Low-orbit satellite communication link switching system based on machine learning

Abstract

The present invention discloses a low-orbit satellite communication link switching system based on machine learning, which belongs to the field of low-orbit satellite technology and includes: a data collection module, configured to collect real-time data of low-orbit satellite communication links; a data processing module, configured to process real-time data of low-orbit satellite communication links; a model training module, configured to predict changes in the quality of low-orbit satellite communication links and determine the prediction results of the quality of low-orbit satellite communication links; a switching decision module, configured to select the best switching time and target satellite according to the prediction results; and a switching execution module, configured to optimize the switching process of low-orbit satellite communication links. The present invention solves the problem that the existing low-orbit satellite communication links cannot be effectively switched based on machine learning, resulting in low continuity and stability of satellite communications. The present invention can achieve accurate prediction of link status and precise grasp of switching timing, reduce the number of link switching times, and improve communication continuity and stability.

Description

Technical Field

[0001] The present invention relates to the field of low-orbit satellite technology, and in particular to a low-orbit satellite communication link switching system based on machine learning. Background Art

[0002] Low-orbit satellite communications, with their wide coverage, low latency, and strong scalability, hold enormous potential for remote area communications and global internet coverage. However, due to the high-speed motion of satellites and the complex link environment, traditional link switching methods struggle to meet the demands of dynamic scenarios.

[0003] The Chinese patent with the announcement number CN116094623B discloses a transmission line monitoring terminal, system and method based on low-orbit satellite communication. The transmission line monitoring terminal includes: a sensing device and a control center; the control center includes an MCU control unit, a connection module, a low-orbit satellite communication module and a GPS module; the MCU control unit is connected to the sensing device through the connection module, and the sensing device sends the acquired transmission line monitoring data to the MCU control unit through the connection module and stores it in a register; the MCU control unit is respectively connected to the low-orbit satellite communication module and the GPS module, and is used to send the transmission line monitoring data at different locations through the low-orbit satellite communication module. It breaks through the geographical use restrictions of the terminal and expands the use scenarios of the terminal. However, the patent has the following defects:

[0004] Existing technologies cannot effectively switch low-orbit satellite communication links based on machine learning, resulting in poor continuity and stability of satellite communications. Summary of the Invention

[0005] The purpose of the present invention is to provide a low-orbit satellite communication link switching system based on machine learning, which can accurately predict the link status and accurately grasp the switching timing, reduce the number of link switching times, improve communication continuity and stability, and solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The low-orbit satellite communication link switching system based on machine learning includes:

[0008] a data collection module configured to collect real-time data from a low-orbit satellite communication link;

[0009] a data processing module configured to process the collected real-time data of the low-orbit satellite communication link and determine characteristic data of the low-orbit satellite communication link;

[0010] Among them, the feature extraction model is used to identify the basic related information of the real-time data of the low-orbit satellite communication link, and then the data analysis and calculation are performed according to the calculation rules to obtain the target feature data information;

[0011] a model training module configured to construct a low-orbit satellite communication link quality prediction model, predict a change in the low-orbit satellite communication link quality, and determine a low-orbit satellite communication link quality prediction result;

[0012] a handover decision module configured to select an optimal handover timing and a target satellite based on a low-orbit satellite communication link quality prediction result;

[0013] The switching execution module is configured to optimize the low-orbit satellite communication link switching process according to the selected optimal switching timing and target satellite.

[0014] Preferably, collecting real-time data of low-orbit satellite communication links includes:

[0015] Real-time monitoring and collection of satellite position, speed, orbit information, signal strength, signal-to-noise ratio, and Doppler shift during communication between satellite and ground terminal to obtain satellite data;

[0016] Real-time monitoring and collection of user location, movement speed, terminal type, and antenna direction during communication between satellites and ground terminals to obtain user data;

[0017] Real-time monitoring and collection of weather conditions, interference sources, and network loads during communications between satellites and ground terminals to obtain environmental data;

[0018] Among them, the real-time data of the low-orbit satellite communication link is determined based on satellite data, user data and environmental data.

[0019] Preferably, processing the collected low-orbit satellite communication link real-time data includes:

[0020] Clean the real-time data of the low-orbit satellite communication link to remove the noise data that is useless for the switching of the low-orbit satellite communication link;

[0021] Check the real-time data of the low-orbit satellite communication link, identify duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link, and process the duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link;

[0022] Remove duplicate values ​​from real-time data of low-orbit satellite communication links and retain unique data records;

[0023] Determine whether missing values ​​and outliers in the real-time data of the low-orbit satellite communication link are useful for the low-orbit satellite communication link switching. If so, fill the missing values ​​with the mean and replace the outliers with the median. Otherwise, delete the missing values ​​and outliers directly.

[0024] Normalize the low-orbit satellite communication link real-time data, convert the low-orbit satellite communication link real-time data of different dimensions into a unified range, remove the dimension differences in the low-orbit satellite communication link real-time data, and determine the standardized low-orbit satellite communication link real-time data;

[0025] Feature extraction is performed on the real-time data of the low-orbit satellite communication link. Features that are useful for low-orbit satellite communication link switching and for predicting link quality changes are extracted from the real-time data of the low-orbit satellite communication link. The characteristic data of the low-orbit satellite communication link is determined, including the average signal strength, the change trend of the link quality, the frequency distribution of the Doppler shift, the relative angle change rate between the satellite and the user, the relative distance and speed between the satellite and the user, and the signal attenuation rate.

