A method and system for avoiding frequency interference by combining AI and digital twins

Abstract

The present application discloses a method and system for avoiding frequency interference by combining AI with digital twins, which relates to the field of satellite communication technology. The method includes: establishing a big data satellite frequency interference avoidance system, and collecting the current orbit information and frequency information of the communication satellite; simulating through a preset interference scenario prediction model to determine the satellite adjustment attitude, the preset interference scenario prediction model is based on the AI algorithm and the digital twin model to train the normal scenario and interference scenario of the satellite system, and the digital twin model construction includes an orbit prediction module and an attitude adjustment prediction module connected in sequence, the orbit prediction module is used to output the orbit prediction value according to the orbit information and frequency information, and the attitude adjustment prediction module outputs the satellite adjustment attitude according to the orbit prediction value; thereby, the interference scenario of the communication satellite can be predicted in advance through the preset interference scenario prediction model to adjust the beam pointing and realize automatic early warning and processing of communication satellite interference.

Description

Technical Field

[0001] The present application relates to the field of satellite communication technology, and in particular to a method and system for avoiding frequency interference by combining AI with digital twins. Background Art

[0002] Due to limitations on the satellite system's orbital position and communication frequency resources, different communication satellites typically need to share communication frequencies, resulting in a high probability of harmful interference between systems. To avoid frequency interference with geostationary satellites, low-orbit communication satellites in low-orbit orbit (LEO) are designed to shut down their payload beams when they approach the equator. These beams are then reactivated after the satellite leaves the shutdown zone.

[0003] However, when in the shutdown area, the normal business of the communication satellite will be affected. At the same time, the shutdown of the communication satellite will also affect the tracking accuracy and stability of the communication satellite. Summary of the Invention

[0004] The main purpose of this application is to provide a method and system for avoiding frequency interference by combining AI with digital twins, aiming to solve the technical problem that when a communication satellite is shut down in a shutdown area to avoid frequency interference, it will affect the normal business of the communication satellite. At the same time, the shutdown of the communication satellite will also affect the tracking accuracy and stability of the communication satellite.

[0005] To achieve the above objectives, this application proposes a method for avoiding frequency interference by combining AI with digital twins, the method comprising:

[0006] Establish a big data satellite frequency interference avoidance system, and use the system to collect the current orbit and frequency information of communication satellites;

[0007] The orbital information and the frequency information are simulated by a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite, wherein the preset interference scenario prediction model is obtained by training the normal scenario and the interference scenario of the satellite system based on the AI algorithm and the digital twin model, and the digital twin model is constructed including an orbit prediction module and an attitude adjustment prediction module connected in sequence, wherein the orbit prediction module is used to output an orbit prediction value according to the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs the satellite adjustment attitude according to the orbit prediction value;

[0008] The beam pointing of the communication satellite is adjusted according to the satellite adjustment attitude.

[0009] In one embodiment, before the step of establishing a big data satellite frequency interference avoidance system and collecting the current orbit information and frequency information of the communication satellite through the system, the following steps are further included:

[0010] Simulate and build a digital twin model corresponding to the satellite system;

[0011] Acquiring communication link data of the satellite system in a normal scenario and an interference scenario;

[0012] Based on the communication link data, the digital twin model is trained through an AI algorithm to obtain a preset interference scenario prediction model.

[0013] In one embodiment, the specific training process of the preset interference scenario prediction model includes:

[0014] Step 1: Standardize the communication link data of the satellite system in normal scenarios and interference scenarios recorded in history to obtain a standardized data set;

[0015] Step 2: The digital twin model construction includes a trajectory prediction module and an attitude adjustment prediction module connected in sequence;

[0016] Step 3: Train the trajectory prediction module based on the normal scene dataset and calculate the loss function of the trajectory prediction module until the loss function converges to the preset value;

[0017] Step 4: Train the posture adjustment prediction module according to the interference scene dataset until the target accuracy is achieved, then stop the training process and obtain the preset interference scene prediction model.

[0018] In one embodiment, the standardization process includes:

[0019] For the raw data x in the communication link data, calculate the mean μ and standard deviation σ of the communication link data; wherein the raw data includes the eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate of each communication satellite in the satellite system; the mean is the average value of the same type of data in each communication satellite, including the mean eccentricity, the mean orbital inclination, the mean right ascension of the ascending node, the mean argument of perigee, the mean true anomaly, the mean signal interference intensity, the mean signal interference frequency, and the mean bit error rate; the standard deviation is a statistical value that measures the degree of dispersion of the same type of data, indicating the degree of deviation of each data from the mean, including the standard deviation of eccentricity, the standard deviation of orbital inclination, the standard deviation of right ascension of the ascending node, the standard deviation of argument of perigee, the standard deviation of true anomaly, the standard deviation of signal interference intensity, the standard deviation of signal interference frequency, and the standard deviation of bit error rate;

[0020] Subtract the mean μ from the original data x and divide it by the standard deviation σ to obtain the standardized data point z. The calculation formula is as follows:

[0021]

[0022] Where x is the original data in the communication link data, μ is the mean of the communication link data, and σ is the standard deviation of the communication link data;

[0023] The standardized dataset includes a normal scene dataset and an interference scene dataset, and both the normal scene dataset and the interference scene dataset are divided into a training set and a test set in a ratio of 2:1.

[0024] In one embodiment, the training process of the trajectory prediction module includes:

[0025] Step 3.1: Train the track prediction module based on the training set of the normal scene dataset and calculate the loss function of the track prediction module;

[0026] Step 3.2: Test and train the trained model based on the test set of the normal scene dataset until the loss function converges to the preset value;

[0027] Among them, the loss function is:

[0028] θ1=θ SE +θ PE +θ C ;

[0029] Among them, θ1 is the loss function value, θ SE is the root mean square error (RMS) between the predicted value and the actual value, θ PE is the average relative percentage error between the predicted value and the actual value; θ C The coefficient of determination is a measure of the degree of correlation between the predicted value and the actual value. The larger the value, the higher the degree of correlation and the better the prediction effect.

