Satellite-ground integrated full-process data intelligent monitoring and early warning method and system

Abstract

The invention discloses a satellite-ground integrated full-process data intelligent monitoring and early warning method and system, and relates to the technical field of data processing. Performing data change analysis on the plurality of historical data sets to obtain a data analysis result, and configuring a plurality of data acquisition frequencies; executing real-time data acquisition according to the plurality of data acquisition frequencies to obtain a plurality of time series data sets; constructing a data exception identification channel based on the plurality of historical data sets, synchronizing the plurality of time series data sets to the data exception identification channel, and obtaining a plurality of data exception identification results; and performing anomaly tracing based on an anomaly identification result, determining a data anomaly information set, and executing early warning reporting. According to the method and the device, the technical problem of insufficient data exception identification accuracy caused by fixed data acquisition frequency and lack of deep data analysis means in satellite-ground integrated data acquisition and monitoring in the prior art is solved, and the accuracy and the flexibility of exception identification and early warning are improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and specifically to a method and system for intelligent monitoring and early warning of full-process data in a satellite-ground integrated manner. Background Art

[0002] With the continuous development of satellite technology, satellites have been widely used in many fields such as earth observation, communication, navigation, etc., especially in space weather monitoring and satellite health status monitoring, playing a vital role. Satellite data acquisition and monitoring systems can collect key information such as satellite orbits, payload conditions, satellite-to-ground link status, etc. in real time, and by analyzing these data, timely identify potential faults and anomalies to ensure the normal progress of satellite missions. However, the acquisition frequency of existing satellite data acquisition and monitoring systems is usually statically set and cannot be flexibly adjusted according to the data change rate and importance, resulting in the inability to capture key information in a timely manner when the data fluctuates greatly or in abnormal situations. Secondly, in terms of anomaly identification, existing systems mostly rely on simple threshold settings, ignoring the complex relationship between data, resulting in the inability to accurately identify potential problems and increasing the risk of false alarms and missed reports. Summary of the invention

[0003] The present application provides a method and system for intelligent monitoring and early warning of the entire process of satellite-ground integrated data, which solves the technical problem in the prior art of satellite-ground integrated data collection and monitoring, which leads to insufficient accuracy in identifying data anomalies due to the fixed data collection frequency and lack of in-depth data analysis methods, and achieves the technical effect of improving the accuracy and flexibility of anomaly identification and early warning.

[0004] In view of the above problems, on the one hand, the present application provides a satellite-ground integrated full-process data intelligent monitoring and early warning method, the method comprising:

[0005] Traverse multiple data sources to obtain multiple historical data sets; perform data change analysis on the multiple historical data sets to obtain data analysis results, and configure multiple data collection frequencies based on the analysis results; perform real-time data collection of the multiple data sources according to the multiple data collection frequencies to obtain multiple time series data sets; perform deep learning based on the multiple historical data sets, build a data anomaly identification channel, synchronize the multiple time series data sets to the data anomaly identification channel, and obtain multiple data anomaly identification results; perform anomaly tracing based on the anomaly identification results, determine the data anomaly information set, and perform early warning alarm for the data anomaly information set.

[0006] On the other hand, the present application also provides a satellite-ground integrated full-process data intelligent monitoring and early warning system, the system including: a historical data extraction module, used to traverse multiple data sources and obtain multiple historical data sets; a collection frequency configuration module, used to perform data change analysis on the multiple historical data sets, obtain data analysis results, and configure multiple data collection frequencies based on the analysis results; a real-time data collection module, used to perform real-time data collection of the multiple data sources according to the multiple data collection frequencies, and obtain multiple time series data sets; an anomaly identification module, used to perform deep learning based on the multiple historical data sets, build a data anomaly identification channel, synchronize the multiple time series data sets to the data anomaly identification channel, and obtain multiple data anomaly identification results; an anomaly early warning module, used to trace the anomaly based on the anomaly identification results, determine the data anomaly information set, and perform early warning on the data anomaly information set.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] By collecting comprehensive historical data related to satellites and space weather from multiple independent sources, sufficient data support is provided for subsequent data analysis and modeling. By performing data change analysis on the multiple historical data sets, data analysis results are obtained, and multiple data collection frequencies are configured based on the analysis results. The data collection frequency can be dynamically adjusted according to the volatility and importance of different data sources, thereby improving data collection efficiency and avoiding redundant collection and omissions. According to the multiple data collection frequencies, real-time data collection of the multiple data sources is performed to obtain multiple time series data sets, which improves the timeliness and integrity of the data and provides real-time data support for subsequent anomaly identification. By performing deep learning on multiple historical data sets, learning anomaly discrimination in historical data, building a data anomaly identification channel, synchronizing the multiple time series data sets to the data anomaly identification channel, identifying abnormal patterns and trends in the data, and obtaining multiple data anomaly identification results, the accuracy of anomaly detection is improved. Based on the anomaly identification results, anomaly tracing is performed to determine the specific source of the data anomaly, and early warning is issued, which improves the efficiency and accuracy of anomaly handling.

