Ammonia gas pipeline purging method for urea hydrolysis system

Abstract

The invention discloses an ammonia gas pipeline purging method for a urea hydrolysis system. The method comprises the following steps: analyzing a historical blockage log of an ammonia gas pipeline, extracting historical related parameters and historical blockage characteristics of the historical blockage log, analyzing the historical related parameters and the historical blockage characteristics, and determining characteristic related parameters; setting a plurality of data intervals of the feature-related parameters, combining the feature-related parameters of different data intervals to obtain a plurality of feature-related parameter combinations, and constructing a congestion prediction model based on the feature-related parameter combinations; the method comprises the following steps: acquiring a real-time characteristic related parameter combination, inputting the real-time characteristic related parameter combination into a blockage prediction model to obtain predicted blockage characteristics, analyzing the predicted blockage characteristics based on a preset analysis model, and determining an ammonia gas pipeline purging strategy according to an analysis result, so that the pipeline monitoring accuracy and the blockage identification accuracy are improved, and the purging efficiency is improved. And stable operation of a urea hydrolysis system and a denitration unit is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of urea hydrolysis systems, and particularly to a method for purging ammonia pipelines in a urea hydrolysis system. Background Art

[0002] Currently, the urea hydrolysis process for ammonia production is increasingly used in denitration systems. To supply ammonia to the denitration unit through the urea hydrolysis system, it is necessary to purge the pipelines that supply ammonia to the outside in a timely manner to prevent ammonia from condensing and blocking the pipelines along the way. Therefore, there is an urgent need for a method for purging ammonia pipelines in a urea hydrolysis system to accurately determine whether the pipeline is blocked and formulate a reasonable purging method to improve the purging efficiency. Summary of the Invention

[0003] To solve the above technical problems, the present application provides a method for purging ammonia pipelines in a urea hydrolysis system. By determining the characteristic-related parameters of historical blockage characteristics, setting data intervals for each characteristic-related parameter, constructing multiple combinations of characteristic-related parameters based on the characteristic-related parameters in different data intervals, and constructing a blockage prediction model, the predicted blockage characteristics of the real-time characteristic-related parameter combination are determined based on the blockage prediction model, and a reasonable ammonia pipeline purging strategy is formulated to improve the accuracy of pipeline monitoring and blockage identification, thereby improving the purging efficiency and ensuring the stable operation of the urea hydrolysis system and the denitration unit.

[0004] In some embodiments of the present application, a method for purging ammonia pipelines in a urea hydrolysis system is provided, including: Analyze the historical blockage logs of the ammonia pipeline, extract the historical-related parameters and historical blockage characteristics of the historical blockage logs, and analyze the historical-related parameters and historical blockage characteristics to determine the characteristic-related parameters; Set multiple data intervals for the characteristic-related parameters, combine the characteristic-related parameters in different data intervals to obtain multiple combinations of characteristic-related parameters, and construct a blockage prediction model based on the combinations of characteristic-related parameters; Obtain the real-time characteristic-related parameter combination and input it into the blockage prediction model to obtain the predicted blockage characteristics, analyze the predicted blockage characteristics based on a preset analysis model, and determine the ammonia pipeline purging strategy according to the analysis results.

[0005] In some embodiments of the present application, analyzing the historical-related parameters and historical blockage characteristics to determine the characteristic-related parameters includes: The historical blockage characteristics include historical blockage type, historical blockage location, and historical blockage degree; Classify the historical blockage logs of the same historical blockage type once to obtain multiple first historical blockage log sets at different historical blockage locations of the same historical blockage type; Reclassify the set of historical blockage logs at the same historical blockage position of the same historical blockage type to obtain a second set of historical blockage logs with different historical blockage degrees at the same historical blockage position of the same historical blockage type; Obtain the historical data corresponding to multiple historical related parameters of the historical blockage logs in each second set of historical blockage logs, and arrange the historical data corresponding to the same historical related parameter in the order of the magnitude of the historical blockage degree in the corresponding historical blockage log to obtain a historical data sequence for each historical related parameter; Analyze the historical data sequence of each historical related parameter to determine whether the historical data in the historical data sequence changes with the corresponding historical blockage degree and whether the data difference between adjacent historical data is greater than a preset data difference; If so, select the corresponding historical related parameter as a feature related parameter and set the weight coefficient of the corresponding feature related parameter.

