Anti-crystallization alarm system of urea denitration system
By constructing a crystal state analysis model of the urea denitrification system and conducting multi-dimensional data monitoring and analysis, the problem of accurate judgment of crystallization phenomena in the urea denitrification system is solved to ensure the stable operation of the system.
Patent Information
- Application Number
- CN202510617093.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-14
- Publication Date
- 2025-08-26
AI Technical Summary
The urea denitrification system is prone to crystallization during operation. A single temperature monitoring cannot accurately determine the crystallization state, resulting in low anti-crystallization efficiency and affecting the stable operation of the system.
By obtaining the historical operation log of the urea denitrification system, constructing a crystal state analysis model, conducting multi-dimensional data monitoring, judging the crystal state and generating corresponding alarms or regulation strategies, improving the accuracy of crystal state judgment.
The stable operation of the urea denitrification system is achieved. Through multi-dimensional data monitoring and analysis models, crystallization phenomena are timely discovered and processed or regulated, improving the accuracy of crystallization state judgment and system stability.
Smart Images
Figure CN120544359A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of urea denitration, and in particular to an anti-crystallization alarm system for a urea denitration system. Background Art
[0002] Urea denitrification systems are prone to crystallization during operation, but the crystallization problem will seriously affect the stable operation of the system. Currently, a single temperature monitoring method is usually used to determine whether crystallization occurs. However, the crystallization phenomenon is not only caused by temperature. A single monitoring method cannot accurately determine the crystallization state, and thus cannot be discovered and regulated in time, reducing the anti-crystallization efficiency and failing to ensure the stable operation of the urea denitrification system. Therefore, there is an urgent need for a urea denitrification system anti-crystallization alarm system to conduct multi-dimensional data monitoring, improve the accuracy of crystallization state judgment, and process or regulate it to ensure the stable operation of the urea denitrification system. Summary of the Invention
[0003] In order to solve the above technical problems, the present application provides a urea denitrification system anti-crystallization alarm system, which determines the relevant monitoring data and constructs a crystallization state analysis model to accurately judge the crystallization state. It determines whether to send an alarm instruction based on the crystallization state. If so, it determines the crystallization treatment strategy. If not, it determines the control strategy. The urea denitrification system is monitored for multi-dimensional data, the accuracy of crystallization state judgment is improved, and processing or control is performed to ensure the stable operation of the urea denitrification system.
[0004] In some embodiments of the present application, a urea denitrification system anti-crystallization alarm system is provided, comprising: An acquisition module is used to obtain the historical operation log of the urea denitrification system, and divide it into historical crystallization logs and historical non-crystallization logs, compare the historical monitoring data of each link in the historical crystallization logs and the historical non-crystallization logs, and determine the relevant monitoring data based on the comparison results; Establish a module for in-depth mining of relevant monitoring data, historical control parameters of each link of urea denitrification coefficient, and corresponding historical crystallization state coefficients, and establish a crystallization state analysis model; The judgment module is used to determine the real-time crystallization state coefficient of the real-time related monitoring data based on the crystallization state analysis model, and judge whether to send an alarm instruction according to the real-time crystallization state coefficient. If so, generate a crystallization processing strategy and issue a processing instruction; if not, generate a control strategy and issue a control instruction.
[0005] In some embodiments of the present application, determining relevant monitoring data based on the comparison results includes: Obtaining a historical crystallization state coefficient in each historical operation log, and dividing the historical operation log into a historical crystallization log and a historical non-crystallization log according to the historical crystallization state coefficient; Presetting a number of first-type crystallization state coefficient intervals and a number of second-type crystallization state coefficient intervals; Analyzing the historical crystallization state coefficients in the historical crystallization logs according to a number of first-type crystallization state coefficient intervals, and subdividing the plurality of historical crystallization logs according to the analysis results to obtain a number of historical crystallization logs corresponding to each first-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical crystallization logs in the same first type crystallization state coefficient interval to obtain a first similarity; Partially eliminating historical crystallization logs having a first similarity greater than a preset first similarity threshold, and constructing a first log sequence based on the remaining historical crystallization logs in the first type of crystallization state coefficient intervals, wherein the first log sequence includes a plurality of first log subsequences; The plurality of first log subsequences are arranged in the order of the first type of crystallization state coefficient intervals, and the historical crystallization logs in each first log subsequence correspond one-to-one to the remaining historical crystallization logs in each first type of crystallization state coefficient interval; performing difference analysis on the same historical monitoring data in a plurality of historical crystallization logs of each first log subsequence to obtain a first difference value; If the first difference value is less than the preset first difference value threshold, the historical data mean of the corresponding historical monitoring data is calculated; if the first difference value is greater than the preset first difference value threshold, the abnormal value is removed and the historical data mean of the corresponding historical monitoring data is calculated; Comparing the historical data mean of each historical monitoring data in several historical crystallization logs of the first first log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallization logs of other first log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a first historical data mean difference matrix of the first log sequence; Analyzing the historical crystallization state coefficients in the historical non-crystallization logs according to a number of second-type crystallization state coefficient intervals, and re-dividing the multiple historical non-crystallization logs according to the analysis results to obtain a number of historical non-crystallization logs corresponding to each second-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical non-crystallized logs in the same second type crystalline state coefficient interval to obtain a second similarity; Partially eliminating historical non-crystallized logs whose second similarity is greater than a preset second similarity threshold, and constructing a second log sequence based on the remaining historical non-crystallized logs in the second type of crystalline state coefficient intervals, wherein the second log sequence includes a plurality of second log subsequences; The plurality of second log subsequences are arranged in the order of the second type of crystalline state coefficient intervals, and the historical non-crystallization logs in each second log subsequence correspond one-to-one to the remaining historical non-crystallization logs in each second type of crystalline state coefficient interval; calculating the historical data mean of the same historical monitoring data in a plurality of historical non-crystallized logs of each second log subsequence; Comparing the historical data mean of each historical monitoring data in several historical non-crystallized logs of the first second log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallized logs of other second log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a second historical data mean difference matrix of the second log sequence; The relevant monitoring data and the attention coefficient of each relevant monitoring data are determined according to the first historical data mean difference matrix and the second historical data mean difference matrix.
