Method and system for recovering satellite state data through historical enhancement

Abstract

The invention relates to the technical field of satellite data processing, and discloses a method and system for recovering satellite state data through historical enhancement, and the method comprises the steps: calculating the similarity between current satellite data and multiple segments of historical data, and selecting a plurality of segments of most similar historical data as matched historical data; designing a tensor fusion filling scheme, determining the similarity between the current satellite data and the historical data, respectively distributing different weights to the current satellite data and the matched neighbor data, fusing the current satellite data and the neighbor data with the distributed weights, training a data completion model, and obtaining a data fusion filling model; outputting complete satellite data through the data completion model; according to the method, the precision of the completion model is effectively improved, the distance between sparse data and historical data is efficiently calculated by adopting a subspace clustering quantization method and utilizing a distance table, the calculation efficiency is improved while the high retrieval precision is kept, and the retrieval performance is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite data processing, and particularly relates to a method and system for restoring satellite state data through historical enhancement. Background Art

[0002] With the rapid development of space technology, satellite networks and satellite computing have shown great potential. Real-time monitoring of satellite status is crucial for ensuring stable operation and mission success. Ensuring the accuracy of satellite data and the reliability of system performance is the foundation for advancing these fields. However, due to factors such as space environment interference (such as electromagnetic interference, cosmic rays), hardware failures, and energy limitations, satellite state data is often incomplete, and the loss of satellite state monitoring data often leads to various problems: (1) Resource management failure: The lack of satellite state data may prevent the accurate assessment of the remaining power, storage capacity, or other resource utilization rates of the satellite, thus affecting effective resource allocation and optimization, and resulting in the inability to complete some tasks as expected; (2) Delayed anomaly diagnosis: Without complete satellite state data, the ground control center may not be able to detect equipment failures or other anomalies in a timely manner, missing the best intervention window, which increases the risk of fault propagation and may even lead to satellite failures.

[0003] Therefore, it is crucial to ensure the accuracy and integrity of satellite state monitoring data. In recent years, neural network-based missing data completion models have received extensive attention due to their advantages in capturing complex data patterns. Existing methods are mainly divided into two categories: neural network models based on multi-dimensional tensor decomposition and completion frameworks based on generative diffusion models: (1) Neural network models based on multi-dimensional tensor decomposition: The core idea of this method is to extend the linear assumption of traditional tensor decomposition through neural networks and explicitly model the non-linear interactions in multi-dimensional data. NTC constructs a three-dimensional interaction graph through outer product and extracts features through three-dimensional convolution. However, its cubic parameters increase with the feature dimension, posing a risk of overfitting. In sparse data, as seen in sensor networks with limited node observations, it weakens global learning. NTF combines LSTM and MLP for temporal and non-linear capture. However, due to the gradient propagation problem of LSTM and the parameter redundancy of MLP, NTF performs poorly in high-dimensional sparse sequences, especially in heterogeneous data. NTM adds a generalized CP decomposition layer and a tensorized MLP to achieve better performance through multi-layer interactions. However, in sparse data, its local attention conflicts with the global pattern, such as losing key nodes in social networks. CoSTCo uses CNN to compress tensor dimensions and applies low-rank constraints, performing well in regular grid data. In heterogeneous sparse data, the fixed kernel cannot adapt well, and the low-rank constraint may distort the results in ultra-low density cases; (2) Completion frameworks based on generative diffusion models: To overcome the limitations of tensor decomposition methods, recent research has turned to generative diffusion models, which learn the missing data distribution through probabilistic modeling. TimesNet uses Fourier transform to convert time series data into 2D tensors for periodic capture. In sparse aperiodic data (such as emergency monitoring), its forced periodicity assumption may distort the signal. CSDI combines diffusion and conditional attention to generate completions through weighted sampling. In high-dimensional sparse scenarios, the maintenance of context vectors in its bidirectional attention leads to exponential growth of memory, and the weight calculation cannot reflect the true correlation of discontinuous data. PriSTI adds a geospatial prior to improve diffusion robustness. However, in non-spatial sparse data such as text, the geocoding module distorts features by mapping non-spatial data to space, and its strict spatial check in multi-step diffusion limits diversity. DAIN selects key entities with influence functions for data enhancement. However, its local importance assessment (e.g., relying on gradient magnitude) may miss global structure nodes, thus reducing the diversity of generated data.

