Satellite traffic volume prediction method and apparatus

Abstract

The present specification relates to the technical field of satellite, and particularly relates to a satellite traffic flow prediction method and device. The method comprises: acquiring geographical feature data of a wave position; acquiring coverage time data of the satellite for the wave position in a historical period; constructing first feature data of the satellite according to the geographical feature data and the coverage time data; the first feature data is used for representing geographical features cumulatively covered by the satellite in the historical period; acquiring second feature data, the second feature data is used for representing traffic flow of the satellite in the historical period; and predicting future traffic flow of the satellite according to the first feature data and the second feature data. The embodiment of the present specification can improve prediction accuracy.

Description

Technical Field

[0001] This specification relates to the field of satellite technology, and in particular to a method and apparatus for predicting satellite traffic. Background Technology

[0002] With the rapid advancement of aerospace and communication technologies, satellite internet has become an important component of the integrated space-air-ground network. Satellite internet is widely used in military, aerospace, communications, and meteorological fields, providing crucial support for remote sensing data transmission, emergency communications, and global coverage. Predicting satellite internet traffic can help optimize satellite network performance, resource allocation, and security management. Therefore, predicting satellite traffic is essential.

[0003] In related technologies, the approach to predicting terrestrial network traffic is typically adopted to predict satellite service traffic. For example, machine learning algorithms are used to predict satellite service traffic directly based on historical traffic data.

[0004] However, due to the highly dynamic nature of satellite motion, satellite internet differs significantly from terrestrial networks in terms of service characteristics and traffic patterns. The spatiotemporal characteristics and fluctuations of satellite internet traffic are more complex than those of terrestrial networks. The aforementioned technologies cannot accurately predict satellite service traffic. Summary of the Invention

[0005] This specification provides a satellite traffic prediction method and apparatus to improve prediction accuracy.

[0006] According to a first aspect of the embodiments of this specification, a satellite service traffic prediction method is provided, comprising:

[0007] Obtain the geographical features of the wave position;

[0008] Obtain the coverage time data of the satellite for the said wavelength within a historical period;

[0009] Based on the geographic feature data and the coverage time data, the first feature data of the satellite is constructed; the first feature data is used to represent the geographic features cumulatively covered by the satellite during the historical period.

[0010] Acquire second feature data, which represents the satellite's service traffic during the historical period;

[0011] Based on the first feature data and the second feature data, the future service traffic of the satellite is predicted.

[0012] According to a second aspect of the embodiments of this specification, a satellite traffic prediction apparatus is provided, comprising:

[0013] The first acquisition unit is used to acquire the geographical feature data of the wave position;

[0014] The second acquisition unit is used to acquire the coverage time data of the satellite for the wave position within a historical time period;

[0015] A construction unit is configured to construct first feature data of the satellite based on the geographic feature data and the coverage time data; the first feature data is used to represent the geographic features cumulatively covered by the satellite during the historical period.

[0016] The third acquisition unit is used to acquire second feature data, which represents the satellite's service traffic during the historical period.

[0017] The prediction unit is used to predict the future service traffic of the satellite based on the first feature data and the second feature data.

[0018] The technical solution of this specification, addressing the highly dynamic nature of satellite motion, constructs first characteristic data of the satellite based on the geographical features of the wave positions and the coverage time data of the wave positions during satellite motion. This first characteristic data represents the cumulative geographical features covered by the satellite over a historical period. Additionally, second characteristic data of the satellite can be obtained. This second characteristic data represents the service traffic of the satellite over a historical period. Based on the first and second characteristic data, future service traffic of the satellite can be predicted. Therefore, addressing the highly dynamic nature of satellite motion, this specification's embodiments predict future service traffic of the satellite by comprehensively considering the geographical features covered during satellite motion and the characteristics of the service traffic it carries. This significantly improves the accuracy and effectiveness of satellite internet traffic prediction. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the satellite traffic prediction method in the embodiments of this specification;

[0021] Figure 2 This is a schematic diagram of the wave positions covered by the satellite during a historical period in the embodiments of this specification;

[0022] Figure 3This is a schematic diagram illustrating the construction process of the time series matrix in the embodiments of this specification;

[0023] Figure 4 This is a schematic diagram of the time series matrix structure in the embodiments of this specification;

[0024] Figure 5 This is a functional structure diagram of the satellite traffic prediction device in the embodiments of this specification. Detailed Implementation

[0025] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. The specific embodiments described herein are only used to explain this disclosure, and not to limit this disclosure. All other embodiments obtained by those skilled in the art based on the described embodiments of this disclosure are within the scope of protection of this disclosure. In addition, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0026] This specification provides an improved method for predicting satellite service traffic. The method can be applied to computer devices such as portable computers, desktop computers, and servers. Please refer to... Figure 1 The method may include the following steps.