[0006] In some embodiments of the present application, determining the relevant monitoring data and the attention coefficient of each relevant monitoring data according to the first historical data difference matrix and the second historical data difference matrix includes: The first historical data mean difference matrix and the second historical data mean difference matrix are: ; ; Among them, W1 is the first historical data mean difference matrix, is the historical data mean difference between the j1th first log subsequence and the i1th historical monitoring data of the first first log subsequence, m1 is the total number of first log subsequences in the first log sequence, n1 is the total number of historical monitoring data, j1=1,2,…m1, i1=1,2,…n1, W2 is the second historical data mean difference matrix, is the difference between the historical data mean of the j2th second log subsequence and the i1th historical monitoring data of the first second log subsequence, m2 is the total number of second log subsequences in the second log sequence, j2=1, 2, ...m2; Vertically splitting the first historical data mean difference matrix and the second historical data mean difference matrix to obtain a first historical data mean difference sequence and a second historical data mean difference sequence of a plurality of historical monitoring data; Calculating a first crystallization state coefficient mean value for each first type crystallization state coefficient interval and a second crystallization state coefficient mean value for each second type crystallization state coefficient interval; Calculating the first coefficient mean difference between the first crystallization state coefficient mean of the first first type crystallization state coefficient interval and the first coefficient mean of other first type crystallization state coefficient intervals, and constructing a first coefficient mean difference sequence; Calculating the second coefficient mean difference between the second crystallization state coefficient mean of the first second type crystallization state coefficient interval and the second coefficient mean of other second type crystallization state coefficient intervals, and constructing a second coefficient mean difference sequence; Based on the same subsequence comparison principle, the first historical data mean difference sequence and the first coefficient mean difference sequence, as well as the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data are subjected to synchronization change degree analysis to obtain the first synchronization change degree and the second synchronization change degree; generating a comprehensive synchronization change degree according to the first synchronization change degree and the second synchronization change degree; If the comprehensive synchronization change degree is greater than the preset synchronization change degree threshold, the corresponding historical monitoring data is set as the relevant monitoring data; Calculate the degree of influence of the historical data mean of the relevant monitoring data on the coefficient mean difference, and set the attention coefficient of the corresponding relevant monitoring data according to the degree of influence and the degree of comprehensive synchronous change.
[0007] In some embodiments of the present application, a first sequence reference line is constructed based on a plurality of first log subsequences, and a second sequence reference line is constructed based on a plurality of second log subsequences; Mapping the first historical data mean difference sequence and the first coefficient mean difference sequence, and the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data to the first sequence reference line and the second sequence reference line, respectively, to obtain a first change analysis graph and a second change analysis graph; The first change analysis graph includes a first historical data mean difference change curve and a first coefficient mean difference change curve of a plurality of historical monitoring data, and the second change analysis graph includes a second historical data mean difference change curve and a second coefficient mean difference change curve of a plurality of historical monitoring data; Obtaining a first curve similarity feature of a first historical data mean difference change curve and a first coefficient mean difference change curve, and a second curve similarity feature of a second historical data mean difference change curve and a second coefficient mean difference change curve for each historical monitoring data; A first synchronous change degree is generated according to the first curve similarity feature, and a second synchronous change degree is generated according to the second curve similarity feature, wherein the curve similarity features include curve fluctuation similarity, curve shape similarity, and curve slope similarity.
[0008] In some embodiments of the present application, establishing a crystallization state analysis model includes: Presetting a number of historical designated time nodes, obtaining historical monitoring information of the urea denitrification system at the number of historical designated time nodes, wherein the historical monitoring information includes a number of relevant monitoring data at the same historical designated time node and a corresponding attention coefficient; Recalculate the historical crystallization state coefficient at the same historical specified time node based on a number of relevant monitoring data at the same historical specified time node and the corresponding attention coefficient; Comparing and matching the historical control parameters of the urea denitrification system at each historical designated time node and the recalculated historical crystallization state coefficient with the corresponding historical monitoring information to obtain a first training data set; A neural network training is performed based on the first training data set to obtain a crystallization state analysis model.
