Low earth orbit satellite system flow prediction method, device, equipment and storage medium

Abstract

The invention provides a low-earth-orbit satellite system flow prediction method, device and equipment and a storage medium, and the method comprises the steps: obtaining historical flow data of a low-earth-orbit satellite system, and carrying out the preprocessing of the historical flow data, and obtaining the preprocessing data; constructing a traffic feature matrix based on the preprocessed data; constructing an adjacent matrix of a graph structure corresponding to the low earth orbit satellite system; correlation coefficients of the satellite pairs are calculated pairwise based on the preprocessed data in the graph structure, and a correlation coefficient matrix is obtained; inputting the traffic characteristic matrix, the adjacent matrix and the correlation coefficient matrix into a pre-established dual-channel space extraction model to obtain spatial characteristics of the traffic; inputting the spatial features into a pre-established time extraction model to obtain time features of the traffic; and inputting the time characteristics into a pre-established full connection layer to obtain a flow prediction result of each satellite. According to the overall prediction method, the spatial correlation and the temporal correlation of the satellite traffic are considered at the same time, and the satellite traffic is predicted by fusing the spatial-temporal characteristics.

Description

Technical Field

[0001] The present invention relates to the field of satellite traffic prediction technology, and in particular to a low-orbit satellite system traffic prediction method, device, equipment and storage medium. Background Art

[0002] The traditional terrestrial Internet of Things has a complete system, while the satellite Internet of Things is in a stage of rapid development. It is mainly composed of low-orbit satellites and medium-orbit satellites, and has the advantages of wide coverage, large communication capacity, no geographical restrictions, high dynamics and strong expansion capabilities.

[0003] In related technologies, the essence of satellite traffic prediction is time series prediction. The basic method is to establish a corresponding mathematical model by analyzing historical traffic data for prediction. It is mainly divided into traditional prediction methods, machine learning prediction methods and deep learning prediction methods. The above technical means can obtain satisfactory prediction accuracy, but the low-orbit satellite constellation system is a complex network topology structure. The traffic of a single satellite will be affected by other satellites in the topology, and the above technical means do not take into account this correlation effect.

[0004] Based on the above analysis of the development status of this technical field, the existing technology lacks a solution for dynamically predicting the traffic of low-orbit satellite systems based on considering the correlation coefficient between a single satellite and other surrounding satellites. Summary of the invention

[0005] The purpose of the present invention is to provide a low-orbit satellite system traffic prediction method, device, equipment and storage medium, aiming to solve the above-mentioned problems in the prior art.

[0006] According to a first aspect of an embodiment of the present invention, a low-orbit satellite system traffic prediction method is provided, comprising:

[0007] Acquire historical traffic data of the low-orbit satellite system and pre-process it to obtain pre-processed data;

[0008] Construct a traffic feature matrix based on the preprocessed data; construct an adjacency matrix corresponding to the graph structure of the low-orbit satellite system; in the graph structure, calculate the correlation coefficients of satellite pairs based on the preprocessed data to obtain a correlation coefficient matrix;

[0009] Input the traffic feature matrix, adjacency matrix and correlation coefficient matrix into the pre-established dual-channel space extraction model to obtain the spatial features of the traffic; input the spatial features into the pre-established time extraction model to obtain the time features of the traffic;

[0010] The time features are input into the pre-established fully connected layer to obtain the traffic prediction results of each satellite.

[0011] According to a second aspect of an embodiment of the present invention, a low-orbit satellite system traffic prediction device is provided, comprising:

[0012] A preprocessing module is used to obtain historical flow data of the low-orbit satellite system and perform preprocessing to obtain preprocessed data;

[0013] The matrix calculation module is used to construct a traffic feature matrix based on preprocessed data; construct an adjacency matrix corresponding to the graph structure of the low-orbit satellite system; in the graph structure, the correlation coefficients of satellite pairs are calculated based on the preprocessed data to obtain a correlation coefficient matrix;

[0014] The spatiotemporal prediction module is used to input the traffic feature matrix, adjacency matrix and correlation coefficient matrix into the pre-established dual-channel spatial extraction model to obtain the spatial features of the traffic; and input the spatial features into the pre-established temporal extraction model to obtain the temporal features of the traffic;

[0015] The result acquisition module is used to input the time features into the pre-established fully connected layer to obtain the traffic prediction results of each satellite.

[0016] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the low-orbit satellite system traffic prediction method provided in the first aspect of the present disclosure are implemented.