[0030] In one embodiment, θ SE ,θ PE and θ C The specific formula is as follows:

[0031]

[0032] Where a represents the number of samples in the training set; z' i is the orbit prediction value, x i is the actual value in the training set, μ i is the average of the actual values.

[0033] In one embodiment, the training process of the posture adjustment prediction module includes:

[0034] Step 4.1: Using T re represents the training time of the posture adjustment prediction module, F represents the total number of training rounds of the posture adjustment prediction module, V represents the total number of target data aggregated by the posture adjustment prediction module, Indicates every The wheel posture adjustment prediction module aggregates the target data once, and then:

[0035] Step 4.2: Select the optimal aggregation frequency through K-Means clustering algorithm;

[0036] Step 4.3: At training time T re The posture adjustment prediction module is based on the optimal aggregation frequency Towards target data training aggregation, if the training time T is reached re , then proceed to step 4.4, otherwise go to step 4.2;

[0037] Step 4.4: Aggregate the model weights through the FedProx algorithm. If the target accuracy is achieved, stop the training process and obtain the preset interference scenario prediction model. Otherwise, go to step 4.2.

[0038] In one embodiment, the optimal aggregation frequency The steps include:

[0039] Step 4.2.1: Establish the time consumption constraint of the posture adjustment prediction module. The expression of the time consumption constraint is:

[0040]

[0041] in, It represents the time it takes for the posture adjustment prediction module to train the interference scene dataset for the nth round. represents the time it takes for the posture adjustment prediction module to achieve aggregation in the nth round, T re Indicates the training time,

[0042] Step 4.2.2: Use the K-Means clustering algorithm to find the connection between the posture adjustment prediction module and the time consumption constraint, and determine the optimal aggregation frequency

[0043] In addition, to achieve the above objectives, this application also proposes a system that utilizes AI and digital twins to avoid frequency interference, the system comprising:

[0044] The information collection module is used to establish a big data satellite frequency interference avoidance system and collect the current orbit and frequency information of the communication satellite through the system;

[0045] A model prediction module is used to simulate the orbit information and the frequency information through a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite, wherein the preset interference scenario prediction model is obtained by training the normal scenario and interference scenario of the satellite system based on the AI algorithm and the digital twin model, and the digital twin model is constructed including an orbit prediction module and an attitude adjustment prediction module connected in sequence, wherein the orbit prediction module is used to output an orbit prediction value based on the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs a satellite adjustment attitude based on the orbit prediction value;

[0046] The beam adjustment module is used to adjust the beam pointing of the communication satellite according to the satellite adjustment attitude.

[0047] One or more technical solutions proposed in this application have at least the following technical effects: this application first establishes a big data satellite frequency interference avoidance system, and collects the current orbit information and frequency information of the communication satellite through the system; then simulates the orbit information and the frequency information through a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite. The preset interference scenario prediction model is obtained by training the normal scenario and interference scenario of the satellite system based on the AI algorithm and the digital twin model. The digital twin model construction includes an orbit prediction module and an attitude adjustment prediction module connected in sequence. The orbit prediction module is used to output an orbit prediction value according to the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs the satellite adjustment attitude according to the orbit prediction value; finally, the beam pointing of the communication satellite is adjusted according to the satellite adjustment attitude. Since this application introduces a preset interference scenario prediction model to pre-predict the current orbit prediction value and satellite adjustment attitude of the communication satellite, the interference scenario of the communication satellite is predicted in advance, and the beam pointing of the communication satellite is adjusted. By identifying potential frequency interference situations, it can avoid the problem that traditional communication satellites are shut down in the shutdown area to avoid frequency interference, affecting the tracking accuracy and stability of the communication satellite, thereby realizing automatic early warning and processing of frequency interference, adaptively reducing frequency interference, and ensuring the normal operation of the communication satellite. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0050] Figure 1 A flowchart illustrating a method for avoiding frequency interference by combining AI and digital twins is provided in Example 1 of this application;

[0051] Figure 2 A flowchart illustrating a method for avoiding frequency interference by combining AI and digital twins is provided in Example 2 of this application;

[0052] Figure 3 This is a schematic diagram of the module structure of a system that uses AI and digital twins to avoid frequency interference in an embodiment of the present application.

[0053] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0055] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0056] It should be noted that the execution subject of this embodiment can be a communication satellite, ground measurement and control equipment, etc., or an electronic device that can realize the above functions, a device that uses AI and digital twins to avoid frequency interference (referred to as anti-interference equipment), etc., and this embodiment does not limit this. The following takes the anti-interference equipment as an example to illustrate this embodiment and the following embodiments.

[0057] Based on this, the embodiment of the present application provides a method for avoiding frequency interference by combining AI with digital twins. Figure 1 , Figure 1 This is a flowchart of a method for avoiding frequency interference by combining AI and digital twins according to the present application.

[0058] In this embodiment, the method of avoiding frequency interference by combining AI with digital twins includes steps S10 to S30:

[0059] Step S10: Establish a big data satellite frequency interference avoidance system, and collect the current orbit information and frequency information of the communication satellite through the system.

[0060] It should be noted that the big data satellite frequency interference avoidance system is a system that uses big data technology to deal with satellite frequency interference problems.

[0061] Specifically, this system can be built by collecting information on frequency usage, signal characteristics, and orbital parameters of different satellites. This data can be used for comprehensive analysis and modeling, leading to the construction of a big data satellite frequency interference avoidance system. This system monitors and identifies potential frequency interference in real time, providing early warnings and implementing appropriate measures to avoid or mitigate the impact. This can include adjusting satellite frequency allocations and optimizing communication strategies, thereby ensuring the normal operation and service quality of satellite communication systems and minimizing the adverse consequences of interference.

[0062] It should be noted that a communications satellite is an artificial satellite in space that functions as a communications relay. It receives signals from earth stations (communication sites on the ground, at sea, or in the air), amplifies and frequency-converts the signals, and then forwards them to other earth stations, enabling long-distance communication.

[0063] It should be noted that orbital information refers to the specific orbital parameters of a satellite in space or in low-Earth orbit, including orbital altitude, orbital shape (such as circular or elliptical), and orbital inclination. Orbital information describes the satellite's path and position relative to the Earth.

