Satellite link saturation evolution analysis method based on spatiotemporal graph convolution

Abstract

The present invention relates to a satellite link saturation evolution analysis method based on spatiotemporal graph convolution, comprising: step S1, processing and cleaning satellite data, retaining satellite attribute information, service information and satellite link information as characteristic information of satellite nodes; step S2, static topology modeling of satellite link network according to different time slices, and sending every two adjacent frames of the modeled spatiotemporal graph as a training sample to the spatiotemporal graph convolution network training; step S3, sending the convolved features to the prediction head, and outputting the load probability of each frequency band until the model training converges; step S4, using the trained model to predict the satellite link saturation. The present invention can effectively capture the spatiotemporal dependency of satellite links, making the prediction and saturation analysis of link load more accurate and reliable, and can better adapt to complex and changeable communication environments, providing more powerful support for network planning, resource optimization and troubleshooting.

Description

Technical Field

[0001] The present invention relates to the fields of satellite links and spatiotemporal graph convolutional neural technology, and in particular to a satellite link saturation evolution analysis method based on spatiotemporal graph convolution. Background Art

[0002] With the rapid development of communication technology, satellite links play a vital role in modern communication and information transmission. Satellite links are not only used for international satellite communications, but are also widely used in multiple fields such as satellite Internet, satellite broadcasting, satellite navigation, and earth remote sensing. In such a diverse range of application scenarios, load balancing and performance analysis of satellite links have become important issues to ensure communication quality and service reliability. The present invention defines the saturation of satellite links according to the characteristics of satellite links, and designs an effective satellite link saturation analysis method, which can help communication operators, government agencies and enterprises to better plan and manage satellite link resources, improve communication coverage, and meet the growing communication needs. This is an important task in the field of satellite communications and has broad value for various applications and fields.

[0003] The application of deep learning in the satellite field is also becoming more and more extensive. For example, deep learning plays an important role in satellite image processing. Large-scale images obtained through satellite remote sensing data need to be efficiently processed and analyzed to extract useful geographic information. Deep learning methods perform well in image classification, target detection, change detection, etc., and can help accurately identify information such as urban planning, land use, resource distribution, etc., and provide strong support for urban planning, agricultural management, natural resource protection and other fields. Secondly, deep learning also has broad application prospects in the field of satellite communications. With the continuous expansion and upgrading of satellite communication networks, more intelligent network management and spectrum allocation methods are needed to cope with the growing communication needs of users. The spatiotemporal graph convolution technology of deep learning can be used to optimize satellite link resources, improve communication quality, improve network performance, and reduce communication costs.

[0004] Therefore, there is an urgent need for a satellite link saturation evolution analysis method based on spatiotemporal graph convolution. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a satellite link saturation evolution analysis method based on spatiotemporal graph convolution, which can effectively capture the spatiotemporal dependence of satellite links, making the prediction of link load and saturation analysis more accurate and reliable.

[0006] To achieve the above-mentioned object of the invention, the present invention provides a satellite link saturation evolution analysis method based on spatiotemporal graph convolution, comprising the following steps:

[0007] Step S1, processing and cleaning the satellite data, retaining satellite attribute information, service information and satellite link information as characteristic information of the satellite node;

[0008] Step S2, static topology modeling of the satellite link network is performed according to different time slices, and every two adjacent frames of the modeled space-time graph are sent as a training sample to the space-time graph convolution network training;

[0009] Step S3: Send the convolved features to the prediction head and output the load probability of each frequency band until the model training converges;

[0010] Step S4: Use the trained model to predict the satellite link saturation.

[0011] According to a technical solution of the present invention, in step S1, the satellite attribute information includes at least a satellite identifier, a satellite communication frequency band, satellite orbit parameters, satellite status parameters and satellite quality; the service information includes at least a service identifier, a service arrival time, a service data volume and a service destination station; the satellite link information includes at least a link frequency band, a link status and a link channel capacity.

[0012] According to a technical solution of the present invention, in step S2, the spatiotemporal graph convolutional network is constructed by a spatiotemporal convolutional module with residual connections of an eight-layer graph convolution module and a temporal convolution module.