[0004] Both of these two types of methods face significant challenges in dealing with high-dimensional sparse data, and their core bottlenecks can be attributed to insufficient parameter efficiency, poor global pattern learning ability, and weak adaptability to heterogeneity. Summary of the Invention

[0005] The object of the present invention is to solve the above problems, and a method and system for restoring satellite state data through historical enhancement are designed to solve the problem that the existing missing data completion method is limited by high missing rate data and long-term data missing, making it difficult for the completion model to learn accurate data features from limited information during training, thus restricting the accuracy of the missing data completion result.

[0006] In the first aspect of the present invention, a method for restoring satellite state data through historical enhancement is provided. The method includes the following steps: S1. Calculate the similarity between the current satellite data and multiple segments of historical data, and select several segments of the most similar historical data as the matched historical data; S2. Design a tensor fusion filling scheme to determine the similarity between the current satellite data and the historical data, and assign different weights to the current satellite data and the matched neighbor data respectively. Then fuse the current satellite data and the neighbor data with assigned weights, train the data completion model, and output complete satellite data through the data completion model, where the neighbor data is the historical data obtained in step S1.

[0007] The specific implementation process of step S1 includes: S1-1. Input the historical data with missing values into the missing data completion model to obtain complete historical data; S1-2. Set a sliding window with the same size as the current satellite data, and start matching from the first time node of the complete historical data to obtain window data; S1-3. Adopt the method of subspace clustering quantization to calculate the similarity between the current satellite data and window data in turn, and select the neighbor data that is most similar to the current satellite data.

[0008] Optionally, in the first implementation manner of the first aspect of the present invention, the specific implementation process of step S1-3 includes: Perform subspace partitioning on each window data, perform clustering on the corresponding subspaces of all window data to construct a codebook, and then quantize the original space according to the codebook; Construct a distance table for the current satellite data, quickly query the distance between the current satellite data and each window tensor according to the distance table as a similarity measurement standard, and select the nearest ones as neighbor data.

[0009] Optionally, in the second implementation manner of the first aspect of the present invention, the performing subspace partitioning on each window data, performing clustering on the corresponding subspaces of all window data to construct a codebook, and then quantizing the original space according to the codebook includes: Perform a slicing operation on each high-dimensional window data, which is divided along the time dimension into slices, and the size of the time dimension of each slice is , and the empty positions are filled with 0. Each slice is a subspace; Perform K-means clustering on the th slice of each window data in sequence to obtain the th clustering center slice of the clustering centers. After times of clustering, finally obtain clustering center slices consisting of clustering centers, which are the codebooks; For the th slice of each window tensor, compare its distance with the th clustering center slice of the clustering centers respectively, then record the clustering center index corresponding to the nearest distance. Finally, convert each high-dimensional window data into one-dimensional data composed of index values.

[0010] Optionally, in the third implementation manner of the first aspect of the present invention, constructing a distance table for the current satellite data, quickly querying the distance between the current satellite data and each window tensor according to the distance table, using it as a similarity measurement standard, and selecting the nearest ones as neighbor data, including: Perform subspace partitioning on the current satellite data to obtain slices, then calculate the distance between each slice and each clustering center slice respectively, and sequentially place the calculated distances in the corresponding positions of the distance table; For the distance calculation between the current satellite data and each window data, query the corresponding elements in the distance table and sum them; Use the distance table to obtain the distances between the current satellite data and all window data, and select the nearest distances. The window data corresponding to these distances are the neighbor data.

[0011] Optionally, in the fourth implementation manner of the first aspect of the present invention, the distance table includes rows columns. The th row of the distance table represents the distance between the th slice of the current satellite data and the th clustering center slice of the th clustering center. The th column of the distance table represents the distance between the th slice of the current satellite data and The distance between the th slice of the

[0012] Optional. In the fifth implementation of the first aspect of the present invention, weights are assigned according to the distances between neighbor data and current satellite data, and the assigned weights are normalized so that they are distributed between and

[0013] The second aspect of the present invention provides a system for restoring satellite state data through historical enhancement. The system includes: A historical data matching module, configured to calculate the similarity between current satellite data and multiple segments of historical data, select several segments of the most similar historical data as the matched historical data: input the historical data with missing values into a missing data completion model to obtain complete historical data; set a sliding window with the same size as the current satellite data, start matching from the first time node of the complete historical data, and obtain window data; adopt the method of subspace clustering quantization to calculate the similarity between the current satellite data and window data in sequence, and select the neighbor data that is most similar to the current satellite data from them; A tensor fusion filling module, configured to design a tensor fusion filling scheme, determine the similarity between the current satellite data and the historical data, and assign different weights to the current satellite data and the matched neighbor data respectively, fuse the current satellite data and the neighbor data with assigned weights, train a data completion model, and output complete satellite data through the data completion model, where the neighbor data is the historical data obtained in step S1.