[0064] The orbit information may be determined based on data such as the semi-major axis, eccentricity, orbit inclination, and right ascension of the ascending node of the satellite, and this embodiment does not impose any limitation thereto.

[0065] It should be noted that frequency information refers to the radio frequency range used by communication satellites for signal transmission. Different communication services generally use specific frequency bands to avoid mutual interference. However, there are also situations where shared communication frequencies are necessary, and the probability of harmful interference is extremely high. Therefore, frequency information is necessary to determine whether interference exists.

[0066] In one embodiment, the anti-interference device can pre-collect information such as frequency usage, signal characteristics, and orbital parameters of different satellites to establish a big data satellite frequency interference avoidance system. The system can collect current orbital information of communication satellites, such as orbital altitude, orbital shape (e.g., circular, elliptical, etc.), orbital inclination, and frequency information. This frequency information can be used to determine whether interference exists. If so, orbital information can be used to plan routes to avoid mutual interference.

[0067] Step S20: The orbital information and the frequency information are simulated through a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite. The preset interference scenario prediction model is obtained by training the normal scenario and interference scenario of the satellite system based on the AI algorithm and the digital twin model. The digital twin model construction includes an orbit prediction module and an attitude adjustment prediction module connected in sequence. The orbit prediction module is used to output an orbit prediction value according to the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs the satellite adjustment attitude according to the orbit prediction value.

[0068] It should be noted that the preset interference scenario prediction model is to first digitally simulate the satellite system to form a virtual digital twin model corresponding to the real satellite system, and then use the AI algorithm to learn and train this digital twin model to obtain the model.

[0069] It should be noted that in the process of learning and training the digital twin model, the communication link data of the satellite system in the interference scenario can be used as the core for training and learning, so that the interference situation of the communication satellite can be quickly determined through the preset interference scenario prediction model.

[0070] It is understandable that the preset interference scenario prediction model can analyze the frequency information of the communication satellite based on the knowledge and patterns obtained through training to determine whether there is a potential interference scenario; if there is a potential interference situation, the orbital information is used to predict the latitude angle of the communication satellite in the future period of time to quickly determine whether the communication satellite is obviously in the interference area in the future period of time.

[0071] It should be noted that the interference zone is an area within the coverage of the beam transmitted by the communication satellite where signal transmission is adversely affected due to various reasons (such as co-channel interference, interference from other signal sources, etc.), resulting in reduced communication quality or inability to communicate normally.

[0072] The interference zone can be determined by interference information from other satellite signals, ground radio equipment, electromagnetic radiation, etc. Signal attenuation, distortion, increased bit error rate, etc. may occur in this area, which will affect the normal operation and services of the communication satellite system.

[0073] It's important to note that satellite attitude adjustment refers to changes in a satellite's position, orientation, or rotation in space. When encountering interference while operating in space, satellites must adjust their attitude to maintain stable communication signals and precise targeting. This can be achieved by controlling rotation, pitch, and yaw angles.

[0074] In one implementation, the anti-interference device can pre-digitize and simulate the satellite system, creating a virtual digital twin model corresponding to the actual satellite system. This digital twin model is then trained using an AI algorithm, centered around the satellite system's communication link data during interference scenarios, to develop a pre-defined interference scenario prediction model. After obtaining orbital and frequency information, the pre-defined interference scenario prediction model can be used to simulate and analyze the frequency information of the communication satellite to determine whether a potential interference scenario exists. If so, the satellite's attitude adjustment is output.

[0075] Step S30: adjusting the beam pointing direction of the communication satellite according to the satellite adjustment attitude.

[0076] For example, satellite attitude adjustment may include the specific angle of satellite attitude adjustment, the value of the direction change, orbit fine-tuning data, etc. This allows the communication satellite to move or adjust along a specific attitude change path, preventing interference from sources such as beam interference areas, thereby ensuring normal and stable communication of the communication satellite.

[0077] It should be noted that beam pointing is the direction in which the signal beam transmitted by the communication satellite is focused.

[0078] In one embodiment, when a communication satellite is in an interference area, the anti-interference device can adjust the beam pointing of the communication satellite by adjusting the satellite's attitude, and plan the attitude angle of the communication satellite with the goal of more stable attitude maneuvering, so that the beam pointing of the communication satellite avoids the beam interference area and accurately points to the target receiving area, ensuring that the signal can be effectively delivered to the location that needs to be covered, thereby ensuring the communication service of the communication satellite.

[0079] In this embodiment, the anti-interference equipment can pre-collect information such as the frequency usage, signal characteristics, and orbital parameters of different satellites to establish a big data satellite frequency interference avoidance system. The satellite system is then digitally simulated and constructed in advance to form a virtual digital twin model corresponding to the real satellite system. Using AI algorithms, with the communication link data of the satellite system in the interference scenario as the core, this digital twin model is studied and trained to obtain a preset interference scenario prediction model. Afterwards, the system can collect the current orbital information of the communication satellite, such as orbital altitude, orbital shape (such as circular, elliptical, etc.), orbital inclination, etc., as well as frequency information. After obtaining the orbital information and frequency information, the preset interference scenario prediction model can be used to simulate and analyze the frequency information of the communication satellite to determine whether there is a potential interference scenario; if there is a potential interference situation, the satellite adjustment attitude is output. When a communication satellite is in an interference area, the anti-interference equipment can adjust the beam pointing of the communication satellite by adjusting the satellite's attitude. With the goal of more stable attitude maneuvering, the attitude angle of the communication satellite is planned so that the beam pointing of the communication satellite avoids the beam interference area and accurately points to the target receiving area, ensuring that the signal can be effectively delivered to the location that needs to be covered, thereby guaranteeing the communication service of the communication satellite.

[0080] In a feasible implementation, step S10 of this embodiment may include the steps of: establishing a big data satellite frequency interference avoidance system; collecting the current semi-major axis, eccentricity, orbital inclination, ascending node right ascension, perigee argument and true anomaly of the communication satellite through the system to determine the current orbit information of the communication satellite; collecting the current antenna pointing and receiving power of the communication satellite through the system to determine the current frequency information of the communication satellite.