[0006] In some embodiments of the present application, set multiple data intervals for the feature related parameters, including: Obtain all the historical data corresponding to each feature related parameter and generate a comprehensive historical data interval, and divide the comprehensive historical data interval into multiple initial historical data intervals according to preset division nodes; Analyze each initial historical data interval of the feature related parameter to obtain the state corresponding to each initial historical data interval, and the state includes a good state, a normal state, a suspected abnormal state, and an abnormal state; When the state corresponding to the initial historical data interval is an abnormal state, determine the number of corresponding historical blockage types, the position deviation value of the historical blockage positions of the same historical blockage type, and the level of the historical blockage degree of each historical blockage type, and the level of the historical blockage degree includes low level, medium level, and high level; If the number of corresponding historical blockage types is not 1, determine the first division node of the corresponding initial historical data interval. If the position deviation value of the historical blockage positions of the same historical blockage type is greater than the preset deviation value, determine the second division node of the corresponding initial historical data interval. If the level of the historical blockage degree of the same historical blockage type is not the same level, determine the third division node of the corresponding initial historical data interval; According to the first division node, the second division node, and the third division node, re-divide the corresponding initial historical data interval to obtain multiple data intervals for the feature related parameter.

[0007] In some embodiments of the present application, combine the feature related parameters in different data intervals to obtain multiple combinations of feature related parameters, including: Configure the corresponding state coefficient according to the state corresponding to each data interval of each feature-related parameter. When the state is an abnormal state, generate a compensation coefficient according to the historical blockage type, historical blockage location, and level of historical blockage degree of the corresponding data interval; Generate the influence evaluation value of each data interval according to the state coefficient, compensation coefficient, and weight coefficient of the corresponding feature-related parameter of each data interval; The calculation formula of the influence evaluation value is: ; Where Y is the influence evaluation value, X is the state coefficient, x0 is the compensation coefficient, b is the weight coefficient of the corresponding feature-related parameter, and y0 is the influence evaluation conversion coefficient; If the influence evaluation value is greater than the influence evaluation value threshold, calculate the influence evaluation value difference, set the acquisition number of the corresponding data interval according to the influence evaluation value difference, and acquire the feature data of the corresponding data interval according to the acquisition number; If the influence evaluation value is less than the influence evaluation value threshold, calculate the data mean in the corresponding data interval and set it as the feature data of the corresponding data interval; Construct a feature data sequence T, T(t1, t2,..., tn) according to the feature data of multiple data intervals of each feature-related parameter, where ti is the i-th feature data and n is the total number of feature data of the feature-related parameter; Randomly combine the feature data in the feature data sequence of all feature-related parameters to obtain multiple feature-related parameter combinations.

[0008] In some embodiments of the present application, a blockage prediction model is constructed based on the feature-related parameter combination, including: Construct a historical feature-related parameter combination based on the historical data corresponding to all feature-related parameters in the historical blockage log, and use the historical feature-related parameter combination and the corresponding historical blockage feature as a labeled data set; Train a preset deep learning model according to the labeled data set to obtain an initial blockage prediction model; Use the feature-related parameter combination as an unlabeled data set, and generate the first predicted blockage feature of each feature-related parameter in the unlabeled data set according to the initial blockage prediction model; Obtain the state of the data interval where the feature data corresponding to each feature-related parameter in the feature-related parameter combination is located, and determine the second predicted blockage feature of the corresponding feature-related parameter combination according to the data interval in the abnormal state; Compare the first predicted blockage feature and the second predicted blockage feature of the same feature-related parameter combination. If the first predicted blockage feature and the second predicted blockage feature are the same, divide the feature-related parameter combination into the labeled data set; Train and iterate the initial blockage prediction model based on the re - partitioned labeled dataset until all the feature - related parameter combinations in the unlabeled dataset are partitioned into the labeled dataset, and set the initial blockage prediction model as the blockage prediction model.