[0014] The third aspect of the present invention provides a device for restoring satellite state data through historical enhancement. The device for restoring satellite state data through historical enhancement includes a memory and at least one processor, and instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the device for restoring satellite state data through historical enhancement executes each step of the method for restoring satellite state data through historical enhancement as described in any one of the above.

[0015] The fourth aspect of the present invention provides a computer-readable storage medium, and instructions are stored on the computer-readable storage medium. When the instructions are executed by a processor, each step of the method for restoring satellite state data through historical enhancement as described in any one of the above is implemented.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) To address the challenges posed by high missing rates and structural missing patterns in satellite data, which make existing data completion models difficult to apply, the present invention proposes a general model training method. By finding the most similar data segments from historical data to increase the training dataset, this framework can be applied to any data completion model, effectively improving the accuracy of the completion model. (2) To quickly match historical data similar to current satellite data, the present invention adopts the method of subspace clustering quantization. By using a distance table to efficiently calculate the distance between sparse data and historical data, while maintaining a high retrieval accuracy, the computing efficiency is improved, and the retrieval performance is significantly enhanced. (3) To balance the importance of matching neighbor data in current satellite data and historical data, the present invention designs a fusion completion strategy based on historical data enhancement. By defining weights according to the similarity with current data, the importance of matching neighbor data and current data for model training is balanced, thereby further enhancing the performance of the data completion model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0018] Figure 1 It is the overall process of the historical data matching module and the tensor fusion filling module provided by the embodiment of the present invention. Figure 2 It is a schematic diagram of the process of the historical data matching module provided by the embodiment of the present invention. Figure 3 It is a schematic diagram of the historical data sliding window operation provided by the embodiment of the present invention. Figure 4 It is a schematic diagram of the historical data matching process based on subspace clustering quantization provided by the embodiment of the present invention. Figure 5 It is a schematic diagram of subspace division in the preliminary preparation stage provided by the embodiment of the present invention. Figure 6 It is a schematic diagram of clustering to construct a codebook in the preliminary preparation stage provided by the embodiment of the present invention. Figure 7 It is a schematic diagram of quantization according to the codebook in the preliminary preparation stage provided by the embodiment of the present invention. Figure 8 It is a schematic diagram of constructing a distance table in the query stage provided by the embodiment of the present invention. Figure 9 It is a schematic diagram of querying distances with a distance table in the query stage provided by the embodiment of the present invention. Figure 10Schematic diagram of the device for restoring satellite state data through historical enhancement provided by the embodiments of the present invention. Detailed implementation manners

[0019] The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that illustrated or described herein. In addition, the terms "comprising" or "having" and any variation thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or device comprising a series of steps or units does not necessarily limit to the clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. The embodiments of the present invention provide a method for restoring satellite state data through historical enhancement. It models satellite data into a three-dimensional tensor of sample-time node-attribute. Aiming at the common problems of high data missing rate and long-term data missing in the real scenario, a historical data matching module is designed to select multiple segments of historical data that are most similar to the current time satellite data to supplement the training data, so as to solve the problem of insufficient training data in the process of training a complete model. Calculate the similarity between the current satellite data and multiple segments of historical data, and select several segments of the most similar historical data as the matched historical data. Further, a tensor fusion filling scheme is designed, fully considering the similarity between the current satellite data and the historical data, and different weights are assigned to the current satellite data and the matched historical data respectively to balance the importance of the current satellite data and the historical data for model training. Apply the method of this embodiment to the existing complete data completion model to solve the problem that the existing complete model has low training accuracy or even cannot be trained due to too high data missing rate; specifically includes the following steps: S1. Calculate the similarity between the current satellite data and multiple segments of historical data, and select several segments of the most similar historical data as the matched historical data; S2. Design a tensor fusion filling scheme, determine the similarity between the current satellite data and the historical data, and assign different weights to the current satellite data and the matched neighbor data respectively, and fuse the current satellite data and the neighbor data with assigned weights to train the data completion model.