[0009] In summary, this application achieves comprehensive optimization of satellite data collection and monitoring. By dynamically adjusting the collection frequency, the efficiency and accuracy of data collection are improved, and resource waste and data omissions are avoided. By building a data anomaly recognition channel through deep learning, the accuracy and efficiency of anomaly recognition are improved, and anomalies in the data can be discovered in a timely manner. Through anomaly tracing and early warning, the efficiency and accuracy of anomaly handling are improved, and data anomalies can be discovered and handled in a timely manner to reduce losses. Overall, this solution significantly improves the accuracy and flexibility of satellite data monitoring and early warning, and provides more reliable data support for research and applications in related fields.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A flow chart of a satellite-ground integrated full-process data intelligent monitoring and early warning method provided in an embodiment of the present application.

[0012] Figure 2 A flow chart of configuring multiple data collection frequencies in the satellite-ground integrated full-process data intelligent monitoring and early warning method provided in an embodiment of the present application.

[0013] Figure 3 A schematic diagram of the process of constructing a data anomaly identification channel in the satellite-ground integrated full-process data intelligent monitoring and early warning method provided in an embodiment of the present application.

[0014] Figure 4 A schematic diagram of the structure of the satellite-ground integrated full-process data intelligent monitoring and early warning system provided in an embodiment of the present application.

[0015] Explanation of the reference numerals: historical data extraction module 10 , collection frequency configuration module 20 , real-time data collection module 30 , abnormality identification module 40 , abnormality warning module 50 . DETAILED DESCRIPTION

[0016] The embodiments of the present application provide a method and system for intelligent monitoring and early warning of the entire process of satellite-ground integrated data, thereby solving the technical problem in the prior art of satellite-ground integrated data collection and monitoring, which leads to insufficient accuracy in identifying data anomalies due to the fixed data collection frequency and the lack of in-depth data analysis methods, thereby achieving the technical effect of improving the accuracy and flexibility of anomaly identification and early warning.

[0017] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a satellite-ground integrated full-process data intelligent monitoring and early warning method, the method comprising:

[0018] Step S1: traverse multiple data sources to obtain multiple historical data sets.

[0019] Specifically, multiple data sources include but are not limited to weather stations, satellite sensors, ground monitoring equipment, etc. Historical data is collected from various data sources through satellite communications, ground networks, etc. The collected data is sorted and classified according to the data source to form multiple historical data sets. These historical data include the collection time, collection location, data source, data type, and specific data values. Exemplarily, in a satellite data collection and monitoring system, the following data sources are included: optical remote sensing satellite: provides image data of the surface; synthetic aperture radar (SAR) satellite: provides radar image data of the surface; ground meteorological station: provides meteorological data, such as temperature, humidity, wind speed, etc.; ground sensor: provides environmental data of specific locations, such as soil moisture, water quality, etc.

[0020] By obtaining historical data sets, basic data is provided for subsequently determining the frequency of data collection and building anomaly identification channels.

[0021] Step S2: Perform data change analysis on the multiple historical data sets to obtain data analysis results, and configure multiple data collection frequencies based on the analysis results.

[0022] Specifically, analyze multiple historical data sets one by one, and use statistical analysis tools to identify the changing characteristics of historical data, such as calculating the coefficient of variation, standard deviation, mean and other indicators to determine the changing trend of historical data. According to the identified changing trend, adjust the data collection frequency of each data source. For data sources with large data volatility and high importance, configure a higher data collection frequency, such as once every 5 minutes, to ensure timely acquisition of the latest data and improve the timeliness of the data; for data sources with slow data changes and low importance, configure a lower data collection frequency, such as once every 3 hours, to avoid excessive collection of stable data, thereby saving data collection resources.

[0023] By dynamically configuring the data collection frequency, we can respond to actual data changes more flexibly and efficiently, thereby improving the continuity and accuracy of data monitoring, while also providing an optimized data collection strategy for subsequent real-time data collection and anomaly identification.

[0024] Step S3: Execute real-time data collection of the multiple data sources according to the multiple data collection frequencies to obtain multiple time series data sets.

[0025] Specifically, the data collection task is started according to the data collection frequency configured in step S2, real-time data is obtained from the corresponding multiple data sources, and then the data received from each data source in chronological order are sorted into multiple time series data sets. For example, the vegetation coverage index data of a certain area collected from the satellite in chronological order is sorted into a time series data set, and the temperature, air pressure and other data collected from the ground meteorological station in chronological order are sorted into another time series data set. Real-time data collection is performed according to the set data collection frequency to ensure that the acquired data reflects the current or near-current state.

[0026] Step S4: Perform deep learning based on the multiple historical data sets, build a data anomaly recognition channel, synchronize the multiple time series data sets to the data anomaly recognition channel, and obtain multiple data anomaly recognition results.