[0081] It should be noted that the semi-major axis is half of the major axis of the elliptical orbit and determines the size of the communication satellite orbit.

[0082] It should be noted that the eccentricity describes the degree of oblateness of the elliptical orbit of the communication satellite. The greater the eccentricity, the flatter the orbit.

[0083] It should be noted that orbital inclination is the angle between the satellite orbit plane and the Earth's equatorial plane, which determines the degree of inclination of the satellite orbit relative to the Earth.

[0084] It should be noted that the right ascension of the ascending node is the angle measured eastward from the vernal equinox along the equator to the ascending node, which is used to determine the position of the satellite orbit on the equatorial plane.

[0085] It should be noted that the argument of perigee is the angle measured from the ascending node along the direction of satellite motion to the perigee, which describes the shape and orientation of the orbit.

[0086] It should be noted that the true anomaly angle is the angle measured from the perigee along the direction of satellite motion to the satellite's position, which is used to specifically determine the satellite's position in orbit.

[0087] Specifically, the semi-major axis determines the size and extent of the orbit; the eccentricity reflects the orbit's shape; the orbital inclination indicates the tilt of the satellite's orbit relative to the Earth's equatorial plane; the right ascension of the ascending node determines the orbit's position on the equatorial plane; the argument of perigee further refines the orbit's orientation; and the true anomaly precisely indicates the satellite's exact position in orbit. These parameters can accurately describe the current orbital state and position of a communications satellite.

[0088] It should be noted that antenna pointing refers to the direction or target location that the satellite antenna is pointed at. Antenna pointing can be used to understand the relative position between the satellite and the ground receiving station.

[0089] It should be noted that received power refers to the power of the signal from the communication satellite received by the ground receiving end. The transmission effect of received power can be used to determine whether there is signal attenuation.

[0090] In this embodiment, the system can collect the current semi-major axis, eccentricity, orbital inclination, ascending node right ascension, perigee argument, true anomaly, antenna pointing and receiving power of the communication satellite, so that the current orbital information and frequency information of the communication satellite can be accurately understood through the above parameters.

[0091] This embodiment provides a method for avoiding frequency interference by combining AI with digital twins. The anti-interference equipment can pre-collect information such as frequency usage, signal characteristics, and orbital parameters of different satellites to establish a big data satellite frequency interference avoidance system. The satellite system is then digitally simulated and constructed in advance to form a virtual digital twin model corresponding to the real satellite system. Using an AI algorithm, with the communication link data of the satellite system in the interference scenario as the core, this digital twin model is learned and trained to obtain a preset interference scenario prediction model. Afterwards, the system can collect the current orbital information of the communication satellite, such as orbital altitude, orbital shape (such as circular, elliptical, etc.), orbital inclination, etc., as well as frequency information. After obtaining the orbital information and frequency information, the frequency information of the communication satellite can be simulated and analyzed using the preset interference scenario prediction model to determine whether there is a potential interference scenario; if there is a potential interference situation, the satellite adjustment attitude is output. When a communication satellite is in an interference area, the anti-interference device can adjust the beam pointing of the communication satellite by adjusting the satellite's attitude, and plan the attitude angle of the communication satellite with the goal of more stable attitude maneuvering, so that the beam pointing of the communication satellite avoids the beam interference area and accurately points to the target receiving area, ensuring that the signal can be effectively delivered to the location that needs to be covered, and guaranteeing the communication service of the communication satellite. Since the present application introduces a preset interference scenario prediction model to pre-predict the current orbit prediction value of the communication satellite and the satellite adjustment attitude, the interference scenario of the communication satellite is predicted in advance, and the beam pointing of the communication satellite is adjusted. By identifying potential frequency interference situations, it can avoid the problem that traditional communication satellites are shut down in the shutdown area to avoid frequency interference, affecting the tracking accuracy and stability of the communication satellite, thereby realizing automatic early warning and processing of frequency interference, adaptively reducing frequency interference, and ensuring the normal operation of the communication satellite.

[0092] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 1 and Figure 2 , Figure 2 This is a flow chart of Example 2 of a method for avoiding frequency interference by combining AI with digital twins.

[0093] Before step S10, this example also includes steps S01 to S03:

[0094] Step S01: Simulate and build a digital twin model corresponding to the satellite system.

[0095] It should be noted that the digital twin model is a virtual model corresponding to the real satellite system through the digital simulation of the satellite system. The digital twin model can reflect the status and behavior of the physical entities of the simulated satellite system in real time.

[0096] Step S02: Acquire communication link data of the satellite system in a normal scenario and an interference scenario.

[0097] It should be noted that the communication link data under normal scenarios are various data generated during the communication between the satellite and the ground station or other satellites when there is no abnormal interference or influence from the outside world, such as signal strength, transmission rate, bit error rate, delay and other data.

[0098] Communication link data under interference scenarios is presented when conditions such as electromagnetic interference, human interference, and inter-satellite interference are present. This data differs from normal data and may show changes such as signal weakening, decreased transmission rate, increased bit error rate, and increased latency. By analyzing communication link data under interference scenarios, we can understand the specific extent and method of interference impacting satellite communications.

[0099] Step S03: Based on the communication link data, the digital twin model is trained through an AI algorithm to obtain a preset interference scenario prediction model.

[0100] It should be noted that AI algorithm, or artificial intelligence algorithm, enables the digital twin model to intelligently simulate and learn communication link data.

[0101] Exemplarily, the AI algorithm may include a machine learning algorithm, a federated proximal algorithm, a natural language processing algorithm, etc., which is not limited in this embodiment.

[0102] In this implementation, a digital simulation of the satellite system is first constructed to create a virtual digital twin model corresponding to the actual satellite system. Communication link data for the satellite system in both normal and interference scenarios is then acquired. Finally, based on this communication link data, the digital twin model is trained using an AI algorithm to obtain a pre-set interference scenario prediction model. This pre-set interference scenario prediction model can then be used to rapidly simulate and predict the latitude and angle of a communication satellite within an interference zone over a period of time.

