraint; after completing the catalyst unloading, control the flow rate and temperature of the gas covering the catalyst bed from the gas flow conduit in the denitrification device with the high-temperature flue gas introduction time as a constraint; after completing the thermal decomposition of the remaining catalyst in the catalyst bed with the high-temperature flue gas introduction time as a constraint, add the catalyst to the catalyst bed by opening the top feeding port of the denitrification device with the catalyst loading amount as a constraint.

[0093] Specifically, first, multiple dynamic replacement strategies are analyzed to determine the maintenance and replacement process of the catalyst under the target scenario. Among them, the poisoning caused by different types of pollutants corresponds to different maintenance processes. The catalyst maintenance process involves variables such as the catalyst unloading amount, the high-temperature flue gas introduction time, and the catalyst loading amount.

[0094] For example, when the ammonium sulfate produced by the long-term operation of the reactor affects the efficiency, high-temperature flue gas can be introduced into the reactor for thermal analysis. The ammonium sulfate can be decomposed at high temperature, and the denitrification catalyst can restore the denitrification efficiency. When the catalyst is deactivated by alkali metals or other pollutants and needs to be replaced with a new catalyst, the lower discharge port can be opened to discharge the deactivated catalyst. This process does not affect the normal operation of other devices in the denitrification system, and online regeneration and online replacement are achieved.

[0095] Specifically, the catalyst is dynamically replaced. First, the bottom discharge port of the denitrification device is controlled to gradually discharge the catalyst according to the set catalyst discharge amount. During this process, dynamic monitoring means (such as weight sensors or volume measuring devices) are used to monitor the discharge progress in real time to ensure that the discharge amount is consistent with the set value. After the catalyst is unloaded, the flow rate and temperature of the flue gas are set based on the duration of high-temperature flue gas introduction, and the flue gas flow rate and temperature in the gas flow pipeline are adjusted by controlling valves and heating devices to ensure that the high-temperature flue gas evenly covers the catalyst bed and that the thermal analysis process meets the set duration and temperature requirements. Furthermore, after the thermal analysis is completed, the catalyst loading amount is used as a constraint to control the top feeding port of the denitrification device to gradually add new catalyst according to the set catalyst loading amount, and the filling height or weight of the catalyst bed is monitored in real time to ensure that the loading amount is consistent with the set value to complete the loading of the catalyst.

[0096] By analyzing the rules of dynamic replacement strategies, extracting the catalyst maintenance process, and completing the catalyst unloading, high-temperature flue gas introduction and catalyst loading operations in sequence, it can effectively respond to the catalyst replacement needs under different poisoning states, optimize the operating efficiency and maintenance cost of the denitrification device, and provide guarantee for the long-term stable operation of the equipment.

[0097] In summary, the method for dynamically replacing the catalyst in the denitration device provided by the present invention has the following technical effects:

[0098] By collecting local data of the denitrification device, multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes are obtained; the data association scale is preset, and the data association scale and multiple catalyst replacement nodes are used as constraints to obtain multiple pollutant concentration time series data in the pollutant monitoring log call; based on multiple pollutant concentration time series data and multiple poisoning levels, the catalyst state transfer analysis is performed to complete the construction of the dynamic state transfer model; the real-time pollutant data is obtained through the multi-modal sensor array pre-configured at the flue gas outlet of the denitrification device; the real-time pollutant data is loaded into the dynamic state transfer model for poisoning state prediction, and the predicted poisoning state is output; the predicted poisoning state and multiple dynamic replacement strategies are analyzed for state association, and the real-time replacement strategy is output; the real-time replacement strategy is used to control the denitrification device to perform dynamic replacement of the catalyst. Thus, the technical effects of improving the accuracy of replacement nodes, reducing operating costs, and stabilizing denitrification efficiency are achieved.

[0099] 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 method for dynamically replacing a catalyst in a denitration device, characterized in that: The method comprises: By collecting local data on the denitration device, multiple poisoning levels and multiple dynamic replacement strategies at multiple catalyst replacement nodes are obtained; A data association scale is preset, and multiple pollutant concentration time series data are obtained by calling a pollutant monitoring log with the data association scale and multiple catalyst replacement nodes as constraints; Performing catalyst state transition analysis based on the multiple pollutant concentration time series data and multiple poisoning levels to complete the construction of a dynamic state transition model; Acquire real-time pollutant data through a multi-modal sensor array pre-configured at the flue gas outlet of the denitrification device; Loading the real-time pollutant data into the dynamic state transfer model to predict the poisoning state, and outputting the predicted poisoning state; Performing state correlation analysis on the predicted poisoning state and multiple dynamic replacement strategies, and outputting a real-time replacement strategy; The real-time replacement strategy is used to control the denitration device to dynamically replace the catalyst; Based on the multiple pollutant concentration time series data and multiple poisoning levels, catalyst state transition analysis is performed to complete the construction of a dynamic state transition model. The method includes: Performing principal component analysis on the denitration device to obtain K types of sample pollutants; Using the K types of sample pollutants to split and reorganize the multiple pollutant concentration time series data to obtain K concentration time series parameter sets; Performing catalyst state transition analysis according to the multiple poisoning levels and K concentration time series parameter sets to construct and obtain K poisoning state transition matrices; Perform state transfer confidence inversion based on the K poisoning state transfer matrices and multiple poisoning levels, and output K poisoning state confidence weights; The K sample pollutants, K poisoning state confidence weights and K poisoning state transfer matrices are stored in association based on the knowledge graph to complete the construction of the dynamic state transfer model.