[0009] In some embodiments of the present application, determining the second predicted blockage feature corresponding to the feature - related parameter combination according to the data interval in the abnormal state includes: Screen out the data intervals in the abnormal state and the historical blockage types, historical blockage positions, and historical blockage degrees corresponding to the data intervals; Judge whether the historical blockage types, historical blockage positions, and historical blockage degrees of the data intervals in the abnormal state with the same feature - related parameter combination are the same; If so, set it as the second predicted blockage feature; If not, obtain the weight coefficient of the feature - related parameter corresponding to the data interval in the abnormal state, and set the historical blockage type, historical blockage position, and historical blockage degree of the data interval corresponding to the feature - related parameter with the largest weight coefficient as the second predicted blockage feature.

[0010] In some embodiments of the present application, obtaining the real - time feature - related parameter combination and inputting it into the blockage prediction model to obtain the predicted blockage feature includes: Obtain the real - time data of the feature - related parameter, and judge whether the state of the data interval where the real - time data is located is a suspected abnormal state or an abnormal state; If not, obtain the historical data of the feature - related parameter at the previous monitoring time node, and determine the change trend and change rate of the feature - related parameter according to the historical data and real - time data of the same feature - related parameter. If the change trend is an abnormal change trend and the change rate is greater than the preset rate threshold, generate a warning signal; If so, construct a real - time feature - related parameter combination according to the real - time data of all the feature - related parameters; Input the real - time feature - related parameter combination into the blockage prediction model to obtain the predicted blockage feature, and the predicted blockage feature includes the predicted blockage type, predicted blockage position, and predicted blockage degree.

[0011] In some embodiments of the present application, analyzing the predicted blockage feature based on a preset analysis model and determining the ammonia pipeline purging strategy according to the analysis result includes: Match the corresponding preset analysis model in the preset analysis model reference library according to the predicted blockage type. The preset analysis model reference library includes preset analysis models corresponding to different blockage types. The matched preset analysis model includes multiple preset blockage positions of the predicted blockage type. Each preset blockage position also corresponds to multiple preset blockage degrees, and each preset blockage position and the corresponding preset blockage degree are mapped to a preset purging strategy. Analyze the predicted blockage position and the predicted blockage degree based on the matched preset analysis model to obtain the first similarity between the predicted blockage position and multiple preset blockage positions. Perform similarity analysis on the multiple preset blockage degrees corresponding to the preset blockage position with the largest first similarity and the predicted blockage degree to obtain the second similarity. Set the preset purging strategy mapped by the preset blockage position with the largest first similarity to the predicted blockage position and the preset blockage degree with the largest second similarity to the predicted blockage degree as the current ammonia pipeline purging strategy. Generate a purging instruction according to the ammonia pipeline purging strategy.

[0012] A method for purging an ammonia pipeline in a urea hydrolysis system according to an embodiment of the present application, compared with the prior art, its beneficial effects are as follows: By determining the characteristic related parameters of the historical blockage characteristics, setting the data interval of each characteristic related parameter, constructing multiple combinations of characteristic related parameters according to the characteristic related parameters in different data intervals, constructing a blockage prediction model, determining the predicted blockage characteristics of the real-time characteristic related parameter combination based on the blockage prediction model, and formulating a reasonable ammonia pipeline purging strategy, the accuracy of pipeline monitoring and blockage identification is improved, thereby improving the purging efficiency and ensuring the stable operation of the urea hydrolysis system and the denitration unit. Brief Description of the Drawings

[0013] Figure 1 It is a schematic flow chart of a method for purging an ammonia pipeline in a urea hydrolysis system according to an embodiment of the present application. Detailed Description of the Embodiment

[0014] The following combines the drawings and embodiments to further describe in detail the specific implementation manners of the present application. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.

[0015] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present application.

[0016] The terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0017] In the description of the present application, it should be noted that unless otherwise clearly specified and defined, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0018] As Figure 1 shown, a method for purging an ammonia pipeline in a urea hydrolysis system according to an embodiment of the present application includes: Step S101: Analyze the historical blockage log of the ammonia pipeline, extract the historical relevant parameters and historical blockage characteristics of the historical blockage log, analyze the historical relevant parameters and historical blockage characteristics, and determine the characteristic relevant parameters; Step S1022: Set multiple data intervals for the characteristic relevant parameters, combine the characteristic relevant parameters in different data intervals to obtain multiple combinations of characteristic relevant parameters, and construct a blockage prediction model based on the combinations of characteristic relevant parameters; Step S103: Obtain the real-time combination of characteristic relevant parameters and input it into the blockage prediction model to obtain the predicted blockage characteristics, analyze the predicted blockage characteristics based on a preset analysis model, and determine the ammonia pipeline purging strategy according to the analysis result.

