Satellite multi-modal data analysis method and system for intelligent operation and maintenance

Abstract

The invention discloses a satellite multi-modal data analysis method and system for intelligent operation and maintenance, and relates to the technical field of satellite data analysis, and the method comprises the steps: obtaining multi-modal data in the operation process of a wind cloud satellite; performing standardization preprocessing to obtain standard multi-modal data; integrating the standard multi-modal data, and constructing a unified data view; carrying out real-time monitoring, obtaining real-time monitoring data as input, carrying out target detection and trend prediction, and obtaining a satellite operation state analysis result; and dynamically updating the data view based on a satellite operation state analysis result, and generating a visual satellite monitoring report. The technical problems that in the prior art, satellite data processing is large in multi-source data format protocol difference and poor in data quality, and satellite operation and maintenance efficiency and operation reliability are affected are solved, and the technical effects that the wind cloud satellite operation state is monitored comprehensively, accurately in real time, and satellite operation management efficiency and reliability are effectively improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite data analysis, and in particular to a satellite multimodal data analysis method and system for intelligent operation and maintenance. Background Art

[0002] In the field of satellite technology, various satellites such as Fengyun satellites play an irreplaceable and important role in many aspects such as meteorological observation, environmental monitoring, and resource exploration. With the continuous development of satellite technology, the amount of data generated by satellites has exploded, and the data sources are wide and diverse, including satellite remote sensing data, operation logs, environmental data, image data, and multimodal data such as time series data. However, existing technologies face many challenges in processing these satellite data. On the one hand, the formats and protocols of multi-source satellite data vary greatly. For example, remote sensing data may use a specific image format, and operation logs are in text format, which makes it difficult to directly integrate and uniformly analyze the data, resulting in low data utilization. On the other hand, the data quality is uneven, and noise interference, outliers, time differences between different data sources, and data loss during data collection and transmission seriously affect the accuracy and reliability of data analysis. In addition, traditional satellite operation and maintenance mainly rely on manual experience and lack real-time, intelligent data analysis and prediction methods. During the operation of the satellite, it is difficult to timely warn and effectively prevent potential faults. Measures can often be taken only after the fault occurs, resulting in increased satellite operation and maintenance costs and increased operation risks. Moreover, existing data processing methods are difficult to intuitively present the overall picture of the satellite's operating status, and cannot provide comprehensive, accurate and visual support for operation and maintenance decisions, which is not conducive to the efficient planning and execution of satellite missions.

[0003] The existing technology has technical problems in satellite data processing, such as large differences in multi-source data format protocols and poor data quality, which affect satellite operation and maintenance efficiency and operational reliability. Summary of the invention

[0004] The present application provides a satellite multimodal data analysis method and system for intelligent operation and maintenance, which is used to solve the technical problems faced by satellite data processing in the prior art, such as large differences in multi-source data format protocols and poor data quality, which affect the efficiency and reliability of satellite operation and maintenance.

[0005] In view of the above problems, the present application provides a satellite multimodal data analysis method and system for intelligent operation and maintenance.

[0006] In a first aspect of the present application, a satellite multimodal data analysis method for intelligent operation and maintenance is provided, the method comprising:

[0007] Based on multiple data sources, multimodal data during the operation of Fengyun satellites are obtained; the multimodal data are preprocessed to obtain standard multimodal data; the standard multimodal data are integrated using a data center to build a unified data view, and a machine learning algorithm is used to pre-build an intelligent operation and maintenance model; Fengyun satellites are monitored in real time, and real-time monitoring data is obtained as input, and the intelligent operation and maintenance model is used to perform target detection and trend prediction to obtain satellite operation status analysis results; based on the satellite operation status analysis results, the data view is dynamically updated, and a visual satellite monitoring report is generated.

[0008] The second aspect of the present application provides a satellite multimodal data analysis system for intelligent operation and maintenance, the system comprising:

[0009] A multimodal data acquisition module, which acquires multimodal data during the operation of Fengyun satellites based on multiple data sources; a standard multimodal data acquisition module, which is used to perform standardized preprocessing on the multimodal data to obtain standard multimodal data; an intelligent operation and maintenance model construction module, which is used to integrate the standard multimodal data using a data center, build a unified data view, and pre-build an intelligent operation and maintenance model using a machine learning algorithm; a satellite operation status analysis result acquisition module, which is used to monitor Fengyun satellites in real time, acquire real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results; a satellite monitoring report generation module, which dynamically updates the data view based on the satellite operation status analysis results, and generates a visual satellite monitoring report.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Based on multiple data sources, obtain multimodal data during the operation of Fengyun satellites; perform standardized preprocessing on the multimodal data to obtain standard multimodal data; use the data center to integrate the standard multimodal data and build a unified data view; monitor Fengyun satellites in real time, obtain real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results; based on the satellite operation status analysis results, dynamically update the data view and generate a visual satellite monitoring report. The comprehensive, real-time and accurate monitoring of the operation status of Fengyun satellites is achieved, effectively improving the efficiency and reliability of satellite operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A flowchart of a satellite multimodal data analysis method for intelligent operation and maintenance provided in an embodiment of the present application.

[0014] Figure 2 A schematic diagram of the process of acquiring multimodal data for the satellite multimodal data analysis method for intelligent operation and maintenance provided in an embodiment of the present application.

[0015] Figure 3 A schematic diagram of the structure of a satellite multimodal data analysis system for intelligent operation and maintenance provided in an embodiment of the present application.

[0016] Explanation of reference numerals: multimodal data acquisition module 10 , standard multimodal data acquisition module 20 , intelligent operation and maintenance model construction module 30 , satellite operation status analysis result acquisition module 40 , satellite monitoring report generation module 50 . DETAILED DESCRIPTION

[0017] The present application provides a satellite multimodal data analysis method and system for intelligent operation and maintenance, which is used to solve the technical problems faced by satellite data processing in the prior art, such as large differences in multi-source data format protocols and poor data quality, which affect the efficiency and reliability of satellite operation and maintenance.

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] Embodiment 1, as Figure 1 As shown, the present application provides a satellite multimodal data analysis method for intelligent operation and maintenance, the method comprising:

[0020] Step S100: Based on multiple data sources, multimodal data of the Fengyun satellite during its operation is obtained.