[0009] In some embodiments of the present application, the system further comprises: The standard-related monitoring data of the urea denitrification system is set according to the crystallization state coefficient threshold, and the corresponding standard control parameters are matched to the standard-related monitoring data at the same time node; Comparing and analyzing the standard-related monitoring data, the corresponding standard control parameters, and the crystallization state coefficient threshold with the first training data set to obtain multiple data deviation groups; Generate multiple preset control strategies for each data deviation group, and calculate the control evaluation value of each preset control strategy for the corresponding data deviation group; The preset control strategy with the largest control evaluation value is set as the preferred control strategy for the corresponding data deviation group; Matching the optimal control strategy with the historical control parameters and historical crystallization state coefficients in the first training data set and the corresponding historical monitoring information to obtain a second training data set; The neural network is trained according to the second training data set to obtain a control model.
[0010] In some embodiments of the present application, calculating the control evaluation value of each preset control strategy for the corresponding data deviation group includes: Obtaining the historical control parameters, relevant monitoring data, and corresponding preset control values to be regulated in the corresponding data deviation group for each preset control strategy; Filter out the historical control parameters, relevant monitoring data, and the control difficulty coefficient, control cost, and control success probability of the corresponding preset control value for each preset control strategy based on the historical control log; The first control evaluation value, the second control evaluation value and the third control evaluation value are obtained respectively according to the control difficulty coefficient, the control cost and the control success probability, and the control evaluation value of the corresponding preset control strategy for the corresponding data deviation group is generated.
[0011] In some embodiments of the present application, whether to send an alarm instruction is determined based on the real-time crystallization state coefficient. If so, a crystallization processing strategy is generated and a processing instruction is issued. If not, a control strategy is generated and a control instruction is issued, including: Obtaining real-time related monitoring data and inputting it into a crystallization state analysis model to obtain a real-time crystallization state coefficient; Presetting a preset crystallization state coefficient threshold; If the real-time crystallization state coefficient is greater than the preset crystallization state coefficient threshold, an alarm instruction is sent, a crystallization processing strategy is generated, and a processing instruction is issued; If the real-time crystallization state coefficient is less than the preset crystallization state coefficient threshold, the real-time crystallization state coefficient difference is calculated, and the real-time crystallization state coefficient difference is input into the control model to obtain the control strategy and issue the control instruction.
[0012] In some embodiments of the present application, generating a crystallization process strategy includes: Building a processing strategy reference library based on historical processing logs, wherein the processing strategy reference library includes a plurality of preset crystallization state coefficients, and each preset crystallization state coefficient is mapped to a corresponding preset processing strategy; The real-time crystallization state coefficient is subtracted from the preset crystallization state coefficient in the processing strategy reference library, and the preset processing strategy of the preset crystallization state coefficient with the smallest difference is set as the current crystallization processing strategy.
[0013] Compared with the prior art, the urea denitrification system anti-crystallization alarm system of the embodiment of the present application has the following advantages: By determining relevant monitoring data and building a crystallization state analysis model, the crystallization state can be accurately judged. Based on the crystallization state, it is determined whether to send an alarm instruction. If so, a crystallization treatment strategy is determined. If not, a control strategy is determined. Multi-dimensional data monitoring of the urea denitrification system is performed to improve the accuracy of crystallization state judgment and perform processing or control to ensure the stable operation of the urea denitrification system. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 This is a schematic diagram of a urea denitrification system anti-crystallization alarm system in an embodiment of the present application. DETAILED DESCRIPTION
[0015] The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.
[0016] In the description of this application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application.
[0017] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout this application, unless otherwise specified, "plurality" means two or more.
[0018] In the description of this application, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.
[0019] like Figure 1 As shown, a urea denitrification system anti-crystallization alarm system according to an embodiment of the present application includes: An acquisition module is used to obtain the historical operation log of the urea denitrification system, and divide it into historical crystallization logs and historical non-crystallization logs, compare the historical monitoring data of each link in the historical crystallization logs and the historical non-crystallization logs, and determine the relevant monitoring data based on the comparison results; Establish a module for in-depth mining of relevant monitoring data, historical control parameters of each link of urea denitrification coefficient, and corresponding historical crystallization state coefficients, and establish a crystallization state analysis model; The judgment module is used to determine the real-time crystallization state coefficient of the real-time related monitoring data based on the crystallization state analysis model, and judge whether to send an alarm instruction according to the real-time crystallization state coefficient. If so, generate a crystallization processing strategy and issue a processing instruction; if not, generate a control strategy and issue a control instruction.
