Three-frequency satellite station multi-source data fusion satellite correction method, system and device and storage medium

Abstract

The invention discloses a three-frequency satellite station multi-source data fusion satellite correction method, system and device and a storage medium, and the method comprises the steps: obtaining a plurality of frequency band signals, and setting the transmitting power proportion of each frequency band through a beam forming device; initializing a phase matrix of the beam forming device to obtain an initial beam pointing satellite, establishing a communication link between the satellite and the portable station, and synchronizing timestamps and phase references of a plurality of frequency band signals; the method comprises the following steps: acquiring communication data and environmental data of a portable station, preprocessing the communication data and the environmental data, and performing feature extraction on the preprocessed communication data and environmental data to obtain feature vectors; integrating the feature vectors to obtain a multi-dimensional feature vector; analyzing the multi-dimensional feature vector by using a fusion model, and determining a satellite correction parameter; and according to the satellite correction parameters, dynamically adjusting a phase matrix of the beam forming device, optimizing beam pointing to a satellite, realizing satellite correction, and shortening satellite convergence time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite communication, and in particular to a three-frequency satellite station multi-source data fusion alignment correction method, system, device and storage medium. BACKGROUND

[0002] At present, the alignment accuracy of a satellite portable station in the field of satellite communication is a key factor that can determine the communication quality and response efficiency. The directional function of the existing mainstream equipment is usually realized by using a nine-axis inertial navigation system, which mainly relies on a magnetometer to measure the azimuth angle and simultaneously corrects the magnetic declination angle through a single-antenna Beidou positioning. However, this approach often leads to inaccurate directional accuracy of the system and makes the alignment convergence time too long, ultimately resulting in the inability to meet the application scenarios with high real-time requirements such as power grid emergency communication.

[0003] The most important defect in the prior art is generally caused by the inherent limitations of the magnetometer orientation, which is caused by the magnetic declination and magnetic field interference, such as the difference in magnetic declination and the significant difference in global magnetic declination caused by geographical location. The traditional system currently only relies on a pre-stored magnetic declination table for correction and has no way to dynamically compensate for the accumulated directional error caused by the change in magnetic declination. At the same time, the traditional system may fail to orient when encountering strong magnetic field interference, especially the strong magnetic field generated by high-voltage equipment in the power grid scenario, which can seriously interfere with the magnetometer. In addition, the lack of directional capability of single-antenna Beidou positioning also leads to the inability to obtain azimuth angle information, and single-antenna Beidou positioning can only provide latitude and longitude coordinates and cannot calculate the orientation of the device. The existing system also uses an indirect scheme, and the indirect scheme used in the existing system has the disadvantage of low accuracy of the magnetic declination table, with a pre-stored table value error of ±0.5°. Moreover, this indirect scheme does not take into account local magnetic field interference. When the magnetic declination changes over time, the traditional system has no way to update it in real time, which can lead to a long alignment convergence time. In addition, due to the large directional error of this indirect scheme, the traditional satellite portable station needs to repeatedly adjust the pointing of the antenna when in use, which can lead to an average alignment convergence time of 10-15 minutes, which is much higher than the requirement of less than or equal to 3 minutes for power grid rescue.

[0004] Therefore, it is urgent to propose a three-frequency satellite station multi-source data fusion alignment correction method to meet the requirement of less than or equal to 3 minutes for the alignment time of a three-frequency satellite portable station in a power grid emergency scenario. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] This invention provides a method, system, device, and storage medium for satellite station multi-source data fusion and satellite alignment, which solves the problems of low orientation accuracy and long convergence time in existing technologies that rely on magnetometers and single-antenna BeiDou positioning in power grid emergency scenarios.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for satellite alignment and calibration using multi-source data fusion from a three-frequency satellite station, comprising: Multiple frequency band signals are acquired, and the transmit power ratio of each frequency band is set through a beamforming device; The phase matrix of the beamforming device is initialized to obtain the initial beam pointing to the satellite, a communication link is established between the satellite and the portable station, and the timestamps and phase references of multiple frequency band signals are synchronized. Acquire communication data and environmental data of the portable station, preprocess the communication data and environmental data, extract features from the preprocessed communication data and environmental data to obtain feature vectors; The feature vectors are integrated to obtain a multidimensional feature vector; The multidimensional feature vectors are analyzed using a fusion model to determine the star alignment parameters; Based on the satellite alignment parameters, the phase matrix of the beamforming device is dynamically adjusted to optimize the beam pointing towards the satellite and achieve satellite alignment.

[0008] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, the initialization of the phase matrix of the beamforming device includes: The beam pointing angles corresponding to multiple frequency band signals are calculated based on satellite orbit parameters and the initial position of the portable station; Based on beamforming theory, an initial phase matrix is ​​generated; Based on the wavelength differences of signals in multiple frequency bands, frequency band compensation is performed on the initial phase matrix to obtain the initial beam pointing towards the satellite.