[0021] Specifically, when obtaining multimodal data during the operation of Fengyun satellites, we first determine multiple reliable data sources and obtain remote sensing data transmitted by Fengyun satellites in real time from special satellite data receiving stations. These data contain rich information obtained by satellites from observing the earth's atmosphere, oceans, land, etc., such as cloud distribution, sea surface temperature, vegetation coverage, etc. At the same time, the satellite operation log is extracted from the satellite's own storage system, which records key information such as the working status of each component of the satellite, the execution of operation instructions, and system events. In addition, environmental data such as the radiation intensity and temperature changes of the space environment where the satellite is located are obtained through environmental monitoring sensors connected to the satellite communication link. Image data taken by imaging equipment on board the satellite is also one of the important sources. These images can intuitively reflect the surface characteristics of the satellite observation area. In addition, time series data is collected from the satellite's control system, such as the time series of attitude adjustment and the changes of orbital parameters over time. After obtaining the raw data, the corresponding parsing tools and algorithms are used to parse the data according to the data format (such as binary, text, etc.) and transmission protocol (such as TCP / IP, specific satellite communication protocol, etc.) of different data sources, and convert it into a unified and easy-to-process format. For the time difference between different data sources caused by clock differences or signal transmission delays, as well as the data loss problem that may occur during the data transmission process, appropriate interpolation algorithms (such as linear interpolation, spline interpolation, etc.) are used to repair and ensure the integrity and continuity of the data in the time series. Finally, according to the preset requirements of satellite operation and maintenance, data analysis, etc., labels with clear meanings are added to the repaired original data, such as labeling remote sensing data by observed object classification, and labeling operation logs by event type, so as to successfully obtain multimodal data and provide a rich and organized data foundation for the subsequent comprehensive and in-depth analysis of the satellite operation status.

[0022] Step S200: performing standardization preprocessing on the multimodal data to obtain standard multimodal data.

[0023] Specifically, when the acquired multimodal data is preprocessed for standardization, format conversion is first carried out for format differences, and data from different data sources and formats, such as specific image formats that may be used for satellite remote sensing data, text formats for operation logs, and numerical formats for environmental data, are uniformly converted into a common format suitable for subsequent processing and analysis, such as a common structured data format. In terms of time series alignment, based on the time identification information in the data, data of different modes but with time association are accurately calibrated to ensure that data such as satellite operating status, environmental changes, and task execution can accurately correspond on the time axis to form a time-synchronized data set. Then, denoising is performed, and digital filtering techniques such as low-pass filtering, high-pass filtering, or wavelet transform are used to identify and remove noise components in the data to prevent them from interfering with subsequent analysis. For outliers in the data, they are identified and eliminated by setting a reasonable threshold range or using statistical analysis methods (such as calculating the mean and standard deviation, and treating data exceeding a certain multiple of the standard deviation as outliers). Finally, data normalization is performed and an appropriate normalization method is selected according to the data characteristics, such as mapping the data to a specific interval (such as minimum-maximum normalization in the interval [0, 1] or a normalization method that makes the data mean 0 and the standard deviation 1), so that data features from different sources and different magnitudes are comparable, thereby obtaining standard multimodal data, laying a solid foundation for subsequent accurate data integration and in-depth analysis.

[0024] Step S300: Use the data center to integrate the standard multimodal data, build a unified data view, and use a machine learning algorithm to pre-build an intelligent operation and maintenance model.

[0025] Specifically, the standard multimodal data after standardized preprocessing is uploaded to the data center. With its powerful data management and processing capabilities, the data center stores the remote sensing data, satellite operation logs, environmental data, image data, time series data, etc. in the standard multimodal data in their respective storage areas according to the data type, so as to achieve classified storage and facilitate subsequent operations. Then, the processing module of the data center is used to perform data fusion operations, and the principal component analysis PCA algorithm is adopted. The algorithm regards the multimodal data as a set of data points in a high-dimensional vector space. The PCA algorithm calculates the covariance matrix of the data, finds its eigenvalues ​​and eigenvectors, sorts the eigenvectors according to the size of the eigenvalues, and selects the first several principal components with larger contribution rates. These principal components can retain the variance information of the original data to the greatest extent, thereby realizing data dimensionality reduction and feature extraction. In this way, the originally complex, high-dimensional and interrelated data are converted into a set of mutually orthogonal principal components, which integrate the key information in the multimodal data and construct a unified data view. This data view can present the core features of multimodal data in a concise and effective way, providing a unified data basis for subsequent analysis. At the same time, to ensure the timeliness of the data view, a real-time update mechanism is configured for it so that the latest status of satellite operation can be reflected in time. While building the data view, a machine learning algorithm is used to pre-build an intelligent operation and maintenance model, and a wealth of historical satellite status analysis data is collected, covering various situations such as normal satellite operation and failure. These historical data are carefully divided into training sets, validation sets, and test sets. In-depth data mining is carried out to extract key features that affect satellite health and operation, such as temperature change trends of key satellite components and signal strength fluctuation characteristics. According to the requirements of satellite operation and maintenance tasks, a suitable machine learning model, such as a neural network, is selected. Based on the extracted key features, the training set is used to supervise the training of the neural network model. The principal component data processed by the PCA algorithm can be used as the input of the neural network. By continuously adjusting the weights and biases of the neural network, it can learn the laws in the data and obtain the initial intelligent operation and maintenance model. Then, the cross-validation technology is used to divide the training set into multiple subsets, and one of the subsets is used for validation in turn, and the remaining subsets are used for training to optimize the model. Finally, the model is evaluated with the validation set and the test set, and the parameters are adjusted until convergence. The intelligent operation and maintenance model containing the target detection unit and the trend analysis unit is obtained for monitoring anomalies and predicting trends.

[0026] Step S400: Monitor the Fengyun satellites in real time, obtain real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results.