[0103] The specific training process of the preset interference scenario prediction model includes:

[0104] Step 1: Standardize the communication link data of the satellite system in historically recorded normal and interference scenarios to obtain a standardized data set to unify the data scale and convert data of different scales to the same scale.

[0105] Since the communication link data is continuous, Z-score normalization can be used to convert each data in the communication link data into a multiple of the standard deviation of its mean to eliminate the scale differences between different data.

[0106] Step 1.1: The standardization process includes:

[0107] For the raw data x in the communication link data, first calculate the mean μ and standard deviation σ of the communication link data. The raw data includes the eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate of each communication satellite in the satellite system. The mean is the average value of the same data type across all communication satellites, including the mean eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate. The standard deviation is a statistical measure of the dispersion of the same data type, indicating the degree of deviation of each data type from the mean, including the standard deviation of eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate. Using these mean and standard deviation, we can eliminate the scale differences between the same data types in the raw data across communication satellites.

[0108] Then subtract the mean μ from the original data x and divide it by the standard deviation σ to get the standardized data point z. The calculation formula is as follows:

[0109]

[0110] Where x is the original data in the communication link data, μ is the mean of the communication link data, and σ is the standard deviation of the communication link data.

[0111] Through such an operation, the original data in the communication link data is converted into standard normal distribution data with a mean of 0 and a standard deviation of 1, thereby achieving data standardization.

[0112] Step 1.2: The standardized dataset includes a normal scene dataset and an interference scene dataset. Both the normal scene dataset and the interference scene dataset are divided into a training set and a test set in a ratio of 2:1.

[0113] Among them, the normal scene data set is various data generated during the communication between the satellite and the ground station or other satellites when there is no abnormal external interference, such as signal strength, transmission rate, orbit and satellite attitude data.

[0114] Interference scenario datasets are generated when satellite communications are subject to interference from various sources. This interference can cause signal attenuation and reduced communication quality. The collected data reflects the specific impact of the interference on the communication link, including signal strength reduction, bit error rate increase, adjusted orbit, and satellite attitude data.

[0115] Step 2: The digital twin model construction includes a trajectory prediction module and an attitude adjustment prediction module connected in sequence.

[0116] The orbit prediction module includes a sequentially connected CNN layer, multiple LSTM layers, and multiple fully connected layers. The CNN layer includes multiple sequentially connected convolutional layers and pooling layers. The orbit prediction module inputs the previous orbit and frequency information of the communication satellite, and the orbit prediction module outputs an orbit prediction value (i.e., predicting the satellite orbit for a period of time in the future to determine whether it is in an interference scenario).

[0117] The attitude adjustment module consists of a sequentially connected regularization layer, multiple residual ResNet units, a bidirectional LSTM unit, and a fully connected unit. The residual ResNet unit consists of two sequentially connected convolutional layers. The residual unit extracts features from the satellite's current orbit and frequency information, performs predictions in the bidirectional LSTM unit and the fully connected unit, and obtains the final predicted value (i.e., the satellite's attitude adjustment for a period of time in the future).

[0118] Step 3: Train the trajectory prediction module based on the normal scene dataset and calculate the loss function of the trajectory prediction module until the loss function converges to the preset value.

[0119] The training process of the orbit prediction module includes:

[0120] Step 3.1: Train the track prediction module based on the training set of the normal scene dataset and calculate the loss function of the track prediction module;

[0121] Step 3.2: Then test the trained model based on the test set of the normal scene dataset until the loss function converges to the preset value.

[0122] Among them, the loss function is:

[0123] θ1=θ SE +θ PE +θ C ;

[0124] Among them, θ1 is the loss function value, θ SE is the root mean square error (RMS) between the predicted value and the actual value, θ PE is the average relative percentage error between the predicted value and the actual value. C The coefficient of determination is a measure of the degree of correlation between the predicted value and the actual value. The larger the value, the higher the degree of correlation and the better the prediction effect.

[0125] Among them, θ SE ,θ PE and θ CThe specific formula is as follows:

[0126]

[0127]

[0128] Where a represents the number of samples in the training set; z' i is the orbit prediction value, x i is the actual value in the training set, μ i is the average of the actual values.

[0129] Step 4: Train the posture adjustment prediction module according to the interference scene dataset until the target accuracy is achieved, then stop the training process and obtain the preset interference scene prediction model.

[0130] The training process of the posture adjustment prediction module includes:

[0131] Step 4.1: First use T re represents the training time of the attitude adjustment prediction module, F represents the total number of training rounds of the attitude adjustment prediction module, V represents the total number of target data aggregated by the attitude adjustment prediction module (i.e., the data of the orbit or satellite attitude adjusted in the interference scene dataset aggregated by the model), Indicates every The wheel posture adjustment prediction module aggregates the target data once, and then:

[0132] Step 4.2: Select the optimal aggregation frequency through K-Means Clustering Algorithm

[0133] Among them, the optimal aggregation frequency The steps include:

[0134] Step 4.2.1: First, a time consumption constraint condition for the posture adjustment prediction module can be established. The expression of the time consumption constraint condition is:

[0135]

[0136] in, It represents the time it takes for the posture adjustment prediction module to train the interference scene dataset for the nth round. represents the time it takes for the posture adjustment prediction module to achieve aggregation in the nth round, T re Indicates the training time,

[0137] Step 4.2.2: Then use the K-Means clustering algorithm to find the connection between the posture adjustment prediction module and the time consumption constraint, and determine the optimal aggregation frequency

[0138] Among them, the K-Means clustering aggregation algorithm is an unsupervised learning algorithm, which aims to continuously adjust the attribution of data points and the position of cluster centers so that data points in the same cluster have high similarity, while data points between different clusters have large differences.

[0139] Step 4.3: At training time T re The posture adjustment prediction module is based on the optimal aggregation frequency Towards target data training aggregation, if the training time T is reached re , then proceed to step 4.4, otherwise go to step 4.2.

[0140] Step 4.4: Aggregate the model weights using the FedProx algorithm (Federated Proximal Algorithm). If the target accuracy is achieved, stop the training process and obtain the preset interference scenario prediction model. Otherwise, go to step 4.2.