[0021] In this embodiment, by using the existing trained missing data completion model, the historical data with missing values is fed into the model to obtain complete historical data; a sliding window with the same size as the current satellite data is set, and starting from the first time node of the complete historical data, window data are obtained; by using the method of subspace clustering quantization, the similarity between the current satellite data and window data is calculated in sequence, and the neighbor data that is most similar to the current satellite data is selected from them.

[0022] In this embodiment, the calculation formula for completing the historical data is: ; where is the historical data, and can be any completion model based on deep learning.

[0023] In this embodiment, the calculation formula for the number of window data is: ; where is the number of time nodes of the historical data, is the size of the sliding window, and is the step size of the sliding window.

[0024] In this embodiment, (1) Preliminary preparation stage: perform subspace partitioning on each window data, cluster and construct a codebook for the corresponding subspaces of all window data, and then quantize the original space according to the codebook; (2) Query stage: construct a distance table for the current satellite data, quickly query the distance between the current satellite data and each window tensor according to the distance table, use it as a similarity measurement standard, and select the with the closest distance as neighbor data.

[0025] In this embodiment, the specific implementation process of the preliminary preparation stage includes: (1) Subspace partitioning: perform slicing operations on each high-dimensional window data, which is divided into slices along the time dimension, the time dimension size of each slice is , the empty positions are filled with , and each slice is a subspace; (2) Clustering to construct a codebook: perform K-means clustering on the th slice of each window data in sequence to obtain the th clustering center slice of clustering centers. After times of clustering, finally obtain clustering centers composed of clustering center slices, which is the codebook; (3) Quantization based on the codebook: for the th slice of each window tensor, we compare it with The distance between the th slice of the th cluster center, then record the index of the cluster center corresponding to the nearest distance, and finally convert each high-dimensional window data into one-dimensional data composed of

[0026] In this embodiment, the formula for constructing one-dimensional data based on codebook quantization is: ; where represents the th element of the one-dimensional data , that is, the index of the cluster center slice closest to the th slice of the th window data. represents the th slice of the th window data, represents the th slice of the th cluster center.

[0027] In this embodiment, the specific implementation process in the query stage includes: (1) Constructing a distance table: We perform subspace partitioning on the current satellite data to obtain slices, and then calculate the distance between each slice and each slice of the cluster center respectively, and put the calculated distances into the corresponding positions of the distance table in turn; (2) Querying the distance using the distance table: For the calculation of the distance between the current satellite data and each window data, we only need to query the corresponding elements in the distance table and then sum them; (3) Determining the neighbor data: Using the distance table to obtain the distances between the current satellite data and all window data, select the nearest distances from them, and the window data corresponding to these distances is the so-called neighbor data.

[0028] In this embodiment, a distance table with C rows and P columns is constructed, where each element of the distance table is represented by , The calculation formula of is: represents the th slice of the current satellite data, represents the th slice of the th cluster center. After calculating all elements, finally obtain a distance table with rows and columns. The th row of the distance table represents the th slice of the current satellite data and the th cluster center of the The distance between the cluster center slices. The th column of the distance table represents the th slice of the current satellite data and the th cluster center's th cluster center slice.

[0029] In this embodiment, the formula for distance query is: ; where represents the distance between the current satellite data and the th window data, represents the element corresponding to the th row and the th column of the distance table.

[0030] In this embodiment, (1) the weights of the current satellite data are assigned in a predefined manner ; (2) weights are assigned to the neighboring data ; (3) the above weights are normalized so that they are distributed between .

[0031] In this embodiment, the formula for weight assignment of the th neighboring data is: , where represents the distance between the th neighboring data and the current satellite data.

[0032] In this embodiment, the formula for normalizing the weights is: ; ; where represents the sum of the weights.