[0013] According to a technical solution of the present invention, in step S2, the spatiotemporal graph convolution network establishes a time link between each node of the topological graph at time t and time t-1, and each time the graph convolution is performed, the topological graph at time t and time t-1 is convolved as a whole;

[0014] Among them, t-1 is the previous moment of t.

[0015] According to a technical solution of the present invention, in step S2, the formula of the spatiotemporal graph convolutional network is:

[0016]

[0017] Among them, A n is the adjacency matrix of the input graph, whose nodes have a time link to themselves at the last moment. The adjacency matrix A of the graph of two nodes v1 and v2 is Then A n for The newly added rows and columns are time links, Λ n Same reason.

[0018] According to a technical solution of the present invention, in step S3, the loss function adopts a mean square error regression loss function, and the mean square error between the network prediction value and the actual saturation of the next frame is averaged as the network loss function.

[0019] According to a technical solution of the present invention, the formula of the loss function is:

[0020]

[0021] Among them, Y is the actual satellite frequency band occupancy at a certain moment, its value is a K*1-dimensional matrix, which is a one-hot encoding, and F(x) is the load probability of each of the K frequency bands predicted by the network, both of which are values ​​0-1.

[0022] According to a technical solution of the present invention, in step S3, the output of the spatiotemporal graph convolutional network is:

[0023] F(x)=[p1,p2,…,p k ]

[0024] Among them, p i It is the load probability of each frequency band in the K frequency bands predicted by the network, and both are values ​​of 0-1.

[0025] According to a technical solution of the present invention, the satellite link saturation is defined as:

[0026]

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] This paper proposes a satellite link saturation evolution analysis method based on spatiotemporal graph convolution, which provides a more accurate, adaptive, efficient and practical solution in load analysis. The introduction of spatiotemporal graph convolution enhances the modeling of complex spatiotemporal relationships, improves the accuracy of prediction, and makes the analysis method more practical. Compared with traditional methods that require manual adjustment of parameters, it has higher adaptability and generalization performance, which is conducive to significantly improving efficiency and quality in satellite communication operations, bringing technological progress and operational improvements to the industry.

[0029] Compared with routing algorithms that only consider static spatiotemporal weighted graphs, the present invention significantly improves the generalization ability of the algorithm. Traditional methods cannot cope with real-time changes in satellite links, while algorithms based on spatiotemporal graph convolutional neural networks can more intelligently capture dynamic topological information. In addition, this scheme takes into account link saturation, which helps reduce congestion and improve communication reliability. This scheme provides a more efficient and reliable solution for satellite link routing.

[0030] Furthermore, the application of deep learning has brought a higher level of intelligence and automation to the satellite field, further improving the performance and benefits of satellite remote sensing, communication, navigation, etc. At the same time, it can promote the development of satellite technology in a more intelligent, reliable and efficient direction. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 The overall flow chart of the satellite link saturation evolution analysis method based on spatiotemporal graph convolution according to the present invention is schematically shown;

[0033] Figure 2 A schematic diagram showing a basic convolution module structure according to an embodiment of the present invention;

[0034] Figure 3 Schematically illustrating a graph convolution process according to an embodiment of the present invention;

[0035] Figure 4 The flowchart of a satellite link saturation evolution analysis method based on spatiotemporal graph convolution according to an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0036] The description of the embodiments of this specification should be combined with the corresponding drawings, which should be considered as part of the complete specification. In the drawings, the shape or thickness of the embodiments may be enlarged and indicated for simplification or convenience. Furthermore, the parts of each structure in the drawings will be described separately. It is worth noting that the elements not shown in the drawings or not described in words are in a form known to ordinary technicians in the relevant technical field.

[0037] The description of the embodiments herein and any reference to directions and orientations are only for the convenience of description and are not to be construed as any limitation on the scope of protection of the present invention. The following description of the preferred embodiments may involve combinations of features, which may exist independently or in combination, and the present invention is not particularly limited to the preferred embodiments. The scope of the present invention is defined by the claims.