[0019] In this embodiment, the historical relevant parameters refer to the parameters that change when the ammonia pipeline supplies ammonia, including but not limited to ammonia supply flow rate, temperature, pressure, moisture content, etc., and the characteristic relevant parameters refer to the parameters that have a greater impact on the historical blockage characteristics.

[0020] In this embodiment, multiple data intervals for setting feature-related parameters are set according to different states corresponding to the feature-related parameters, different historical blockage types, historical blockage positions, and historical blockage degrees in the abnormal state. Multiple feature data are determined according to the multiple data intervals of each feature-related parameter, and the feature data of all feature-related parameters are randomly combined to obtain possible operating conditions in the ammonia pipeline, thereby constructing a blockage prediction model to improve the accuracy and timeliness of subsequent determination of predicted blockage features and ammonia pipeline purging strategies, and ensure the stable operation of the urea hydrolysis system and the denitration unit.

[0021] In some embodiments of the present application, historical-related parameters and historical blockage features are analyzed to determine feature-related parameters, including: The historical blockage features include historical blockage type, historical blockage position, and historical blockage degree; Perform a first classification on the historical blockage logs of the same historical blockage type to obtain multiple first historical blockage log sets at different historical blockage positions of the same historical blockage type; Perform a second classification on the historical blockage log set at the same historical blockage position of the same historical blockage type to obtain a second historical blockage log set at different historical blockage degrees at the same historical blockage position of the same historical blockage type; Obtain the historical data corresponding to multiple historical-related parameters of the historical blockage logs in each second historical blockage log set, and arrange the historical data corresponding to the same historical-related parameter in the order of the size of the historical blockage degree in the corresponding historical blockage log to obtain the historical data sequence of each historical-related parameter; Analyze the historical data sequence of each historical-related parameter to determine whether the historical data in the historical data sequence changes with the corresponding historical blockage degree, and whether the data difference between adjacent historical data is greater than a preset data difference; If so, select the corresponding historical-related parameter as a feature-related parameter and set the weight coefficient of the corresponding feature-related parameter.

[0022] In this embodiment, the historical data corresponding to the same historical-related parameter is arranged in the order of the size of the historical blockage degree in the corresponding historical blockage log. The historical blockage degree in the corresponding historical blockage log is screened, and the difference between adjacent historical blockage degrees is greater than the preset degree difference. When the historical data in the historical data sequence changes with the corresponding historical blockage degree and the data difference between adjacent historical data is greater than the preset data difference, it indicates that the change in the historical data of the historical-related parameter has an impact on the historical blockage degree, that is, the historical-related parameter is set as a feature-related parameter. When a small data difference of the feature-related parameter causes a large change in the historical blockage degree, the weight coefficient of the corresponding feature-related parameter is larger.

[0023] In some embodiments of the present application, a plurality of data intervals for setting feature-related parameters are set, including: Obtain all historical data corresponding to each feature-related parameter and generate a comprehensive historical data interval, and divide the comprehensive historical data interval into a plurality of initial historical data intervals according to preset division nodes; Analyze each initial historical data interval of the feature-related parameter to obtain the state corresponding to each initial historical data interval, and the state includes a good state, a normal state, a suspected abnormal state, and an abnormal state; When the state corresponding to the initial historical data interval is an abnormal state, determine the number of corresponding historical blockage types, the position deviation value of the historical blockage positions of the same historical blockage type, and the level of the historical blockage degree of each historical blockage type, and the level of the historical blockage degree includes low level, medium level, and high level; If the number of corresponding historical blockage types is not 1, determine the first division node of the corresponding initial historical data interval. If the position deviation value of the historical blockage positions of the same historical blockage type is greater than the preset deviation value, determine the second division node of the corresponding initial historical data interval. If the levels of the historical blockage degrees of the same historical blockage type are not of the same level, determine the third division node of the corresponding initial historical data interval; According to the first division node, the second division node, and the third division node, re-divide the corresponding initial historical data interval to obtain a plurality of data intervals for the feature-related parameter.