[0027] Step 11: Obtain the geographical feature data of the wave position.

[0028] Step 12: Obtain satellite coverage time data for wave positions within historical time periods.

[0029] Step 13: Construct the first feature data of the satellite based on the geographic feature data and coverage time data; the first feature data is used to represent the geographic features cumulatively covered by the satellite during historical periods.

[0030] Step 14: Obtain the second feature data, which is used to represent the satellite's service traffic during historical periods.

[0031] Step 15: Based on the first feature data and the second feature data, predict the future service traffic of the satellite.

[0032] In some embodiments, the number of satellites is one or more. Each satellite has a corresponding satellite identifier. The satellite identifier is used to identify the satellite. The one or more satellites may include Low Earth Orbit (LEO) satellites, Medium Earth Orbit (MEO) satellites, and any combination thereof. LEO and MEO satellites move relative to the Earth while in orbit. Due to the high dynamism of satellite motion, the spatiotemporal characteristics and fluctuations of satellite internet traffic are more complex. Therefore, the accuracy of predicting satellite service traffic using the same approach as terrestrial network traffic prediction is relatively low. This embodiment of the specification can construct first feature data of the satellite based on the geographical feature data of the wave position and the coverage time data of the wave position during satellite motion. The first feature data is used to represent the geographical features cumulatively covered by the satellite within a historical period. Additionally, second feature data of the satellite can also be obtained. The second feature data is used to represent the service traffic of the satellite within a historical period. Future service traffic of the satellite can be predicted based on the first feature data and the second feature data. Therefore, considering the high dynamism of satellite motion, this embodiment of the specification predicts future service traffic of the satellite by comprehensively considering the geographical features covered during satellite motion and the characteristics of the service traffic it carries. This greatly improves the accuracy and effectiveness of satellite internet traffic prediction.

[0033] In some embodiments, please refer to Figure 2 A beam position, also known as a ground beam position, refers to a geographical unit with specific geographical scope and communication characteristics, defined based on factors such as service requirements and beam characteristics. A beam position is the spatial unit for satellite resource allocation and service scheduling. One or more beam positions can be obtained by dividing a preset geographical area. Each beam position has a corresponding beam position identifier. The beam position identifier is used to identify the beam position. Satellites form signal coverage areas on the ground through beams. The beam coverage area includes one or more beam positions. The periodic movement of satellites in their orbits causes changes in the beam positions covered by the satellite.

[0034] In some embodiments, satellite services are strongly correlated with geographic location. For this purpose, position planning data can be acquired. The position planning data includes geographic feature data for one or more positions. Geographic feature data is used to represent the geographic characteristics of the position. The geographic feature data for a position can include indicator data under one or more geographic features. Geographic features can include landform type, specific services, etc. Landform type can include one or more of marine landforms, desert landforms, urban landforms, forest landforms, etc. Indicator data for landform type can include landform area. For example, the surface area corresponding to the landform type can be calculated using methods such as NDWI (Normalized Difference Water Index) and NDVI (Normalized Difference Vegetation Index) by referring to telemetry data. Specific services can include key services. Key services refer to core services that require priority in ensuring communication continuity and stability during satellite operation, such as forest fire prevention communication, maritime search and rescue communication, and earthquake emergency command communication. Indicator data for specific services can include the number of specific services, such as the number of key services.

[0035] For example, a preset geographical area can be divided into W1, W2, ..., W i ... W w There are w wave positions. w is an integer greater than or equal to 0. Each wave position (e.g., wave position W) i The geographic feature data can include geographic feature data A. Geographic feature data A can include five data elements: a1, a2, a3, a4, and a5. a1 represents the area of ​​marine landforms, a2 represents the area of ​​desert / Gobi landforms, a3 represents the area of ​​urban landforms, a4 represents the area of ​​forest landforms, and a5 represents the number of key protected businesses.

[0036] In some embodiments, the duration of a historical period can be, for example, 5 minutes, 6 minutes, 30 minutes, etc. The number of historical periods can be one or more. The durations of the one or more historical periods can be the same or different. Each historical period can be understood as a statistical cycle used to collect characteristic data of the satellite within that historical period, such as collecting first characteristic data, second characteristic data, etc. The multiple historical periods can be consecutive. For example, the historical time interval before a certain moment can be divided into multiple consecutive historical periods. Based on the characteristic data of the satellite in the multiple historical periods, future service traffic of the satellite after that moment can be predicted. For example, the historical time interval is 15 days. The duration of a historical period can be 5 minutes. The historical time interval can be divided into 4320 consecutive historical periods. Based on the characteristic data of the 4320 historical periods, the service traffic of the satellite for 864 periods within the next 3 days can be predicted.