[0038] like Figures 1 to 4 As shown, a satellite link saturation evolution analysis method based on spatiotemporal graph convolution of the present invention comprises the following steps:

[0039] Step S1, configuring the equipment composition of the equipment system, configuring the internal composition of the equipment system;

[0040] Step S2, generating and setting system internal level operations, generating and setting system level operations;

[0041] Step S3, generating and configuring the mission tasks of the equipment system;

[0042] Step S4, generating and setting the system internal level working mode, generating and setting the system level working mode;

[0043] Step S5: construct and configure the task mode of the equipment system.

[0044] In this embodiment, the complex satellite data is first processed and cleaned, and the satellite attribute information, service information and satellite link information are retained as the characteristic information of the satellite node. Then, according to different time slice divisions, the satellite link is statically topologically modeled to form a satellite link space-time graph, and this space-time graph matrix is ​​sent as training data to the space-time graph convolution network in the present invention for training, and the network model is optimized through special space-time graph convolution and space-time loss calculation. Finally, the saturation of the satellite link is defined, and the satellite link network evolution analysis is performed according to the saturation.

[0045] Specifically, spatiotemporal graph convolution allows the satellite link saturation analysis model to transfer information and learn features in the spatiotemporal dimension, thereby better understanding the topological structure of the communication network and the relationships between nodes. By learning node embeddings on the spatiotemporal graph, the deep learning method can effectively capture the spatiotemporal dependencies of satellite links, making the prediction and saturation analysis of link load more accurate and reliable. This comprehensive modeling approach enables satellite link saturation analysis to better adapt to complex and changing communication environments, providing stronger support for network planning, resource optimization, and troubleshooting.

[0046] In one embodiment of the present invention, preferably, the satellite link saturation evolution analysis method based on spatiotemporal graph convolution of the present invention includes three parts: a satellite link spatiotemporal graph modeling method, a spatiotemporal graph convolutional neural network algorithm, and a satellite link saturation definition and analysis method.

[0047] 1. Satellite data processing and spatiotemporal graph convolutional network design

[0048] In the early stage of data collection, it may be affected by various interference sources, such as sensor errors, communication channel interference or environmental noise. Therefore, the first task is to clean and denoise the raw data to eliminate outliers and bad data points. It helps to ensure that the satellite data used is of high quality and provides reliable feature information. Then, satellite attribute information, service information and satellite link information are extracted as feature information of satellite nodes. Among them, satellite attribute information includes satellite identifier Sid, satellite communication frequency band Srange, satellite orbit parameter Sop, satellite status parameter Ssp and satellite quality Sq; service information includes service identifier Bid, service arrival time Bstart, service data volume Bdata and service destination station Bdes; satellite link information includes link frequency band Lrange, link status Lstatus and link channel capacity Lcap.

[0049] The above process is called the data preprocessing stage. Next, the satellite link network will be modeled in detail based on different time slices. The key goal of this modeling process is to capture the complexity of the spatiotemporal relationships in the satellite link network. The spatiotemporal graph convolutional network will be trained to understand these relationships so as to better predict the saturation of satellite links. Data conversion between different time slices is used so that the network can adapt to changes in link topology and maintain accuracy. Then, the data of each two adjacent time slices is converted into a training batch. This training batch will be used as input and fed into the spatiotemporal graph convolutional network for training. At the same time, a portion of the data in the spatiotemporal graph is reserved as test data to test the performance of the spatiotemporal graph convolutional network model.