[0024] In this embodiment, the first division node refers to the data division node with different historical blockage types in the initial historical data interval. The second division node refers to the data division node with a large deviation in the historical blockage positions of the same historical blockage type. The second division node divides the initial historical data interval after the first division node. The third division node refers to the data division node with different historical blockage degree levels at the same historical blockage position of the same historical blockage type. The third division node divides the initial historical data interval after the second division node.

[0025] In this embodiment, by setting a plurality of data intervals for the feature-related parameter, it lays a foundation for subsequently obtaining a plurality of feature-related parameter combinations and constructing a blockage prediction model, improves the prediction accuracy of the blockage characteristics of different operating conditions of the ammonia pipeline, lays a foundation for subsequently formulating a reasonable ammonia pipeline purging strategy for different blockage characteristics, improves the purging efficiency, and ensures the stable operation of the urea hydrolysis system and the denitration unit.

[0026] In some embodiments of the present application, combine the feature-related parameters of different data intervals to obtain a plurality of feature-related parameter combinations, including: Configure the corresponding state coefficient according to the state corresponding to each data interval of each feature-related parameter. When the state is an abnormal state, generate a compensation coefficient according to the historical blockage type, historical blockage location, and level of historical blockage degree in the corresponding data interval; Generate the influence evaluation value of each data interval according to the state coefficient, compensation coefficient, and weight coefficient of the corresponding feature-related parameter of each data interval; The calculation formula of the influence evaluation value is: ; Among them, Y is the influence evaluation value, X is the state coefficient, x0 is the compensation coefficient, b is the weight coefficient of the corresponding feature-related parameter, and y0 is the influence evaluation conversion coefficient; If the influence evaluation value is greater than the influence evaluation value threshold, calculate the influence evaluation value difference, set the acquisition number of the corresponding data interval according to the influence evaluation value difference, and collect the feature data of the corresponding data interval according to the acquisition number; If the influence evaluation value is less than the influence evaluation value threshold, calculate the data mean value in the corresponding data interval and set it as the feature data of the corresponding data interval; Construct a feature data sequence T, T(t1, t2,..., tn) according to the feature data of multiple data intervals of each feature-related parameter, where ti is the i-th feature data, and n is the total number of feature data of the feature-related parameter; Randomly combine the feature data in the feature data sequences of all feature-related parameters to obtain multiple feature-related parameter combinations.

[0027] In this embodiment, the influence evaluation value refers to the influence evaluation of each data interval of the feature-related parameter on the blockage characteristics of the ammonia pipeline. The greater the influence evaluation value, the more likely the corresponding data interval of the feature-related parameter will cause blockage of the ammonia pipeline and the more serious the blockage situation. Set the acquisition number of the data interval according to the influence evaluation value, so as to lay a foundation for subsequent construction of multiple feature-related parameter combinations and blockage prediction models, and improve the judgment accuracy of different feature data of feature-related parameters for blockage of the ammonia pipeline.

[0028] In some embodiments of the present application, constructing a blockage prediction model based on the feature-related parameter combination includes: Construct a historical feature-related parameter combination based on the historical data corresponding to all feature-related parameters in the historical blockage log, and use the historical feature-related parameter combination and the corresponding historical blockage characteristics as a labeled data set; Train a preset deep learning model according to the labeled data set to obtain an initial blockage prediction model; Take the combination of feature-related parameters as the unlabeled dataset, and generate the first predicted clogging feature of each feature-related parameter in the unlabeled dataset according to the initial clogging prediction model; Obtain the state of the data interval where the feature data corresponding to each feature-related parameter in the combination of feature-related parameters is located, and determine the second predicted clogging feature of the corresponding combination of feature-related parameters according to the data interval in the abnormal state; Compare the first predicted clogging feature and the second predicted clogging feature of the same combination of feature-related parameters. If the first predicted clogging feature and the second predicted clogging feature are the same, divide the combination of feature-related parameters into the labeled dataset; Train and iterate the initial clogging prediction model according to the re-divided labeled dataset until all the combinations of feature-related parameters in the unlabeled dataset are divided into the labeled dataset, and set the initial clogging prediction model as the clogging prediction model.

