System database data processing method of nitrogen saving device

Abstract

The invention discloses a system database data processing method of a nitrogen saving device, and relates to the technical field of data processing. Comprising the following steps: storing processed data to a database group by utilizing a preset storage strategy, screening historical information by adopting write-in data of a database, crawling information keywords of target information by adopting a crawler technology, and importing the target information into the database according to a classification result and the information keywords, and analyzing the evaluation index of the power data corresponding to each section of nitrogen device, and clearing the target data to which each section of nitrogen device belongs according to the corresponding suitable replacement period. According to the system database data processing method of the nitrogen-saving device, unified acquisition of multi-source data can be realized based on database management of a plurality of nitrogen-saving devices, and data conversion before data acquisition is not needed, so that the complexity of system construction is reduced, and the problem of authority potential safety hazards of data entries is avoided to a certain extent.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for processing system database data of a nitrogen-saving device. Background Art

[0002] In modern agricultural production, nitrogen-saving devices are used more and more widely. Nitrogen-saving devices are agricultural equipment that can effectively reduce the use of fertilizers and increase crop yields and quality. The nitrogen-saving device database usually stores a lot of relevant data. By collecting and organizing the design parameters, operating data and maintenance records of various nitrogen-saving devices, it is convenient for staff to have an in-depth understanding of the working principles of nitrogen-saving devices and analyze the working performance of nitrogen-saving devices under different conditions, thereby providing a scientific basis for the design, selection and use of nitrogen-saving devices. If the system database of the nitrogen-saving device is not stored reasonably, it is easy to add additional burden to the system database of the nitrogen-saving device, making it impossible to select typical data for subsequent staff inquiries. Therefore, it is extremely necessary to analyze the storage and transmission of the system database of the nitrogen-saving device.

[0003] The invention patent with patent number CN117573715A discloses a system database storage and transmission method for a nitrogen-saving device, including: step one, obtaining data related to the nitrogen-saving device, step two, obtaining historical data related to the nitrogen-saving device, step three, analyzing information related to the operation of the nitrogen-saving device, step four, storing information related to the operation of the nitrogen-saving device, step five, transmitting information related to the operation of the nitrogen-saving device and step six, updating information related to the operation of the nitrogen-saving device.

[0004] The existing system database storage method for nitrogen-saving devices improves the accuracy of judging whether the operation-related information of the nitrogen-saving device during the detection cycle is stored in the system database of the nitrogen-saving device, thereby avoiding the phenomenon of erroneous storage. However, in the case of long-term data collection and storage, the stored data are of many types and the amount of data is huge, which easily leads to the loss of original data and the omission of some information hidden in the data. At the same time, when collecting and storing data from different nitrogen-saving devices, the data format must be uniformly converted first, resulting in a more complicated system construction, which may be repeated with the data in the database. The storage of duplicate data will cause additional memory usage and low efficiency of subsequent data retrieval, affecting the overall management reliability of the database. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a system database data processing method for a nitrogen-saving device, which solves the existing problems.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A system database data processing method for a nitrogen-saving device comprises the following steps: Step 1: pre-processing the nitrogen-saving device status data collected in real time to obtain target data, and storing it in the database according to a predetermined storage strategy; Step 2: Based on the coincidence comparison algorithm, the historical information is screened using the written data of the database, the target information is generated and the coincidence database is updated; Step 3: Use crawler technology to crawl information keywords of target information, input the information keywords into the preset target classification model to output the corresponding classification results; Step 4: Import the target information into the database according to the classification results and information keywords; Step 5: Analyze the storage evaluation index of the power data corresponding to each nitrogen-saving device, the storage evaluation index of the basic data, and the storage evaluation index of the obstacle data; Step 6: Clear the target data of each nitrogen-saving device according to the corresponding appropriate replacement cycle.

[0007] Preferably, the database group consists of a real-time database, a relational database and a file database.