[0026] Preferably, constructing a low-orbit satellite communication link quality prediction model includes:

[0027] According to the switching requirements of low-orbit satellite communication links, historical data of low-orbit satellite communication links are collected, and the collected historical data of low-orbit satellite communication links are divided to determine the training set and test set;

[0028] Based on machine learning technology, a training set is used to train the machine learning model, so that the machine learning model can autonomously learn the low-orbit satellite communication link quality prediction behavior from the training set, and determine the low-orbit satellite communication link quality prediction model based on machine learning;

[0029] The performance of the low-orbit satellite communication link quality prediction model based on machine learning was tested using a test set. The accuracy, recall rate, and F1 score were used to evaluate whether the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the quality of low-orbit satellite communication links.

[0030] When the low-orbit satellite communication link quality prediction model based on machine learning cannot achieve the effect of predicting the low-orbit satellite communication link quality, the parameters of the low-orbit satellite communication link quality prediction model based on machine learning are adjusted, and the low-orbit satellite communication link quality prediction model based on machine learning is continuously optimized until the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the low-orbit satellite communication link quality, and then the optimal low-orbit satellite communication link quality prediction model is determined.

[0031] Preferably, predicting the change in the quality of the low-orbit satellite communication link includes:

[0032] Obtain the best low-orbit satellite communication link quality prediction model and deploy the best low-orbit satellite communication link quality prediction model in the actual low-orbit satellite communication link quality prediction environment;

[0033] The low-orbit satellite communication link characteristic data is input into the optimal low-orbit satellite communication link quality prediction model, the low-orbit satellite communication link characteristic data is analyzed according to the optimal low-orbit satellite communication link quality prediction model, and the link quality change between the satellite and the user is predicted to determine the low-orbit satellite communication link quality prediction result.

[0034] Preferably, selecting the optimal switching timing and target satellite according to the low-orbit satellite communication link quality prediction result includes:

[0035] According to the prediction results of the low-orbit satellite communication link quality, and combined with the status and switching cost of the candidate satellites, the critical point at which the low-orbit satellite communication link quality deteriorates is determined, the optimal switching time is selected, and the satellite with high signal strength and stable link quality is selected as the target satellite for switching.

[0036] Preferably, the low-orbit satellite communication link switching process is optimized according to the selected optimal switching timing and target satellite, including:

[0037] After selecting the optimal switching time and target satellite, smooth switching technology is used to dynamically adjust switching parameters and switch the low-orbit satellite communication link to adapt to the dynamic changes of the low-orbit satellite communication link;

[0038] According to the switching situation of the low-orbit satellite communication link, the low-orbit satellite communication link is updated, and the status of the low-orbit satellite communication link after switching is monitored in real time. The monitoring results are fed back to the low-orbit satellite communication link quality prediction model, so that the low-orbit satellite communication link quality prediction model forms a closed-loop optimization, and then the low-orbit satellite communication link is continuously optimized and switched.

[0039] Preferably, feature extraction is performed on the real-time data of the low-orbit satellite communication link, including:

[0040] Analyze the low-orbit satellite communication link switching, determine the associated data features of the low-orbit satellite communication link switching, and obtain the first data feature;

[0041] Analyze the link quality change prediction, determine the data features related to the link quality change prediction, and obtain the second data feature;

[0042] The first feature of the data is combined with the second feature of the data to perform common feature analysis to determine the target feature, and the target feature is analyzed to determine the calculation rules and basic correlation information of the target feature;

[0043] Establish a feature extraction model based on the calculation rules of the target features and basic correlation information to obtain a feature extraction model;

[0044] The feature extraction model is used to identify the basic correlation information of the real-time data of the low-orbit satellite communication link, and then data analysis and calculation are performed according to the calculation rules to obtain the target feature data information.

[0045] Preferably, the machine learning model is trained using a training set, including:

[0046] Perform model training on the training samples in the training set through the machine learning model to obtain model training data;

[0047] Analyze whether the machine learning model achieves the expected effect based on the model training data and obtain the model analysis results;

[0048] When the model analysis result shows that the machine learning model does not achieve the expected effect, the deviation data of the training samples is analyzed according to the model training data, and the training samples are divided for the training set according to the deviation data to obtain a first training sample subset and a second training sample subset; wherein the first training sample subset is a sample set composed of unbiased training samples, and the second training sample subset is a sample set composed of biased training samples;

[0049] Performing training sample feature analysis on the second training sample subset to obtain training sample features;

[0050] Randomly select a preset number of training samples from the first training sample subset as first target training samples, and at the same time, obtain a preset number of new training samples based on the characteristics of the training samples to obtain second target training samples;

[0051] Deleting training samples from the first training sample subset according to the first target training sample to obtain a first training sample update subset, and adding training samples from the second training sample subset according to the second target training sample to obtain a second training sample update subset;

[0052] The first training sample update subset and the second training sample update subset together constitute an updated training set, so as to continue model training for the machine learning model based on the updated training set, and repeat multiple cycles until the model analysis results show that the machine learning model achieves the expected effect.