[0027] Specifically, after completing the construction of a unified data view and the pre-construction of an intelligent operation and maintenance model, the Fengyun satellite is monitored in real time. With the help of various high-precision monitoring instruments, such as high-precision attitude sensors, orbit measurement radars, etc., as well as the satellite's own state monitoring system, real-time satellite status data covering satellite attitude, orbital parameters, and the working status of each component are obtained. Real-time environmental monitoring data such as the radiation intensity and temperature of the satellite's surrounding environment are obtained through the space environment monitor. At the same time, real-time task execution data such as satellite data collection and transmission task progress are recorded, thereby forming complete real-time monitoring data, which are provided as input to the intelligent operation and maintenance model. The target detection unit in the model uses its built-in deep learning algorithm and the pattern recognition ability formed by historical data training to conduct a comprehensive analysis of the input data. For example, by comparing the key parameters in the real-time satellite status data with the standard parameter range under normal operating conditions, target abnormal states such as abnormal reduction in solar panel output power and unstable operation of attitude control engines can be quickly identified. Based on the preset abnormal judgment logic, these abnormal states are deeply evaluated to obtain accurate abnormal judgment results. Once an abnormality is determined, the early warning mechanism is immediately activated, and a fault risk warning is issued in a variety of ways (such as sending instant messages to operation and maintenance personnel, popping up eye-catching alarm prompts in the monitoring system, etc.), detailing the abnormal situation. At the same time, real-time monitoring data is input into the trend analysis unit, which uses time series analysis methods to predict the future trend of satellite operation status changes, such as orbital attenuation trends, energy consumption trends, etc., based on the long-term operation data patterns of satellites. Finally, the results of abnormality determination and trend prediction are combined to generate satellite operation status analysis results, which provide key basis for the subsequent dynamic update of data views and the formulation of operation and maintenance strategies, ensuring the stable operation of Fengyun satellites.

[0028] Step S500: Based on the satellite operation status analysis result, dynamically update the data view and generate a visual satellite monitoring report.

[0029] Specifically, after obtaining the satellite operation status analysis results, the real-time update mechanism of the data view is activated based on this, and key information in the analysis results is automatically extracted, such as detected abnormal conditions (component failures, signal interference, etc.), predicted operation trends (orbit changes, energy consumption trends, etc.), and these information are converted into a specific data format, matched and integrated with the corresponding parts in the data view, to achieve dynamic update of the data view, ensuring that it always reflects the latest operation status of the Fengyun satellite. Then, a visual satellite monitoring report is generated based on the updated data view. Using professional visualization tools, the satellite operation status analysis results are presented in the form of a combination of intuitive and easy-to-understand charts, graphics and text. The report includes dynamic charts of key indicators of the satellite's real-time operating status (such as attitude, orbital parameters, etc.), and displays their changing trends in the form of curves, bar charts, etc.; it uses eye-catching colors to mark areas where faults have occurred or are potential, and details the type of fault, possible causes, and scope of impact; at the same time, it uses time series charts to present predicted trends in the long-term operating status of satellites, such as orbital attenuation curves in the future, energy reserve changing trends, etc., to provide the satellite operation and maintenance team with comprehensive and clear satellite operating status information so that they can make timely and accurate decisions to ensure the safe and stable operation of Fengyun satellites.

[0030] In one possible implementation, Figure 2 As shown, step S100 also includes:

[0031] Step S110: Acquire raw data from multiple satellite data sources, including remote sensing data, satellite operation logs, environmental data, image data, and time series data.

[0032] Step S120: parsing the original data according to the formats and protocols of different data sources, and performing data interpolation and repair according to the time difference and data loss between different data sources.

[0033] Step S130: labeling the repaired original data according to preset requirements to obtain the multimodal data.

[0034] Specifically, in order to obtain comprehensive and accurate information on the operation of Fengyun satellites, data is collected from multiple satellite data sources. Through the ground receiving station that directly communicates with the satellite, the remote sensing data transmitted by the satellite in real time is obtained. These remote sensing data contain rich information obtained by the satellite from observing the earth's surface and atmospheric environment, such as cloud distribution, ocean temperature, land vegetation coverage, etc. At the same time, the satellite operation log is extracted from the storage system inside the satellite, which records in detail the working status of each subsystem of the satellite, the execution of operation instructions, the occurrence time and description of various events, and other key information, which is crucial for understanding the operation history and state change process of the satellite. In addition, environmental monitoring sensors deployed around the satellite or communicating with the satellite are used to obtain environmental data of the space environment in which the satellite is located, including radiation intensity, temperature changes, magnetic field intensity, etc. These environmental factors have an important impact on the operation and performance of the satellite. Image data taken by the imaging equipment on board the satellite is also one of the important data sources. These images can intuitively present the surface characteristics and meteorological phenomena of the satellite observation area, providing visual information for subsequent analysis. In addition, time series data are collected from the satellite's control system, such as the time series of satellite attitude adjustment and the change of orbital parameters over time. These time series data can reflect the dynamic process of satellite operation. Through the comprehensive collection of multiple satellite data sources, original data including remote sensing data, satellite operation logs, environmental data, image data, and time series data are obtained, laying a data foundation for subsequent analysis.

[0035] Since different satellite data sources use different formats and protocols, it is necessary to parse the acquired raw data. For specific image formats (such as HDF format), text formats (such as XML format) of satellite operation logs, and numerical formats (such as binary format) of environmental data that may be used by remote sensing data, corresponding parsing tools and algorithms are used to convert them into a unified format that is convenient for subsequent processing. For example, the data is parsed into a structured table form, in which each column represents a data attribute and each row represents a data record. In the process of data parsing, the time difference between different data sources also needs to be solved. Due to the fact that the clocks of the data acquisition devices of various satellite data sources may not be completely synchronized, as well as factors such as delays in the data transmission process, the data may have deviations on the time axis. By analyzing the time identification information in the data, a time synchronization algorithm is used, such as based on the network time protocol (NTP) or the reference satellite's own high-precision clock signal, to calibrate the data from different data sources so that they are consistent in time. At the same time, data interpolation and repair technology is used to deal with possible data loss during data transmission. According to the characteristics and distribution of the data, select the appropriate interpolation method, such as linear interpolation, spline interpolation, or mean interpolation based on neighboring data points. For missing data points in time series data, the estimated value of the missing point is calculated by linear interpolation method based on the known values ​​and time intervals of adjacent data points; for missing pixels in image data, spline interpolation or mean interpolation method can be used to repair them based on the color and brightness information of surrounding pixels. Through data analysis and repair, the quality and availability of the original data are improved.

[0036] After the original data is repaired, it is labeled according to the preset requirements. According to the specific goals of satellite operation and maintenance, data analysis and application, labels with clear meanings are assigned to the data. For example, for remote sensing data, labels are classified according to its observation objects (such as clouds, oceans, land, etc.), observation bands (such as visible light, infrared, etc.), observation areas (such as specific longitude and latitude ranges) and other attributes; for satellite operation logs, labels are marked according to event types (such as system startup, shutdown, fault alarm, task switching, etc.) and event severity (such as emergency, important, general, etc.); for environmental data, labels are divided according to environmental factor categories (such as radiation, temperature, magnetic field, etc.) and data collection locations (such as satellite bodies, solar wings, etc.); for image data, labels are identified according to image content (such as landform types, meteorological phenomena, etc.), shooting time and location, etc.; for time series data, labels are set according to the satellite operation process represented by the data (such as orbital control process, attitude adjustment process, etc.) and time series characteristics (such as periodic changes, trend changes, etc.). Through data labeling, the originally heterogeneous and unorganized raw data is converted into multimodal data with semantic information and structural characteristics, which facilitates subsequent data management, analysis and mining, and provides rich and orderly data resources for in-depth understanding of the operating status of Fengyun satellites.