2. The method for dynamically replacing the catalyst in the denitration device according to claim 1, characterized in that: The real-time replacement strategy is used to control the denitration device to dynamically replace the catalyst, and the method includes: Analyzing the replacement rules of the multiple dynamic replacement strategies to obtain a catalyst maintenance process; According to the catalyst maintenance process, the catalyst discharge amount, the high-temperature flue gas introduction time and the catalyst loading amount are sequentially extracted from the real-time replacement strategy; Taking the catalyst discharge amount as a constraint, the catalyst in the catalyst bed is discharged by opening the bottom discharge port of the denitration device; After the catalyst is unloaded, the flow rate and temperature of the gas covering the catalyst bed from the gas flow pipeline in the denitration device are controlled based on the duration of the high-temperature flue gas introduction; After the thermal decomposition of the remaining catalyst in the catalyst bed is completed under the constraint of the high-temperature flue gas introduction time, the catalyst is added to the catalyst bed by opening the top feeding port of the denitration device under the constraint of the catalyst loading amount.

3. The method for dynamically replacing the catalyst in the denitration device according to claim 1, characterized in that: Performing catalyst state transition analysis according to the multiple poisoning levels and K concentration time series parameter sets to construct and obtain K poisoning state transition matrices, the method comprising: Acquire a first concentration time series parameter set corresponding to a first sample pollutant from the K concentration time series parameter sets, wherein the first concentration time series parameter set includes a plurality of concentration time series parameters; Calculate and obtain multiple concentration change rates based on the multiple concentration time series parameters; Obtaining M concentration change intervals according to the multiple concentration change rates; Performing catalyst state transition characteristic analysis according to the M concentration variation intervals and the multiple poisoning levels to construct a first poisoning state transfer matrix; By analogy, catalyst state transfer analysis is performed according to the multiple poisoning levels and K concentration time series parameter sets to construct and obtain the K poisoning state transfer matrices.

4. The method for dynamically replacing the catalyst in the denitration device according to claim 3, characterized in that: Performing catalyst state transition characteristic analysis according to the M concentration variation intervals and multiple poisoning levels to construct a first poisoning state transition matrix, the method comprising: According to the mapping relationship between the multiple concentration time series parameters and the multiple poisoning levels, the multiple poisoning levels are classified into M groups of poisoning level information corresponding to the M concentration change intervals; Calculating the frequency of recurrence of poisoning levels for the M group of poisoning level information to obtain the level probability of the M group of samples; Interactively obtain multiple sample state level spans, and construct the first poisoning state transfer matrix according to the multiple sample state level spans and M concentration change intervals; The M groups of sample level probabilities are used to fill in data of the first poisoning state transfer matrix, thereby completing the localization of the first poisoning state transfer matrix.

5. The method for dynamically replacing the catalyst in the denitration device according to claim 3, characterized in that: Based on the K poisoning state transfer matrices and multiple poisoning levels, state transfer confidence inversion is performed to output K poisoning state confidence weights, and the method includes: Inputting the plurality of pollutant concentration time series data into the K poisoning state transfer matrices to obtain a plurality of groups of sample level data; Performing state consistency analysis on the multiple groups of sample level data and multiple poisoning levels to obtain K state consistency frequencies of the K types of sample pollutants; A weighted calculation is performed on the K state consistency frequencies to obtain the K poisoning state confidence weights.

6. The method for dynamically replacing the catalyst in the denitration device according to claim 4, characterized in that: The real-time pollutant data is loaded into the dynamic state transfer model to predict the poisoning state, and the predicted poisoning state is output. The method includes: According to the K types of sample pollutants, a multi-modal sensor array is configured at the flue gas inlet of the denitrification device; Collecting the real-time pollutant data based on the multimodal sensor array, wherein the real-time pollutant data includes K pollution time-varying data; The K poisoning state transfer matrices in the dynamic state transfer model are used to map and analyze the K contaminated time-varying data to obtain K poisoning state probability distributions; Using the K poisoning state confidence weights to perform data update on the K poisoning state probability distributions to obtain K poisoning state update distributions; Performing state aggregation of the K poisoning state update distributions with the multiple sample state level spans as constraints to obtain multiple standby level probability coefficients; After the plurality of backup level probability coefficients are serialized, the predicted poisoning state is located at the plurality of sample state levels according to the sequence maximum value.

7. The method for dynamically replacing the catalyst in the denitration device according to claim 1, characterized in that: Performing state correlation analysis on the predicted poisoning state and multiple dynamic replacement strategies, and outputting a real-time replacement strategy, the method includes: Aggregating the multiple dynamic replacement strategies according to the multiple poisoning levels to obtain N groups of dynamic replacement strategies corresponding to the poisoning levels of the N samples; interactively obtaining multiple maintenance costs and multiple steady-state durations of the multiple dynamic replacement strategies; Evaluate the effectiveness of the N groups of dynamic replacement strategies based on the multiple maintenance costs and the multiple steady-state durations, and obtain N standby replacement strategies based on the evaluation results; The N sample poisoning levels and the N backup replacement strategies are stored in association to complete the construction of a catalyst replacement analysis model; The predicted poisoning state is synchronized to the catalyst replacement analysis model for state correlation analysis, and the real-time replacement strategy is called and output.

Images