[0053] Preferably, real-time monitoring of the low-orbit satellite communication link status after switching includes:

[0054] Perform feature analysis on low-orbit satellite communication links, determine link transmission characteristics, and determine monitoring nodes based on the link transmission characteristics;

[0055] Monitoring the low-orbit satellite communication link according to the monitoring node to obtain low-orbit satellite communication link node monitoring information;

[0056] The monitoring nodes are divided into first monitoring nodes and second monitoring nodes based on their positions in the low-orbit satellite communication link, wherein the first monitoring nodes are the two end nodes of the low-orbit satellite communication link and the second monitoring nodes are the middle nodes of the low-orbit satellite communication link;

[0057] Extracting information from low-orbit satellite communication link node monitoring information according to the first monitoring node to obtain first monitoring node monitoring information, and extracting information from low-orbit satellite communication link node monitoring information according to the second monitoring node to obtain second monitoring node monitoring information;

[0058] Performing a preliminary analysis and judgment based on the monitoring information of the first monitoring node to determine whether the low-orbit satellite communication link is in an idle state, and obtaining a preliminary analysis and judgment result;

[0059] When the preliminary analysis determines that the low-orbit satellite communication link is in an idle state, an abnormality analysis of the second monitoring node is performed based on the monitoring information of the second monitoring node to determine whether an abnormality occurs in the second monitoring node, obtain a monitoring analysis result, and issue an abnormality reminder based on the monitoring analysis result;

[0060] When the preliminary analysis determines that the low-orbit satellite communication link is in a non-idle state, segmentation is performed according to the monitoring nodes. Based on the segmentation results, the monitoring information of the first monitoring node is combined with the monitoring information of the second monitoring node to analyze the transmission performance by segment, determine the low-orbit satellite communication link performance data, and obtain the low-orbit satellite communication link monitoring information.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] The present invention collects real-time data of low-orbit satellite communication links, processes the collected real-time data of low-orbit satellite communication links, determines characteristic data of the low-orbit satellite communication links, constructs a low-orbit satellite communication link quality prediction model according to the switching requirements of the low-orbit satellite communication links, analyzes the characteristic data of the low-orbit satellite communication links, predicts changes in link quality between satellites and users, determines a low-orbit satellite communication link quality prediction result, selects an optimal switching timing and a target satellite according to the low-orbit satellite communication link quality prediction result, dynamically adjusts switching parameters, switches the low-orbit satellite communication link, and monitors the status of the low-orbit satellite communication link after switching in real time, and feeds back the monitoring results to the low-orbit satellite communication link quality prediction model, so that the low-orbit satellite communication link quality prediction model forms a closed-loop optimization, and then continuously optimizes and switches the low-orbit satellite communication link to achieve accurate prediction of the link status and precise grasp of the switching timing, reduce the number of link switching times, reduce switching delays, and improve communication continuity and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a module diagram of the low-orbit satellite communication link switching system based on machine learning of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0065] To solve the problem that the existing low-orbit satellite communication link cannot be effectively switched based on machine learning, resulting in low continuity and stability of satellite communication, please refer to Figure 1 , this embodiment provides the following technical solutions:

[0066] A low-orbit satellite communication link switching system based on machine learning includes: a data collection module, a data processing module, a model training module, a switching decision module and a switching execution module.

[0067] Specifically, through the interactive communication between the data collection module, data processing module, model training module, switching decision module and switching execution module, accurate prediction of link status and precise grasp of switching timing can be achieved, which reduces the number of link switches, reduces switching delay, and improves communication continuity and stability.

[0068] The data collection module is configured to collect real-time data of low-orbit satellite communication links;

[0069] In this embodiment, collecting real-time data of low-orbit satellite communication links includes:

[0070] Real-time monitoring and collection of satellite position, speed, orbit information, signal strength, signal-to-noise ratio, and Doppler shift during communication between satellite and ground terminal to obtain satellite data;

[0071] It should be noted that the satellite position is the latitude, longitude and altitude of the satellite, and the relative distance between the satellite and the ground terminal; the satellite speed is the instantaneous speed of the satellite, and the relative speed between the satellite and the ground terminal, which is used to calculate the Doppler shift; the signal strength is an indication of the strength of the received signal, reflecting the signal quality of the current link; the signal-to-noise ratio is the ratio of the signal to the noise, which measures the stability of the link quality; the Doppler shift is the frequency change caused by the relative motion of the satellite and the user, which affects signal reception.

[0072] Real-time monitoring and collection of user location, movement speed, terminal type, and antenna direction during communication between satellites and ground terminals to obtain user data;

[0073] It should be noted that the user location refers to the user's latitude, longitude and altitude; the moving speed refers to the user's moving speed and direction (such as stationary, low-speed movement, and high-speed movement); and the terminal type refers to the terminal's hardware capabilities (such as antenna gain and transmit power).

[0074] Real-time monitoring and collection of weather conditions, interference sources, and network loads during communications between satellites and ground terminals to obtain environmental data;

[0075] It should be noted that weather conditions refer to the impact of rain, fog, snow and other weather conditions on signal propagation; interference sources refer to the interference intensity of other communication systems or equipment.

[0076] Among them, the real-time data of the low-orbit satellite communication link is determined based on satellite data, user data and environmental data.