[0037] In some embodiments, coverage time data for each satellite within each historical time period can be acquired. A satellite can cover one or more spectral positions within a historical time period. Coverage time data includes the percentage of time a satellite covers the spectral positions. The percentage of coverage time includes the ratio of the satellite's coverage time for the spectral position within a historical time period to the duration of the historical time period. Coverage duration includes the duration the satellite is present at the spectral position. The percentage of coverage time characterizes the intensity of satellite resource allocation for spectral positions.

[0038] For each satellite, the coverage duration for one or more spectral positions within each historical time period can be obtained; the coverage duration for each spectral position can be divided by the duration of the historical time period to obtain the coverage duration percentage for that spectral position. The coverage time data for that satellite within the historical time period may include the coverage duration percentage for the one or more spectral positions.

[0039] Each satellite can cover one or more of the planned wave positions within a historical time period, and the duration of the satellite's coverage on those wave positions can be the same or different. Therefore, as an example, for each satellite, the one or more wave positions covered by the satellite in each historical time period can be determined; the coverage duration of each wave position can be divided by the duration of the historical time period to obtain the coverage duration percentage of that wave position. The coverage time data of the satellite in that historical time period includes the coverage duration percentage of the one or more wave positions. As another example, for each satellite, the coverage duration of the satellite for all planned wave positions in each historical time period can be obtained; the coverage duration of each wave position can be divided by the duration of the historical time period to obtain the coverage duration percentage of that wave position. The coverage time data of the satellite in that historical time period can include the coverage duration percentage of all planned wave positions. Among all planned wave positions, for wave positions covered by the satellite, the coverage duration of the satellite for that wave position can be obtained; for wave positions not covered by the satellite, the coverage duration of that wave position is 0.

[0040] In some scenario examples, the satellites to be predicted include S1, S2, ..., S... i S s There are s satellites. The total planned wavelengths include W1, W2, ..., W... i ... W w There are w wave positions. Historical time periods include T1, T2, ..., T... i ... T t Wait for t historical time periods. Satellite S can be obtained. i In historical period T i The coverage duration for each of the w wavelengths is calculated. For each of the w wavelengths, the coverage duration can be compared with the historical time period T. i Divide the duration by the coverage duration to obtain the proportion of coverage time for that wavelength. Satellite S i In historical period T i The coverage time data includes the percentage of coverage time for the w wavelengths.

[0041] Optionally, a matrix SWS can also be constructed. The matrix SWS represents the coverage time data of s satellites over t historical time periods. The dimensions of the matrix SWS can be represented as (s, t, w), where w represents the number of wavelengths. The data elements in the matrix SWS are SWS. ijk For satellite S i In historical period T j Internal relative to wave position W k The percentage of coverage time. 1≤i≤s, 1≤j≤t, 1≤k≤w.

[0042] In some embodiments, first feature data for each satellite in each historical time period can be acquired. A satellite can cover one or more wave positions within a historical time period. The one or more wave positions can correspond to geographic feature data. The first feature data can include the geographic features cumulatively covered by the satellite within the historical time period. The geographic feature data of the wave position can include indicator data under one or more geographic features. Thus, the first feature data can include the cumulative indicator data of the satellite under one or more geographic features. The cumulative indicator data can characterize the actual cumulative indicator data for geographic features during the statistical period as the satellite moves. The coverage duration ratio can characterize the satellite's resource allocation intensity for wave positions. Therefore, the resource allocation intensity can be used as a weight to statistically analyze the actual cumulative indicator data for various geographic features during the satellite's movement, obtaining the cumulative indicator data for various geographic features. In addition, the first feature data of the satellite in multiple historical time periods can also characterize the dynamic changes of the geographic features of the wave positions covered by the satellite during its movement. For example, it can characterize the dynamic changes of one or more geographic features of the wave positions covered by the satellite.

[0043] For each satellite, coverage time data for each historical period can be obtained. This coverage time data includes the coverage duration percentage of one or more wavebands. The index data for each waveband under each geographic feature is multiplied by the coverage duration percentage of that waveband to obtain the product corresponding to that geographic feature. The products of the one or more wavebands corresponding to the same geographic feature are summed to obtain the cumulative index data for that geographic feature. The first feature data for the satellite within this historical period may include the cumulative index data for one or more geographic features.