[0141] Among them, the federated proximal algorithm is an optimization algorithm used in federated learning scenarios. Since communication link data is distributed across multiple different nodes or devices and the data cannot be centralized, in federated learning, this algorithm can solve the problem of how to effectively train and optimize the model in such a distributed environment to update and optimize the model parameters and achieve better learning results.

[0142] By continuously aggregating and learning the data of the adjusted operating orbit or satellite attitude in the interference scenario dataset, the attitude adjustment prediction module can determine whether it is an interference scenario based on the predicted operating orbit output by the orbit prediction module, and then predict the satellite attitude adjustment of the communication satellite.

[0143] This embodiment first digitally simulates the satellite system to create a virtual digital twin model corresponding to the actual satellite system. Communication link data for the satellite system in both normal and interference scenarios is then acquired. Finally, based on this communication link data, the digital twin model is trained using an AI algorithm to obtain a preset interference scenario prediction model. This preset interference scenario prediction model can then be used to rapidly simulate and predict the orbit of the communication satellite over a period of time. This ultimately trained preset interference scenario prediction model can quickly identify potential interference scenarios for the communication satellite.

[0144] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the present application's method of avoiding frequency interference by combining AI with digital twins. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0145] This application also provides a system that uses AI and digital twins to avoid frequency interference. Please refer to Figure 3 , Figure 3 This is a schematic diagram of the module structure of a system for avoiding frequency interference by combining AI with digital twins according to an embodiment of the present application; the system for avoiding frequency interference by combining AI with digital twins includes:

[0146] Information collection module 401 is used to establish a big data satellite frequency interference avoidance system and collect the current orbit information and frequency information of the communication satellite through the system;

[0147] A model prediction module 402 is configured to simulate the orbit information and the frequency information using a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite. The preset interference scenario prediction model is obtained by training the normal scenario and interference scenario of the satellite system based on an AI algorithm and a digital twin model. The digital twin model construction includes an orbit prediction module and an attitude adjustment prediction module connected in sequence. The orbit prediction module is configured to output an orbit prediction value based on the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs a satellite adjustment attitude based on the orbit prediction value.

[0148] The beam adjustment module 403 is configured to adjust the beam pointing direction of the communication satellite according to the satellite's attitude adjustment.

[0149] Furthermore, the system for avoiding frequency interference by combining AI with digital twins also includes a model construction module 404, which is used to simulate the construction of a digital twin model corresponding to the satellite system; obtain the communication link data of the satellite system in normal scenarios and interference scenarios; based on the communication link data, the digital twin model is trained through an AI algorithm to obtain a preset interference scenario prediction model.

[0150] Furthermore, in the model building module 404, the specific training process of the preset interference scenario prediction model includes:

[0151] Step 1: Standardize the communication link data of the satellite system in normal scenarios and interference scenarios recorded in history to obtain a standardized data set;

[0152] Step 2: The digital twin model construction includes a trajectory prediction module and an attitude adjustment prediction module connected in sequence;

[0153] Step 3: Train the trajectory prediction module based on the normal scene dataset and calculate the loss function of the trajectory prediction module until the loss function converges to the preset value;

[0154] Step 4: Train the posture adjustment prediction module according to the interference scene dataset until the target accuracy is achieved, then stop the training process and obtain the preset interference scene prediction model.

[0155] Furthermore, in the model building module 404, the standardization process includes:

[0156] For the raw data x in the communication link data, calculate the mean μ and standard deviation σ of the communication link data; wherein the raw data includes the eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate of each communication satellite in the satellite system; the mean is the average value of the same type of data in each communication satellite, including the mean eccentricity, the mean orbital inclination, the mean right ascension of the ascending node, the mean argument of perigee, the mean true anomaly, the mean signal interference intensity, the mean signal interference frequency, and the mean bit error rate; the standard deviation is a statistical value that measures the degree of dispersion of the same type of data, indicating the degree of deviation of each data from the mean, including the standard deviation of eccentricity, the standard deviation of orbital inclination, the standard deviation of right ascension of the ascending node, the standard deviation of argument of perigee, the standard deviation of true anomaly, the standard deviation of signal interference intensity, the standard deviation of signal interference frequency, and the standard deviation of bit error rate;

[0157] Subtract the mean μ from the original data x and divide it by the standard deviation σ to obtain the standardized data point z. The calculation formula is as follows:

[0158]

[0159] Where x is the original data in the communication link data, μ is the mean of the communication link data, and σ is the standard deviation of the communication link data;

[0160] The standardized dataset includes a normal scene dataset and an interference scene dataset, and both the normal scene dataset and the interference scene dataset are divided into a training set and a test set in a ratio of 2:1.

[0161] Furthermore, in the model building module 404, the training process of the orbit prediction module includes:

[0162] Step 3.1: Train the track prediction module based on the training set of the normal scene dataset and calculate the loss function of the track prediction module;

[0163] Step 3.2: Test and train the trained model based on the test set of the normal scene dataset until the loss function converges to the preset value;

[0164] Among them, the loss function is:

[0165] θ1=θ SE +θ PE +θ C ;

[0166] Among them, θ1 is the loss function value, θ SE is the root mean square error (RMS) between the predicted value and the actual value, θ PE is the average relative percentage error between the predicted value and the actual value; θ C The coefficient of determination is a measure of the degree of correlation between the predicted value and the actual value. The larger the value, the higher the degree of correlation and the better the prediction effect.

[0167] Furthermore, in the model building module 404, θ SE ,θ PE and θ C The specific formula is as follows:

[0168]

[0169] Where a represents the number of samples in the training set; z' i is the orbit prediction value, x i is the actual value in the training set, μ i is the average of the actual values.