[0037] In a possible implementation, step S200 further includes:

[0038] Step S210: performing format conversion and time alignment on the multimodal data to obtain multi-source aligned data.

[0039] Step S220: performing denoising, outlier removal and data normalization processing on the multi-source aligned data to obtain the standard multimodal data.

[0040] Specifically, the acquired multimodal data is first converted into a format. Since multimodal data comes from different satellite data sources, its format is diverse. For example, remote sensing data uses a specific image format (such as GeoTIFF, etc.), satellite operation logs are in text format (such as TXT or CSV format), environmental data may be in binary format, image data has common formats such as JPEG, and time series data is stored in a specific data structure. In order to facilitate subsequent unified processing and analysis, it is necessary to convert these data in different formats into a universal format suitable for data processing, such as converting all data into a structured table or matrix form, so that the data is consistent in structure and convenient for subsequent operations. After completing the format conversion, the time series alignment process is performed. Although different types of data in multimodal data are related in time, due to differences in data sources, their time tags may be inconsistent or have deviations. By analyzing the timestamp information in the data and using a time synchronization algorithm, such as based on a time reference signal or a logical time relationship between data, the data of different modalities are accurately aligned on the time axis. For example, the event time in the satellite operation log is matched one by one with the collection time of remote sensing data, the measurement time of environmental data, and the shooting time of image data, ensuring that different types of data at the same time point can be accurately matched, thereby obtaining multi-source aligned data and providing a time-synchronized data foundation for subsequent in-depth processing.

[0041] De-noising is performed on the obtained multi-source aligned data to improve data quality. Multimodal data may be interfered by various noises during the collection, transmission and storage process, such as sensor noise, communication channel noise, etc. For the noise in time series data and environmental data, filtering techniques such as moving average filtering and Kalman filtering can be used to smooth the fluctuations in the data and remove the noise components by weighted averaging of adjacent data points or estimating based on the state space model. For the noise in image data, image filtering algorithms such as median filtering and Gaussian filtering can be used to correct the noise pixels according to the neighborhood information of the image pixels. After the denoising is completed, outlier removal is performed. Through statistical analysis methods, statistical indicators such as the mean and standard deviation of the data are calculated, and a reasonable threshold range (such as mean ± 3 times the standard deviation) is set. Data exceeding the threshold range is regarded as outliers and removed. Outliers may be caused by sensor failures, data transmission errors, etc., and their existence will have a great impact on the subsequent analysis results. Finally, data normalization is performed. Since data of different modalities may have large differences in numerical range and dimension, for example, the numerical range of remote sensing data may be between [0, 10000], while the numerical range of temperature in environmental data is between [-50, 50]. In order to make different types of data have equal importance and comparability in subsequent analysis, normalization methods are used, such as minimum-maximum normalization to map data to the interval [0, 1], or zero-mean normalization to make the mean of the data 0 and the standard deviation 1. After denoising, outlier removal and data normalization, standard multimodal data is obtained, which has the advantages of high quality, consistency and comparability, and provides a reliable data foundation for subsequent data integration, model construction and analysis.

[0042] In a possible implementation, step S300 further includes:

[0043] Step S310: Upload the standard multimodal data to the data center and classify and store them according to different data types.

[0044] Step S320: Based on the processing module of the data center, the standard multimodal data is fused to generate a unified data view, and a real-time update mechanism is configured for the data view.

[0045] Specifically, after obtaining the standard multimodal data, it is uploaded to the data center. As a powerful data management and processing platform, the data center has efficient data reception and storage capabilities. According to different data types in the standard multimodal data, such as remote sensing data, satellite operation logs, environmental data, image data, time series data, etc., the corresponding storage areas are divided in the data center for classified storage. For remote sensing data, it is stored in a storage area dedicated to storing large-scale images and geographic information data to ensure efficient access and management of the data; satellite operation logs are stored in an area suitable for text data storage and retrieval, which is convenient for subsequent query and analysis of satellite operation historical events; environmental data is stored in an area suitable for numerical data storage according to its data structure and characteristics, so as to perform numerical calculations and statistical analysis; image data is stored in a special image repository, which can be quickly retrieved and processed by using the characteristics of the image database; time series data is stored in a storage structure optimized for time series data, which is convenient for time series analysis and trend prediction. Through this classified storage method, the management efficiency of data can be improved, the subsequent data processing and call can be facilitated, and a good data organization foundation is laid for building a unified data view.

[0046] Based on the processing module of the data center, the ontology-based data fusion algorithm is used to fuse the standard multimodal data to generate a unified data view. First, a satellite data ontology model is constructed. The ontology model defines the concepts, attributes and relationships related to satellite operation, such as satellite components, operating status, observation tasks, environmental factors, and their mutual relationships (such as the operating status of satellite components affects the overall operating status of the satellite, and environmental factors interfere with satellite observation tasks, etc.). For each data type in the standard multimodal data, its data elements are mapped to the corresponding concepts and attributes of the ontology model. For example, the pixel value and band information in the remote sensing data are mapped to the attributes related to the observation data in the ontology model; the events and timestamps in the satellite operation log are mapped to the concepts and attributes related to the operating status; the radiation intensity, temperature, etc. in the environmental data are mapped to the attributes related to the environmental factors. Through this mapping, different types of data establish semantic associations under the framework of the ontology model. Then, based on the reasoning rules of the ontology model, the deep-level relationships between the data are mined, such as the deviation between the actual trajectory of the satellite and the expected trajectory is inferred based on the time series data of the satellite attitude change and the displacement change of the observation area in the remote sensing image. Finally, the fused data is presented in an intuitive and structured form to generate a unified data view that shows the full picture of satellite operation. At the same time, to ensure the timeliness of the data view, a real-time update mechanism is configured. This mechanism monitors the update status of the data source in real time. When there is new data, it automatically triggers the above data fusion process and integrates the new data into the existing data view to ensure that the data view always reflects the latest operation status of the satellite.