[0077] The data processing module is configured to process the collected real-time data of the low-orbit satellite communication link to determine the characteristic data of the low-orbit satellite communication link;

[0078] In this embodiment, the collected real-time data of the low-orbit satellite communication link is processed, including:

[0079] Clean the real-time data of the low-orbit satellite communication link to remove the noise data that is useless for the switching of the low-orbit satellite communication link;

[0080] Check the real-time data of the low-orbit satellite communication link, identify duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link, and process the duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link;

[0081] Remove duplicate values ​​from real-time data of low-orbit satellite communication links and retain unique data records;

[0082] Determine whether missing values ​​and outliers in the real-time data of the low-orbit satellite communication link are useful for the low-orbit satellite communication link switching. If so, fill the missing values ​​with the mean and replace the outliers with the median. Otherwise, delete the missing values ​​and outliers directly.

[0083] Normalize the low-orbit satellite communication link real-time data, convert the low-orbit satellite communication link real-time data of different dimensions into a unified range, remove the dimension differences in the low-orbit satellite communication link real-time data, and determine the standardized low-orbit satellite communication link real-time data;

[0084] Feature extraction is performed on the real-time data of the low-orbit satellite communication link. Features that are useful for low-orbit satellite communication link switching and for predicting link quality changes are extracted from the real-time data of the low-orbit satellite communication link. The characteristic data of the low-orbit satellite communication link is determined, including the average signal strength, the change trend of the link quality, the frequency distribution of the Doppler shift, the relative angle change rate between the satellite and the user, the relative distance and speed between the satellite and the user, and the signal attenuation rate.

[0085] It should be noted that by processing the collected real-time data of the low-orbit satellite communication link and determining the characteristic data of the low-orbit satellite communication link, it can facilitate the subsequent analysis of the characteristic data of the low-orbit satellite communication link and better predict the changes in the quality of the low-orbit satellite communication link.

[0086] The model training module is configured to construct a low-orbit satellite communication link quality prediction model, predict changes in the low-orbit satellite communication link quality, and determine a low-orbit satellite communication link quality prediction result;

[0087] In this embodiment, building a low-orbit satellite communication link quality prediction model includes:

[0088] According to the switching requirements of low-orbit satellite communication links, historical data of low-orbit satellite communication links are collected, and the collected historical data of low-orbit satellite communication links are divided to determine the training set and test set;

[0089] Based on machine learning technology, a training set is used to train the machine learning model, so that the machine learning model can autonomously learn the low-orbit satellite communication link quality prediction behavior from the training set, and determine the low-orbit satellite communication link quality prediction model based on machine learning;

[0090] The performance of the low-orbit satellite communication link quality prediction model based on machine learning was tested using a test set. The accuracy, recall rate, and F1 score were used to evaluate whether the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the quality of low-orbit satellite communication links.

[0091] When the low-orbit satellite communication link quality prediction model based on machine learning cannot achieve the effect of predicting the low-orbit satellite communication link quality, the parameters of the low-orbit satellite communication link quality prediction model based on machine learning are adjusted, and the low-orbit satellite communication link quality prediction model based on machine learning is continuously optimized until the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the low-orbit satellite communication link quality, and then the optimal low-orbit satellite communication link quality prediction model is determined.

[0092] In this embodiment, predicting the change in the quality of the low-orbit satellite communication link includes:

[0093] Obtain the best low-orbit satellite communication link quality prediction model and deploy the best low-orbit satellite communication link quality prediction model in the actual low-orbit satellite communication link quality prediction environment;

[0094] The low-orbit satellite communication link characteristic data is input into the optimal low-orbit satellite communication link quality prediction model, the low-orbit satellite communication link characteristic data is analyzed according to the optimal low-orbit satellite communication link quality prediction model, and the link quality change between the satellite and the user is predicted to determine the low-orbit satellite communication link quality prediction result.

[0095] It should be noted that the low-orbit satellite communication link quality prediction model is used to predict the changing trend of link quality, make switching decisions in advance, and avoid communication interruption.

[0096] The handover decision module is configured to select the best handover time and target satellite according to the low-orbit satellite communication link quality prediction result;

[0097] In this embodiment, selecting the optimal switching timing and target satellite based on the low-orbit satellite communication link quality prediction result includes:

[0098] Based on the low-orbit satellite communication link quality prediction results, and combined with the status of candidate satellites and the switching cost (switching delay, resource allocation), the critical point at which the low-orbit satellite communication link quality deteriorates is determined, the optimal switching time is selected, and satellites with high signal strength and stable link quality are selected as the target satellites for switching.

[0099] Specifically, a reinforcement learning method is used to dynamically adjust the switching timing, balance the switching frequency and communication quality, and determine the optimal switching timing and target satellite.

[0100] Among them, the switching execution module is configured to optimize the low-orbit satellite communication link switching process based on the selected optimal switching timing and target satellite.

[0101] In this embodiment, the low-orbit satellite communication link switching process is optimized based on the selected optimal switching timing and target satellite, including:

[0102] After selecting the optimal switching time and target satellite, smooth switching technology is used to dynamically adjust switching parameters and switch the low-orbit satellite communication link to adapt to the dynamic changes of the low-orbit satellite communication link;

[0103] According to the switching situation of the low-orbit satellite communication link, the low-orbit satellite communication link is updated, and the status of the low-orbit satellite communication link after switching is monitored in real time. The monitoring results are fed back to the low-orbit satellite communication link quality prediction model, so that the low-orbit satellite communication link quality prediction model forms a closed-loop optimization, and then the low-orbit satellite communication link is continuously optimized and switched.