[0027] Specifically, the data anomaly identification channel is a logical structure or model channel built based on deep learning for identifying data anomalies. Features are extracted from multiple historical data sets and used to train deep learning models. These models may be convolutional neural networks, recurrent neural networks, long short-term memory networks, or other models suitable for time series data analysis. After training, the data anomaly identification channel is built based on these models. As new data continues to flow in, the deep learning model will analyze this data in real time and compare it with the patterns learned during training to identify anomalies in these monitoring data.

[0028] Through data anomaly recognition, real-time monitoring and anomaly detection of satellite-ground integrated data can be achieved. Compared with traditional rule-based systems, the anomaly recognition method based on deep learning can more accurately capture complex patterns and subtle changes in the data, greatly improving the accuracy of identifying data anomalies and the flexibility of early warning.

[0029] Step S5: Tracing the source of the anomaly based on the anomaly identification result, determining a data anomaly information set, and issuing a pre-alarm for the data anomaly information set.

[0030] Specifically, the anomaly identification results are analyzed in detail to trace the root causes of the data anomalies to determine the specific information of the abnormal data, including the source of the abnormal data, the type of abnormal data, the time point of the data anomaly, etc., thereby generating a data anomaly information set.

[0031] Obtain the preset warning rules in the satellite data acquisition and monitoring system, which is a mapping table of data anomaly information and warning level. Match the data anomaly information set in the warning rules to determine the corresponding warning level. For example, if the severity of the anomaly is "serious" and the impact range is "whole system", it may match a "high level" warning. Then, according to the matched warning level, configure a multi-level warning channel. For low-level warnings, it may only update the dashboard to provide information prompts; for medium-level warnings, it may be necessary to notify relevant responsible personnel through SMS or email; and for high-level warnings, it may trigger the emergency response system and quickly convey it through real-time push and other methods. Generate corresponding warning information based on the data anomaly information set and warning level, which can include the abnormal location, type, severity and recommended operations, etc., and report the warning information through the configured warning channel to ensure that relevant personnel can understand the abnormal situation in time and take corresponding actions.

[0032] Through abnormal tracing and early warning output, we can respond to different abnormal situations more flexibly and efficiently, improving the practicality and effectiveness of the entire monitoring and early warning process.

[0033] Further, such as Figure 2 As shown, step S2 of the embodiment of the present application includes:

[0034] Step S21: extracting a first historical data set from the multiple historical data sets, performing data segmentation on the first historical data set according to a preset time interval to obtain multiple first historical data subsets, wherein the first historical data set corresponds to a first data source.

[0035] Step S22: traverse the multiple first historical data subsets to perform data fluctuation analysis, and calculate multiple first data variation coefficients, where the multiple first data variation coefficients correspond to the first historical data subsets one by one.

[0036] Step S23: Calculate the mean of the plurality of first data variation coefficients to obtain the first data variation coefficient mean.

[0037] Step S24: performing importance evaluation on the first data source to obtain a first importance coefficient.

[0038] Step S25: Determine a weight coefficient of the first data source based on the first data variation coefficient mean and the first importance coefficient, and adjust the original collection frequencies of the multiple data sources based on the weight coefficient to obtain a first data collection frequency.

[0039] Specifically, the preset time interval is a pre-set time length, such as 1 hour, 1 day or 1 week. A specific historical data set is randomly selected from multiple historical data sets and recorded as a first historical data set. The first historical data set is divided into multiple smaller data subsets according to a preset time interval (such as every hour, every day or every week) to obtain multiple first historical data subsets. Through historical data segmentation, data in different time periods can be analyzed separately to more accurately capture the changing trend and fluctuation of data over time.

[0040] The coefficient of variation is a statistic that measures the degree of data dispersion. Check the data of each first historical data subset one by one and analyze the fluctuation of the data in the subset. The coefficient of variation of each data subset is obtained by statistical calculation method. This coefficient reflects the fluctuation of the data in the subset relative to the data mean. By performing fluctuation analysis on multiple first historical data subsets, multiple first data coefficients of variation corresponding to multiple first historical data subsets can be obtained.

[0041] The calculated coefficients of variation of all first data are averaged, and the average value can comprehensively reflect the overall fluctuation of the first historical data set after being divided according to the preset time interval.

[0042] The importance of the first data source is evaluated, and the accuracy and completeness of the data provided by the data source, as well as the criticality of the data content, are considered. The first importance coefficient corresponding to the first data source is determined through expert evaluation. This coefficient reflects the importance of the data source to the entire monitoring system. The higher the importance coefficient, the greater the impact of the data source on the system.