[0047] In a possible implementation, step S300 further includes:

[0048] Step S330: Acquire historical satellite status analysis data and divide it into a training set, a validation set, and a test set.

[0049] Step S340: Extract key features that affect satellite health and operation, and select a machine learning model based on mission requirements.

[0050] Step S350: Based on the key features, use the training set to perform supervised training on the machine learning model to obtain an initial intelligent operation and maintenance model.

[0051] Step S360: Optimize the initial intelligent operation and maintenance model through cross-validation, and use the validation set and the test set to evaluate the initial intelligent operation and maintenance model until the model converges to obtain the intelligent operation and maintenance model, wherein the intelligent operation and maintenance model includes a target detection unit and a trend analysis unit.

[0052] Specifically, in order to build an effective intelligent operation and maintenance model, it is necessary to first obtain rich historical satellite status analysis data. These data come from a wide range of sources, including a large amount of status information recorded by Fengyun satellites during normal operation, various faults, and different mission executions, covering data from all aspects of the satellite system, such as performance parameters of satellite components, operation data under different environmental conditions, data collection and transmission in various observation tasks, etc. After obtaining the data, it is divided into training set, validation set, and test set according to a certain ratio. Usually, the training set accounts for the largest proportion and is used in the model training process to enable it to learn the laws and characteristics in the data; the validation set is used to conduct a preliminary evaluation of the performance of the model during the model training process, adjust the model's hyperparameters, and prevent the model from overfitting; the test set is used to objectively and comprehensively evaluate the performance of the final model after the model training is completed to ensure that the model has good generalization ability. For example, the data is divided into training set, validation set, and test set according to the ratio of 70%, 15%, and 15% to ensure the scientificity and accuracy of the model construction and evaluation process.

[0053] In-depth analysis of historical satellite status analysis data to extract features that have a key impact on satellite health and operation. These features include performance indicator change trends of key satellite components (such as solar panels, communication antennas, attitude control engines, etc.), such as the stability of panel output power, fluctuations in antenna signal strength, and changes in engine thrust; correlation characteristics between satellite operating environment factors (such as space radiation intensity, temperature changes, magnetic field interference, etc.) and satellite operating status; task-related features such as data transmission efficiency and data quality when satellites perform different tasks (such as meteorological observation, earth resource exploration, etc.). According to the specific task requirements of satellite operation and maintenance, such as fault prediction, performance optimization, and task planning, select the most suitable model from a number of machine learning models. If the task focuses on classifying the satellite operating status and determining whether it is normal or the type of fault, select classification models such as decision trees and support vector machines; if it is to predict the future performance trend or operating status change of the satellite, neural networks, time series prediction models, etc. may be more appropriate. For example, if the main task is to predict the remaining life of the satellite energy system, a regression model based on a neural network is a better choice because it can handle complex nonlinear relationships and effectively learn the relationship between factors such as energy consumption and time.

[0054] According to the extracted key features, the training set data is input into the selected machine learning model for supervised training. During the training process, the model continuously adjusts its own model parameters based on the feature values ​​(input) of the data in the training set and the corresponding satellite status labels (output) to minimize the error between the predicted results and the actual labels. For example, for the neural network model, the gradient of each layer of neurons is calculated according to the prediction error through the back propagation algorithm, and then the connection weights and biases between neurons are updated, so that the model gradually learns the mapping relationship between the features in the data and the satellite status. After multiple iterations of training, when the error of the model on the training set reaches a certain threshold or no longer decreases significantly, the initial intelligent operation and maintenance model is obtained. The model initially has the ability to analyze and predict the operating status of the satellite, but there may still be problems such as overfitting or underfitting, which need to be further optimized.

[0055] The initial intelligent operation and maintenance model is optimized by using cross-validation technology, and the training set is further divided into multiple subsets (such as k-fold cross-validation, where k is usually 5 or 10). One of the subsets is used as the validation set, and the remaining subsets are used as the training set. The model is trained and evaluated multiple times. In this way, the model can be trained and verified on different data subsets, making full use of limited data resources, improving the generalization ability of the model, and avoiding overfitting. In each cross-validation process, the model's hyperparameters, such as the number of hidden layers in the neural network and the learning rate, are adjusted according to the evaluation indicators on the validation set (such as accuracy, recall, mean square error, etc.) to optimize the model performance. After multiple cross-validation optimizations, the model is fully evaluated using independent validation and test sets. When the evaluation indicators of the model on the validation and test sets reach a stable state, that is, when the model converges, it indicates that the model has learned the effective rules in the data and has good generalization ability, and finally the intelligent operation and maintenance model is obtained. The model includes a target detection unit that can detect abnormal target status during satellite operation in real time, and a trend analysis unit that can accurately predict the future trend of satellite operation status changes, providing strong support for satellite operation and maintenance management.

[0056] In a possible implementation, step S400 further includes:

[0057] Step S410: Perform real-time monitoring on the Fengyun satellites to obtain real-time monitoring data, where the real-time monitoring data includes real-time satellite status data, real-time environmental monitoring data, and real-time mission execution data.

[0058] Step S420: input the real-time monitoring data into the intelligent operation and maintenance model, use the target detection unit to perform anomaly detection, and identify and obtain the target abnormal state.

[0059] Step S430: Based on the target abnormal state, perform abnormality determination and obtain an abnormality determination result.

[0060] Step S440: If the abnormality determination result is that an abnormality exists, the early warning mechanism is triggered to issue a fault risk early warning.

[0061] Specifically, a series of advanced monitoring equipment and technical means are used to conduct all-round, uninterrupted real-time monitoring of Fengyun satellites. The high-precision sensors carried on the satellites can collect satellite attitude information in real time, including real-time satellite status data such as pitch angle, yaw angle, and roll angle. These data are crucial to ensure that the satellite accurately points to the target observation area; at the same time, the satellite orbit parameters such as orbit altitude, orbit inclination, and right ascension of the ascending node are monitored to grasp the orbital status of the satellite in real time. Environmental monitoring sensors distributed around the satellite are used to obtain real-time environmental monitoring data of the space environment in which the satellite is located, such as the real-time value of space radiation intensity, whose changes may affect the performance of satellite electronic equipment; monitor ambient temperature, extreme temperature may cause satellite components to expand and contract, affecting their structural stability; measure magnetic field strength, and magnetic field anomalies may interfere with satellite communication and attitude control. In addition, pay close attention to the satellite's mission execution and obtain real-time mission execution data, such as the working status of sensors, acquisition frequency, data transmission rate, etc. in data acquisition tasks, as well as signal strength, signal quality, and bit error rate of data transmission in communication tasks, to ensure the smooth progress of satellite missions. Through the real-time collection of these different types of data, we can obtain comprehensive real-time monitoring data of the Fengyun satellite during its operation, providing a basis for subsequent analysis.