[0008] Preferably, in step 1, the device status data collected in real time by a preset target sensor is cleaned to obtain target data, and the process includes: S1: extracting a preset amount of historical device status data from a real-time database and dividing it into historical normal status data and historical abnormal status data, and performing dimension reduction and normalization processing on the historical normal status data to obtain first data; S2: Combine the Adam and SGD algorithms, use the first data as the input vector to train the neural network to obtain a parameter optimization model to find the optimal values ​​of all parameters and output the optimal feature vector; S3: Compare the input vector with the optimal feature vector to obtain a first reconstruction error under normal state data of the device; S4: inputting the device status data collected in real time as a first input vector into the parameter optimization model, outputting a first eigenvector, and then comparing the first eigenvector with the first input vector to obtain a second reconstruction error; S5: Compare the first reconstruction error and the second reconstruction error at each time point to determine the data type; S6: Based on the determined data type, all noise data and missing values ​​are automatically repaired into their corresponding reconstructed values ​​to complete data cleaning.

[0009] Preferably, the first reconstruction error and the second reconstruction error at each time point are compared to determine the data type, specifically: When the first reconstruction error is equal to the second reconstruction error at a time point, the data is judged to be a normal value; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is no such same situation around the time point, the data is judged to be noise data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is data around the time point with a fixed value equal to the first reconstruction error, the data is judged to be missing value data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and the data around the time point have a similar change trend, the data is judged to be a truly useful outlier.

[0010] Preferably, the target data includes power data, basic data and obstacle data; The power data includes the current, voltage and power corresponding to each detection time point, the basic data includes the total amount of input air, the total amount of output ammonia and the purity of ammonia corresponding to each detection time period, and the fault data includes the type and maintenance time corresponding to each fault; The abnormal power data include the current, voltage and power corresponding to each abnormal time point of each abnormal detection cycle; the abnormal basic data include the total input air volume, the total output ammonia volume and the ammonia purity corresponding to each risk detection period of each risk detection cycle; the fault data include the type and maintenance time corresponding to each fault of each fault detection cycle; the normal power data include the normal current range, the normal voltage range and the normal power range; the normal basic data include the appropriate output ammonia volume and the appropriate ammonia purity corresponding to each total input air volume.

[0011] Preferably, the overlap comparison algorithm is used to compare the initial information and the written data to output overlap information, and to perform a deduplication operation on the initial information based on the overlap information to generate target information.

[0012] Preferably, the target classification model is based on a Monte Carlo algorithm, and the training data is used to construct an initial classification model. The test data is used to verify the initial classification model, and the verification result is output. If the verification result is standard, the initial classification model is used as the target classification model. If the verification result is non-standard, the training data is used to train and optimize the initial classification model, and the step of using the test data to verify the initial classification model and outputting the verification result is jumped to execute.

[0013] Preferably, the information keywords stored in the specific storage location are reconstructed, and the construction includes at least a power data database, a basic data database and a barrier data database set, and each database in the database set is used to classify and store the key information. Beneficial Effects

[0014] The present invention provides a method for processing system database data of a nitrogen-saving device. Compared with the prior art, it has the following beneficial effects: 1. The system database data processing method of the nitrogen-saving device uses the written data of the database to deduplicate the initial information to generate target information and update the overlap database, and uses the information keywords of the target information to determine the classification result based on the target classification model, and imports the target information into the corresponding storage location of the database according to the classification result. The present invention is based on the database management of multiple nitrogen-saving devices, and can realize the unified collection of multi-source data without the need to convert data before data collection, thereby reducing the complexity of system construction and avoiding the problem of data entry permission security risks to a certain extent.