[0043] The first data variation coefficient mean and the first importance coefficient are calculated comprehensively according to certain rules to obtain the weight coefficient of the first data source. This weight coefficient can reflect the comprehensive considerations of the first data source in adjusting the data collection frequency. For example, the weight coefficient can be obtained by a simple weighted average. According to the weight coefficient of the first data source, its original data collection frequency is adjusted. If the weight coefficient is high, the collection frequency will be increased to obtain more data; if the weight coefficient is low, the collection frequency will be appropriately reduced to save resources. For example, the original collection frequency of each data source is obtained, and the original collection frequency is weighted by the weight coefficient to obtain a new data collection frequency.

[0044] According to the same operation as step S21 to step S25, data change analysis is performed on multiple historical data sets to obtain corresponding multiple data analysis results, and corresponding data collection frequencies are configured for multiple data sources according to multiple data analysis structures.

[0045] The above steps can dynamically adjust the frequency of data collection according to the data fluctuation characteristics and importance, so that the collection strategy can be flexibly adjusted according to the actual changes and importance of the data, thereby improving the efficiency and response speed of data monitoring.

[0046] Furthermore, the method described in the embodiment of the present application also includes:

[0047] Step S2-1: setting an update evaluation time window, traversing the multiple data sources to collect data based on the update evaluation time window, and extracting multiple update history data sets.

[0048] Step S2-2: Perform data change analysis based on the multiple update history data sets to obtain update data analysis results, and configure multiple update data collection frequencies based on the update data analysis results.

[0049] Step S2-3: performing updated real-time data collection of the multiple data sources according to the multiple updated data collection frequencies.

[0050] Specifically, the update evaluation time window is a set time range used to periodically re-evaluate the data collection frequency. Starting from the last configuration of the monitoring collection frequency, the update evaluation time window is started. Based on this time window, all data sources are re-traversed to collect new data to form an update history data set. This update history data set is a new history data set collected within the update evaluation time window.

[0051] According to the same process as step S21 to step S25, data change analysis is performed on the collected multiple update history data sets to obtain update data analysis results, and the data collection frequency is reconfigured according to the update data analysis results to determine multiple update data collection frequencies to adapt to new changes in data.

[0052] Real-time data collection for the multiple data sources begins again according to the reconfigured update data collection frequency.

[0053] Through the above steps, the data collection frequency can be dynamically adjusted and the real-time data can be continuously updated, thereby improving the adaptability and accuracy of the entire monitoring and early warning system. This periodic evaluation and update mechanism can flexibly respond to data changes and maintain the timeliness and effectiveness of monitoring and early warning.

[0054] Further, such as Figure 3 As shown, step S4 of the embodiment of the present application includes:

[0055] Step S41: extracting data types of the plurality of historical data sets, dividing the historical data sets according to the data types, and determining a plurality of historical data classification sets, wherein each historical data classification set is marked with a data type identifier.

[0056] Step S42: Based on the data type identifier, traverse the deep learning model sequence to perform model matching, and determine multiple target models, where the multiple target models are marked with the data type identifier.

[0057] Step S43: training the multiple target models according to the multiple historical data classification sets to obtain multiple data anomaly recognition models.

[0058] Step S44: connecting the multiple data anomaly recognition models in parallel to construct the abnormal data recognition channel.

[0059] Specifically, historical data sets contain many different types of data, including numerical data (such as satellite orbit data, payload condition data, satellite-to-ground link status data), image data (such as surface image data taken by satellites), text data (such as descriptive reports related to space weather data), etc. Data types are extracted from multiple historical data sets, and the data sets are divided into multiple historical data classification sets based on these types. Each classification set is marked with the corresponding data type identifier.

[0060] A deep learning model sequence refers to a series of predefined deep learning models, each of which is labeled with its applicable data type. For example, convolutional neural networks are suitable for processing image data, and recurrent neural networks and their variants are suitable for processing sequence data. Based on the data type identifier, the deep learning model sequence is traversed for model matching to determine the target model applicable to each historical data classification set.

[0061] Use the data in each historical data classification set as training samples to train the corresponding target model. For example, the data in the numerical historical data classification set is used to train the matched target model suitable for numerical data processing. During the training process, the model learns the normal patterns and abnormal features in the data. After multiple iterations of training, a model that can identify whether the data of this type is abnormal is obtained, that is, a data anomaly recognition model. Since there are multiple historical data classification sets and corresponding target models, multiple data anomaly recognition models will eventually be obtained.

[0062] Multiple data anomaly recognition models are connected in parallel to build a comprehensive abnormal data recognition channel. This channel can process data from different data sources and types at the same time to achieve comprehensive monitoring and recognition of data anomalies. When inputting data, different types of data will be sent to the data anomaly recognition model corresponding to their data types for processing. Each model independently determines whether the data is abnormal, and then combines the results of each model to achieve anomaly recognition of multiple types of data.

[0063] Further, step S43 includes:

[0064] Step S431: Select a first target model from the multiple target models, where the first target model is marked with a first data type identifier.

[0065] Step S432: using the first data type identifier as an index, traverse the multiple historical data classification sets to extract multiple first training data sets.