[0170] Furthermore, in the model building module 404, the training process of the posture adjustment prediction module includes:

[0171] Step 4.1: Using T re represents the training time of the posture adjustment prediction module, F represents the total number of training rounds of the posture adjustment prediction module, V represents the total number of target data aggregated by the posture adjustment prediction module, Indicates every The wheel posture adjustment prediction module aggregates the target data once, and then:

[0172] Step 4.2: Select the optimal aggregation frequency through K-Means clustering algorithm;

[0173] Step 4.3: At training time T re The posture adjustment prediction module is based on the optimal aggregation frequency Towards target data training aggregation, if the training time T is reached re , then proceed to step 4.4, otherwise go to step 4.2;

[0174] Step 4.4: Aggregate the model weights through the FedProx algorithm. If the target accuracy is achieved, stop the training process and obtain the preset interference scenario prediction model. Otherwise, go to step 4.2.

[0175] Furthermore, in the model building module 404, the optimal aggregation frequency The steps include:

[0176] Step 4.2.1: Establish the time consumption constraint of the posture adjustment prediction module. The expression of the time consumption constraint is:

[0177]

[0178] in, It represents the time it takes for the posture adjustment prediction module to train the interference scene dataset for the nth round. represents the time it takes for the posture adjustment prediction module to achieve aggregation in the nth round, T re Indicates the training time,

[0179] Step 4.2.2: Use the K-Means clustering algorithm to find the connection between the posture adjustment prediction module and the time consumption constraint, and determine the optimal aggregation frequency

[0180] Other embodiments or specific implementation methods of the system of the present application that utilizes a combination of AI and digital twins to avoid frequency interference can refer to the above-mentioned method embodiments and will not be repeated here.

[0181] The present application provides a system that utilizes a combination of AI and digital twins to avoid frequency interference. The method of utilizing a combination of AI and digital twins to avoid frequency interference in the above-mentioned embodiment is adopted to solve the technical problem that when a communication satellite is shut down in a shutdown area to avoid frequency interference, it will affect the normal operation of the communication satellite. At the same time, the shutdown of the communication satellite will also affect the tracking accuracy and stability of the communication satellite. Compared with the prior art, the beneficial effects of the system that utilizes a combination of AI and digital twins to avoid frequency interference provided by the present application are the same as the beneficial effects of the method of utilizing a combination of AI and digital twins to avoid frequency interference provided by the above-mentioned embodiment, and other technical features of the system that utilizes a combination of AI and digital twins to avoid frequency interference are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0182] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, and the computer-readable program instructions are used to execute a method in the above-mentioned embodiment of combining AI with digital twins to avoid frequency interference.

[0183] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0184] The above-mentioned computer-readable storage medium may be included in a device that utilizes a combination of AI and digital twins to avoid frequency interference; or it may exist independently without being assembled into a device that utilizes a combination of AI and digital twins to avoid frequency interference.

[0185] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by a device that uses AI and digital twins to avoid frequency interference, the device that uses AI and digital twins to avoid frequency interference: establishes a big data satellite frequency interference avoidance system, and collects the current orbit information and frequency information of the communication satellite through the system; simulates the orbit information and the frequency information through a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite. The preset interference scenario prediction model is obtained by training the normal scenario and interference scenario of the satellite system based on the AI algorithm and the digital twin model. The digital twin model construction includes an orbit prediction module and an attitude adjustment prediction module connected in sequence. The orbit prediction module is used to output an orbit prediction value according to the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs a satellite adjustment attitude according to the orbit prediction value; adjusts the beam pointing of the communication satellite according to the satellite adjustment attitude.

[0186] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0187] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of code, and the module, program segment or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0188] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0189] The readable storage medium provided by this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., a computer program) for executing the above-mentioned method of avoiding frequency interference by combining AI with digital twins. It can solve the technical problem that when a communication satellite is shut down in a shutdown area to avoid frequency interference, it will affect the normal business of the communication satellite. At the same time, the shutdown of the communication satellite will also affect the tracking accuracy and stability of the communication satellite. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by this application are the same as the beneficial effects of the method of avoiding frequency interference by combining AI with digital twins provided in the above-mentioned embodiment, and will not be repeated here.

[0190] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned method of avoiding frequency interference by combining AI with digital twins.

[0191] The computer program product provided in this application can address the technical issue of communications satellites shutting down in shutdown zones to avoid frequency interference, which can affect the normal operation of the communications satellite and, at the same time, the tracking accuracy and stability of the communications satellite due to the shutdown. Compared to the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the method for avoiding frequency interference using AI and digital twins provided in the above-mentioned embodiment, and are not further elaborated here.

[0192] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for avoiding frequency interference by combining AI with digital twins, characterized in that: The method includes: Establish a big data satellite frequency interference avoidance system, and use the system to collect the current orbit and frequency information of communication satellites; The orbital information and the frequency information are simulated by a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite, wherein the preset interference scenario prediction model is obtained by training normal scenarios and interference scenarios of the satellite system based on an AI algorithm and a digital twin model, and the digital twin model includes an orbit prediction module and an attitude adjustment prediction module connected in sequence, wherein the orbit prediction module is used to output an orbit prediction value according to the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs a satellite adjustment attitude according to the orbit prediction value; Adjusting the beam pointing of the communication satellite according to the satellite attitude adjustment; Wherein, before the step of establishing a big data satellite frequency interference avoidance system and collecting the current orbit information and frequency information of the communication satellite through the system, the method further includes: Simulate and build a digital twin model corresponding to the satellite system; Acquiring communication link data of the satellite system in a normal scenario and an interference scenario; Based on the communication link data, the digital twin model is trained using an AI algorithm to obtain a preset interference scenario prediction model; The training process of the orbit prediction module includes: Step 3.1: Train the track prediction module based on the training set of the normal scene dataset and calculate the loss function of the track prediction module; Step 3.2: Test and train the trained model based on the test set of the normal scene dataset until the loss function converges to the preset value; Among them, the loss function is: θ1=θ SE +θ PE +θ C ; Among them, θ1 is the loss function value, θ SE is the root mean square error (RMS) between the predicted value and the actual value, θ PE is the average relative percentage error between the predicted value and the actual value; θ C The coefficient of determination is a measure of the degree of correlation between the predicted value and the actual value. The larger the value, the higher the degree of correlation and the better the prediction effect. The training process of the posture adjustment prediction module includes: Step 4.1: Using T re represents the training time of the posture adjustment prediction module, F represents the total number of training rounds of the posture adjustment prediction module, V represents the total number of target data aggregated by the posture adjustment prediction module, Indicates every The wheel posture adjustment prediction module aggregates the target data once, and then: Step 4.2: Select the optimal aggregation frequency through K-Means clustering algorithm Step 4.3: At training time T re The posture adjustment prediction module is based on the optimal aggregation frequency Towards target data training aggregation, if the training time T is reached re , then proceed to step 4.4, otherwise go to step 4.2; Step 4.4: Aggregate the model weights through the FedProx algorithm. If the target accuracy is achieved, stop the training process and obtain the preset interference scenario prediction model. Otherwise, go to step 4.