[0104] In this embodiment, feature extraction is performed on real-time data of a low-orbit satellite communication link, including:

[0105] Analyze the low-orbit satellite communication link switching, determine the associated data features of the low-orbit satellite communication link switching, and obtain the first data feature;

[0106] Analyze the link quality change prediction, determine the data features related to the link quality change prediction, and obtain the second data feature;

[0107] The first feature of the data is combined with the second feature of the data to perform common feature analysis to determine the target feature, and the target feature is analyzed to determine the calculation rules and basic correlation information of the target feature;

[0108] Establish a feature extraction model based on the calculation rules of the target features and basic correlation information to obtain a feature extraction model;

[0109] The feature extraction model is used to identify the basic correlation information of the real-time data of the low-orbit satellite communication link, and then data analysis and calculation are performed according to the calculation rules to obtain the target feature data information.

[0110] Among them, basic related information refers to information that can be directly determined from the real-time data of the low-orbit satellite communication link. Basic related information includes: the frequency of the transmitted and received signals, the signal power, the signal transmission and reception time, the noise power, the number of bits that have errors during the transmission process, the total number of transmitted bits, etc.

[0111] The target features are those that are useful for switching low-orbit satellite communication links and are used to predict changes in link quality, including: average signal strength, link quality change trend, Doppler shift frequency distribution, relative angle change rate between satellite and user, relative distance and speed between satellite and user, and signal attenuation rate.

[0112] Calculation rules for target features include: signal strength mean calculation formula, link quality indicator calculation and comparison formula, relative angle calculation formula, relative distance and speed calculation formula, signal attenuation rate calculation formula, etc.

[0113] The feature extraction model is used to identify basic related information of the real-time data of the low-orbit satellite communication link, and then data analysis and calculation are performed according to the calculation rules, including:

[0114] Identify real-time data of low-orbit satellite communication links and obtain basic correlation information from the real-time data of low-orbit satellite communication links;

[0115] Analyze the calculation rules based on the target features and determine the calculation variables;

[0116] Perform matching analysis on basic correlation information according to calculated variables, including:

[0117] Performing vectorization processing on the calculation variables and the basic correlation information to obtain vectorized information of the calculation variables and vectorized information of the basic correlation information;

[0118] Normalization processing is performed on the vectorized information of the calculation variables and the vectorized information of the basic association information. If the vectorized information of the calculation variables or the vectorized information of the basic association information is the same as the vector length of the standard, then the vectorized information of the calculation variables or the vectorized information of the basic association information is the standard vector of the calculation variables or the standard vector of the basic association information. If the vectorized information of the calculation variables or the vectorized information of the basic association information is different from the vector length of the standard, then the end of the vectorized information of the calculation variables or the vectorized information of the basic association information is added with zeros to make the vectorized information of the calculation variables or the vectorized information of the basic association information the same as the vector length of the standard, and the standard vector of the calculation variables or the standard vector of the basic association information is obtained; wherein, the vector length of the standard can be determined according to the vectorized information of the variable, and the longest vector in the vectorized information of the variable is used as the vector length of the standard. It can also be set according to needs. However, when setting it by yourself, the vector length of the standard is greater than or equal to the longest vector length in the vectorized information of the variable.

[0119] The canonical vector of the calculated variable or the canonical vector of the basic association information is determined by matching the analysis data using the following formula:

[0120] ;

[0121] In the above formula, For the The calculated variables and Matching analysis data between basic related information, For the The normalized vector of the calculated variables Component values, For the The first Component values;

[0122] Determine whether there is a matching correspondence between the calculated variable and the basic correlation information based on the matching analysis data, and if there is a matching correspondence between the calculated variable and the basic correlation information, retrieve the corresponding basic correlation information, obtain the value of the basic correlation information, and obtain the data information of the calculated variable;

[0123] The data information of the calculation variables is combined with the calculation rules to calculate and obtain the target feature data information.

[0124] The above analysis respectively determines the associated data features of low-orbit satellite communication link switching for low-orbit satellite communication link switching and determines the data features related to link quality change prediction for link quality change prediction, providing data support for common feature analysis, so that common feature analysis can be performed based on more comprehensive feature data, thereby improving the accuracy of target features. Moreover, by establishing a feature extraction model based on the calculation rules of target features and basic associated information, efficient feature extraction of low-orbit satellite communication link real-time data can be achieved by using the feature extraction model, thereby improving the accuracy of feature extraction of low-orbit satellite communication link real-time data, providing a guarantee for predicting low-orbit satellite communication link quality changes, and thereby improving the communication continuity and stability of the low-orbit satellite communication link switching system based on machine learning. In addition, when the feature extraction model is used to identify the basic correlation information of the real-time data of the low-orbit satellite communication link and the data analysis and calculation are performed according to the calculation rules, matching analysis is performed in the form of vectors, which not only facilitates the matching analysis, but also can directly determine whether there is a matching correspondence between the calculation variables and the basic correlation information with objective data, providing convenience for determining the data information of the calculation variables, so that the data information of the variables can be quickly locked. Moreover, through normalization processing, the unity of the normalized vector of the calculation variable and the normalized vector of the basic correlation information is guaranteed, and the feasibility of the matching analysis data calculation is guaranteed, avoiding the inability to calculate the matching analysis data, so that subsequent calculations can be performed according to the calculation rules, and then the results can be output based on the feature extraction model, ensuring the operation of the feature extraction model.