[0062] The acquired real-time monitoring data is fully input into the pre-built intelligent operation and maintenance model. The target detection unit in the intelligent operation and maintenance model starts working immediately, and uses its built-in deep learning algorithm and pre-training model to conduct in-depth analysis of the real-time monitoring data. For real-time satellite status data, the current attitude angle, orbital parameters, etc. are compared with the standard range under normal operating conditions. If the attitude angle deviation is found to exceed the allowable range, or the orbital parameter changes do not conform to the expected orbital dynamics model, it is identified as a possible abnormal state; for real-time environmental monitoring data, analyze whether the values ​​of radiation intensity, temperature, magnetic field intensity, etc. exceed the environmental threshold that the satellite can withstand for normal operation. Once exceeded, it is marked as a potential abnormality; for real-time task execution data, check various indicators of data collection and transmission, such as a sudden decrease in data collection frequency, too slow transmission rate, or too high bit error rate, and determine it as a task execution abnormality. Through this multi-dimensional and refined analysis, target abnormal states such as satellite attitude loss of control, orbit disturbance, abnormal environmental factors affecting satellite performance, and task execution failure in satellite operation are identified and acquired, providing a basis for subsequent abnormality determination.

[0063] Based on the identified abnormal state of the target, rigorous abnormality judgment is performed according to the pre-set abnormality judgment rules and logic. These judgment rules comprehensively consider the complexity of the satellite system and the possibility of various abnormal situations. For example, for satellite attitude abnormalities, not only the current attitude angle deviation is considered, but also the working status of the attitude adjustment system, historical attitude change trends and other factors; for abnormalities caused by environmental factors, the duration and rate of environmental changes and the sensitivity of satellite components to the environmental changes are analyzed. Through a comprehensive assessment of the factors related to the abnormal state of the target, using logical judgment, threshold comparison, probability statistics and other methods, accurate abnormality judgment results are obtained. The abnormality judgment results are expressed in a clear and unambiguous form, such as normal, slight abnormality, severe abnormality, etc., and detailed records of abnormality type, abnormality degree, possible impact range and other information are recorded to provide detailed reference for subsequent decision-making.

[0064] If the abnormality judgment result indicates that the satellite has an abnormality, the early warning mechanism in the intelligent operation and maintenance model will be triggered immediately. The early warning mechanism sends fault risk warnings to satellite operation and maintenance personnel through multiple communication channels to ensure that the information can be conveyed in a timely and accurate manner. In the satellite control center, a striking pop-up alarm will pop up, showing detailed information about the abnormality, including the type of abnormality (such as abnormal satellite attitude, communication failure, etc.), time of occurrence, severity, etc.; at the same time, SMS notifications will be sent to the mobile devices of the operation and maintenance personnel so that they can get the alarm information in the first time when they are not in the control center; in addition, an email will be sent to elaborate on the abnormal situation and possible countermeasures in detail, so as to facilitate the subsequent analysis and processing by the operation and maintenance personnel. The timely triggering of the early warning mechanism enables the operation and maintenance personnel to respond quickly and take corresponding measures, such as adjusting the satellite attitude, checking the communication link, analyzing the cause of the failure, etc., to minimize the impact of abnormal conditions on satellite operation and ensure the safe and stable operation of Fengyun satellites.

[0065] In a possible implementation, step S400 further includes:

[0066] Step S450: input the real-time monitoring data into the trend analysis unit in the intelligent operation and maintenance model.

[0067] Step S460: The trend analysis unit performs time series prediction analysis to obtain the long-term operating status change trend of the Fengyun satellites.

[0068] Step S470: Generate a satellite operation status analysis result by combining the abnormality determination result and the long-term operation status change trend.

[0069] Specifically, after the abnormality detection and judgment are completed, the real-time monitoring data is input into the trend analysis unit in the intelligent operation and maintenance model. These real-time monitoring data contain rich operation information of the satellite at the current moment, such as the precise attitude angle and instantaneous value of orbital parameters in the real-time satellite status data, the actual measurement values ​​of the current space radiation intensity, ambient temperature and magnetic field intensity in the real-time environmental monitoring data, and the current working status of the data acquisition equipment and the real-time rate of data transmission in the real-time task execution data. The trend analysis unit receives this data and prepares for the subsequent time series prediction analysis. Accurate data input is the basis for effective trend analysis, ensuring that the trend analysis unit can obtain the latest status of satellite operation, so as to more accurately predict its future change trend.

[0070] The trend analysis unit uses advanced time series prediction and analysis technology to process the input data. It first extracts and identifies the time series features in the real-time monitoring data. For example, for the satellite attitude angle data, it analyzes its change pattern in the past period of time, including periodic fluctuations (such as periodic adjustment of attitude caused by the rotation and revolution of the earth), long-term trends (such as gradual changes in attitude due to orbital decay) and random fluctuation components. For orbital parameter data, its evolution pattern over time is studied to predict the rate of decrease of orbital altitude and the possible direction of change of orbital inclination. At the same time, the impact of time series changes of environmental factors on satellite operation is considered, such as the seasonal changes in space radiation intensity, the long-term cold and warm trends of ambient temperature, etc. on the performance and life of satellite components. By establishing a suitable time series prediction model, such as the autoregressive moving average model (ARIMA), seasonal decomposition time series model (STL) or recurrent neural network (RNN) based on deep learning, long short-term memory network (LSTM), etc., the long-term operation status change trend of the satellite is predicted. The prediction results cover key aspects such as the satellite's attitude change trend, orbit stability trend, energy consumption trend, mission execution capability trend, etc. in the future (such as hours, days or even months, depending on the model's prediction capability and mission requirements), providing forward-looking information for a comprehensive understanding of the satellite's future operating status.

[0071] The abnormality determination results are organically combined with the long-term operation status change trend to generate a comprehensive satellite operation status analysis result. The abnormality determination results provide information such as whether the satellite currently has an abnormality and the specific type and severity of the abnormality. For example, if the temperature of a key component of the satellite is abnormally high, it is a serious abnormality. The long-term operation status change trend shows the future development direction of the satellite. For example, it is predicted that the satellite energy reserve will drop to a dangerous level within the next week, and the orbital altitude will drop by a certain value within the next month. Combining the two, the satellite operation status analysis results can clearly present the current health status of the satellite and the risks and challenges it may face in the future. This result not only provides a decision-making basis for satellite operation and maintenance personnel, such as deciding whether to take immediate measures to repair abnormal components and adjust satellite mission plans to cope with energy crises or orbital changes, but also provides important references for subsequent satellite operation management, resource allocation and mission planning, which helps to ensure the safe, stable and efficient operation of Fengyun satellites throughout the entire operation cycle.