[0044] In some scenario examples, satellite S i In historical period T i The coverage time data includes the coverage duration percentage for each of the w wavelengths. Each wavelength (e.g., wavelength W) iThe geographic feature data can include a1, a2, a3, a4, a5, etc. a1 represents the area of ​​marine landforms, a2 represents the area of ​​desert / Gobi landforms, a3 represents the area of ​​urban landforms, a4 represents the area of ​​forest landforms, and a5 represents the number of key service areas. For each of the w wave positions, a1 of that wave position can be multiplied by its coverage time percentage to obtain the product corresponding to a1; a2 of that wave position can be multiplied by its coverage time percentage to obtain the product corresponding to a2; a3 of that wave position can be multiplied by its coverage time percentage to obtain the product corresponding to a3; a4 of that wave position can be multiplied by its coverage time percentage to obtain the product corresponding to a4; and a5 of that wave position can be multiplied by its coverage time percentage to obtain the product corresponding to a5. The products of w wave positions corresponding to a1 can be added together to obtain the cumulative indicator data corresponding to a1 (cumulative marine landform area); the products of w wave positions corresponding to a2 can be added together to obtain the cumulative indicator data corresponding to a2 (cumulative desert / Gobi landform area); the products of w wave positions corresponding to a3 can be added together to obtain the cumulative indicator data corresponding to a3 (cumulative urban landform area); the products of w wave positions corresponding to a4 can be added together to obtain the cumulative indicator data corresponding to a4 (cumulative forest landform area); and the products of w wave positions corresponding to a5 can be added together to obtain the cumulative indicator data corresponding to a5 (cumulative number of key service items). Satellite S i In historical period T i The first characteristic data may include the cumulative marine landform area, cumulative desert / Gobi landform area, cumulative urban landform area, cumulative forest landform area, and cumulative number of key protected services. Among these, the cumulative marine landform area can represent the area of ​​key protected services in historical time period T. i Inside, with satellite S i The movement of [the landform] actually cumulatively covers the area of ​​marine landforms. The cumulative area of ​​desert and Gobi landforms can characterize the area covered during the historical period T. i Inside, with satellite S i The movement of [the city / region] actually cumulatively covers the area of ​​desert and Gobi landforms. The cumulative urban landform area can represent the area covered during the historical period T. i Inside, with satellite S i The movement of [the forest] actually cumulatively covers the urban landform area. The cumulative forest landform area can characterize the area covered during the historical period T. i Inside, with satellite S i The movement, and the actual cumulative forest landform area covered. The cumulative number of key protection services can represent the historical period T. i Inside, with satellite S i The movement of [the satellite], and the actual cumulative number of key services guaranteed. Additionally, satellite S... iThe first feature data over t historical periods can also characterize the dynamic changes of various geographical features covered by the satellite during its movement.

[0045] Optionally, a matrix SW can also be constructed. Matrix SW represents geographic feature data for w wave positions. The dimension of matrix SW can be represented as (w, a), where a represents the number of geographic feature types, for example, 5. The data elements SW in matrix SW are... km Wave position W k Indicator data under geographic feature m. 1≤k≤w, 1≤m≤a.

[0046] Matrix SW can be multiplied by matrix SWS to obtain matrix GFT. Matrix GFT represents the first feature data of s satellites over t historical time periods. The dimension of matrix GFT can be represented as (s, t, a). The data elements in matrix GFT are GFT... ijm For satellite S i In historical period T j Cumulative indicator data under the intrinsic geographical feature m. .

[0047] In some embodiments, second feature data for each satellite can be acquired for each historical time period. The second feature data represents the service traffic of that satellite during that historical time period. Second feature data for a satellite across multiple historical time periods can also characterize the dynamic changes in service traffic of one or more service types during the satellite's movement.

[0048] Based on the characteristic analysis of traffic service applications, the internet services to which the traffic belongs can be categorized to obtain one or more service types for satellite internet services. Service types can include one or more of the following: broadband access, mobile communication, emergency communication, professional applications, and narrowband access. Broadband access can include enterprise private networks and mobile broadband. Mobile communication can include voice services and SMS services. Emergency communication can include emergency rescue. Professional applications can include remote sensing mapping, environmental monitoring, and meteorological monitoring. Narrowband access can include IoT services. Therefore, the second characteristic data can include the first traffic data of one or more service types.

[0049] Optionally, a matrix FL can be constructed. Matrix FL represents the second feature data of s satellites over t historical time periods. The dimensions of matrix FL can be represented as (s, t, b), where b represents the number of service types. The data elements FL in matrix FL are... ijn For satellite S i In historical period T j The first traffic data under the intrinsic business type n. 1≤i≤s, 1≤j≤t, 1≤n≤b.

[0050] In some embodiments, future satellite traffic can be predicted based on first feature data and second feature data. The first feature data represents the cumulative geographical features covered by the satellite over a historical period. The second feature data represents the satellite's traffic over that historical period. Therefore, by combining geographical features and traffic characteristics, future satellite traffic can be predicted, significantly improving the accuracy and effectiveness of satellite internet traffic prediction.