[0125] In this embodiment, the machine learning model is trained using a training set, including:

[0126] Perform model training on the training samples in the training set through the machine learning model to obtain model training data;

[0127] Analyze whether the machine learning model achieves the expected effect based on the model training data and obtain the model analysis results;

[0128] When the model analysis result shows that the machine learning model does not achieve the expected effect, the deviation data of the training samples is analyzed according to the model training data, and the training samples are divided for the training set according to the deviation data to obtain a first training sample subset and a second training sample subset; wherein the first training sample subset is a sample set composed of unbiased training samples, and the second training sample subset is a sample set composed of biased training samples;

[0129] Performing training sample feature analysis on the second training sample subset to obtain training sample features;

[0130] Randomly select a preset number of training samples from the first training sample subset as first target training samples, and at the same time, obtain a preset number of new training samples based on the characteristics of the training samples to obtain second target training samples;

[0131] Deleting training samples from the first training sample subset according to the first target training sample to obtain a first training sample update subset, and adding training samples from the second training sample subset according to the second target training sample to obtain a second training sample update subset;

[0132] The first training sample update subset and the second training sample update subset together constitute an updated training set, so as to continue model training for the machine learning model based on the updated training set, and repeat multiple cycles until the model analysis results show that the machine learning model achieves the expected effect.

[0133] The preset number can be adjusted according to needs, and is usually determined in proportion to the size of the second training sample subset.

[0134] The above-mentioned update of the training set makes it possible to adjust the training set when training the machine learning model using the training set, thereby improving the efficiency of model training, enabling the machine learning model to achieve the expected effect in a shorter time, reducing the time of model training, and at the same time ensuring the accuracy of the low-orbit satellite communication link quality prediction model based on machine learning. When the machine learning model fails to achieve the expected effect, the training set is updated with the biased training samples as the focus by analyzing the deviation data of the training samples, so that the updated training set can optimize the improvement of the machine learning model on the biased training samples, improve the effect of the machine learning model, and then improve the accuracy of the low-orbit satellite communication link quality prediction model based on machine learning, so that the low-orbit satellite communication link quality prediction model based on machine learning can more accurately predict the link status.

[0135] In this embodiment, real-time monitoring of the low-orbit satellite communication link status after switching includes:

[0136] Perform feature analysis on low-orbit satellite communication links, determine link transmission characteristics, and determine monitoring nodes based on the link transmission characteristics;

[0137] Monitoring the low-orbit satellite communication link according to the monitoring node to obtain low-orbit satellite communication link node monitoring information;

[0138] The monitoring nodes are divided into first monitoring nodes and second monitoring nodes based on their positions in the low-orbit satellite communication link, wherein the first monitoring nodes are the two end nodes of the low-orbit satellite communication link and the second monitoring nodes are the middle nodes of the low-orbit satellite communication link;

[0139] Extracting information from low-orbit satellite communication link node monitoring information according to the first monitoring node to obtain first monitoring node monitoring information, and extracting information from low-orbit satellite communication link node monitoring information according to the second monitoring node to obtain second monitoring node monitoring information;

[0140] Performing a preliminary analysis and judgment based on the monitoring information of the first monitoring node to determine whether the low-orbit satellite communication link is in an idle state, and obtaining a preliminary analysis and judgment result;

[0141] When the preliminary analysis determines that the low-orbit satellite communication link is in an idle state, an abnormality analysis of the second monitoring node is performed based on the monitoring information of the second monitoring node to determine whether an abnormality occurs in the second monitoring node, obtain a monitoring analysis result, and issue an abnormality reminder based on the monitoring analysis result;

[0142] When the preliminary analysis determines that the low-orbit satellite communication link is in a non-idle state, segmentation is performed according to the monitoring nodes. Based on the segmentation results, the monitoring information of the first monitoring node is combined with the monitoring information of the second monitoring node to analyze the transmission performance by segment, determine the low-orbit satellite communication link performance data, and obtain the low-orbit satellite communication link monitoring information.

[0143] Among them, transmission performance includes: signal-to-noise ratio, delay, bit error rate, etc.

[0144] The above-mentioned multi-node monitoring of the low-orbit satellite communication link is realized through the monitoring node, and the monitoring node is determined based on the link transmission characteristics, and can monitor according to different low-orbit satellite communication links, thereby improving the reliability of low-orbit satellite communication link status monitoring. In addition, the preliminary analysis of the low-orbit satellite communication link is realized through the first monitoring node, so that different analyses are performed when the low-orbit satellite communication link is in a non-idle state and an idle state. Not only can abnormal phenomena be discovered in time in the idle state to avoid the impact of abnormal phenomena on the use of the low-orbit satellite communication link, but also monitoring information of the low-orbit satellite communication link can be obtained in the non-idle state to understand the specific status of the low-orbit satellite communication link. In addition, the use of segment-by-segment analysis of transmission performance can effectively improve the efficiency of transmission performance analysis and the efficiency of determining low-orbit satellite communication link performance data, so that low-orbit satellite communication link monitoring information can be obtained in a shorter time. At the same time, it can also improve the accuracy of transmission performance and reduce the error of low-orbit satellite communication link monitoring information.