[0066] Step S433: Divide the first training data set into a first training set and a first validation set.

[0067] Step S434: Train a first target model based on the first training set, and verify the first target model based on the first verification set to obtain a first data anomaly recognition model.

[0068] Step S435: Similarly, the multiple target models are trained according to the multiple historical data classification sets to obtain multiple data anomaly recognition models.

[0069] Specifically, a target model is randomly selected from multiple target models and defined as a first target model. The first target model is marked with a first data type identifier, which clarifies the data type that the model can process. For example, if the first data type identifier is "satellite image data", then the first target model is a model specifically matched for processing satellite image data.

[0070] Since each historical data classification set has its data type identifier, the first data type identifier of the first target model is used as a basis for searching among multiple historical data classification sets, and all historical data classification sets marked with the first data type identifier are extracted. These historical data classification sets are used as the first training data set for training the first target model.

[0071] In order to effectively train and evaluate the first target model, the first training data set needs to be divided into two parts, namely the first training set and the first validation set. The first training set is used to train the model so that the model can learn the features and patterns in the data. The first validation set is used to verify the performance of the model during the training process to ensure that the model is not overfitted and can generalize on unseen data. Data partitioning can be randomly divided according to a fixed ratio, such as 80% of the data as a training set and 20% of the data as a validation set.

[0072] The data in the first training set is input into the first target model, and the calculation is performed according to the structure and algorithm of the model. For example, if the first target model is a convolutional neural network (CNN), the data will be sequentially passed through the convolution layer, pooling layer, fully connected layer, etc. for feature extraction and classification calculation. The model continuously adjusts its own parameters (such as the weight of the convolution kernel, etc.) according to the input data so that the output result matches the true label (normal or abnormal) in the data as much as possible. After multiple rounds of iterative training, the performance of the model on the training set meets the expected requirements. After each round of training or at regular intervals, the data of the first validation set is input into the first target model being trained to obtain the prediction results of the model. Then, the prediction results are compared with the true labels of the data in the validation set to calculate evaluation indicators such as accuracy, recall rate, and F1 value. These indicators can reflect the performance of the model on unseen data (validation set), so as to determine whether the model is overfitting or whether it needs to continue training. When the performance of the first target model on the training set reaches a high level and the performance on the validation set is relatively stable (for example, the accuracy reaches a certain threshold and no longer fluctuates significantly), a trained and validated first data anomaly recognition model is obtained. This model can perform anomaly recognition on the data type (such as image data) corresponding to the first data type identifier.

[0073] According to the method of step S431 to step S434, each target model is operated in turn. For each target model, the corresponding historical data classification set is selected, and the training set and the verification set are divided, and training and verification are performed, and finally multiple data anomaly recognition models are obtained. Each data anomaly recognition model can perform anomaly recognition on a specific type of data.

[0074] Through the above steps, a corresponding anomaly recognition model is trained for each type of data, thereby improving the accuracy and pertinence of abnormal data detection.

[0075] Furthermore, step S5 of the embodiment of the present application includes:

[0076] Step S51: performing data source backtracking on the abnormal identification result to determine multiple abnormal data sources and multiple abnormal collection time points.

[0077] Step S52: performing abnormal classification identification on the abnormal identification result, and determining a plurality of abnormal data types based on data features and abnormal patterns.

[0078] Step S53: matching a plurality of abnormal data processing strategies from a preset strategy library according to the plurality of abnormal data sources and the plurality of abnormal data types.

[0079] Step S54: Integrate the multiple abnormal data sources, multiple abnormal collection time points, multiple abnormal data types, and multiple abnormal data processing strategies to generate a data anomaly information set.

[0080] Specifically, after obtaining the abnormal data identification results, the data source of the abnormal identification results is traced back. By checking the metadata information in the data record, it is found from which data sources these abnormal data are obtained and when they are collected, and multiple abnormal data sources and multiple abnormal collection time points corresponding to multiple abnormal data identification results are determined. For example, in the satellite monitoring system, the abnormal data of the satellite-to-ground link status data is detected through the abnormal data identification channel, which is manifested as a sudden drop in the data transmission rate. By querying the data management system, it can be determined that the abnormal data originated from a specific satellite sensor, and the abnormal collection time point is 14:00 on a certain day of a certain month of a certain year. By tracing back the data source, the satellite sensor with the problem can be quickly located, and the data changes before and after that time point can be viewed.

[0081] The features and patterns of abnormal data are analyzed, the anomalies are classified into different categories, and multiple abnormal data types corresponding to multiple abnormal data identification results are determined. For example, by analyzing the data features and abnormal patterns, it is identified that this anomaly belongs to the "data transmission interruption" category. The abnormality classification helps to determine the specific cause of the abnormality in the future, such as satellite sensor failure, ground receiving station problems, or signal interference.