[0051] In some embodiments, machine learning models can be used to predict future satellite traffic. The machine learning model can be trained using sample data to obtain a trained model. This trained model can then be used to predict future satellite traffic. Machine learning models can include time series models. A time series model is a machine learning model used to analyze and predict time series data. It can uncover patterns in data over time and infer future data based on these patterns. Time series models can include ARIMA models, LSTM models, Informer models, etc. Among these, the Informer model is a machine learning model for efficient time series forecasting. By improving the self-attention mechanism, the Informer model can effectively handle long-term series data, overcoming the high computational complexity and poor performance of traditional methods when dealing with long series.

[0052] The first and second feature data can be input into a machine learning model to obtain its output. The output of the machine learning model can include the satellite's service traffic over one or more future time periods. The duration of each future time period can be the same or different. The service traffic for each future time period can include multiple first traffic data points from a single satellite. Each first traffic data point corresponds to a service type. Alternatively, the service traffic for each future time period can also include second traffic data from a single satellite. The second traffic data can include the sum of first traffic data points across multiple service types.

[0053] As an example, first and second feature data of a satellite over a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model can include the satellite's traffic over one or more future periods. The first and second feature data can be input into the machine learning model separately. Alternatively, the first and second feature data can be fused; the fused feature data can then be input into the machine learning model. Fusion methods can include concatenating the first and second feature data, etc.

[0054] As another example, first and second feature data of a satellite across multiple historical time periods can be input into a machine learning model to obtain its output. The output of the machine learning model can include the satellite's service traffic over one or more future time periods. Specifically, time-series data of the satellite can be constructed based on the first and second feature data from multiple historical time periods; this time-series data can then be input into the machine learning model. For example, the first and second feature data of each historical time period can be fused to obtain fused feature data for that historical time period; or the fused feature data from multiple historical time periods can be fused to obtain the satellite's time-series data. Fusion methods can include concatenation. For example, the first and second feature data can be concatenated; or the fused feature data from multiple historical time periods can be concatenated. Alternatively, the first and second feature data of multiple historical time periods can be input into the machine learning model separately. Or, the first and second feature data of each historical time period can be fused; or the fused feature data from multiple historical time periods can be input into the machine learning model separately.

[0055] As another example, first and second feature data from multiple satellites over a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model can include traffic data for multiple satellites over one or more future periods. For each satellite, one or more future periods' traffic data can be output.

[0056] As another example, first and second feature data from multiple satellites over multiple historical time periods can be input into a machine learning model to obtain its output. The output of the machine learning model includes the traffic flow of multiple satellites over one or more future time periods. Each of the multiple satellites has first and second feature data from multiple historical time periods. The machine learning model can output the traffic flow of each satellite over one or more future time periods.

[0057] Time-series data for satellites can be constructed based on the first and second feature data of multiple satellites across multiple historical time periods. This time-series data can then be input into a machine learning model. For example, the first and second feature data of each satellite in each historical time period can be fused to obtain sub-fused feature data for that satellite in that historical time period; sub-fused feature data of the satellite across multiple historical time periods can be fused to obtain fused feature data for that satellite; or fused feature data from multiple satellites can be fused to obtain time-series data. Alternatively, the first and second feature data of multiple satellites across multiple historical time periods can be input into the machine learning model separately.

[0058] Optionally, the GFT and FL matrices can be concatenated to obtain a concatenated matrix. This concatenated matrix can be understood as time-series data. The concatenated matrix can then be input into a machine learning model to obtain its output.

[0059] In some embodiments, third characteristic data of the satellite may also be acquired. This third characteristic data represents second traffic data of the satellite over a historical period. The second traffic data includes the sum of first traffic data for each service type. The second characteristic data of the satellite over multiple historical periods can characterize the dynamic changes in overall service traffic during the satellite's movement.

[0060] It is possible to obtain the third characteristic data for each satellite in each historical time period. For example, it is possible to obtain the second characteristic data for each satellite in each historical time period. The second characteristic data may include first traffic data for one or more service types. The individual first traffic data within the second characteristic data can be added together to obtain the second traffic data for that satellite in that historical time period.

[0061] Optionally, a matrix Ff can be constructed. Matrix Ff represents the third feature data of s satellites over t historical time periods. The dimension of matrix Ff can be (s, t). The data elements Ff in matrix Ff... ij For satellite S i In historical period T j The second flow data within.

[0062] Based on the first, second, and third feature data, future satellite traffic can be predicted. Therefore, by integrating geographical features, traffic characteristics under various service types, and overall traffic characteristics, future satellite traffic can be predicted, thereby further improving the accuracy and effectiveness of satellite internet traffic prediction.

[0063] Machine learning models can be used to predict future satellite traffic. First, second, and third feature data can be input into the machine learning model to obtain its output. The output can include the satellite's traffic over one or more future time periods. The durations of these future time periods can be the same or different. Each future time period's traffic can include multiple first traffic data points for a single satellite. Each first traffic data point corresponds to a specific service type. Alternatively, each future time period's traffic can also include second traffic data for a single satellite. The second traffic data can be the sum of first traffic data points across multiple service types.