[0015] 2. The system database data processing method of the nitrogen-saving device uses a preset storage strategy to store the processed data in a database group, ensuring that the data is complete while facilitating subsequent data queries. Through deduplication operations, the storage content of the database is more streamlined, which also helps to improve the efficiency of data search. At the same time, the target classification model is used to intelligently classify data keywords into categories, thereby improving the overall management reliability of the database. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the method flow of the present invention; Figure 2 It is a schematic diagram of the flow chart of the preprocessing of the status data of the nitrogen saving device of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1-2 , the present invention provides two technical solutions: The first embodiment: a method for processing system database data of a nitrogen-saving device, comprising the following steps: Step 1: pre-processing the nitrogen-saving device status data collected in real time to obtain target data, and storing it in a database according to a predetermined storage strategy, wherein the database is composed of a real-time database, a relational database, and a file database; Step 2: Based on the coincidence comparison algorithm, the historical information is screened using the written data of the database, the target information is generated and the coincidence database is updated; Step 3: Use crawler technology to crawl information keywords of target information, input the information keywords into the preset target classification model to output the corresponding classification results; Step 4: Import the target information into the database according to the classification results and information keywords; Step 5: Analyze the storage evaluation index of the power data corresponding to each nitrogen-saving device, the storage evaluation index of the basic data, and the storage evaluation index of the obstacle data; Step 6: Clear the target data of each nitrogen-saving device according to the corresponding appropriate replacement cycle.

[0019] In this embodiment, the target data includes power data, basic data and obstacle data; The power data includes the current, voltage and power corresponding to each detection time point, the basic data includes the total amount of input air, the total amount of output ammonia and the purity of ammonia corresponding to each detection time period, and the fault data includes the type and maintenance time corresponding to each fault; The abnormal power data include the current, voltage and power corresponding to each abnormal time point of each abnormal detection cycle; the abnormal basic data include the total input air volume, the total output ammonia volume and the ammonia purity corresponding to each risk detection period of each risk detection cycle; the fault data include the type and maintenance time corresponding to each fault of each fault detection cycle; the normal power data include the normal current range, the normal voltage range and the normal power range; the normal basic data include the appropriate output ammonia volume and the appropriate ammonia purity corresponding to each total input air volume.

[0020] In this embodiment, the overlap comparison algorithm can specifically be one or a combination of TF-IDF algorithm, cosine similarity, vector space model algorithm, SVD algorithm, LSI algorithm or LDA algorithm. The overlap comparison algorithm is used to perform overlap comparison on the initial information and the written data to output the overlap information, determine that there is initial information that is essentially the same as the written data content as the overlap information, filter out the overlap information from the initial information and output the target information that needs to be written into the database, perform deduplication operation on the initial information according to the overlap information to generate the target information, perform keyword crawling on the filtered overlap information through crawler technology, construct its corresponding overlap keyword as a new overlap keyword, and import the new overlap keyword and the associated overlap information as new written overlap information into the preset overlap database for updating. After receiving the target information, perform keyword extraction on the target information based on the crawler technology to generate the corresponding information keyword.

[0021] In this embodiment, the target classification model is based on a Monte Carlo algorithm, and the training data is used to construct an initial classification model. The test data is used to verify the initial classification model, and the verification result is output. If the verification result is standard, the initial classification model is used as the target classification model. If the verification result is non-standard, the training data is used to train and optimize the initial classification model, and the step of using the test data to verify the initial classification model and outputting the verification result is jumped to execute.

[0022] In this embodiment, category sample data is constructed based on the written data of the database, wherein the category sample data includes the classification categories of the database and the corresponding classification keywords, and the classification categories of the database include main classification categories and subordinate classification categories. The category sample data is divided into training data and test data according to a certain ratio, and the initial classification model is constructed based on the learning mechanism of the Monte Carlo algorithm using the training data, and then the model is verified by inputting the test data into the initial classification model. When the output verification result meets the expectation, that is, it is the standard, the initial classification model is used as the target classification model; when the output verification result is not standard, the training data is continued to be used to train and optimize the initial classification model based on the Monte Carlo algorithm and perform model verification until the output verification result is standard to obtain the target classification model.