[0020] In some embodiments of the present application, determining relevant monitoring data based on the comparison results includes: Obtaining a historical crystallization state coefficient in each historical operation log, and dividing the historical operation log into a historical crystallization log and a historical non-crystallization log according to the historical crystallization state coefficient; Presetting a number of first-type crystallization state coefficient intervals and a number of second-type crystallization state coefficient intervals; Analyzing the historical crystallization state coefficients in the historical crystallization logs according to a number of first-type crystallization state coefficient intervals, and subdividing the plurality of historical crystallization logs according to the analysis results to obtain a number of historical crystallization logs corresponding to each first-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical crystallization logs in the same first type crystallization state coefficient interval to obtain a first similarity; Partially eliminating historical crystallization logs having a first similarity greater than a preset first similarity threshold, and constructing a first log sequence based on the remaining historical crystallization logs in the first type of crystallization state coefficient intervals, wherein the first log sequence includes a plurality of first log subsequences; The plurality of first log subsequences are arranged in the order of the first type of crystallization state coefficient intervals, and the historical crystallization logs in each first log subsequence correspond one-to-one to the remaining historical crystallization logs in each first type of crystallization state coefficient interval; performing difference analysis on the same historical monitoring data in a plurality of historical crystallization logs of each first log subsequence to obtain a first difference value; If the first difference value is less than the preset first difference value threshold, the historical data mean of the corresponding historical monitoring data is calculated; if the first difference value is greater than the preset first difference value threshold, the abnormal value is removed and the historical data mean of the corresponding historical monitoring data is calculated; Comparing the historical data mean of each historical monitoring data in several historical crystallization logs of the first first log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallization logs of other first log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a first historical data mean difference matrix of the first log sequence; Analyzing the historical crystallization state coefficients in the historical non-crystallization logs according to a number of second-type crystallization state coefficient intervals, and re-dividing the multiple historical non-crystallization logs according to the analysis results to obtain a number of historical non-crystallization logs corresponding to each second-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical non-crystallized logs in the same second type crystalline state coefficient interval to obtain a second similarity; Partially eliminating historical non-crystallized logs whose second similarity is greater than a preset second similarity threshold, and constructing a second log sequence based on the remaining historical non-crystallized logs in the second type of crystalline state coefficient intervals, wherein the second log sequence includes a plurality of second log subsequences; The plurality of second log subsequences are arranged in the order of the second type of crystalline state coefficient intervals, and the historical non-crystallization logs in each second log subsequence correspond one-to-one to the remaining historical non-crystallization logs in each second type of crystalline state coefficient interval; calculating the historical data mean of the same historical monitoring data in a plurality of historical non-crystallized logs of each second log subsequence; Comparing the historical data mean of each historical monitoring data in several historical non-crystallized logs of the first second log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallized logs of other second log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a second historical data mean difference matrix of the second log sequence; The relevant monitoring data and the attention coefficient of each relevant monitoring data are determined according to the first historical data mean difference matrix and the second historical data mean difference matrix.
[0021] In this embodiment, historical operation logs whose historical crystallization coefficients are greater than a preset crystallization coefficient threshold are classified as historical crystallization logs, and historical operation logs whose historical crystallization coefficients are less than the preset crystallization coefficient threshold are classified as historical non-crystallization logs.
[0022] In this embodiment, several first-class crystalline state coefficient intervals are set based on the historical crystalline state coefficients in the historical crystallization log, and several first-class crystalline state coefficient intervals are arranged in order from small to large, which means that the crystalline state appears and the degree of the crystalline state is getting higher and higher. Several second-class crystalline state coefficient intervals are set based on the historical crystalline state coefficients in the historical non-crystallization log, and several second-class crystalline state coefficient intervals are arranged in order from small to large, which means that the probability of no crystalline state and the possible occurrence of the crystalline state is getting higher and higher.
[0023] In this embodiment, by constructing a first historical data mean difference matrix and a second historical data mean difference matrix, an accurate comparison of the data differences of the historical monitoring data of different crystallization intensities in the historical crystallization log and the historical monitoring data of different crystallization probabilities in the historical non-crystallization log is achieved, laying the foundation for the subsequent determination of relevant monitoring data, improving the accuracy of the judgment of relevant monitoring data, thereby determining the monitoring data associated with the crystallization state, reducing the amount of data monitoring and analysis, and improving the accuracy of the judgment of the crystallization state.