[0050] The spatiotemporal graph convolutional network of the present invention adopts a highly complex architecture, consisting of eight layers of graph convolution modules and residual connections of temporal convolution modules. The core idea of ​​this design is to combine spatial information and temporal information to more comprehensively understand the evolution of the satellite link network. Specifically, each graph convolution module is responsible for extracting important features from the spatial graph, while the temporal convolution module captures the temporal dependencies. These two aspects of information are organically integrated and transmitted through residual connections to ensure the integrity and coherence of the information. Finally, after passing through these complex modules, the features are fed into the prediction head, which is the last part of the network. The task of the prediction head is to convert these features into load probabilities for each frequency band. Once the model is trained and converged, it will be able to predict the saturation of the satellite link with high accuracy, making the analysis of satellite link saturation more efficient and accurate. It not only helps to identify the performance bottlenecks of the network in a timely manner, but also provides strong data support for decision makers to better plan and manage satellite communication networks. The spatiotemporal graph convolutional network of the present invention provides an advanced tool for monitoring and predicting satellite link saturation.

[0051] The key principle of the satellite link spatiotemporal graph modeling method is to extract the key features of satellite links and perform static topological modeling to better understand the dynamics and complexity of the link network. First, in the data preprocessing stage, the satellite data is cleaned and denoised to ensure the quality and reliability of the data. Then, satellite attributes, service information, and satellite link information are extracted from the satellite data, which will be used as features to describe the status and performance of the satellite link network. Subsequently, data in different time slices are used to construct a static topological model of the satellite link. This modeling process aims to capture the structure and evolution of the satellite link network over time. By modeling this topological structure, the performance and saturation of the satellite link network can be better understood. The importance of this method lies in that it provides an effective way to analyze the saturation of satellite links. By cleaning data, extracting features, and building a topological model, the performance of the satellite link network can be better understood, and corresponding measures can be taken to improve its performance and reliability.

[0052] 2. Detailed explanation of spatiotemporal graph convolution and loss function design

[0053] like Figure 2 As shown in the figure, the basic convolution module has a relatively simple structure but is the core of the entire graph convolutional network module. First, the input feature undergoes a 1*1 two-dimensional convolution operation, which helps to learn the nonlinear mapping of the feature. Then, the sum of the feature matrix after convolution is calculated using the Einstein summation convention, which helps to process the feature as a whole. Subsequently, a DropOut layer is used to randomly close a part of the neurons, thereby preventing the model from being overly dependent on specific neurons and improving the generalization ability. Next, the feature is passed to the temporal convolutional network module, which is an important part of the entire convolutional network. In the temporal convolutional module, the feature is first sent to the BN (Batch Normalization) layer for feature normalization, which helps to accelerate the training process. Then, the ReLU activation function is used for nonlinear mapping to further extract the complexity of the feature. The feature then undergoes a 9*1 two-dimensional convolution operation, which helps to capture the association between time series information and frequency domain information. Finally, the feature is normalized and the DropOut layer is used again through the BN layer to improve the generalization ability of the model and reduce the risk of overfitting.

[0054] The core of the entire graph convolutional network consists of this basic convolutional module repeatedly connected 8 times to fully extract and integrate feature information. Finally, these features will be fed into a prediction head to output the load probability of each frequency band and complete the saturation prediction task of the satellite link. The structure and hierarchical distribution of this model help to effectively capture spatiotemporal information and improve model performance.

[0055] The formula for traditional graph convolution is as follows:

[0056]

[0057] Λ ii =∑ j (A ij +I ij )

[0058] Among them, A is the adjacency matrix of the input graph, I is the unit matrix, and A+I adds a self-loop to the input graph to ensure the effectiveness of data transmission. W is composed of weight vectors of multiple output channels, f in is the input feature map, f out is the output feature map. ii are the diagonal elements of the adjacency matrix, and their values ​​are calculated by summing the elements at corresponding positions in the input graph and the identity matrix.

[0059] The graph convolution formula of the present invention is as follows:

[0060]

[0061] Among them A n It is still the adjacency matrix of the input graph, but its nodes have a time link to their last moment, which is equivalent to the expansion of the graph. Taking the graph with two nodes v1 and v2 as an example, its adjacency matrix A is Then A n for The newly added rows and columns are time links, Λ n Same reason.

[0062] The loss function uses mean square error loss, and its specific function formula is as follows:

[0063]

[0064] Where Y is the actual satellite frequency band occupancy at a certain moment, and its value is a K*1-dimensional matrix, which is a one-hot encoding. F(x) is the load probability of each of the K frequency bands predicted by the network, and both are 0-1 values. The difference between these two values ​​and the sum of squares is the mean square error, which is the loss function of the present invention.