[0082] The preset strategy library is a pre-established library containing various response strategies for abnormal situations of different data sources and data types. For example, for anomalies of the "data transmission interruption" type, the strategy library may include the following strategies: restarting the satellite sensor, switching to an alternate receiving station, or adjusting the signal transmission frequency to reduce interference. Based on the multiple abnormal data sources and multiple abnormal data types that have been determined, the corresponding processing strategy is searched in the preset strategy library, such as deciding to try restarting the satellite sensor first.

[0083] The identified multiple abnormal data sources, multiple abnormal collection time points, multiple abnormal data types, and multiple abnormal data processing strategies are aggregated and integrated to generate a data anomaly information set. This information set contains all key information about data anomalies and can be used for subsequent decision-making, reporting, or further processing, such as generating an abnormality report to send to relevant departments or as a basis to trigger an automated processing flow.

[0084] Furthermore, the method described in the embodiment of the present application also includes:

[0085] Step S61: Setting a feedback monitoring window based on the data anomaly information set.

[0086] Step S62: Acquire feedback monitoring data according to the feedback monitoring window and the data collection frequency, input the feedback monitoring data into the data anomaly identification channel for abnormal feedback identification, and acquire abnormal feedback data information.

[0087] Step S63: updating the early warning alarm based on the abnormal feedback data information.

[0088] Specifically, the feedback monitoring window refers to a set time or data range used to monitor and evaluate the data status after exception processing. A feedback monitoring window is set based on the data anomaly information set to evaluate the effectiveness of exception processing and the overall health of the system. For example, if the type of abnormal data is a sudden drop in a key performance indicator, it may be necessary to set a shorter feedback monitoring window (such as 1 to 2 hours) to timely monitor the effect of the processing strategy after implementation; if it is a slight deviation in a long-term statistical data, the feedback monitoring window can be set longer (such as 1 to 2 days). By setting the feedback monitoring window, it is possible to actively monitor whether the exception handling measures are effective and whether further adjustments or optimizations are needed.

[0089] Feedback monitoring data refers to data collected within the feedback monitoring window, which is used to evaluate the effectiveness of exception handling. Feedback monitoring data is obtained based on the feedback monitoring window and data collection frequency. The data collection frequency determines the data collection interval, while the feedback monitoring window limits the time range for obtaining feedback monitoring data. For example, if the data collection frequency is once every 15 minutes and the feedback monitoring window is set to 2 hours, then within these 2 hours, data is collected at intervals of every 15 minutes, and this data is the feedback monitoring data. These feedback monitoring data are input into the aforementioned data anomaly recognition channel for abnormal feedback recognition. Each data anomaly recognition model in the channel will judge the data based on its own algorithm and training results to identify whether there are still anomalies and generate abnormal feedback data information.

[0090] Based on the latest abnormal feedback data information, the current warning status is evaluated and the warning signal is updated accordingly. If the feedback data shows that the abnormality has been resolved, the warning signal may be lifted or lowered; if the abnormality still exists or new problems arise, the warning signal may be updated to reflect the new situation. The updated warning signal is sent to the relevant maintenance personnel or management system through the configured warning channel. This process ensures the dynamic and adaptable nature of the warning system, and can adjust the warning level and response measures in a timely manner based on real-time data.

[0091] In summary, the satellite-ground integrated full-process data intelligent monitoring and early warning method provided in the embodiment of the present application has the following technical effects:

[0092] The embodiment of the present application obtains historical data from multiple data sources and dynamically adjusts the acquisition frequency according to the change analysis. This adaptive data acquisition frequency mechanism can ensure that the acquisition frequency is increased when the data changes dramatically, and the acquisition frequency is reduced when the data changes slowly, thereby optimizing resource allocation and improving the efficiency and accuracy of data acquisition. The application of deep learning models not only improves the accuracy of data anomaly identification, but also realizes the accurate prediction and timely tracing of potential problems through the analysis of real-time data. Further, through the feedback monitoring and update mechanism, the scheme enhances the intelligent early warning effect on the basis of anomaly detection, thereby improving the accuracy of satellite data monitoring and the early warning response speed in actual operation, and can timely discover and handle data anomalies, improving the efficiency and accuracy of anomaly handling. Overall, the embodiment of the present application significantly improves the accuracy and flexibility of satellite data monitoring and early warning, and provides more reliable data support for research and application in related fields.

[0093] Embodiment 2, as Figure 4 As shown, the embodiment of the present application provides a satellite-ground integrated full-process data intelligent monitoring and early warning system, the system comprising:

[0094] The historical data extraction module 10 is used to traverse multiple data sources and obtain multiple historical data sets.

[0095] The collection frequency configuration module 20 is used to perform data change analysis on the multiple historical data sets, obtain data analysis results, and configure multiple data collection frequencies based on the analysis results.

[0096] The real-time data collection module 30 is used to perform real-time data collection of the multiple data sources according to the multiple data collection frequencies to obtain multiple time series data sets.