[0024] In some embodiments of the present application, determining the relevant monitoring data and the attention coefficient of each relevant monitoring data according to the first historical data difference matrix and the second historical data difference matrix includes: The first historical data mean difference matrix and the second historical data mean difference matrix are: ; ; Among them, W1 is the first historical data mean difference matrix, is the historical data mean difference between the j1th first log subsequence and the i1th historical monitoring data of the first first log subsequence, m1 is the total number of first log subsequences in the first log sequence, n1 is the total number of historical monitoring data, j1=1,2,…m1, i1=1,2,…n1, W2 is the second historical data mean difference matrix, is the difference between the historical data mean of the j2th second log subsequence and the i1th historical monitoring data of the first second log subsequence, m2 is the total number of second log subsequences in the second log sequence, j2=1, 2, ...m2; Vertically splitting the first historical data mean difference matrix and the second historical data mean difference matrix to obtain a first historical data mean difference sequence and a second historical data mean difference sequence of a plurality of historical monitoring data; Calculating a first crystallization state coefficient mean value for each first type crystallization state coefficient interval and a second crystallization state coefficient mean value for each second type crystallization state coefficient interval; Calculating the first coefficient mean difference between the first crystallization state coefficient mean of the first first type crystallization state coefficient interval and the first coefficient mean of other first type crystallization state coefficient intervals, and constructing a first coefficient mean difference sequence; Calculating the second coefficient mean difference between the second crystallization state coefficient mean of the first second type crystallization state coefficient interval and the second coefficient mean of other second type crystallization state coefficient intervals, and constructing a second coefficient mean difference sequence; Based on the same subsequence comparison principle, the first historical data mean difference sequence and the first coefficient mean difference sequence, as well as the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data are subjected to synchronization change degree analysis to obtain the first synchronization change degree and the second synchronization change degree; generating a comprehensive synchronization change degree according to the first synchronization change degree and the second synchronization change degree; If the comprehensive synchronization change degree is greater than the preset synchronization change degree threshold, the corresponding historical monitoring data is set as the relevant monitoring data; Calculate the degree of influence of the historical data mean of the relevant monitoring data on the coefficient mean difference, and set the attention coefficient of the corresponding relevant monitoring data according to the degree of influence and the degree of comprehensive synchronous change.
[0025] In this embodiment, the first synchronous change degree and the second synchronous change degree are both for evaluating whether the change of each historical monitoring data causes the change of the historical crystallization state coefficient and the degree of change, accurately screening out relevant monitoring data, laying the foundation for the subsequent construction of the crystallization state analysis model, improving the accuracy of the crystallization state coefficient, timely discovering and issuing control instructions, and ensuring the stable operation of the urea denitrification system.
[0026] In this embodiment, if the historical data mean of the relevant monitoring data is smaller but the historical coefficient mean difference of the crystallization state coefficient caused is larger, the corresponding relevant monitoring data has a greater degree of influence. The greater the degree of influence and the greater the degree of comprehensive synchronous change, the greater the attention coefficient of the corresponding relevant monitoring data, and vice versa.
[0027] In some embodiments of the present application, a first sequence reference line is constructed based on a plurality of first log subsequences, and a second sequence reference line is constructed based on a plurality of second log subsequences; Mapping the first historical data mean difference sequence and the first coefficient mean difference sequence, and the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data to the first sequence reference line and the second sequence reference line, respectively, to obtain a first change analysis graph and a second change analysis graph; The first change analysis graph includes a first historical data mean difference change curve and a first coefficient mean difference change curve of a plurality of historical monitoring data, and the second change analysis graph includes a second historical data mean difference change curve and a second coefficient mean difference change curve of a plurality of historical monitoring data; Obtaining a first curve similarity feature of a first historical data mean difference change curve and a first coefficient mean difference change curve, and a second curve similarity feature of a second historical data mean difference change curve and a second coefficient mean difference change curve for each historical monitoring data; A first synchronous change degree is generated according to the first curve similarity feature, and a second synchronous change degree is generated according to the second curve similarity feature, wherein the curve similarity features include curve fluctuation similarity, curve shape similarity, and curve slope similarity.
[0028] In this embodiment, the curve fluctuation similarity, curve shape similarity and curve slope similarity are converted into numerical values with the same dimension as the degree of synchronous change. The greater the curve fluctuation similarity, curve shape similarity and curve slope similarity, the greater the corresponding degree of synchronous change, and vice versa.
[0029] In this embodiment, by calculating the degree of synchronous change between historical monitoring data and historical crystallization state coefficients, the degree of influence of each historical monitoring data on the historical crystallization state coefficient is accurately obtained, thereby determining the relevant monitoring data and laying the foundation for the subsequent construction of a crystallization state analysis model.
[0030] In some embodiments of the present application, establishing a crystallization state analysis model includes: Presetting a number of historical designated time nodes, obtaining historical monitoring information of the urea denitrification system at the number of historical designated time nodes, wherein the historical monitoring information includes a number of relevant monitoring data at the same historical designated time node and a corresponding attention coefficient; Recalculate the historical crystallization state coefficient at the same historical specified time node based on a number of relevant monitoring data at the same historical specified time node and the corresponding attention coefficient; Comparing and matching the historical control parameters of the urea denitrification system at each historical designated time node and the recalculated historical crystallization state coefficient with the corresponding historical monitoring information to obtain a first training data set; A neural network training is performed based on the first training data set to obtain a crystallization state analysis model.
[0031] In this embodiment, the relevant monitoring data is compared with the corresponding standard relevant monitoring data interval, and the historical crystallization state coefficient at the historical specified time node is recalculated according to the comparison result and the corresponding attention coefficient to improve the accuracy of the crystallization state analysis model.