[0033] In this embodiment, by using the existing trained missing data completion model, the historical data with missing values is input into the model to obtain complete historical data. Through this step, the original historical data with missing values is repaired, providing a complete data basis for the subsequent matching work and ensuring the availability of historical data. The missing data completion model used here can be any deep learning-based completion model, whose purpose is to reasonably fill the missing values in the historical data so that the historical data can more accurately reflect the actual state of the satellite. For example, if the satellite power value at a certain time node in the historical data is missing, after being processed by the model, a reasonable estimated value can be obtained to complete the data at that time node. A sliding window with the same size as the current satellite data is set, and the matching starts from the first time node of the complete historical data. According to the number of time nodes in the historical data, the size of the sliding window, and the sliding window step size, the number of window data is calculated. The specific calculation formula is: the number of window data = (the number of time nodes in the historical data - the size of the sliding window) ÷ the sliding window step size + 1, and the calculation result is rounded down. In this way, the complete historical data is divided into multiple window data with the same scale as the current satellite data, so as to find the historical data segment similar to the current satellite data later. Taking the current satellite data with 10 time nodes, the sliding window size of 5, and the sliding window step size of 2 as an example, if the historical data has 20 time nodes, then the number of window data is (20 - 5) ÷ 2 + 1 = 8.5, and then rounding down to get 8. Each window data contains 5 consecutive time node data, corresponding to the time scale of the current satellite data.

[0034] The following combines Figures 1 to 9 to describe the fast and general framework for restoring satellite state data through historical enhancement provided by the present invention.

[0035] As Figure 1 shown, it presents the overall processes of the historical data matching module and the tensor fusion filling module, clearly showing the whole process from historical data processing to data fusion filling, which is convenient for understanding the overall architecture and data flow of the invention, including: (1) Historical data matching module: For the satellite data matching problem in different missing data scenarios, a historical data matching module is designed to effectively extract similar data from the historical satellite data for data enhancement; (2) Tensor fusion filling module: Through the above method, it is found that A neighbor data that matches the current satellite data. However, the similarity between each neighbor data and the current satellite data is different. If all data is treated equally, the prediction accuracy will be limited. Intuitively, the training model needs to be set closer to the data characteristics at the current moment. Therefore, the present invention designs a tensor fusion filling module to assign different weights to the data and balance the importance between the data, thereby further improving the filling accuracy.

[0036] As Figure 2 shown, the historical data matching module includes: (1) Historical data pre-filling: The present invention uses an existing missing data completion model to estimate the missing values in the historical data to ensure its integrity; (2) Historical data partitioning: A historical data partitioning method based on a sliding window is designed to partition the historical data into multiple window data with the same scale as the current data, ensuring that the historical data segment belonging to the same time period as the current satellite data can be identified; (3) Fast historical data matching: The present invention proposes a fast historical data matching method based on subspace clustering quantization, which can effectively match the neighbor data most similar to the current satellite data from the window data. This method can not only minimize the storage overhead, but also significantly reduce the calculation cost.

[0037] The above-mentioned historical data pre-filling is specifically implemented as: using a simple missing data filling model to estimate the missing values in the historical data. The formula is: ; where is the historical data, can be any deep learning-based completion model.

[0038] As Figure 3 shown, the historical data partitioning is specifically implemented as: Given a complete historical data, which contains time nodes, for the current satellite data, it contains time nodes, where . Set a sliding window with the same size as the current satellite data, and divide the entire historical data along the dimension into window data. The sliding window step size is , and the calculation formula for the number of window data is: .

[0039] As Figure 4 shown, the fast historical data matching includes the following two stages: (1) Preliminary preparation stage: This stage includes subspace partitioning, clustering to construct a codebook, and quantization according to the codebook; (2) Query phase: This phase includes three parts: constructing a distance table, querying distances using the distance table, and determining neighbor data.

[0040] The subspace partitioning mentioned above is as Figure 5 shown. The specific implementation is as follows: Direct quantization of high-dimensional data encounters the curse of dimensionality, resulting in a sharp increase in computational and storage costs. Therefore, we propose to decompose the high-dimensional space into multiple low-dimensional subspaces, and each subspace is quantized independently, significantly reducing the computational amount. The present invention performs slicing operations on high-dimensional window data. Along the dimension is divided into slices, and the time dimension size of each slice is , and the empty positions are filled with . Each slice is a subspace.

[0041] For example, consider three-dimensional window data , whose size is 4×5×2. When we partition the subspace along the time dimension, we get three slices, namely , and the size of each slice is 4×2×2. It should be noted that the empty positions in the second dimension of the third slice are filled with the element 0.

[0042] Clustering to construct a codebook is as Figure 6 shown. The specific implementation is as follows: The present invention sequentially performs K-means clustering on the th slice of each window data to obtain the th clustering center slice of clustering centers. After times of clustering, finally, clustering center slices are obtained to form the clustering centers, which are the codebook.