[0145] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0146] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A low-orbit satellite communication link switching system based on machine learning, characterized in that: include: a data collection module configured to collect real-time data from a low-orbit satellite communication link; a data processing module configured to process the collected real-time data of the low-orbit satellite communication link and determine characteristic data of the low-orbit satellite communication link; Among them, the feature extraction model is used to identify the basic related information of the real-time data of the low-orbit satellite communication link, and then the data analysis and calculation are performed according to the calculation rules to obtain the target feature data information; a model training module configured to construct a low-orbit satellite communication link quality prediction model, predict a change in the low-orbit satellite communication link quality, and determine a low-orbit satellite communication link quality prediction result; a handover decision module configured to select an optimal handover timing and a target satellite based on a low-orbit satellite communication link quality prediction result; a handover execution module configured to optimize a low-orbit satellite communication link handover process based on a selected optimal handover timing and target satellite; Feature extraction of real-time data from low-orbit satellite communication links, including: Analyze the low-orbit satellite communication link switching, determine the associated data features of the low-orbit satellite communication link switching, and obtain the first data feature; Analyze the link quality change prediction, determine the data features related to the link quality change prediction, and obtain the second data feature; The first feature of the data is combined with the second feature of the data to perform common feature analysis to determine the target feature, and the target feature is analyzed to determine the calculation rules and basic correlation information of the target feature; Establish a feature extraction model based on the calculation rules of the target features and basic correlation information to obtain a feature extraction model; The feature extraction model is used to identify basic related information of the real-time data of the low-orbit satellite communication link, and then data analysis and calculation are performed according to the calculation rules, including: Identify real-time data of low-orbit satellite communication links and obtain basic correlation information from the real-time data of low-orbit satellite communication links; Analyze the calculation rules based on the target features and determine the calculation variables; Perform matching analysis on basic correlation information according to calculated variables, including: Performing vectorization processing on the calculation variables and the basic correlation information to obtain vectorized information of the calculation variables and vectorized information of the basic correlation information; The normative vector of the calculated variable and the normative vector of the basic association information are determined by matching the analysis data using the following formula: ； In the above formula, For the calculated variables With the Basic related information Matching analysis data between For the The normalized vector of the calculated variables Component values, For the The first component values, m is the number of component values ​​in the normative vector; Determine whether there is a matching correspondence between the calculated variable and the basic correlation information based on the matching analysis data, and if there is a matching correspondence between the calculated variable and the basic correlation information, retrieve the corresponding basic correlation information, obtain the value of the basic correlation information, and obtain the data information of the calculated variable; The data information of the calculation variables is combined with the calculation rules to calculate and obtain the target feature data information.

2. The low-orbit satellite communication link switching system based on machine learning according to claim 1, characterized in that: Collect real-time data of low-orbit satellite communication links, including: Real-time monitoring and collection of satellite position, speed, orbit information, signal strength, signal-to-noise ratio, and Doppler shift during communication between satellite and ground terminal to obtain satellite data; Real-time monitoring and collection of user location, movement speed, terminal type, and antenna direction during communication between satellites and ground terminals to obtain user data; Real-time monitoring and collection of weather conditions, interference sources, and network loads during communications between satellites and ground terminals to obtain environmental data; Among them, the real-time data of the low-orbit satellite communication link is determined based on satellite data, user data and environmental data.

3. The low-orbit satellite communication link switching system based on machine learning according to claim 1, characterized in that: Processing of collected real-time data of low-orbit satellite communication links, including: Clean the real-time data of the low-orbit satellite communication link to remove the noise data that is useless for the switching of the low-orbit satellite communication link; Check the real-time data of the low-orbit satellite communication link, identify duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link, and process the duplicate values, missing values ​​and abnormal values ​​in the real-time data of the low-orbit satellite communication link; Remove duplicate values ​​from real-time data of low-orbit satellite communication links and retain unique data records; Determine whether missing values ​​and outliers in the real-time data of the low-orbit satellite communication link are useful for the low-orbit satellite communication link switching. If so, fill the missing values ​​with the mean and replace the outliers with the median. Otherwise, delete the missing values ​​and outliers directly. Normalize the low-orbit satellite communication link real-time data, convert the low-orbit satellite communication link real-time data of different dimensions into a unified range, remove the dimension differences in the low-orbit satellite communication link real-time data, and determine the standardized low-orbit satellite communication link real-time data; Feature extraction is performed on the real-time data of the low-orbit satellite communication link. Features that are useful for low-orbit satellite communication link switching and for predicting link quality changes are extracted from the real-time data of the low-orbit satellite communication link. The characteristic data of the low-orbit satellite communication link is determined, including the average signal strength, the change trend of the link quality, the frequency distribution of the Doppler shift, the relative angle change rate between the satellite and the user, the relative distance and speed between the satellite and the user, and the signal attenuation rate.

4. The low-orbit satellite communication link switching system based on machine learning according to claim 1, characterized in that Construct a low-orbit satellite communication link quality prediction model, including: According to the switching requirements of low-orbit satellite communication links, historical data of low-orbit satellite communication links are collected, and the collected historical data of low-orbit satellite communication links are divided to determine the training set and test set; Based on machine learning technology, a training set is used to train the machine learning model, so that the machine learning model can autonomously learn the low-orbit satellite communication link quality prediction behavior from the training set, and determine the low-orbit satellite communication link quality prediction model based on machine learning; The performance of the low-orbit satellite communication link quality prediction model based on machine learning was tested using a test set. The accuracy, recall rate, and F1 score were used to evaluate whether the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the quality of low-orbit satellite communication links. When the low-orbit satellite communication link quality prediction model based on machine learning cannot achieve the effect of predicting the low-orbit satellite communication link quality, the parameters of the low-orbit satellite communication link quality prediction model based on machine learning are adjusted, and the low-orbit satellite communication link quality prediction model based on machine learning is continuously optimized until the low-orbit satellite communication link quality prediction model based on machine learning can achieve the effect of predicting the low-orbit satellite communication link quality, and then the optimal low-orbit satellite communication link quality prediction model is determined.