[0032] In some embodiments of the present application, the system further comprises: The standard-related monitoring data of the urea denitrification system is set according to the crystallization state coefficient threshold, and the corresponding standard control parameters are matched to the standard-related monitoring data at the same time node; Comparing and analyzing the standard-related monitoring data, the corresponding standard control parameters, and the crystallization state coefficient threshold with the first training data set to obtain multiple data deviation groups; Generate multiple preset control strategies for each data deviation group, and calculate the control evaluation value of each preset control strategy for the corresponding data deviation group; The preset control strategy with the largest control evaluation value is set as the preferred control strategy for the corresponding data deviation group; Matching the optimal control strategy with the historical control parameters and historical crystallization state coefficients in the first training data set and the corresponding historical monitoring information to obtain a second training data set; The neural network is trained according to the second training data set to obtain a control model.
[0033] In this embodiment, the data deviation group includes data deviations between several relevant monitoring data and standard relevant monitoring data, parameter deviations between historical control parameters and standard control parameters, and coefficient deviations between recalculated historical crystallization state coefficients and crystallization state coefficient thresholds.
[0034] In this embodiment, the control evaluation value is used to evaluate the control effect of each preset control strategy on the data deviation group, so as to screen out the preferred control strategy and construct a control model, laying the foundation for subsequent timely discovery and control, ensuring the timeliness of control and the stable operation of the urea denitrification system.
[0035] In some embodiments of the present application, calculating the control evaluation value of each preset control strategy for the corresponding data deviation group includes: Obtaining the historical control parameters, relevant monitoring data, and corresponding preset control values to be regulated in the corresponding data deviation group for each preset control strategy; Filter out the historical control parameters, relevant monitoring data, and the control difficulty coefficient, control cost, and control success probability of the corresponding preset control value for each preset control strategy based on the historical control log; The first control evaluation value, the second control evaluation value and the third control evaluation value are obtained respectively according to the control difficulty coefficient, the control cost and the control success probability, and the control evaluation value of the corresponding preset control strategy for the corresponding data deviation group is generated.
[0036] In this embodiment, when the control difficulty coefficient is greater, the control cost is higher, and the control success probability is lower, the corresponding first control evaluation value, second control evaluation value, and third control evaluation value are smaller, that is, the control evaluation value is smaller, and vice versa.
[0037] In this embodiment, by calculating the control evaluation value, the optimal control strategy for each data deviation group is selected, laying the foundation for building a control model, improving the control efficiency and accuracy, and ensuring the stable operation of the urea denitrification system.
[0038] In some embodiments of the present application, whether to send an alarm instruction is determined based on the real-time crystallization state coefficient. If so, a crystallization processing strategy is generated and a processing instruction is issued. If not, a control strategy is generated and a control instruction is issued, including: Obtaining real-time related monitoring data and inputting it into a crystallization state analysis model to obtain a real-time crystallization state coefficient; Presetting a preset crystallization state coefficient threshold; If the real-time crystallization state coefficient is greater than the preset crystallization state coefficient threshold, an alarm instruction is sent, a crystallization processing strategy is generated, and a processing instruction is issued; If the real-time crystallization state coefficient is less than the preset crystallization state coefficient threshold, the real-time crystallization state coefficient difference is calculated, and the real-time crystallization state coefficient difference is input into the control model to obtain the control strategy and issue the control instruction.
[0039] In some embodiments of the present application, generating a crystallization process strategy includes: Building a processing strategy reference library based on historical processing logs, wherein the processing strategy reference library includes a plurality of preset crystallization state coefficients, and each preset crystallization state coefficient is mapped to a corresponding preset processing strategy; The real-time crystallization state coefficient is subtracted from the preset crystallization state coefficient in the processing strategy reference library, and the preset processing strategy of the preset crystallization state coefficient with the smallest difference is set as the current crystallization processing strategy.
[0040] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and replacements can be made without departing from the technical principles of the present application. These improvements and replacements should also be regarded as the scope of protection of the present application.
Claims
1. A urea denitrification system anti-crystallization alarm system, characterized in that: include: An acquisition module is used to obtain the historical operation log of the urea denitrification system, and divide it into historical crystallization logs and historical non-crystallization logs, compare the historical monitoring data of each link in the historical crystallization logs and the historical non-crystallization logs, and determine the relevant monitoring data based on the comparison results; Establish a module for in-depth mining of relevant monitoring data, historical control parameters of each link of urea denitrification coefficient, and corresponding historical crystallization state coefficients, and establish a crystallization state analysis model; The judgment module is used to determine the real-time crystallization state coefficient of the real-time related monitoring data based on the crystallization state analysis model, and judge whether to send an alarm instruction according to the real-time crystallization state coefficient. If so, generate a crystallization processing strategy and issue a processing instruction; if not, generate a control strategy and issue a control instruction.