[0064] As an example, the first, second, and third feature data of a satellite over a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model can include the satellite's traffic over one or more future periods. Alternatively, the first, second, and third feature data can be input into the machine learning model separately. Or, the first, second, and third feature data can be fused; the fused feature data can then be input into the machine learning model.

[0065] As another example, the first, second, and third feature data of a satellite across multiple historical time periods can be input into a machine learning model to obtain its output. The output of the machine learning model can include the satellite's service traffic over one or more future time periods. Specifically, time-series data of the satellite can be constructed based on the first, second, and third feature data from multiple historical time periods; this time-series data can then be input into the machine learning model. For example, the first, second, and third feature data from each historical time period can be fused to obtain fused feature data for that historical period; or the fused feature data from multiple historical time periods can be fused to obtain the satellite's time-series data. Fusion methods can include methods such as stitching.

[0066] As another example, the first, second, and third feature data of multiple satellites over a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model can include the traffic data of multiple satellites over one or more future periods. For each satellite, the traffic data for one or more future periods can be output.

[0067] As another example, first feature data, second feature data, and third feature data from multiple satellites over multiple historical time periods can be input into a machine learning model to obtain the model's output. The output of the machine learning model includes the service traffic of the multiple satellites over one or more future time periods. Each of the multiple satellites has first feature data, second feature data, and third feature data from multiple historical time periods. The machine learning model can output the service traffic of each satellite over one or more future time periods.

[0068] Time-series data for satellites can be constructed based on the first, second, and third feature data from multiple satellites across multiple historical time periods. This time-series data can then be input into a machine learning model. For example, the first, second, and third feature data for each satellite in each historical time period can be fused to obtain sub-fused feature data for that satellite in that historical time period; sub-fused feature data for that satellite across multiple historical time periods can be fused to obtain fused feature data for that satellite; or fused feature data from multiple satellites can be fused to obtain time-series data. Fusion methods can include splicing. Alternatively, the first, second, and third feature data from multiple satellites across multiple historical time periods can be input into the machine learning model separately.

[0069] Optionally, the GFT, FL, and Ff matrices can be concatenated to obtain a concatenated matrix. This concatenated matrix can be understood as time-series data. The concatenated matrix can then be input into a machine learning model.

[0070] In some embodiments, fourth feature data of the satellite can also be acquired. The fourth feature data represents the number of wavelengths covered by the satellite within a historical time period. The fourth feature data of the satellite over multiple historical time periods can characterize the dynamic changes in the number of wavelengths covered by the satellite during its motion. The fourth feature data of each satellite for each historical time period can be acquired. For example, for each satellite, one or more wavelengths covered by the satellite in each historical time period can be determined; the number of wavelengths covered by the satellite in that historical time period can be counted. Optionally, a matrix Fwn can be constructed. Matrix Fwn represents the fourth feature data of s satellites over t historical time periods. The dimension of matrix Fwn can be represented as (s, t). The data element Fwn in matrix Fwn... ij For satellite S i In historical period T j Number of wave positions covered internally.

[0071] Based on the first, second, and fourth feature data, future satellite traffic can be predicted. Therefore, by integrating geographical features, traffic characteristics under various service types, and overall wavenumber characteristics, future satellite traffic prediction can be made, thereby further improving the accuracy and effectiveness of satellite internet traffic prediction.

[0072] Machine learning models can be used to predict future satellite traffic. First, second, and fourth feature data can be input into the machine learning model to obtain its output. The output can include the satellite's traffic over one or more future time periods. The durations of these future time periods can be the same or different. Each future time period's traffic can include multiple first traffic data points for a single satellite. Each first traffic data point corresponds to a specific service type. Alternatively, each future time period's traffic can also include second traffic data points for a single satellite. The second traffic data can be the sum of first traffic data points across multiple service types.

[0073] The process of making a prediction based on the first, second, and fourth feature data is similar to the process of making a prediction based on the first, second, and third feature data. They can be used as a reference and will not be repeated here.

[0074] In some embodiments, future satellite traffic can also be predicted based on first feature data, second feature data, third feature data, and fourth feature data. This allows for the comprehensive prediction of future satellite traffic by integrating geographical features, traffic characteristics under various service types, overall traffic characteristics, and overall wavenumber characteristics. This further improves the accuracy and effectiveness of satellite internet traffic prediction.