[0029] In this embodiment, the first predicted clogging feature is obtained according to the initial clogging prediction model, and the second predicted clogging feature is determined according to the data interval where the feature data of each feature-related parameter in the combination of feature-related parameters is located.

[0030] In this embodiment, the combinations of feature-related parameters are continuously divided into the labeled dataset, and the initial clogging prediction model is iteratively optimized to improve the clogging prediction accuracy of the initial clogging prediction model for the remaining combinations of feature-related parameters, so as to improve the accuracy of predicting the clogging feature of the real-time combination of feature-related parameters in the future, accurately judge whether clogging occurs and the type, location, and degree of clogging, and improve the rationality and purging efficiency of the ammonia pipeline purging strategy.

[0031] In some embodiments of the present application, determining the second predicted clogging feature of the corresponding combination of feature-related parameters according to the data interval in the abnormal state includes: Screen out the data intervals in the abnormal state and the historical clogging type, historical clogging location, and historical clogging degree of the corresponding data intervals; Judge whether the historical clogging type, historical clogging location, and historical clogging degree of the data intervals in the abnormal state of the same combination of feature-related parameters are the same; If so, set it as the second predicted clogging feature; If not, obtain the weight coefficient of the feature-related parameter corresponding to the data interval in the abnormal state, and set the historical clogging type, historical clogging location, and historical clogging degree of the data interval corresponding to the feature-related parameter with the largest weight coefficient as the second predicted clogging feature.

[0032] In some embodiments of the present application, obtaining the real-time combination of feature-related parameters and inputting it into the clogging prediction model to obtain the predicted clogging feature includes: Obtain real-time data of feature-related parameters, and determine whether the status of the data interval where the real-time data is located is a suspected abnormal status or an abnormal status; If not, obtain historical data of the feature-related parameters at the previous monitoring time node, and determine the change trend and change rate of the feature-related parameters based on the historical data and real-time data of the same feature-related parameters. If the change trend is an abnormal change trend and the change rate is greater than the preset rate threshold, generate a warning signal; If so, construct a real-time feature-related parameter combination based on the real-time data of all feature-related parameters; Input the real-time feature-related parameter combination into the blockage prediction model to obtain predicted blockage features, where the predicted blockage features include predicted blockage types, predicted blockage positions, and predicted blockage degrees.

[0033] In this embodiment, the abnormal change trend means that the status corresponding to the data interval where the historical data is located changes to the data interval where the real-time data is located is a downward trend. For example, the status corresponding to the data interval changes from a good status to a normal status, or from a normal status to a suspected abnormal status.

[0034] In some embodiments of the present application, analyze the predicted blockage features based on a preset analysis model, and determine the ammonia pipeline purging strategy according to the analysis result, including: Match the corresponding preset analysis model in the preset analysis model reference library according to the predicted blockage type. The preset analysis model reference library includes preset analysis models corresponding to different blockage types. The matched preset analysis model includes multiple preset blockage positions of the predicted blockage type, and each preset blockage position also corresponds to multiple preset blockage degrees, and each preset blockage position and the corresponding preset blockage degree are mapped to a preset purging strategy; Analyze the predicted blockage position and the predicted blockage degree based on the matched preset analysis model to obtain the first similarity between the predicted blockage position and multiple preset blockage positions; Perform similarity analysis on the multiple preset blockage degrees corresponding to the preset blockage position with the largest first similarity and the predicted blockage degree to obtain the second similarity; Set the preset purging strategy mapped to the preset blockage position with the largest first similarity to the predicted blockage position and the preset blockage degree with the largest second similarity to the predicted blockage degree as the current ammonia pipeline purging strategy; Generate a purging instruction according to the ammonia pipeline purging strategy.

[0035] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present application, several improvements and replacements can still be made, and these improvements and replacements should also be regarded as the protection scope of the present application.

Claims

1. A method for purging an ammonia pipeline in a urea hydrolysis system, characterized in that: include: Analyze the historical blockage logs of the ammonia pipeline, extract the historical related parameters and historical blockage features of the historical blockage logs, analyze the historical related parameters and historical blockage features, and determine the feature related parameters; Setting multiple data intervals of feature-related parameters, combining feature-related parameters of different data intervals to obtain multiple feature-related parameter combinations, and constructing a congestion prediction model based on the feature-related parameter combinations; Obtain a combination of real-time feature-related parameters and input them into the blockage prediction model to obtain predicted blockage features. Analyze the predicted blockage features based on a preset analysis model, and determine the ammonia pipeline purge strategy based on the analysis results.