[0065] The spatiotemporal graph convolutional neural network plays a key role in satellite link saturation analysis. Its core principle is to combine the complexity of spatiotemporal data and encode the spatiotemporal relationship into the neural network, so that the network can better understand the evolution and load distribution of satellite links. This deep learning method that integrates spatiotemporal information has great advantages over traditional methods. On the one hand, the spatiotemporal graph convolution can capture spatiotemporal dependencies. The performance of satellite links is related to time and space. The network can effectively capture hidden spatiotemporal dependencies through the spatiotemporal graph learning module, which helps to more accurately predict the saturation of satellite links. On the other hand, the spatiotemporal graph convolutional neural network provides more accurate saturation analysis results through the powerful feature extraction and spatiotemporal modeling capabilities of the deep learning model, which helps to take appropriate adjustment measures in advance to ensure the normal operation of the network. In addition, the spatiotemporal graph convolutional neural network is adaptable, which makes the management of satellite links more flexible and reliable. The importance of the spatiotemporal graph convolutional neural network in satellite link saturation analysis lies in that its deep learning model based on spatiotemporal relationships can more comprehensively, accurately and adaptively analyze and predict the load status of satellite links, which helps to improve the performance and reliability of satellite communication networks.

[0066] 3. Satellite link saturation calculation and analysis

[0067] F(x)=[p1,p2,…,p k ]

[0068] The above formula is the final output of the spatiotemporal graph convolutional network, where p i is the load probability of each of the K frequency bands predicted by the network, all of which are values ​​between 0 and 1. The saturation of the satellite link is defined as:

[0069]

[0070] High saturation indicates that the satellite link network is extremely busy, which means that many satellite communication bands are close to or exceed their load capacity. In this case, network performance may be significantly affected, such as increased packet loss rate, increased latency, and even increased risk of communication interruption. This is a serious problem for satellite communication networks because these networks are often used to transmit important data and communications in areas such as emergency rescue, remote monitoring and communication. In contrast, low saturation indicates that the satellite link network is relatively lightly loaded. This may mean that the network has enough spare capacity to handle additional communication traffic, providing faster response times and more reliable communication services. However, too low saturation may also indicate that resources are not fully utilized, resulting in inefficient network performance. Therefore, maintaining moderate saturation is crucial for the effective operation of satellite link networks.

[0071] The core principle of the satellite link saturation analysis method is to define the saturation of the satellite link by predicting the load probabilities of K satellite communication frequency bands and accumulating them. Saturation reflects the load condition of the satellite link network. High saturation means that the network load is heavy, and low saturation means that the load is relatively light. It can monitor and manage satellite communication systems in real time to ensure the continuity of data transmission and communication services. Accurate saturation analysis helps to take measures in advance to prevent network congestion or waste of resources, thereby maintaining the performance and reliability of satellite communications. Therefore, the satellite link saturation analysis method is a key tool to ensure the smooth operation of satellite communication systems and has important practical application significance.

[0072] Predicting and monitoring the saturation of satellite links through spatiotemporal graph convolutional networks can help network operators better understand the status of satellite links and take timely measures to optimize network performance, allocate resources, and ensure high quality and reliability of communications, which is crucial to improving the performance of satellite link networks in different application scenarios.

[0073] According to one aspect of the present invention, an electronic device is provided, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory, and when the electronic device is running, the processor executes the one or more computer programs stored in the memory, so that the electronic device executes a satellite link saturation evolution analysis method based on spatiotemporal graph convolution as any one of the above technical solutions.

[0074] According to one aspect of the present invention, a computer-readable storage medium is provided for storing computer instructions. When the computer instructions are executed by a processor, a satellite link saturation evolution analysis method based on spatiotemporal graph convolution as described in any of the above technical solutions is implemented.