[0023] In this embodiment, the classification categories and classification keywords of the target classification model correspond one-to-one with the classification categories and classification keywords set in the database. Then, in the database, multiple corresponding main folders and multiple subordinate folders are created according to the classification categories, including the main classification categories and the subordinate classification categories, wherein the main folders and the subordinate folders can be sorted according to the first letter of the folder name. At the same time, the classification results output by the target classification model include the main classification category results and the subordinate classification category results to which the information keyword belongs. According to the classification results, the data storage module can import the target information into the main folders and subordinate folders corresponding to the database step by step. In this embodiment, the information keywords stored in a specific storage location are reconstructed, and the construction includes at least a power data database, a basic data database and a barrier data database set, and each database in the database set is used to classify and store the key information.

[0024] The second implementation method: using the device status data collected in real time by the preset target sensor to perform data cleaning to obtain the target data, the process includes: S1: extracting a preset amount of historical device status data from a real-time database and dividing it into historical normal status data and historical abnormal status data, and performing dimension reduction and normalization processing on the historical normal status data to obtain first data; S2: Combine the Adam and SGD algorithms, use the first data as the input vector to train the neural network to obtain a parameter optimization model to find the optimal values ​​of all parameters and output the optimal feature vector; S3: Compare the input vector with the optimal feature vector to obtain a first reconstruction error under normal state data of the device; S4: inputting the device status data collected in real time as a first input vector into the parameter optimization model, outputting a first eigenvector, and then comparing the first eigenvector with the first input vector to obtain a second reconstruction error; S5: Compare the first reconstruction error and the second reconstruction error at each time point to determine the data type; S6: Based on the determined data type, all noise data and missing values ​​are automatically repaired into their corresponding reconstructed values ​​to complete data cleaning.

[0025] In this embodiment, the first reconstruction error and the second reconstruction error at each time point are compared to determine the data type, specifically: When the first reconstruction error is equal to the second reconstruction error at a time point, the data is judged to be a normal value; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is no such same situation around the time point, the data is judged to be noise data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is data around the time point with a fixed value equal to the first reconstruction error, the data is judged to be missing value data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and the data around the time point have a similar change trend, the data is judged to be a truly useful outlier.

[0026] In this embodiment, the target sensor is a device that is pre-installed on the power plant equipment for collecting data, such as a temperature sensor; the equipment status data refers to the data generated by the equipment during operation, such as air temperature, current, and voltage; the target data is the data obtained after cleaning the equipment status data, and its purpose is to improve data quality by identifying useful abnormal data and filling missing values ​​and noise data, which is helpful for accurate data analysis; the target features include data distribution characteristics, data types, data discreteness, data concentration, etc.

[0027] In this embodiment, the real-time database is used to store device status data and corresponding feature data; the preset amount is set in advance; the device status data refers to the data generated by the device during operation, such as air temperature, current, and voltage; the first data is the data obtained after dimension reduction and normalization of the historical normal state data; the parameter optimization model is based on the Adam and SGD algorithms, and the model obtained by training the neural network with the first data as the input vector can directly extract the hidden features of the data and correctly distinguish between useless data and abnormal data reflecting equipment failure; The relational database is used to store the results of data logic analysis, and while saving the data, it also saves the logical relationships between the data, such as linear relationships and non-linear relationships, which is conducive to subsequent data query and application; the file database is used to store semi-structured data and unstructured data such as charts, reports, videos and audio; the actual equipment operation data refers to the equipment operation video captured by the camera.

[0028] In this embodiment, the optimal values ​​of all parameters include the number of hidden layers, learning rate, and noise ratio; the optimal feature vector is the vector output by using the first data as the input vector to train the neural network to obtain a parameter optimization model to seek the optimal values ​​of all parameters; the first reconstruction error is the degree of difference between the optimal feature vector and the input vector; the first input vector refers to the device status data collected in real time; the second reconstruction error is the degree of difference between the first feature vector and the first input vector; the data types include normal data, abnormal data, missing value data, and noise data.

[0029] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art, and the model parameters of each electrical appliance are not specifically limited, and conventional equipment can be used.