[0017] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps of the low-orbit satellite system traffic prediction method provided in the first aspect of the present disclosure are implemented.

[0018] The technical solution provided by the embodiment of the present invention includes the following beneficial effects: a dynamic prediction reference for low-orbit satellite system traffic prediction is provided. Overall, the prediction method takes into account the spatial correlation and temporal correlation of satellite traffic, and integrates spatiotemporal features to predict satellite traffic; and the correlation coefficient between satellites is taken into account when extracting spatial features to take into account that the traffic of a single satellite will be affected by other satellites in the topology; overall, the accuracy of low-orbit satellite system traffic prediction is improved.

[0019] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 is a flow chart of a low-orbit satellite system traffic prediction method according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of a low-orbit satellite system flow prediction device according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0025] Method Embodiment

[0026] According to an embodiment of the present invention, a method for predicting flow of a low-orbit satellite system is provided. Figure 1 : is a flow chart of a method for predicting flow of a low-orbit satellite system according to an embodiment of the present invention. Figure 1 As shown, the low-orbit satellite system traffic prediction method according to an embodiment of the present invention specifically includes:

[0027] In step S110, historical traffic data of the low-orbit satellite system is obtained and preprocessed to obtain preprocessed data, which specifically includes:

[0028] Through space-based, air-based or ground-based equipment, satellites within the visible range can be communicated with to obtain historical traffic data with temporal and spatial characteristics;

[0029] Use the isolation forest to check abnormal and invalid data in the historical traffic data. Preferably, additional features such as timestamp, traffic change rate, weather conditions, bandwidth, delay and packet loss rate can be obtained to assist random segmentation in the isolation forest; remove abnormal and invalid data from the historical traffic data to obtain a first data set;

[0030] Fill the missing data in the first data set using the mean difference filling method to obtain a second data set;

[0031] The second data set is normalized. In the embodiment of the present invention, the Min-Max method is used for normalization to obtain preprocessed data.

[0032] In step S120, a traffic feature matrix is ​​constructed based on the preprocessed data; an adjacency matrix corresponding to the graph structure of the low-orbit satellite system is constructed; in the graph structure, correlation coefficients of satellite pairs are calculated based on the preprocessed data to obtain a correlation coefficient matrix, which specifically includes:

[0033] Based on the preprocessed data including N satellites and D step time, that is, N represents the number of satellites, D represents the length of the traffic feature matrix, that is, the feature data, the traffic feature matrix is ​​constructed using Formula 1:

[0034]

[0035] Among them, X represents the traffic feature matrix, represents the flow value of the nth satellite at time d, 1≤n≤N, 1≤d≤D, and D represents the length of the characteristic matrix.

[0036] The traffic prediction problem can be abstracted as follows:

[0037]

[0038] Among them, T f Indicates the length of time required for prediction; T h Indicates the time length of the historical sequence used for prediction; G is the graph structure abstracted from the low-orbit satellite constellation; F represents the correspondence between the historical traffic and the future traffic to be predicted;

[0039] A single satellite is abstracted as a node in a graph structure, and the links between satellites are abstracted as edges in the graph structure. The graph structure is an undirected graph G = (V, E), and the adjacency matrix A∈R corresponding to the graph structure is established. N×N , R represents the matrix form, and the adjacency matrix is ​​established as shown in Formula 7:

[0040]

[0041] Then, the correlation coefficient matrix is ​​obtained based on the preprocessed data. If the satellite pair has no connected edges in the graph structure, the correlation coefficient of the satellite pair is zero. Otherwise, the correlation coefficient is calculated using Formula 2 based on the satellite traffic data sequence in the preprocessed data:

[0042]

[0043] in, represents the correlation coefficient between satellite 1 and satellite 2, E represents the mathematical expectation, S1 and S2 represent the traffic data sequences of the corresponding satellites, l represents the sequence length, and They respectively represent the average values ​​of the flow data sequence of the corresponding satellite, and S1(i) and S2(i) respectively represent the i-th flow value in the flow data sequence of the corresponding satellite.