2. The urea denitration system anti-crystallization alarm system according to claim 1, characterized in that: Determine relevant monitoring data based on the comparison results, including: Obtaining a historical crystallization state coefficient in each historical operation log, and dividing the historical operation log into a historical crystallization log and a historical non-crystallization log according to the historical crystallization state coefficient; Presetting a number of first-type crystallization state coefficient intervals and a number of second-type crystallization state coefficient intervals; Analyzing the historical crystallization state coefficients in the historical crystallization logs according to a number of first-type crystallization state coefficient intervals, and subdividing the plurality of historical crystallization logs according to the analysis results to obtain a number of historical crystallization logs corresponding to each first-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical crystallization logs in the same first type crystallization state coefficient interval to obtain a first similarity; Partially eliminating historical crystallization logs having a first similarity greater than a preset first similarity threshold, and constructing a first log sequence based on the remaining historical crystallization logs in the first type of crystallization state coefficient intervals, wherein the first log sequence includes a plurality of first log subsequences; The plurality of first log subsequences are arranged in the order of the first type of crystallization state coefficient intervals, and the historical crystallization logs in each first log subsequence correspond one-to-one to the remaining historical crystallization logs in each first type of crystallization state coefficient interval; performing difference analysis on the same historical monitoring data in a plurality of historical crystallization logs of each first log subsequence to obtain a first difference value; If the first difference value is less than the preset first difference value threshold, the historical data mean of the corresponding historical monitoring data is calculated; if the first difference value is greater than the preset first difference value threshold, the abnormal value is removed and the historical data mean of the corresponding historical monitoring data is calculated; Comparing the historical data mean of each historical monitoring data in several historical crystallization logs of the first first log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallization logs of other first log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a first historical data mean difference matrix of the first log sequence; Analyzing the historical crystallization state coefficients in the historical non-crystallization logs according to a number of second-type crystallization state coefficient intervals, and re-dividing the multiple historical non-crystallization logs according to the analysis results to obtain a number of historical non-crystallization logs corresponding to each second-type crystallization state coefficient interval; Performing similarity analysis on historical monitoring data of several historical non-crystallized logs in the same second type crystalline state coefficient interval to obtain a second similarity; Partially eliminating historical non-crystallized logs whose second similarity is greater than a preset second similarity threshold, and constructing a second log sequence based on the remaining historical non-crystallized logs in the second type of crystalline state coefficient intervals, wherein the second log sequence includes a plurality of second log subsequences; The plurality of second log subsequences are arranged in the order of the second type of crystalline state coefficient intervals, and the historical non-crystallization logs in each second log subsequence correspond one-to-one to the remaining historical non-crystallization logs in each second type of crystalline state coefficient interval; calculating the historical data mean of the same historical monitoring data in a plurality of historical non-crystallized logs of each second log subsequence; Comparing the historical data mean of each historical monitoring data in several historical non-crystallized logs of the first second log subsequence with the historical data mean of the corresponding historical monitoring data in several historical crystallized logs of other second log subsequences, obtaining several historical data mean differences of the same historical monitoring data, and constructing a second historical data mean difference matrix of the second log sequence; The relevant monitoring data and the attention coefficient of each relevant monitoring data are determined according to the first historical data mean difference matrix and the second historical data mean difference matrix.
3. The urea denitration system anti-crystallization alarm system according to claim 2, characterized in that: Determining relevant monitoring data and an attention coefficient of each relevant monitoring data according to the first historical data difference matrix and the second historical data difference matrix includes: The first historical data mean difference matrix and the second historical data mean difference matrix are: ； ； Among them, W1 is the first historical data mean difference matrix, is the historical data mean difference between the j1th first log subsequence and the i1th historical monitoring data of the first first log subsequence, m1 is the total number of first log subsequences in the first log sequence, n1 is the total number of historical monitoring data, j1=1,2,…m1, i1=1,2,…n1, W2 is the second historical data mean difference matrix, is the difference between the historical data mean of the j2th second log subsequence and the i1th historical monitoring data of the first second log subsequence, m2 is the total number of second log subsequences in the second log sequence, j2=1, 2, ...m2; Vertically splitting the first historical data mean difference matrix and the second historical data mean difference matrix to obtain a first historical data mean difference sequence and a second historical data mean difference sequence of a plurality of historical monitoring data; Calculating a first crystallization state coefficient mean value for each first type crystallization state coefficient interval and a second crystallization state coefficient mean value for each second type crystallization state coefficient interval; Calculating the first coefficient mean difference between the first crystallization state coefficient mean of the first first type crystallization state coefficient interval and the first coefficient mean of other first type crystallization state coefficient intervals, and constructing a first coefficient mean difference sequence; Calculating the second coefficient mean difference between the second crystallization state coefficient mean of the first second type crystallization state coefficient interval and the second coefficient mean of other second type crystallization state coefficient intervals, and constructing a second coefficient mean difference sequence; Based on the same subsequence comparison principle, the first historical data mean difference sequence and the first coefficient mean difference sequence, as well as the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data are subjected to synchronization change degree analysis to obtain the first synchronization change degree and the second synchronization change degree; generating a comprehensive synchronization change degree according to the first synchronization change degree and the second synchronization change degree; If the comprehensive synchronization change degree is greater than the preset synchronization change degree threshold, the corresponding historical monitoring data is set as the relevant monitoring data; Calculate the degree of influence of the historical data mean of the relevant monitoring data on the coefficient mean difference, and set the attention coefficient of the corresponding relevant monitoring data according to the degree of influence and the degree of comprehensive synchronous change.