[0072] In a possible implementation, step S500 further includes:

[0073] Step S510: Based on the satellite operation status analysis result, activate the real-time update mechanism to dynamically update the data view.

[0074] Step S520: Generate a visual satellite monitoring report including trend prediction and fault analysis based on the updated data view.

[0075] Specifically, based on the results of the satellite operation status analysis, the pre-configured data view real-time update mechanism is automatically activated. The satellite operation status analysis results contain rich information, such as abnormal conditions such as satellite component failures and signal interference identified by abnormality judgment, as well as future operation status change trends such as satellite orbit changes and energy consumption trends predicted by the trend analysis unit. The update mechanism first extracts these key information and converts them into a data format that matches the structure and format of the data view. Then, through data fusion and association technology, the new abnormal conditions and trend information are integrated with the existing satellite operation data in the data view. For example, if a fault is found that the output power of the satellite solar panel is abnormally reduced, the update mechanism will mark the abnormality in the corresponding solar panel performance data area in the data view, and associate the relevant historical data and real-time monitoring data to fully display the fault situation. For the predicted orbit change trend, the update mechanism will update the trend curve in the orbit parameter related area to show the expected change path of the future orbit. In this way, the data view can reflect the latest status of the satellite operation in real time, providing timely and accurate data support for subsequent monitoring and decision-making.

[0076] Based on the updated data view, a visual satellite monitoring report is generated using professional visualization tools. The report presents satellite operation status information in an intuitive and easy-to-understand form, which is convenient for the satellite operation and maintenance team to quickly understand and analyze. In the report, dynamic charts are used to display key indicators of the satellite's real-time operation status, such as a line chart depicting the change curve of the satellite attitude angle over time, which intuitively reflects the stability of the satellite attitude; a bar chart is used to display the energy consumption of each component of the satellite, and the energy consumption difference of different components is compared. For areas where faults have occurred or are potential, they are marked with eye-catching colors (such as red for serious faults and yellow for potential risks), and the report details the fault type (such as communication failure, attitude control failure, etc.), possible causes (such as component aging, external interference, etc.) and impact range (such as affecting data transmission, reducing observation accuracy, etc.). At the same time, a time series chart is used to present the predicted trend of changes in the long-term operation status of the satellite, such as drawing a decay curve of the satellite orbit height in the future period of time, to help operation and maintenance personnel plan orbit adjustment strategies in advance; a consumption trend chart of energy reserves is displayed to arrange energy replenishment plans in a timely manner. In addition, the report can also include text descriptions and analysis to summarize and evaluate the satellite's operating status, providing the operation and maintenance team with comprehensive and clear satellite operating status information, enabling them to make timely and accurate decisions based on the report to ensure the safe and stable operation of Fengyun satellites.

[0077] Embodiment 2, based on the same inventive concept as the satellite multimodal data analysis method for intelligent operation and maintenance in the above embodiment, Figure 3 As shown, the present application provides a satellite multimodal data analysis system for intelligent operation and maintenance, and the system and method embodiments in the embodiments of the present application are based on the same inventive concept. Among them, the system includes:

[0078] The multimodal data acquisition module 10 acquires multimodal data during the operation of the Fengyun satellite based on multiple data sources.

[0079] The standard multimodal data acquisition module 20 is used to perform standard preprocessing on the multimodal data to obtain standard multimodal data.

[0080] The intelligent operation and maintenance model construction module 30 is used to integrate the standard multimodal data using the data center, build a unified data view, and pre-build an intelligent operation and maintenance model using a machine learning algorithm.

[0081] The satellite operation status analysis result acquisition module 40 is used to perform real-time monitoring of Fengyun satellites, obtain real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results.

[0082] The satellite monitoring report generating module 50 dynamically updates the data view based on the satellite operation status analysis result and generates a visual satellite monitoring report.

[0083] Furthermore, the multimodal data acquisition module 10 further includes:

[0084] The raw data acquisition unit is used to acquire raw data from multiple satellite data sources, including remote sensing data, satellite operation logs, environmental data, image data, and time series data.

[0085] The data interpolation and repair unit is used to parse the original data according to the formats and protocols of different data sources, and to perform data interpolation and repair according to the time difference and data loss between different data sources.

[0086] A multimodal data acquisition unit is used to label the repaired original data according to preset requirements to obtain the multimodal data.

[0087] Furthermore, the standard multimodal data acquisition module 20 also includes:

[0088] A multi-source alignment data acquisition unit is used to perform format conversion and time sequence alignment on the multimodal data to obtain multi-source alignment data.

[0089] A standard multimodal data acquisition unit is used to perform denoising, outlier removal and data normalization on the multi-source aligned data to obtain the standard multimodal data.

[0090] Furthermore, the intelligent operation and maintenance model building module 30 also includes:

[0091] A classification storage unit, wherein the classification storage unit is used to upload the standard multimodal data to the data center and classify and store the data according to different data types.

[0092] A real-time update mechanism configuration unit, wherein the real-time update mechanism configuration unit performs data fusion on the standard multimodal data based on the processing module of the data middle platform, generates a unified data view, and configures a real-time update mechanism for the data view.

[0093] Furthermore, the intelligent operation and maintenance model building module 30 also includes:

[0094] The historical satellite state analysis data acquisition unit is used to acquire historical satellite state analysis data and divide it into a training set, a verification set and a test set.

[0095] A machine learning model selection unit is used to extract key features that affect satellite health and operation and select a machine learning model based on mission requirements.

[0096] An initial intelligent operation and maintenance model acquisition unit is used to perform supervised training on the machine learning model using the training set according to the key features to obtain an initial intelligent operation and maintenance model.

[0097] A model evaluation unit, wherein the model evaluation unit is used to optimize the initial intelligent operation and maintenance model through cross-validation, and use the validation set and the test set to evaluate the initial intelligent operation and maintenance model until the model converges, thereby obtaining the intelligent operation and maintenance model, wherein the intelligent operation and maintenance model includes a target detection unit and a trend analysis unit.