2. The ammonia pipeline purging method for a urea hydrolysis system according to claim 1, characterized in that: Analyze historical related parameters and historical congestion characteristics to determine characteristic related parameters, including: The historical congestion characteristics include historical congestion type, historical congestion location and historical congestion degree; Classifying the historical congestion logs of the same historical congestion type once to obtain a plurality of first historical congestion log sets of different historical congestion locations of the same historical congestion type; Reclassifying the historical congestion log set of the same historical congestion type and the same historical congestion location to obtain a second historical congestion log set of the same historical congestion type and the same historical congestion location with different historical congestion degrees; Obtaining historical data corresponding to a plurality of historical related parameters of the historical congestion log in each second historical congestion log set, and arranging the historical data corresponding to the same historical related parameter in order of magnitude of the historical congestion degree in the corresponding historical congestion log, to obtain a historical data sequence for each historical related parameter; Analyze the historical data sequence of each historical related parameter to determine whether the historical data in the historical data sequence changes with the corresponding historical congestion degree, and whether the data difference between adjacent historical data is greater than a preset data difference; If so, the corresponding history-related parameter is selected as the feature-related parameter, and the weight coefficient of the corresponding feature-related parameter is set.

3. The ammonia pipeline purging method for a urea hydrolysis system according to claim 2, characterized in that: Set multiple data intervals for feature-related parameters, including: Obtain all historical data corresponding to each feature-related parameter and generate a comprehensive historical data interval, and divide the comprehensive historical data interval into multiple initial historical data intervals according to preset division nodes; Analyze each initial historical data interval of the feature-related parameters to obtain a state corresponding to each initial historical data interval, wherein the state includes a good state, a normal state, a suspected abnormal state, and an abnormal state; When the state corresponding to the initial historical data interval is an abnormal state, determining the number of corresponding historical congestion types, the position deviation value of the historical congestion position of the same historical congestion type, and the level of the historical congestion degree of each historical congestion type, wherein the levels of the historical congestion degree include low, medium and high; If the number of the corresponding historical congestion types is not 1, determine the first division node corresponding to the initial historical data interval; if the position deviation value of the historical congestion position of the same historical congestion type is greater than the preset deviation value, determine the second division node corresponding to the initial historical data interval; if the levels of the historical congestion degrees of the same historical congestion type are not the same level, determine the third division node corresponding to the initial historical data interval; The corresponding initial historical data interval is divided again according to the first division node, the second division node and the third division node to obtain multiple data intervals of feature-related parameters.

4. The ammonia pipeline purging method for a urea hydrolysis system according to claim 3, characterized in that: The feature-related parameters of different data intervals are combined to obtain multiple feature-related parameter combinations, including: A corresponding state coefficient is configured according to the state corresponding to each data interval of each characteristic-related parameter, and when the state is an abnormal state, a compensation coefficient is generated according to the historical congestion type, historical congestion location and historical congestion degree level of the corresponding data interval; Generate an impact evaluation value for each data interval according to the state coefficient, compensation coefficient and weight coefficient of the corresponding feature-related parameters of each data interval; The calculation formula of the impact evaluation value is: ； Among them, Y is the impact evaluation value, X is the state coefficient, x0 is the compensation coefficient, b is the weight coefficient of the corresponding feature-related parameter, and y0 is the impact evaluation conversion coefficient; If the impact evaluation value is greater than the impact evaluation value threshold, the impact evaluation value difference is calculated, the number of collections of the corresponding data interval is set according to the impact evaluation value difference, and the characteristic data of the corresponding data interval is collected according to the number of collections; If the impact evaluation value is less than the impact evaluation value threshold, the data mean in the corresponding data interval is calculated and set as the characteristic data of the corresponding data interval; Construct a feature data sequence T, T (t1, t2, ..., tn) according to the feature data of multiple data intervals of each feature-related parameter, where ti is the i-th feature data and n is the total number of feature data of the feature-related parameter; The feature data in the feature data sequence of all feature-related parameters are randomly combined to obtain a plurality of feature-related parameter combinations.