[0097] The anomaly identification module 40 is used to perform deep learning based on the multiple historical data sets, build a data anomaly identification channel, synchronize the multiple time series data sets to the data anomaly identification channel, and obtain multiple data anomaly identification results.

[0098] The abnormality warning module 50 is used to trace the abnormality source based on the abnormality identification result, determine the data abnormality information set, and perform early warning on the data abnormality information set.

[0099] Furthermore, the acquisition frequency configuration module 20 of the embodiment of the present application is also used to perform the following steps:

[0100] Extract a first historical data set from the multiple historical data sets, perform data segmentation on the first historical data set according to a preset time interval to obtain multiple first historical data subsets, the first historical data set corresponds to a first data source; traverse the multiple first historical data subsets to perform data fluctuation analysis, calculate multiple first data variation coefficients, the multiple first data variation coefficients correspond one-to-one to the first historical data subsets; calculate the mean of the multiple first data variation coefficients to obtain the mean of the first data variation coefficients; perform importance assessment on the first data source to obtain a first importance coefficient; determine a weight coefficient of the first data source by combining the mean of the first data variation coefficient and the first importance coefficient, and adjust the original collection frequencies of the multiple data sources based on the weight coefficients to obtain a first data collection frequency.

[0101] Furthermore, the system described in the embodiment of the present application is also used to perform the following steps:

[0102] An update evaluation time window is set, and based on the update evaluation time window, the multiple data sources are traversed to perform data collection, and multiple update history data sets are extracted; data change analysis is performed based on the multiple update history data sets to obtain update data analysis results, and multiple update data collection frequencies are configured based on the update data analysis results; and update real-time data collection of the multiple data sources is executed according to the multiple update data collection frequencies.

[0103] Furthermore, the abnormality identification module 40 in the embodiment of the present application is also used to perform the following steps:

[0104] Extract data types of the multiple historical data sets, divide the historical data sets according to the data types, and determine multiple historical data classification sets, wherein each historical data classification set is labeled with a data type identifier; based on the data type identifier, traverse the deep learning model sequence for model matching to determine multiple target models, wherein the multiple target models are labeled with the data type identifier; train the multiple target models according to the multiple historical data classification sets to obtain multiple data anomaly recognition models; connect the multiple data anomaly recognition models in parallel to construct the abnormal data recognition channel.

[0105] Furthermore, the abnormality identification module 40 in the embodiment of the present application is also used to perform the following steps:

[0106] A first target model is selected from the multiple target models, wherein the first target model is marked with a first data type identifier; taking the first data type identifier as an index, the multiple historical data classification sets are traversed to extract multiple first training data sets; the first training data sets are divided into a first training set and a first validation set; the first target model is trained based on the first training set, and the first target model is validated based on the first validation set to obtain a first data anomaly recognition model; and so on, the multiple target models are trained according to the multiple historical data classification sets to obtain multiple data anomaly recognition models.

[0107] Furthermore, the abnormal warning module 50 of the embodiment of the present application is also used to perform the following steps:

[0108] The data source of the anomaly identification result is traced back to determine multiple anomaly data sources and multiple anomaly collection time points; the anomaly identification result is classified and identified, and multiple anomaly data types are determined based on data features and anomaly patterns; multiple anomaly data processing strategies are matched from a preset strategy library according to the multiple anomaly data sources and the multiple anomaly data types; the multiple anomaly data sources, multiple anomaly collection time points, multiple anomaly data types, and multiple anomaly data processing strategies are integrated to generate a data anomaly information set.

[0109] Furthermore, the system described in the embodiment of the present application is also used to perform the following steps:

[0110] A feedback monitoring window is set based on the data anomaly information set; feedback monitoring data is acquired according to the feedback monitoring window and the data collection frequency, the feedback monitoring data is input into the data anomaly identification channel for abnormal feedback identification, and abnormal feedback data information is acquired; and an early warning alarm is updated based on the abnormal feedback data information.

[0111] Through the above-mentioned detailed description of the satellite-ground integrated full-process data intelligent monitoring and early warning method, technical personnel in this specification can clearly understand the satellite-ground integrated full-process data intelligent monitoring and early warning system in this embodiment. For the system disclosed in Example 2, since it corresponds to the method disclosed in Example 1 and has corresponding functional modules and beneficial effects, please refer to the method part for relevant details.

[0112] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A satellite-ground integrated full-process data intelligent monitoring and early warning method, characterized in that: include: Traverse multiple data sources to obtain multiple historical data sets; Performing data change analysis on the multiple historical data sets to obtain data analysis results, and configuring multiple data collection frequencies based on the analysis results; Execute real-time data collection of the multiple data sources according to the multiple data collection frequencies to obtain multiple time series data sets; Perform deep learning based on the multiple historical data sets, build a data anomaly recognition channel, synchronize the multiple time series data sets to the data anomaly recognition channel, and obtain multiple data anomaly recognition results; Based on the anomaly identification result, anomaly tracing is performed to determine a data anomaly information set, and a pre-alarm is issued for the data anomaly information set.

2. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 1 is characterized in that: Performing data change analysis on the multiple historical data sets to obtain data analysis results, and configuring multiple data collection frequencies based on the analysis results, including: Extracting a first historical data set from the multiple historical data sets, performing data segmentation on the first historical data set according to a preset time interval to obtain multiple first historical data subsets, wherein the first historical data set corresponds to a first data source; Traversing the plurality of first historical data subsets to perform data fluctuation analysis, and calculating a plurality of first data variation coefficients, wherein the plurality of first data variation coefficients correspond one-to-one to the first historical data subsets; Calculate the mean of the plurality of first data variation coefficients to obtain the first data variation coefficient mean; Performing importance evaluation on the first data source to obtain a first importance coefficient; The weight coefficient of the first data source is determined by comprehensively considering the first data variation coefficient mean and the first importance coefficient, and the original collection frequencies of the multiple data sources are adjusted based on the weight coefficient to obtain the first data collection frequency.

3. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 1 is characterized in that: The method further comprises: Setting an update evaluation time window, traversing the multiple data sources to collect data based on the update evaluation time window, and extracting multiple update history data sets; Performing data change analysis based on the multiple update history data sets to obtain update data analysis results, and configuring multiple update data collection frequencies based on the update data analysis results; Updated real-time data collection of the plurality of data sources is performed according to the plurality of updated data collection frequencies.

4. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 1 is characterized in that: Based on the multiple historical data sets, deep learning is performed to build a data anomaly recognition channel, including: Extracting data types of the plurality of historical data sets, dividing the historical data sets according to the data types, and determining a plurality of historical data classification sets, wherein each historical data classification set is marked with a data type identifier; Extracting data distribution features of the multiple historical data sets, and based on the data type identifier and the data distribution features, traversing a deep learning model sequence to perform model matching, and determining multiple target models, wherein the multiple target models are annotated with the data type identifier; Training the multiple target models according to the multiple historical data classification sets to obtain multiple data anomaly recognition models; The multiple data anomaly recognition models are connected in parallel to construct the abnormal data recognition channel.

5. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 4 is characterized in that: The multiple target models are trained according to the multiple historical data classification sets to obtain multiple data anomaly recognition models, including: Selecting a first target model from the multiple target models, wherein the first target model is marked with a first data type identifier; Using the first data type identifier as an index, traverse the multiple historical data classification sets to extract multiple first training data sets; Dividing the first training data set into a first training set and a first validation set; Training a first target model based on the first training set, and validating the first target model based on the first validation set to obtain a first data anomaly recognition model; By analogy, the multiple target models are trained according to the multiple historical data classification sets to obtain multiple data anomaly recognition models.

6. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 1 is characterized in that: Based on the anomaly identification result, anomaly tracing is performed to determine a data anomaly information set, including: Perform data source backtracking on the anomaly identification result to determine multiple anomaly data sources and multiple anomaly collection time points; Performing abnormal classification and identification on the abnormal identification results, and determining a plurality of abnormal data types based on data features and abnormal patterns; Matching a plurality of abnormal data processing strategies from a preset strategy library according to the plurality of abnormal data sources and the plurality of abnormal data types; The multiple abnormal data sources, multiple abnormal collection time points, multiple abnormal data types, and multiple abnormal data processing strategies are integrated to generate a data anomaly information set.

7. The satellite-ground integrated full-process data intelligent monitoring and early warning method according to claim 1 is characterized in that: The method further comprises: Setting a feedback monitoring window based on the data anomaly information set; Acquire feedback monitoring data according to the feedback monitoring window and the data collection frequency, input the feedback monitoring data into the data anomaly identification channel for abnormal feedback identification, and acquire abnormal feedback data information; An updated early warning alarm is issued based on the abnormal feedback data information.

8. The satellite-ground integrated full-process data intelligent monitoring and early warning system is characterized by: The system is used to execute the satellite-ground integrated full-process data intelligent monitoring and early warning method according to any one of claims 1 to 7, comprising: A historical data extraction module is used to traverse multiple data sources and obtain multiple historical data sets; A collection frequency configuration module, used to perform data change analysis on the multiple historical data sets, obtain data analysis results, and configure multiple data collection frequencies based on the analysis results; A real-time data collection module, used to perform real-time data collection of the multiple data sources according to the multiple data collection frequencies, and obtain multiple time series data sets; An anomaly recognition module, used to perform deep learning based on the multiple historical data sets, build a data anomaly recognition channel, synchronize the multiple time series data sets to the data anomaly recognition channel, and obtain multiple data anomaly recognition results; The abnormality warning module is used to trace the abnormality source based on the abnormality identification result, determine the data abnormality information set, and perform early warning on the data abnormality information set.

Images