[0075] Machine learning models can be used to predict future satellite traffic. First, second, third, and fourth feature data can be input into the machine learning model to obtain its output. The output can include the satellite's traffic over one or more future time periods. The durations of these future time periods can be the same or different. Each future time period's traffic can include multiple first traffic data points for a single satellite. Each first traffic data point corresponds to a specific service type. Alternatively, each future time period's traffic can also include second traffic data for a single satellite. The second traffic data can include the sum of first traffic data points across multiple service types.

[0076] As an example, the first, second, third, and fourth feature data of a satellite over a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model includes the satellite's traffic over one or more future periods. The first, second, third, and fourth feature data can be input into the machine learning model separately. Alternatively, the first, second, third, and fourth feature data can be fused; the fused feature data can then be input into the machine learning model.

[0077] As another example, the first, second, third, and fourth feature data of a satellite across multiple historical time periods can be input into a machine learning model to obtain its output. The output of the machine learning model includes the satellite's service traffic over one or more future time periods. Specifically, time-series data of the satellite can be constructed based on the first, second, third, and fourth feature data from multiple historical time periods; this time-series data can then be input into the machine learning model. For example, the first, second, third, and fourth feature data for each historical time period can be fused to obtain fused feature data for that historical period; or the fused feature data from multiple historical time periods can be fused to obtain the satellite's time-series data.

[0078] As another example, the first, second, third, and fourth feature data of multiple satellites during a historical period can be input into a machine learning model to obtain its output. The output of the machine learning model can include the traffic data of multiple satellites for one or more future periods. For each satellite, the traffic data for one or more future periods can be output.

[0079] As another example, first feature data, second feature data, third feature data, and fourth feature data from multiple satellites over multiple historical time periods can be input into a machine learning model to obtain the model's output. The output of the machine learning model can include the service traffic of the multiple satellites over one or more future time periods. Each of the multiple satellites has first feature data, second feature data, third feature data, and fourth feature data from multiple historical time periods. The machine learning model can output the service traffic of each satellite over one or more future time periods.

[0080] Time-series satellite data can be constructed based on the first, second, third, and fourth feature data of multiple satellites across multiple historical time periods. This time-series data can then be input into a machine learning model. For example, the first, second, third, and fourth feature data of each satellite in each historical time period can be fused to obtain sub-fused feature data for that satellite in that historical time period; sub-fused feature data of the satellite across multiple historical time periods can be fused to obtain fused feature data for that satellite; or fused feature data from multiple satellites can be fused to obtain time-series data. Fusion methods can include splicing, etc. Alternatively, the first, second, third, and fourth feature data of multiple satellites across multiple historical time periods can be input into the machine learning model separately.

[0081] Optionally, please refer to Figure 3Matrix M can be obtained by concatenating matrices GFT, FL, Ff, and Fwn. Matrix M can be time series data. Matrix M can be input into a machine learning model. The dimensions of matrix GFT are (s, t, a). The dimensions of matrix FL are (s, t, b). The dimensions of matrix Ff are (s, t, 1). The dimensions of matrix Fwn are (s, t, 1). Considering that matrices GFT, FL, Ff, and Fwn all have satellite dimensions (the dimension corresponding to s) and historical time period dimensions (the dimension corresponding to t), the geographic feature dimension of matrix GFT (the dimension corresponding to a), the business type dimension of matrix FL (the dimension corresponding to b), the business traffic dimension of matrix Ff (the dimension corresponding to 1), and the wave position dimension of matrix Fwn (the dimension corresponding to 1) can be concatenated to obtain matrix M. The dimensions of matrix M can be represented as (s, t, f). f = a + 1 + 1 + b. For example, the value of a can be 5, the value of b can be 5, and f = 17. Please refer to [link to documentation]. Figure 4 The first dimension of matrix M (the dimension corresponding to s) represents the satellite. The second dimension of matrix M (the dimension corresponding to t) represents the historical time period. The third dimension of matrix M (the dimension corresponding to f) represents the satellite features. Each data element in the third dimension can represent a satellite feature in a historical time period. This satellite feature can be characterized by the first feature data, second feature data, third feature data, and fourth feature data of the satellite in that historical time period.

[0082] This specification also provides a satellite service traffic prediction device in its embodiments. The device can be applied to computer equipment such as portable computers, desktop computers, and servers. Please refer to... Figure 5 The device may include the following units.

[0083] The first acquisition unit 51 is used to acquire the geographical feature data of the wave position;

[0084] The second acquisition unit 52 is used to acquire the coverage time data of the satellite for the wave position within a historical period.

[0085] Construction unit 53 is used to construct first feature data of the satellite based on the geographic feature data and the coverage time data; the first feature data is used to represent the geographic features cumulatively covered by the satellite during the historical period.

[0086] The third acquisition unit 54 is used to acquire second feature data, which represents the service traffic of the satellite during the historical period.

[0087] The prediction unit 55 is used to predict the future service traffic of the satellite based on the first feature data and the second feature data.