5. The ammonia pipeline purging method for a urea hydrolysis system according to claim 4, characterized in that: Construct a congestion prediction model based on a combination of feature-related parameters, including: Based on the historical data corresponding to all feature-related parameters in the historical congestion log, a historical feature-related parameter combination is constructed, and the historical feature-related parameter combination and the corresponding historical congestion feature are used as a labeled data set; The preset deep learning model is trained according to the labeled data set to obtain an initial congestion prediction model; taking the feature-related parameter combination as an unlabeled data set, and generating a first predicted congestion feature for each feature-related parameter in the unlabeled data set according to an initial congestion prediction model; Acquire the state of the data interval where the feature data corresponding to each feature-related parameter in the feature-related parameter combination is located, and determine the second predicted congestion feature corresponding to the feature-related parameter combination according to the data interval in the abnormal state; comparing a first predicted congestion feature and a second predicted congestion feature of the same feature-related parameter combination, and if the first predicted congestion feature and the second predicted congestion feature are the same, dividing the feature-related parameter combination into a labeled data set; The initial congestion prediction model is trained iteratively according to the re-divided labeled data set until all feature-related parameter combinations in the unlabeled data set are divided into the labeled data set, and the initial congestion prediction model is set as the congestion prediction model.

6. The ammonia pipeline purging method for a urea hydrolysis system according to claim 5, characterized in that: Determining a second predicted blocking feature corresponding to a feature-related parameter combination according to a data interval in an abnormal state includes: Filter out the data intervals in abnormal state and the historical congestion types, historical congestion locations and historical congestion degrees of the corresponding data intervals; Determine whether the historical congestion type, historical congestion position and historical congestion degree of the data interval in the abnormal state of the same feature-related parameter combination are the same; If yes, set it as the second predicted blocking feature; If not, obtain the weight coefficient of the feature-related parameter corresponding to the data interval in the abnormal state, and set the historical congestion type, historical congestion position and historical congestion degree of the data interval in the abnormal state corresponding to the feature-related parameter with the largest weight coefficient as the second predicted congestion feature.

7. The ammonia pipeline purging method for a urea hydrolysis system according to claim 6, characterized in that: Obtain the real-time feature-related parameter combination and input it into the congestion prediction model to obtain the predicted congestion features, including: Acquire real-time data of feature-related parameters, and determine whether the state of the data interval in which the real-time data is located is a suspected abnormal state or an abnormal state; If not, obtain the historical data of the feature-related parameters at the previous monitoring time node, determine the change trend and change rate of the feature-related parameters based on the historical data and real-time data of the same feature-related parameters, and generate an early warning signal if the change trend is an abnormal change trend and the change rate is greater than the preset rate threshold; If so, construct a real-time feature-related parameter combination based on the real-time data of all feature-related parameters; The real-time feature-related parameter combination is input into the congestion prediction model to obtain predicted congestion features, which include predicted congestion type, predicted congestion location and predicted congestion degree.

8. The ammonia pipeline purging method for a urea hydrolysis system according to claim 7, characterized in that: The predicted blockage characteristics are analyzed based on the preset analysis model, and the ammonia pipeline purge strategy is determined based on the analysis results, including: Matching a corresponding preset analysis model in a preset analysis model reference library according to the predicted blockage type, wherein the preset analysis model reference library includes preset analysis models corresponding to different blockage types, and the matched preset analysis model includes a plurality of preset blockage positions of the predicted blockage type, each preset blockage position also corresponds to a plurality of preset blockage degrees, and each preset blockage position and the corresponding preset blockage degree are mapped with a preset purge strategy; Analyzing the predicted congestion position and the predicted congestion degree based on the matched preset analysis model to obtain a first similarity between the predicted congestion position and a plurality of preset congestion positions; Performing similarity analysis on a plurality of preset congestion levels corresponding to the preset congestion position with the greatest first similarity and the predicted congestion level to obtain a second similarity; Setting a preset purging strategy mapped to a preset blocking position having a first maximum similarity to the predicted blocking position and a preset blocking degree having a second maximum similarity to the predicted blocking degree as a current ammonia pipeline purging strategy; Generate purge instructions according to the ammonia pipeline purge strategy.

Images