4. The urea denitration system anti-crystallization alarm system according to claim 3, characterized in that: constructing a first sequence reference line based on the plurality of first log subsequences, and constructing a second sequence reference line based on the plurality of second log subsequences; Mapping the first historical data mean difference sequence and the first coefficient mean difference sequence, and the second historical data mean difference sequence and the second coefficient mean difference sequence of each historical monitoring data to the first sequence reference line and the second sequence reference line, respectively, to obtain a first change analysis graph and a second change analysis graph; The first change analysis graph includes a first historical data mean difference change curve and a first coefficient mean difference change curve of a plurality of historical monitoring data, and the second change analysis graph includes a second historical data mean difference change curve and a second coefficient mean difference change curve of a plurality of historical monitoring data; Obtaining a first curve similarity feature of a first historical data mean difference change curve and a first coefficient mean difference change curve, and a second curve similarity feature of a second historical data mean difference change curve and a second coefficient mean difference change curve for each historical monitoring data; A first synchronous change degree is generated according to the first curve similarity feature, and a second synchronous change degree is generated according to the second curve similarity feature, wherein the curve similarity features include curve fluctuation similarity, curve shape similarity, and curve slope similarity.
5. The urea denitration system anti-crystallization alarm system according to claim 4, characterized in that: Establish a crystallization state analysis model, including: Presetting a number of historical designated time nodes, obtaining historical monitoring information of the urea denitrification system at the number of historical designated time nodes, wherein the historical monitoring information includes a number of relevant monitoring data at the same historical designated time node and a corresponding attention coefficient; Recalculate the historical crystallization state coefficient at the same historical specified time node based on a number of relevant monitoring data at the same historical specified time node and the corresponding attention coefficient; Comparing and matching the historical control parameters of the urea denitrification system at each historical designated time node and the recalculated historical crystallization state coefficient with the corresponding historical monitoring information to obtain a first training data set; A neural network training is performed based on the first training data set to obtain a crystallization state analysis model.
6. The urea denitration system anti-crystallization alarm system according to claim 5, characterized in that: Also includes: According to the crystallization state coefficient threshold, the standard-related monitoring data of the urea denitrification system is set, and the corresponding standard control parameters are matched to the standard-related monitoring data at the same time node; Comparing and analyzing the standard-related monitoring data, the corresponding standard control parameters, and the crystallization state coefficient threshold with the first training data set to obtain multiple data deviation groups; Generate multiple preset control strategies for each data deviation group, and calculate the control evaluation value of each preset control strategy for the corresponding data deviation group; The preset control strategy with the largest control evaluation value is set as the preferred control strategy for the corresponding data deviation group; Matching the optimal control strategy with the historical control parameters and historical crystallization state coefficients in the first training data set and the corresponding historical monitoring information to obtain a second training data set; The neural network is trained according to the second training data set to obtain a control model.
7. The urea denitration system anti-crystallization alarm system according to claim 6, characterized in that: Calculate the control evaluation value of each preset control strategy for the corresponding data deviation group, including: Obtaining the historical control parameters, relevant monitoring data, and corresponding preset control values to be regulated in the corresponding data deviation group for each preset control strategy; Filter out the historical control parameters, relevant monitoring data, and the control difficulty coefficient, control cost, and control success probability of the corresponding preset control value for each preset control strategy based on the historical control log; The first control evaluation value, the second control evaluation value and the third control evaluation value are obtained respectively according to the control difficulty coefficient, the control cost and the control success probability, and the control evaluation value of the corresponding preset control strategy for the corresponding data deviation group is generated.
8. The urea denitration system anti-crystallization alarm system according to claim 7, characterized in that: Determine whether to send an alarm instruction based on the real-time crystallization state coefficient. If so, generate a crystallization processing strategy and issue a processing instruction. If not, generate a control strategy and issue a control instruction, including: Obtaining real-time related monitoring data and inputting it into a crystallization state analysis model to obtain a real-time crystallization state coefficient; Presetting a preset crystallization state coefficient threshold; If the real-time crystallization state coefficient is greater than the preset crystallization state coefficient threshold, an alarm instruction is sent, a crystallization processing strategy is generated, and a processing instruction is issued; If the real-time crystallization state coefficient is less than the preset crystallization state coefficient threshold, the real-time crystallization state coefficient difference is calculated, and the real-time crystallization state coefficient difference is input into the control model to obtain the control strategy and issue the control instruction.
9. The urea denitration system anti-crystallization alarm system according to claim 8, characterized in that: Generate crystallization processing strategies including: Building a processing strategy reference library based on historical processing logs, wherein the processing strategy reference library includes a plurality of preset crystallization state coefficients, and each preset crystallization state coefficient is mapped to a corresponding preset processing strategy; The real-time crystallization state coefficient is subtracted from the preset crystallization state coefficient in the processing strategy reference library, and the preset processing strategy of the preset crystallization state coefficient with the smallest difference is set as the current crystallization processing strategy.