[0044] In step S130, the flow feature matrix, the adjacency matrix and the correlation coefficient matrix are input into a pre-established dual-channel space extraction model to obtain the spatial features of the flow; the spatial features are input into a pre-established time extraction model to obtain the time features of the flow, which specifically includes:

[0045] Obtain a pre-established two-channel graph convolutional network as a two-channel spatial extraction model;

[0046] The essence of graph convolution is the convolution of graph signals in the Fourier domain. The traffic feature matrix X and the adjacency matrix A are input into the first channel of the dual-channel graph convolution network, and the traffic feature matrix X and the correlation coefficient matrix P are input into the second channel of the dual-channel graph convolution network. Formula 3 is used to obtain the output of the dual-channel graph convolution network:

[0047]

[0048] Among them, f(X,A) represents the output of the first channel, f(X,P) represents the output of the second channel, represents the preprocessing result of the adjacency matrix A, which is Represents the preprocessing result of the correlation coefficient matrix P, that is, the symmetric normalization of P In the preprocessing process, D is used as a matrix to represent the degree matrix, σ and ReLU represent commonly used activation functions, and W0∈R D×H represents the weight matrix from the input layer to the hidden layer, D is the length of the feature matrix, W1∈R H×T represents the weight matrix from the hidden layer to the output layer, H represents the number of hidden units in the hidden layer, and T represents the prediction time length;

[0049] The outputs of the first channel and the second channel are concatenated to obtain the spatial characteristics of the traffic, as shown in Formula 8:

[0050]

[0051] A gated recurrent unit of a pre-established attention mechanism is obtained as a temporal extraction model. The temporal extraction model is composed of multiple gated recurrent units. The multiple gated recurrent units transmit information over time to capture the temporal dynamics of the satellite system traffic sequence.

[0052] The specific operation process is as follows:

[0053] According to the hidden state h of the gated recurrent unit at the previous moment t-1 and the current input X t ' to obtain two gating states: reset gate and update gate. In the present invention, the current input X t ′ should be the output f(X) after the spatial features have been extracted through the graph convolutional network t ,A,P), for simplicity, the following formula still uses X t ′ replaces f(Xt,A,P).

[0054] The specific principle of the gated recurrent unit is shown in Formula 9, Formula 10, Formula 11 and Formula 4:

[0055] r t =σ(W r [h t-1 ,X t ′]+b r ) Formula 9;

[0056] z t =σ(W z [h t-1 ,X t ′]+b z ) Formula 10;

[0057]

[0058]

[0059] Among them, r t Indicates the reset gate coefficient; W r Represents the calculation weight of the reset gate; b r represents the calculation bias term of the reset gate; z t represents the update gate coefficient; W z represents the calculation weight of the update gate; b z represents the bias term of the update gate. h′ represents the prepared hidden state after reset; W represents the calculation weight of h′; b represents the calculation bias term of h′; tanh and σ both represent activation functions; represents Hadamard multiplication, that is, the corresponding matrix elements are multiplied.

[0060] The spatial features at different time steps are input into each gated recurrent unit, and the output of the gated recurrent unit is obtained using formula 4:

[0061]

[0062] Among them, h trepresents the hidden state at time step t, h′ represents the prepared hidden state after reset, and z t represents the update gate coefficient, represents Hadamard multiplication, h t-1 represents the hidden state at time step t-1;

[0063] The above formula is the update formula of the hidden layer of the gated recurrent unit, which describes the process of retaining old features and learning new temporal features. The temporal features of the traffic sequence are learned by information transmission in the order of time steps through multiple gated recurrent units. Each gated recurrent unit has a hidden state output, and the corresponding hidden state output contains the temporal features of the input traffic matrix extracted by this gated recurrent unit and the previous gated recurrent unit;

[0064] Use the weighted sum of the attention mechanism of formula 5 to obtain the new hidden state, and use the new hidden state as the time feature:

[0065]

[0066] Among them, h represents the new hidden state, n T represents the number of time steps, α t represents the weight of the hidden state of the corresponding gated recurrent unit at time step t. In this embodiment of the present invention, the weight calculation method of the attention mechanism is shown in Formula 12:

[0067]

[0068] Among them, f(h t ) means h t The relationship coefficient with the final output hidden state h;

[0069] The new hidden state obtained by weighted summing up the outputs of each hidden state through the attention mechanism not only contains the time features of all input sequences, but also pays more attention to key time features and pays less attention to low-importance features, such as key features where satellite traffic surges or decreases sharply, and low-importance features such as where traffic changes smoothly.

[0070] In step S140, the time feature is input into the pre-established fully connected layer to obtain the traffic prediction results of each satellite, which specifically include:

[0071] The embodiment of the present invention does not limit the structure of the fully connected layer;

[0072] After obtaining the flow prediction results, the mean square error, mean absolute error and determination coefficient can be used to describe the accuracy of the prediction results.