5. The low-orbit satellite communication link switching system based on machine learning according to claim 4, characterized in that: Predict changes in low-orbit satellite communication link quality, including: Obtain the best low-orbit satellite communication link quality prediction model and deploy the best low-orbit satellite communication link quality prediction model in the actual low-orbit satellite communication link quality prediction environment; The low-orbit satellite communication link characteristic data is input into the optimal low-orbit satellite communication link quality prediction model, the low-orbit satellite communication link characteristic data is analyzed according to the optimal low-orbit satellite communication link quality prediction model, and the link quality change between the satellite and the user is predicted to determine the low-orbit satellite communication link quality prediction result.

6. The low-orbit satellite communication link switching system based on machine learning according to claim 1, characterized in that: The optimal handover timing and target satellite are selected based on the low-orbit satellite communication link quality prediction results, including: According to the prediction results of the low-orbit satellite communication link quality, and combined with the status and switching cost of the candidate satellites, the critical point at which the low-orbit satellite communication link quality deteriorates is determined, the optimal switching time is selected, and the satellite with high signal strength and stable link quality is selected as the target satellite for switching.

7. The low-orbit satellite communication link switching system based on machine learning according to claim 1, characterized in that: Optimize the low-orbit satellite communication link handover process based on the selected optimal handover timing and target satellite, including: After selecting the optimal switching time and target satellite, smooth switching technology is used to dynamically adjust switching parameters and switch the low-orbit satellite communication link to adapt to the dynamic changes of the low-orbit satellite communication link; According to the switching situation of the low-orbit satellite communication link, the low-orbit satellite communication link is updated, and the status of the low-orbit satellite communication link after switching is monitored in real time. The monitoring results are fed back to the low-orbit satellite communication link quality prediction model, so that the low-orbit satellite communication link quality prediction model forms a closed-loop optimization, and then the low-orbit satellite communication link is continuously optimized and switched.

8. The low-orbit satellite communication link switching system based on machine learning according to claim 4, characterized in that: Use the training set to train the machine learning model, including: Perform model training on the training samples in the training set through the machine learning model to obtain model training data; Analyze whether the machine learning model achieves the expected effect based on the model training data and obtain the model analysis results; When the model analysis result shows that the machine learning model does not achieve the expected effect, the deviation data of the training samples is analyzed according to the model training data, and the training samples are divided for the training set according to the deviation data to obtain a first training sample subset and a second training sample subset; wherein the first training sample subset is a sample set composed of unbiased training samples, and the second training sample subset is a sample set composed of biased training samples; Performing training sample feature analysis on the second training sample subset to obtain training sample features; Randomly select a preset number of training samples from the first training sample subset as first target training samples, and at the same time, obtain a preset number of new training samples based on the characteristics of the training samples to obtain second target training samples; Deleting training samples from the first training sample subset according to the first target training sample to obtain a first training sample update subset, and adding training samples from the second training sample subset according to the second target training sample to obtain a second training sample update subset; The first training sample update subset and the second training sample update subset together constitute an updated training set, so as to continue model training for the machine learning model based on the updated training set, and repeat multiple cycles until the model analysis results show that the machine learning model achieves the expected effect.

9. The low-orbit satellite communication link switching system based on machine learning according to claim 7, characterized in that: Real-time monitoring of the low-orbit satellite communication link status after switching, including: Perform feature analysis on low-orbit satellite communication links, determine link transmission characteristics, and determine monitoring nodes based on the link transmission characteristics; Monitoring the low-orbit satellite communication link according to the monitoring node to obtain low-orbit satellite communication link node monitoring information; The monitoring nodes are divided into first monitoring nodes and second monitoring nodes based on their positions in the low-orbit satellite communication link, wherein the first monitoring nodes are the two end nodes of the low-orbit satellite communication link and the second monitoring nodes are the middle nodes of the low-orbit satellite communication link; Extracting information from low-orbit satellite communication link node monitoring information according to the first monitoring node to obtain first monitoring node monitoring information, and extracting information from low-orbit satellite communication link node monitoring information according to the second monitoring node to obtain second monitoring node monitoring information; Performing a preliminary analysis and judgment based on the monitoring information of the first monitoring node to determine whether the low-orbit satellite communication link is in an idle state, and obtaining a preliminary analysis and judgment result; When the preliminary analysis determines that the low-orbit satellite communication link is in an idle state, an abnormality analysis of the second monitoring node is performed based on the monitoring information of the second monitoring node to determine whether an abnormality occurs in the second monitoring node, obtain a monitoring analysis result, and issue an abnormality reminder based on the monitoring analysis result; When the preliminary analysis determines that the low-orbit satellite communication link is in a non-idle state, segmentation is performed according to the monitoring nodes. Based on the segmentation results, the monitoring information of the first monitoring node is combined with the monitoring information of the second monitoring node to analyze the transmission performance by segment, determine the low-orbit satellite communication link performance data, and obtain the low-orbit satellite communication link monitoring information.

Images