[0098] Furthermore, the satellite operation status analysis result acquisition module 40 also includes:

[0099] A real-time monitoring data acquisition unit is used to perform real-time monitoring on Fengyun satellites and acquire real-time monitoring data, wherein the real-time monitoring data includes real-time satellite status data, real-time environmental monitoring data, and real-time mission execution data.

[0100] A target abnormal state acquisition unit is used to input the real-time monitoring data into the intelligent operation and maintenance model, use the target detection unit to perform abnormality detection, and identify and acquire the target abnormal state.

[0101] The abnormality determination result acquisition unit performs abnormality determination based on the target abnormal state to obtain an abnormality determination result.

[0102] A fault risk warning unit is used to trigger a warning mechanism and issue a fault risk warning if the abnormality determination result is that an abnormality exists.

[0103] Furthermore, the satellite operation status analysis result acquisition module 40 also includes:

[0104] A monitoring data input unit, wherein the monitoring data input unit is used to input the real-time monitoring data into the trend analysis unit in the intelligent operation and maintenance model.

[0105] An operation status change trend acquisition unit is used for the trend analysis unit to perform time series prediction analysis to acquire the long-term operation status change trend of the Fengyun satellite.

[0106] A status analysis result generating unit is used to generate a satellite operation status analysis result by combining the abnormality determination result and the long-term operation status change trend.

[0107] Furthermore, the satellite monitoring report generating module 50 also includes:

[0108] A data view updating unit activates a real-time update mechanism based on the satellite operation status analysis result to dynamically update the data view.

[0109] A visual satellite monitoring report generating unit is used to generate a visual satellite monitoring report including trend prediction and fault analysis according to the updated data view.

[0110] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0112] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A satellite multimodal data analysis method for intelligent operation and maintenance, characterized in that: The method comprises: Based on multiple data sources, obtain multi-modal data during the operation of Fengyun satellites; Performing standardization preprocessing on the multimodal data to obtain standard multimodal data; The standard multimodal data is integrated by using the data center, a unified data view is constructed, and a machine learning algorithm is used to pre-build an intelligent operation and maintenance model; Monitor the Fengyun satellites in real time, obtain real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results; Based on the satellite operation status analysis result, the data view is dynamically updated and a visual satellite monitoring report is generated.

2. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 1, characterized in that: Based on multiple data sources, multi-modal data of Fengyun satellites during operation is obtained, including: Obtain raw data from multiple satellite data sources, including remote sensing data, satellite operation logs, environmental data, image data, and time series data; According to the formats and protocols of different data sources, the raw data is parsed, and data interpolation and repair are performed according to the time difference and data loss between different data sources; The repaired original data is labeled according to preset requirements to obtain the multimodal data.

3. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 1, characterized in that: The multimodal data is subjected to standardization preprocessing to obtain standard multimodal data, including: Performing format conversion and time alignment on the multimodal data to obtain multi-source aligned data; The multi-source aligned data is subjected to denoising, outlier removal and data normalization processing to obtain the standard multimodal data.

4. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 1, characterized in that: The standard multimodal data is integrated using the data center to build a unified data view, including: Upload the standard multimodal data to the data center and classify and store them according to different data types; The processing module based on the data middle platform performs data fusion on the standard multimodal data, generates a unified data view, and configures a real-time update mechanism for the data view.

5. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 1, characterized in that: Pre-build intelligent operation and maintenance models using machine learning algorithms, including: Obtain historical satellite status analysis data and divide it into training set, validation set and test set; Extract key features that affect satellite health and operation and select machine learning models based on mission requirements; According to the key features, using the training set, supervised training is performed on the machine learning model to obtain an initial intelligent operation and maintenance model; The initial intelligent operation and maintenance model is optimized by cross-validation, and the initial intelligent operation and maintenance model is evaluated using the validation set and the test set until the model converges, thereby obtaining the intelligent operation and maintenance model, which includes a target detection unit and a trend analysis unit.

6. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 5, characterized in that: Perform real-time monitoring of Fengyun satellites, obtain real-time monitoring data as input, and use the intelligent operation and maintenance model to perform target detection, including: Perform real-time monitoring on Fengyun satellites to obtain real-time monitoring data, wherein the real-time monitoring data includes real-time satellite status data, real-time environmental monitoring data, and real-time mission execution data; Inputting the real-time monitoring data into the intelligent operation and maintenance model, using the target detection unit to perform anomaly detection, and identifying and acquiring an abnormal state of the target; Based on the target abnormal state, perform abnormality determination and obtain an abnormality determination result; If the abnormality determination result is that an abnormality exists, the early warning mechanism is triggered to issue a fault risk early warning.

7. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 6, characterized in that: The intelligent operation and maintenance model is used to perform target detection and trend analysis to obtain satellite operation status analysis results, including: Input the real-time monitoring data into the trend analysis unit in the intelligent operation and maintenance model; The trend analysis unit performs time series prediction analysis to obtain the long-term operation status change trend of the Fengyun satellites; The satellite operation status analysis result is generated by combining the abnormality determination result and the long-term operation status change trend.

8. The satellite multimodal data analysis method for intelligent operation and maintenance according to claim 4, characterized in that: Based on the satellite operation status analysis results, the data view is dynamically updated and a visual satellite monitoring report is generated, including: Based on the analysis result of the satellite operation status, activating the real-time update mechanism to dynamically update the data view; Generate visual satellite monitoring reports including trend prediction and fault analysis based on the updated data view.

9. Satellite multimodal data analysis system for intelligent operation and maintenance, characterized by: The system is used to implement the satellite multimodal data analysis method for intelligent operation and maintenance according to any one of claims 1 to 8, and the system includes: A multimodal data acquisition module, wherein the multimodal data acquisition module acquires multimodal data during the operation of the Fengyun satellite based on multiple data sources; A standard multimodal data acquisition module, wherein the standard multimodal data acquisition module is used to perform standard preprocessing on the multimodal data to obtain standard multimodal data; An intelligent operation and maintenance model building module, which is used to integrate the standard multimodal data using the data center, build a unified data view, and pre-build an intelligent operation and maintenance model using a machine learning algorithm; A satellite operation status analysis result acquisition module, which is used to perform real-time monitoring of Fengyun satellites, obtain real-time monitoring data as input, use the intelligent operation and maintenance model to perform target detection and trend prediction, and obtain satellite operation status analysis results; A satellite monitoring report generating module dynamically updates the data view based on the satellite operation status analysis result and generates a visual satellite monitoring report.

Images