[0073] To sum up, in response to the existing problems, the low-orbit satellite system traffic prediction method invented this time provides a dynamic prediction reference for low-orbit satellite system traffic prediction. Overall, the prediction method takes into account the spatial correlation and temporal correlation of satellite traffic, and integrates spatiotemporal features to predict satellite traffic; and considers the correlation coefficient between satellites when extracting spatial features to consider that the traffic of a single satellite will be affected by other satellites in the topology; and uses gated recurrent units as time extraction models during time extraction, which can convey the temporal characteristics of the jointly learned traffic sequence; overall, the accuracy of low-orbit satellite system traffic prediction is improved.

[0074] Device Embodiment

[0075] According to an embodiment of the present invention, a low-orbit satellite system flow prediction device is provided. Figure 2 is a schematic diagram of a low-orbit satellite system flow prediction device according to an embodiment of the present invention. Figure 2 As shown, the low-orbit satellite system traffic prediction device according to an embodiment of the present invention specifically includes:

[0076] The preprocessing module 20 is used to obtain the historical flow data of the low-orbit satellite system and preprocess it to obtain preprocessed data, which is specifically used for:

[0077] Using isolation forest to check abnormal and invalid data in the historical traffic data, removing abnormal and invalid data from the historical traffic data to obtain a first data set;

[0078] Fill the missing data in the first data set using the mean difference filling method to obtain a second data set;

[0079] The second data set is normalized to obtain preprocessed data.

[0080] The matrix calculation module 22 is used to construct a traffic feature matrix based on the preprocessed data; construct an adjacency matrix corresponding to the graph structure of the low-orbit satellite system; in the graph structure, the correlation coefficients of the satellite pairs are calculated based on the preprocessed data to obtain a correlation coefficient matrix, which is specifically used for:

[0081] Based on the preprocessed data including N satellites and D time steps, the traffic feature matrix is ​​constructed using Formula 1:

[0082]

[0083] Among them, X represents the traffic feature matrix, represents the flow value of the nth satellite at time d, 1≤n≤N, 1≤d≤D, and D represents the length of the characteristic matrix.

[0084] A single satellite is abstracted as a node in a graph structure, and the links between satellites are abstracted as edges in the graph structure. The graph structure is an undirected graph, and an adjacency matrix corresponding to the graph structure is established.

[0085] If there is no connected edge between the satellite pairs in the graph structure, the correlation coefficient of the satellite pair is zero. Otherwise, the correlation coefficient is calculated using Formula 2 based on the satellite traffic data sequence in the preprocessed data:

[0086]

[0087] in, represents the correlation coefficient between satellite 1 and satellite 2, E represents the mathematical expectation, S1 and S2 represent the traffic data sequences of the corresponding satellites, l represents the sequence length, and They respectively represent the average values ​​of the flow data sequence of the corresponding satellite, and S1(i) and S2(i) respectively represent the i-th flow value in the flow data sequence of the corresponding satellite.

[0088] The spatiotemporal prediction module 24 is used to input the flow feature matrix, the adjacency matrix and the correlation coefficient matrix into a pre-established dual-channel spatial extraction model to obtain the spatial features of the flow; and input the spatial features into a pre-established temporal extraction model to obtain the temporal features of the flow, specifically for:

[0089] Obtain a pre-established two-channel graph convolutional network as a two-channel spatial extraction model;

[0090] The traffic feature matrix X and the adjacency matrix A are input into the first channel of the dual-channel graph convolutional network, and the traffic feature matrix X and the correlation coefficient matrix P are input into the second channel of the dual-channel graph convolutional network. Formula 3 is used to obtain the output of the dual-channel graph convolutional network:

[0091]

[0092] Among them, f(X,A) represents the output of the first channel, f(X,P) represents the output of the second channel, represents the preprocessing result of the adjacency matrix A, represents the preprocessing result of the correlation coefficient matrix P, σ and ReLU represent activation functions, W0∈R D×H Represents the weight matrix from the input layer to the hidden layer, W1∈R H×T represents the weight matrix from the hidden layer to the output layer, H represents the number of hidden units in the hidden layer, and T represents the prediction time length;

[0093] The outputs of the first channel and the second channel are spliced ​​to obtain the spatial characteristics of the flow.