[0030] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0031] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for processing system database data of a nitrogen saving device, characterized in that: The following steps are involved: Step 1: pre-processing the nitrogen-saving device status data collected in real time to obtain target data, and storing it in the database according to a predetermined storage strategy; The target data is obtained by cleaning the equipment status data collected in real time by the preset target sensor. The process includes: S1: extracting a preset amount of historical device status data from a real-time database and dividing it into historical normal status data and historical abnormal status data, and performing dimension reduction and normalization processing on the historical normal status data to obtain first data; S2: Combine the Adam and SGD algorithms, use the first data as the input vector to train the neural network to obtain a parameter optimization model to find the optimal values ​​of all parameters and output the optimal feature vector; S3: Compare the input vector with the optimal feature vector to obtain a first reconstruction error under normal state data of the device; S4: inputting the device status data collected in real time as a first input vector into the parameter optimization model, outputting a first eigenvector, and then comparing the first eigenvector with the first input vector to obtain a second reconstruction error; S5: Compare the first reconstruction error and the second reconstruction error at each time point to determine the data type; S6: Based on the determined data type, all noise data and missing values ​​are automatically repaired into their corresponding reconstructed values ​​to complete data cleaning; Step 2: Based on the coincidence comparison algorithm, the historical information is screened using the written data of the database, the target information is generated and the coincidence database is updated; Step 3: Use crawler technology to crawl information keywords of target information, input the information keywords into the preset target classification model to output the corresponding classification results; Step 4: Import the target information into the database according to the classification results and information keywords; Step 5: Analyze the storage evaluation index of the power data corresponding to each nitrogen-saving device, the storage evaluation index of the basic data, and the storage evaluation index of the obstacle data; Step 6: Clear the target data of each nitrogen-saving device according to the corresponding appropriate replacement cycle.

2. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The database is composed of a real-time database, a relational database and a file database.

3. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The first reconstruction error and the second reconstruction error at each time point are compared to determine the data type, specifically: When the first reconstruction error is equal to the second reconstruction error at a time point, the data is judged to be a normal value; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is no such same situation around the time point, the data is judged to be noise data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and there is data around the time point with a fixed value equal to the first reconstruction error, the data is judged to be missing value data; When the first reconstruction error and the second reconstruction error at a time point are not equal, and the data around the time point have a similar change trend, the data is judged to be a truly useful outlier.

4. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The target data includes power data, basic data and obstacle data; The power data includes the current, voltage and power corresponding to each detection time point, the basic data includes the total amount of input air, the total amount of output ammonia and the purity of ammonia corresponding to each detection time period, and the fault data includes the type and maintenance time corresponding to each fault; The abnormal power data include the current, voltage and power corresponding to each abnormal time point of each abnormal detection cycle; the abnormal basic data include the total input air volume, the total output ammonia volume and the ammonia purity corresponding to each risk detection period of each risk detection cycle; the fault data include the type and maintenance time corresponding to each fault of each fault detection cycle; the normal power data include the normal current range, the normal voltage range and the normal power range; the normal basic data include the appropriate output ammonia volume and the appropriate ammonia purity corresponding to each total input air volume.

5. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The overlap comparison algorithm is used to compare the initial information with the written data to output overlap information, and to perform a de-duplication operation on the initial information based on the overlap information to generate target information.

6. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The target classification model is based on a Monte Carlo algorithm. The training data is used to construct an initial classification model. The test data is used to verify the initial classification model and the verification result is output. If the verification result is standard, the initial classification model is used as the target classification model. If the verification result is non-standard, the training data is used to train and optimize the initial classification model, and the step of using the test data to verify the initial classification model and outputting the verification result is jumped to execute.

7. The method for processing system database data of a nitrogen saving device according to claim 1, characterized in that: The information keywords stored in the specific storage location are reconstructed, and the construction includes at least a power data database, a basic data database and a barrier data database set, and each database in the database set is used to classify and store the key information.

Images