[0075] The present invention discloses a satellite link saturation evolution analysis method based on spatiotemporal graph convolution, comprising: step S1, processing and cleaning satellite data, retaining satellite attribute information, service information and satellite link information as characteristic information of satellite nodes; step S2, static topology modeling of satellite link network according to different time slices, and sending each two adjacent frames of the spatiotemporal graph after modeling as a training sample to the spatiotemporal graph convolution network training; step S3, sending the convolved features to the prediction head, and outputting the load probability of each frequency band until the model training converges; step S4, using the trained model to predict the satellite link saturation; spatiotemporal graph convolution allows the satellite link saturation analysis model to transfer information and learn features in the spatiotemporal dimension, so as to better understand the topological structure of the communication network and the relationship between nodes. By learning the node embedding on the spatiotemporal graph, the deep learning method can effectively capture the spatiotemporal dependency of the satellite link, making the prediction and saturation analysis of the link load more accurate and reliable, and also making the satellite link saturation analysis better adapt to the complex and changeable communication environment, providing more powerful support for network planning, resource optimization and troubleshooting.

[0076] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0077] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0078] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0079] It should also be noted that, in this article, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0080] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A satellite link saturation evolution analysis method based on spatiotemporal graph convolution, characterized in that: The following steps are involved: Step S1, processing and cleaning the satellite data, retaining satellite attribute information, service information and satellite link information as characteristic information of the satellite node; Step S2, static topology modeling of the satellite link network is performed according to different time slices, and every two adjacent frames of the modeled space-time graph are sent as a training sample to the space-time graph convolution network training; Step S3: Send the convolved features to the prediction head and output the load probability of each frequency band until the model training converges; Step S4: using the trained model to predict the satellite link saturation; The satellite link saturation is used to reflect the load status of the satellite link network. A high saturation indicates that the network load is heavy, and a low saturation indicates that the load is relatively light. In step S2, the spatiotemporal graph convolution network is constructed by using a spatiotemporal convolution module with residual connections of an eight-layer graph convolution module and a temporal convolution module; In step S2, the spatiotemporal graph convolution network establishes a time link between each node of the topological graph at time t and time t-1, and each time the graph convolution is performed, the topological graph at time t and time t-1 is convolved as a whole; Among them, t-1 is the previous moment of t; In step S2, the formula of the spatiotemporal graph convolutional network is: Among them, A n is the adjacency matrix of the input graph, whose nodes have a time link to themselves at the last moment. The adjacency matrix A of the graph of two nodes v1 and v2 is Then A n for The newly added rows and columns are time links, Λ n Similarly; I is the unit matrix, W is composed of weight vectors of multiple output channels, f in is the input feature map, Λ ii are the diagonal elements of the adjacency matrix, and their values ​​are calculated by summing the elements of the input graph and the identity matrix at corresponding positions; In step S3, the output of the spatiotemporal graph convolutional network is: F(x)=[p1,p2,…,p k ] Among them, p i is the load probability of each frequency band in the K frequency bands predicted by the network, and both are values ​​of 0-1; The satellite link saturation is defined as:

2. The satellite link saturation evolution analysis method based on spatiotemporal graph convolution according to claim 1, characterized in that: In step S1, the satellite attribute information includes at least a satellite identifier, a satellite communication frequency band, a satellite orbit parameter, a satellite status parameter and a satellite quality; the service information includes at least a service identifier, a service arrival time, a service data volume and a service destination station; The satellite link information at least includes link frequency band, link status and link channel capacity.

3. The satellite link saturation evolution analysis method based on spatiotemporal graph convolution according to claim 1, characterized in that: In step S3, the loss function adopts a mean square error regression loss function, and the mean square error between the network prediction value and the actual saturation of the next frame is averaged as the network loss function.

4. The satellite link saturation evolution analysis method based on spatiotemporal graph convolution according to claim 3 is characterized in that: The formula of the loss function is: Among them, Y is the actual satellite frequency band occupancy at a certain moment, its value is a K*1-dimensional matrix, which is a one-hot encoding, and F(x) is the load probability of each of the K frequency bands predicted by the network, both of which are values ​​0-1.

Images