[0094] Get the pre-built attention mechanism of gated recurrent unit as the temporal abstraction model;

[0095] The spatial features at different time steps are input into each gated recurrent unit, and the output of the gated recurrent unit is obtained using formula 4:

[0096]

[0097] Among them, h t represents the hidden state at time step t, h′ represents the prepared hidden state after reset, and z t express, represents Hadamard multiplication, h t-1 represents the hidden state at time step t-1;

[0098] Use the weighted sum of the attention mechanism of formula 5 to obtain the new hidden state, and use the new hidden state as the time feature:

[0099]

[0100] Among them, h represents the new hidden state, n T represents the number of time steps, α t represents the weight of the hidden state of the corresponding gated recurrent unit at time step t.

[0101] The result acquisition module 26 is used to input the time feature into the pre-established fully connected layer to obtain the traffic prediction result of each satellite.

[0102] To sum up, in response to the existing problems, the low-orbit satellite system traffic prediction device invented this time provides a dynamic prediction reference for low-orbit satellite system traffic prediction. On the whole, the prediction method takes into account the spatial correlation and temporal correlation of satellite traffic at the same time, and integrates spatiotemporal features to predict satellite traffic; and considers the correlation coefficient between satellites when extracting spatial features to consider that the traffic of a single satellite will be affected by other satellites in the topology; and uses a gated recurrent unit as a time extraction model during time extraction, which can convey the temporal characteristics of the jointly learned traffic sequence; and overall improves the accuracy of low-orbit satellite system traffic prediction.

[0103] Electronic device embodiment

[0104] Figure 3 3 is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device 300 may include at least one processor 310 and a memory 320. The processor 310 may execute instructions stored in the memory 320. The processor 310 is connected to the memory 320 through a data bus. In addition to the memory 320, the processor 310 may also be connected to an input device 330, an output device 340, and a communication device 350 through a data bus.

[0105] The processor 310 may be any conventional processor, such as a commercially available CPU. The processor may also include a graphics processor (Graphic Process Unit, GPU), a field programmable gate array (Field Programmable Gate Array, FPGA), a system on chip (System on Chip, SOC), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC) or a combination thereof.

[0106] The memory 320 may be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0107] In the embodiment of the present disclosure, executable instructions are stored in the memory 320, and the processor 310 can read the executable instructions from the memory 320 and execute the instructions to implement all or part of the steps of the low-orbit satellite system traffic prediction method of any of the above exemplary embodiments.

[0108] Computer Readable Storage Medium Embodiments

[0109] In addition to the above-mentioned methods and devices, an exemplary embodiment of the present disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product. The computer product includes computer program instructions, and the computer program instructions can be executed by a processor to implement all or part of the steps described in any of the low-orbit satellite system traffic prediction methods in the above-mentioned exemplary embodiments.

[0110] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, etc., and also conventional procedural programming languages ​​such as "C" language or similar programming languages ​​and scripting languages ​​(e.g., Python). The program code may be executed entirely on the user computing device, partially on the user computing device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0111] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of readable storage media include: a static random access memory (SRAM) with one or more wires electrically connected, an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk, or any suitable combination of the above.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A low-orbit satellite system traffic prediction method, characterized in that: include: Acquire historical traffic data of the low-orbit satellite system and pre-process it to obtain pre-processed data; Constructing a flow characteristic matrix based on the preprocessed data; Constructing an adjacency matrix corresponding to the graph structure of the low-orbit satellite system; In the graph structure, correlation coefficients of satellite pairs are calculated based on the preprocessed data to obtain a correlation coefficient matrix; Inputting the flow characteristic matrix, the adjacency matrix and the correlation coefficient matrix into a pre-established dual-channel space extraction model to obtain the spatial characteristics of the flow; Inputting the spatial characteristics into a pre-established time extraction model to obtain the time characteristics of the flow; The time feature is input into a pre-established fully connected layer to obtain the traffic prediction result of each satellite.

2. The method according to claim 1, characterized in that: The acquisition of historical traffic data of the low-orbit satellite system and preprocessing to obtain preprocessed data specifically includes: Using an isolation forest to check abnormal and invalid data in the historical traffic data, removing the abnormal and invalid data from the historical traffic data, and obtaining a first data set; Fill the missing data in the first data set using the mean difference filling method to obtain a second data set; The second data set is normalized to obtain preprocessed data.