[0043] For example, consider 4 window data belonging to the space 4×5×2. These 4 window data are divided into 3 slices, namely . The goal is to cluster all slices of the 4 window tensors into a codebook consisting of 2 clustering centers. When , perform K-means clustering on the first slice of the 4 window tensors to obtain the first clustering center slice of 2 clustering centers. When and , perform the same operation. Finally, these clustering center slices are restored to the corresponding 2 clustering centers and , where , .

[0044] According to the codebook quantization, as Figure 7 shown, the specific implementation is as follows: For the th slice of each window tensor, we separately compare its distance with the th cluster center's th cluster center slice, and then record the index of the cluster center corresponding to the closest distance. Finally, each high-dimensional window data is transformed into one-dimensional data composed of index values.

[0045] For example, in the clustering stage, two cluster centers are obtained: and . The window data corresponds to three slices . When , calculate the distance between the slice and the cluster center slices and to obtain and . If , then ; when , calculate the distance between the slice and the cluster center slice and to obtain and . If , then ; when , we calculate the distance between the slice and the cluster center slices and to obtain and . If , then . Finally, transform the multi-dimensional window tensor into one-dimensional data .

[0046] According to the codebook quantization, as Figure 7 shown, the specific implementation is as follows: For the p-th subspace of each window tensor, we calculate its distance from C cluster centers in the codebook. Assign the index of the nearest neighbor cluster center, and its mathematical expression is: , where represents the p-th slice of the d-th window, represents the p-th slice of the c-th cluster center. By sequentially calculating the distance between the p-th slice of the current window tensor and the p-th slices of different cluster centers, find the nearest neighbor cluster center and assign its corresponding index value to .

[0047] Construct a distance table, as Figure 8 shown. The specific implementation is as follows: Divide the current satellite tensor into P sub-tensors, and then calculate the distances between these sub-tensors and the centroids of each cluster center to obtain the elements in the distance table. The mathematical expression of each element in the distance table is: .

[0048] Query distances using the distance table, as Figure 9 shown. The specific implementation is as follows: Given the distance table, to calculate the distance between the current satellite tensor and the nth window tensor, we only need to query the corresponding element in the distance table and sum them. The mathematical expression for calculating the distance between the current satellite tensor and the dth window tensor is: , where represents the element value corresponding to the th row and pth column in the distance table, where is calculated through the formula mentioned above.

[0049] Determine neighbor data. To find the N most similar historical window tensors, we use the distance table to sequentially query the distances between D window tensors and the current satellite tensor, and then select the smallest N values from them. The window tensors corresponding to these minimum distances are the final similar tensors.

[0050] Figure 10 FIG. is a schematic structural diagram of a device for restoring satellite state data through historical enhancement provided by an embodiment of the present invention. The device 600 for restoring satellite state data through historical enhancement may vary greatly due to configuration or performance differences, and may include one or more processors (central processing units, CPU) 610 (for example, one or more processors) and a memory 620, and one or more storage media 630 (for example, one or more mass storage devices) for storing application programs 633 or data 632. Among them, the memory 620 and the storage media 630 can be transient storage or persistent storage. The program stored in the storage media 630 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the device 600 for restoring satellite state data through historical enhancement. Further, the processor 610 may be configured to communicate with the storage media 630 and execute a series of instruction operations in the storage media 630 on the device 600 for restoring satellite state data through historical enhancement to implement the method provided by the above embodiment.

[0051] The device 600 for restoring satellite status data through historical enhancement may further include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input / output interfaces 660, and / or one or more operating devices 631, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that Figure 10 The structure of the device for restoring satellite status data through historical enhancement shown does not constitute a limitation on the computer device provided by the present invention, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0052] The present invention also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the method for restoring satellite status data through historical enhancement provided in each of the above embodiments.

[0053] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses, or units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0054] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0055] The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A method for restoring satellite status data through historical enhancement, characterized in that, The method includes the following steps: S1. Calculate the similarity between the current satellite data and multiple segments of historical data, and select several segments of the most similar historical data as the matched historical data; S2. Design a tensor fusion filling scheme, determine the similarity between the current satellite data and the historical data, and assign different weights to the current satellite data and the matched neighbor data respectively. Then fuse the current satellite data and the neighbor data with assigned weights, train a data completion model, and output complete satellite data through the data completion model, where the neighbor data is the historical data obtained in step S1; The specific implementation process of step S1 includes: S1-1. Input the historical data with missing values into the missing data completion model to obtain complete historical data; S1-2. Set a sliding window with the same size as the current satellite data, and start matching from the first time node of the complete historical data to obtain window data; S1-3. Using the method of subspace clustering quantization, calculate the similarity between the current satellite data and the data of each window in turn, and select the neighboring data that is most similar to the current satellite data.