2.

2. The method according to claim 1, wherein The specific training process of the preset interference scenario prediction model includes: Step 1: Standardize the communication link data of the satellite system in normal scenarios and interference scenarios recorded in history to obtain a standardized data set; Step 2: The digital twin model includes a trajectory prediction module and an attitude adjustment prediction module connected in sequence; Step 3: Train the trajectory prediction module based on the normal scene dataset and calculate the loss function of the trajectory prediction module until the loss function converges to the preset value; Step 4: Train the posture adjustment prediction module according to the interference scene dataset until the target accuracy is achieved, then stop the training process and obtain the preset interference scene prediction model.

3. The method according to claim 2, wherein The process of the standardization process includes: For the raw data x in the communication link data, calculate the mean μ and standard deviation σ of the communication link data; wherein the raw data includes the eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, true anomaly, signal interference intensity, signal interference frequency, and bit error rate of each communication satellite in the satellite system; the mean is the average value of the same type of data in each communication satellite, including the mean eccentricity, the mean orbital inclination, the mean right ascension of the ascending node, the mean argument of perigee, the mean true anomaly, the mean signal interference intensity, the mean signal interference frequency, and the mean bit error rate; the standard deviation is a statistical value that measures the degree of dispersion of the same type of data, indicating the degree of deviation of each data from the mean, including the standard deviation of eccentricity, the standard deviation of orbital inclination, the standard deviation of right ascension of the ascending node, the standard deviation of argument of perigee, the standard deviation of true anomaly, the standard deviation of signal interference intensity, the standard deviation of signal interference frequency, and the standard deviation of bit error rate; Subtract the mean μ from the original data x and divide it by the standard deviation σ to obtain the standardized data point z. The calculation formula is as follows: Where x is the original data in the communication link data, μ is the mean of the communication link data, and σ is the standard deviation of the communication link data; The standardized dataset includes a normal scene dataset and an interference scene dataset, and both the normal scene dataset and the interference scene dataset are divided into a training set and a test set in a ratio of 2:

1.

4. The method according to claim 1, wherein θ SE ,θ PE and θ C The specific formula is as follows: Where a represents the number of samples in the training set; z' i is the orbit prediction value, x i is the actual value in the training set, μ i is the average of the actual values.

5. The method according to claim 1, wherein The step 4.2 includes: Step 4.2.1: Establish the time consumption constraint of the posture adjustment prediction module. The expression of the time consumption constraint is: in, It represents the time it takes for the posture adjustment prediction module to train the interference scene dataset for the nth round. represents the time it takes for the posture adjustment prediction module to achieve aggregation in the nth round, T re Indicates the training time, Step 4.2.2: Use the K-Means clustering algorithm to find the connection between the posture adjustment prediction module and the time consumption constraint, and determine the optimal aggregation frequency 6. A system that uses AI and digital twins to avoid frequency interference, characterized in that: The system comprises: The information collection module is used to establish a big data satellite frequency interference avoidance system and collect the current orbit and frequency information of the communication satellite through the system; A model prediction module is used to simulate the orbit information and the frequency information through a preset interference scenario prediction model to determine the satellite adjustment attitude of the communication satellite, wherein the preset interference scenario prediction model is obtained by training normal scenarios and interference scenarios of the satellite system based on an AI algorithm and a digital twin model, and the digital twin model includes an orbit prediction module and an attitude adjustment prediction module connected in sequence, the orbit prediction module is used to output an orbit prediction value based on the orbit information and frequency information of the communication satellite, and the attitude adjustment prediction module outputs a satellite adjustment attitude based on the orbit prediction value; A beam adjustment module, configured to adjust the beam pointing direction of the communication satellite according to the satellite adjustment attitude; A model building module is used to simulate and build a digital twin model corresponding to the satellite system; obtain communication link data of the satellite system in normal scenarios and interference scenarios; based on the communication link data, train the digital twin model through an AI algorithm to obtain a preset interference scenario prediction model; Among them, in the model construction module, the training process of the orbit prediction module includes: Step 3.1: Train the track prediction module based on the training set of the normal scene dataset and calculate the loss function of the track prediction module; Step 3.2: Test and train the trained model based on the test set of the normal scene dataset until the loss function converges to the preset value; Among them, the loss function is: θ1=θ SE +θ PE +θ C ; Among them, θ1 is the loss function value, θ SE is the root mean square error (RMS) between the predicted value and the actual value, θ PE is the average relative percentage error between the predicted value and the actual value; θ C The coefficient of determination is a measure of the degree of correlation between the predicted value and the actual value. The larger the value, the higher the degree of correlation and the better the prediction effect. Among them, in the model construction module, the training process of the posture adjustment prediction module includes: Step 4.1: Using T re represents the training time of the posture adjustment prediction module, F represents the total number of training rounds of the posture adjustment prediction module, V represents the total number of target data aggregated by the posture adjustment prediction module, Indicates every The wheel posture adjustment prediction module aggregates the target data once, and then: Step 4.2: Select the optimal aggregation frequency through K-Means clustering algorithm Step 4.3: At training time T re The posture adjustment prediction module is based on the optimal aggregation frequency Towards target data training aggregation, if the training time T is reached re , then proceed to step 4.4, otherwise go to step 4.2; Step 4.4: Aggregate the model weights through the FedProx algorithm. If the target accuracy is achieved, stop the training process and obtain the preset interference scenario prediction model. Otherwise, go to step 4.2.

Images