3. The method according to claim 1, characterized in that The constructing of the traffic feature matrix based on the preprocessed data specifically includes: Based on the preprocessed data including N satellites and D time steps, the traffic feature matrix is ​​constructed using Formula 1: Among them, X represents the traffic feature matrix, represents the flow value of the nth satellite at time d, 1≤n≤N, 1≤d≤D, and D represents the length of the characteristic matrix.

4. The method according to claim 1, characterized in that: The step of constructing the adjacency matrix of the graph structure corresponding to the low-orbit satellite system specifically includes: A single satellite is abstracted as a node in a graph structure, and links between satellites are abstracted as edges in the graph structure, wherein the graph structure is an undirected graph, and an adjacency matrix corresponding to the graph structure is established.

5. The method according to claim 1, characterized in that In the graph structure, calculating the correlation coefficients of satellite pairs based on the preprocessed data to obtain the correlation coefficient matrix specifically includes: If there is no connected edge between the satellite pair in the graph structure, the correlation coefficient of the satellite pair is zero. Otherwise, the correlation coefficient is calculated using Formula 2 based on the satellite traffic data sequence in the preprocessed data: in, represents the correlation coefficient between satellite 1 and satellite 2, E represents the mathematical expectation, S1 and S2 represent the traffic data sequences of the corresponding satellites, l represents the sequence length, and They respectively represent the average values ​​of the flow data sequence of the corresponding satellite, and S1(i) and S2(i) respectively represent the i-th flow value in the flow data sequence of the corresponding satellite.

6. The method according to claim 3, characterized in that The step of inputting the flow feature matrix, the adjacency matrix and the correlation coefficient matrix into a pre-established dual-channel space extraction model to obtain the spatial features of the flow specifically includes: Obtain a pre-established two-channel graph convolutional network as a two-channel spatial extraction model; The traffic feature matrix X and the adjacency matrix A are input into the first channel of the dual-channel graph convolutional network, and the traffic feature matrix X and the correlation coefficient matrix P are input into the second channel of the dual-channel graph convolutional network. The output of the dual-channel graph convolutional network is obtained using Formula 3: Among them, f(X,A) represents the output of the first channel, f(X,P) represents the output of the second channel, represents the preprocessing result of the adjacency matrix A, represents the preprocessing result of the correlation coefficient matrix P, σ and ReLU represent the activation function, W0∈R D×H Represents the weight matrix from the input layer to the hidden layer, W1∈R H×T represents the weight matrix from the hidden layer to the output layer, H represents the number of hidden units in the hidden layer, and T represents the prediction time length; The outputs of the first channel and the second channel are spliced ​​to obtain the spatial characteristics of the flow.

7. The method according to claim 1, characterized in that The step of inputting the spatial features into a pre-established time extraction model to obtain the time features of the traffic specifically includes: Get the pre-built attention mechanism of gated recurrent unit as the temporal abstraction model; The spatial features at different time steps are input into each gated recurrent unit, and the output of the gated recurrent unit is obtained using formula 4: Among them, h t represents the hidden state at time step t, h′ represents the prepared hidden state after reset, and z t represents the update gate coefficient, represents Hadamard multiplication, h t-1 represents the hidden state at time step t-1; The new hidden state is obtained by weighted summation using the attention mechanism of Formula 5, and the new hidden state is used as the time feature: Among them, h represents the new hidden state, n T represents the number of time steps, α t represents the weight of the hidden state of the corresponding gated recurrent unit at time step t.

8. A low-orbit satellite system traffic prediction device, characterized in that: include: A preprocessing module is used to obtain historical flow data of the low-orbit satellite system and perform preprocessing to obtain preprocessed data; A matrix calculation module, used for constructing a flow characteristic matrix based on the preprocessed data; Constructing an adjacency matrix corresponding to the graph structure of the low-orbit satellite system; In the graph structure, correlation coefficients of satellite pairs are calculated based on the preprocessed data to obtain a correlation coefficient matrix; A spatiotemporal prediction module, used for inputting the flow characteristic matrix, the adjacency matrix and the correlation coefficient matrix into a pre-established dual-channel space extraction model to obtain the spatial characteristics of the flow; Inputting the spatial characteristics into a pre-established time extraction model to obtain the time characteristics of the flow; The result acquisition module is used to input the time feature into a pre-established fully connected layer to obtain the traffic prediction results of each satellite.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the low-orbit satellite system traffic prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by a processor, the steps of the low-orbit satellite system traffic prediction method according to any one of claims 1 to 7 are implemented.

Images