2. The method for restoring satellite status data through historical enhancement as claimed in claim 1, wherein The specific implementation process of step S1-3 includes: Perform subspace partitioning on each window data, cluster the corresponding subspaces of all window data to construct a codebook, and then quantize the original space according to the codebook; Construct a distance table for the current satellite data, and quickly query the distance between the current satellite data and each window tensor according to the distance table. As a similarity measurement criterion, select the ones with the closest distance as neighbor data.

3. The method for restoring satellite status data through historical enhancement according to claim 2, characterized in that The performing subspace partitioning on each window data, clustering the corresponding subspaces of all window data to construct a codebook, and then quantizing the original space according to the codebook includes: Slice each high-dimensional window data, which is divided along the time dimension into slices, and the size of the time dimension of each slice is . The empty positions are filled with 0, and each slice is a subspace; Perform K-means clustering on the th slice of each window data in sequence to obtain th clustering center's th clustering center slice. After times of clustering, finally obtain clustering center slices that form clustering centers, which are the codebooks; For the th slice of each window tensor, respectively compare its distance with the th cluster center's th cluster center slice, then record the index of the cluster center corresponding to the nearest distance. Finally, convert each high-dimensional window data into one-dimensional data composed of index values.

4. The method for restoring satellite status data through historical enhancement as claimed in claim 3, wherein, Construct a distance table for the current satellite data, quickly query the distances between the current satellite data and each window tensor according to the distance table, use them as similarity measurement criteria, and select the ones with the closest distances as neighbor data, including: Perform subspace partitioning on the current satellite data to obtain slices, and then calculate the distances between each slice and the slices of each cluster center respectively, and sequentially place the calculated distances in the corresponding positions of the distance table; For the distance calculation between the current satellite data and each window data, query the corresponding elements in the distance table and sum them; Use the distance table to obtain the distances between the current satellite data and all window data, and select the one with the smallest distance from them. The window data corresponding to these distances is the neighbor data.

5. A method for restoring satellite status data through historical enhancement as claimed in claim 4, wherein, The distance table includes rows and columns. The th row of the distance table represents the th slice of the current satellite data and the th cluster center's th cluster center slice. The th column of the distance table represents the th slice of the current satellite data and the th cluster center's th cluster center slice.

6. The method for restoring satellite status data through historical enhancement as claimed in claim 1, wherein The specific implementation process of step S2 includes: Assign weights according to the distance between neighbor data and current satellite data, and normalize the assigned weights so that they are distributed within between.

7. A system for restoring satellite status data through historical enhancement, characterized in that, The system includes: The historical data matching module is used to calculate the similarity between the current satellite data and multiple segments of historical data, and select several segments of the most similar historical data as the matched historical data: input the historical data with missing values into the missing data completion model to obtain complete historical data; set a sliding window with the same size as the current satellite data, and start matching from the first time node of the complete historical data to obtain window data; adopt the method of subspace clustering quantization to calculate the similarity between the current satellite data and window data in turn, and select the neighbor data that is most similar to the current satellite data; A tensor fusion filling module, which is used to design a tensor fusion filling scheme, determine the similarity between the current satellite data and the historical data, and assign different weights to the current satellite data and the matched neighbor data respectively. Then fuse the current satellite data and the neighbor data with assigned weights, train a data completion model, and output complete satellite data through the data completion model, where the neighbor data is the historical data obtained in step S1.

8. A device for restoring satellite status data through historical enhancement, characterized in that, The device for recovering satellite state data through historical enhancement includes a memory and at least one processor. Instructions are stored in the memory; the at least one processor calls the instructions in the memory to cause the device for recovering satellite state data through historical enhancement to execute each step of the method for recovering satellite state data through historical enhancement as described in any one of claims 1-6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that, When the instructions are executed by the processor, each step of the method for recovering satellite state data through historical enhancement as described in any one of claims 1-6 is implemented.

Images