[0088] This specification also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described satellite service traffic prediction method.

[0089] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described satellite service traffic prediction method.

[0090] This specification also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described satellite service traffic prediction method.

[0091] Those skilled in the art will understand that this specification can be provided as a method, system, or computer program product. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments thereof. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. The computer may be a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0093] The functional units in the embodiments of this specification can be integrated into one processing unit, or each functional unit can exist physically separately, or two or more functional units can be integrated into one processing unit.

[0094] Those skilled in the art will understand that the descriptions of the various embodiments in this specification have different focuses, and parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. Furthermore, it is understood that those skilled in the art, after reading this specification, can conceive of any combination of some or all of the embodiments listed in this specification without creative effort, and such combinations are also within the scope of disclosure and protection of this specification.

[0095] Although this specification has been described through embodiments, those skilled in the art will understand that the above embodiments are merely illustrative of the core ideas of this specification. Those skilled in the art will appreciate that many variations and modifications are possible with this specification. It is intended that the appended claims encompass these variations and modifications without departing from the spirit of this specification.

Claims

1. A method of satellite traffic flow prediction, characterized by, The method comprises: obtaining geographical feature data of wave positions; the number of wave positions is multiple, and the geographical feature data comprises index data of each wave position under multiple geographical features; obtaining coverage time data of the satellite for the wave positions in a historical period; the coverage time data comprises a coverage time length ratio of each wave position, and the coverage time length ratio comprises a ratio of a coverage time length of the satellite for the wave position to a time length of the historical period; constructing first feature data of the satellite according to the geographical feature data and the coverage time data; the first feature data is used to represent geographical features cumulatively covered by the satellite in the historical period; the construction of the first feature data of the satellite comprises: multiplying multiple index data of each wave position and the coverage time length ratio of the wave position; adding products of corresponding same geographical features of multiple wave positions to obtain cumulative index data of the geographical features; and the first feature data comprises cumulative index data of the satellite under various geographical features; obtaining second feature data, the second feature data is used to represent traffic of the satellite in the historical period; predicting future traffic of the satellite according to the first feature data and the second feature data.

2. The method of claim 1, wherein, The method further comprises: determining one or more service types of internet services of the satellite; the second feature data comprises first traffic data of the satellite under each service type.

3. The method of claim 2, wherein, The method further comprises: obtaining third feature data of the satellite, the third feature data is used to represent second traffic data of the satellite in the historical period, and the second traffic data comprises a sum of the first traffic data under each service type; the prediction of the future traffic of the satellite comprises: predicting the future traffic of the satellite according to the first feature data, the second feature data and the third feature data.

4. The method of claim 1, wherein, The method further comprises: obtaining fourth feature data, the fourth feature data is used to represent a number of wave positions covered by the satellite in the historical period; the prediction of the future traffic of the satellite comprises: predicting the future traffic of the satellite according to the first feature data, the second feature data and the fourth feature data.

5. The method of claim 1, wherein, The number of the historical periods is multiple; the prediction of the future traffic of the satellite comprises: constructing time series data according to the first feature data and the second feature data of multiple historical periods; inputting the time series data into a time series model to obtain an output of the time series model; the output of the time series model comprises traffic of the satellite in one or more future periods.

6. The method of claim 5, wherein, The number of the satellites is multiple; the construction of the time series data comprises: constructing time series data according to the first feature data and the second feature data of each satellite in multiple historical periods; the output of the time series model comprises traffic of each satellite in one or more future periods.

7. The method of claim 1, wherein: the satellite comprises one or more of a low earth orbit satellite and a medium earth orbit satellite.

8. A satellite traffic volume prediction apparatus characterized by comprising: ​ The first obtaining unit is configured to obtain geographical feature data of wave positions; the number of the wave positions is multiple, and the geographical feature data comprises index data of each wave position under multiple geographical features; The second obtaining unit is configured to obtain coverage time data of the satellite for the wave positions in a historical period; the coverage time data comprises a coverage time length proportion of each wave position, and the coverage time length proportion comprises a ratio of a coverage time length of the satellite for a wave position to a length of the historical period; The constructing unit is configured to construct first feature data of the satellite according to the geographical feature data and the coverage time data; The first feature data is used to represent geographical features that are cumulatively covered by the satellite in the historical period; the first feature data of the satellite is constructed by multiplying multiple index data of each wave position with the coverage time length proportion of the wave position, adding products of multiple wave positions corresponding to the same geographical feature, and obtaining cumulative index data of the geographical feature; and the first feature data comprises cumulative index data of the satellite under various geographical features; The third obtaining unit is configured to obtain second feature data, which is used to represent a business flow of the satellite in the historical period; The predicting unit is configured to predict a future business flow of the satellite according to the first feature data and the second feature data.

Images