[0009] The beneficial effects of this preferred technical solution are that, through precise calculation and frequency band compensation, the initial beam of multi-frequency signals can be accurately pointed to the satellite, significantly improving satellite alignment efficiency and communication stability.

[0010] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, the preprocessing of the communication data and environmental data includes: The communication data undergoes a first filtering process to separate signals into multiple frequency bands; The environmental data undergoes a second filtering process to remove noise data in the frequency band.

[0011] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, the feature extraction of the preprocessed communication data and environmental data includes: Different feature extraction methods are used to extract features from preprocessed communication data and environmental data; The feature vector includes the intensity fluctuation value of the three-frequency signal, the phase difference value of the three-frequency signal, the time delay change value of the three-frequency signal, the inertial navigation velocity value, the inertial navigation acceleration value, the BeiDou coordinate accuracy value, the ambient temperature value, and the ambient humidity value.

[0012] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, the integration of the feature vectors includes: Normalize the different eigenvectors separately to obtain the normalized eigenvectors; The normalized feature vectors are divided into weight coefficients, and the normalized feature vectors are concatenated according to the weight coefficients to obtain multidimensional feature vectors.

[0013] The beneficial effects of this preferred technical solution are that by normalizing and weighting the splicing, multi-source feature vectors are integrated, improving data consistency and model input quality, and enhancing the accuracy and reliability of model star correction.

[0014] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, determining the satellite alignment parameters includes: Training data from multiple power grid emergency scenarios were acquired, and multidimensional feature vectors were preprocessed, including feature encoding, data standardization, and data partitioning. The fusion model is trained using the preprocessed multidimensional feature vectors. The objective function is to minimize the time required for satellite alignment. The objective function combines the error between the adjustment amount of the alignment parameters and the actual optimal alignment parameters, and introduces physical constraints and system constraints, including physical range constraints of the alignment parameters, signal quality constraints, and upper limit constraints of the adjustment time. The threshold of the fusion model is dynamically adjusted according to the type of power grid emergency scenario and environmental parameters, and the satellite correction parameters are output.

[0015] The beneficial effects of this preferred technical solution are that, through data-driven model training and dynamic adjustment of the fusion model threshold, rapid and accurate satellite alignment correction can be achieved, significantly shortening the satellite alignment time and improving emergency response efficiency.

[0016] As a preferred embodiment of the multi-source data fusion and satellite alignment method for a three-frequency satellite station described in this invention, the dynamic adjustment of the phase matrix of the beamforming device includes: Calculate the adjustment value of the phase matrix based on the star correction parameters; Calculate the corresponding phase adjustment amount for each frequency band signal, and obtain the phase adjustment matrix through weighted fusion; The calibration frequency is dynamically adjusted according to the moving speed of the portable station; The phase adjustment matrix is ​​superimposed on the current phase matrix to obtain the updated phase matrix; The phase value is converted into a bias voltage by a field-programmable gate array controller, which drives each element of the reconfigurable intelligent surface phased array to adjust the beam pointing in real time and achieve star correction.

[0017] Secondly, the present invention provides a three-frequency satellite station multi-source data fusion and satellite alignment system, comprising: The ratio setting module is used to acquire signals from multiple frequency bands and set the transmit power ratio of each frequency band through a beamforming device; An initialization module is used to initialize the phase matrix of the beamforming device, obtain an initial beam pointing to the satellite, establish a communication link between the satellite and the portable station, and synchronize the timestamps and phase references of multiple frequency band signals. The feature extraction module is used to acquire communication data and environmental data of the portable station, preprocess the communication data and environmental data, and extract features from the preprocessed communication data and environmental data to obtain feature vectors. An integration module is used to integrate the feature vectors to obtain multidimensional feature vectors; The parameter determination module is used to analyze the multidimensional feature vector using the fusion model to determine the star correction parameters; The adjustment module is used to dynamically adjust the phase matrix of the beamforming device according to the satellite alignment parameters, optimize the beam pointing to the satellite, and achieve satellite alignment.

[0018] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the described method for multi-source data fusion and satellite alignment of a three-frequency satellite station.

[0019] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the aforementioned method for multi-source data fusion and satellite alignment correction of a three-frequency satellite station.

[0020] Compared with existing technologies, the advantages of this invention are as follows: By integrating nine-axis inertial navigation attitude angle data, single-antenna BeiDou positioning latitude and longitude information, and the three-frequency signal strength characteristics of a reconfigurable smart surface (RIS) phased array, this invention constructs a multi-dimensional input feature set, significantly reducing the impact of regional differences in magnetic declination and strong magnetic field interference, providing a complete data foundation for high-precision positioning. A fusion model using convolutional neural networks (CNN) and long short-term memory networks (LSTM) is used to train multiple multi-source data accumulated in power grid emergency scenarios, constructing a dynamic fusion model. The fusion model in this invention adaptively learns the nonlinear mapping relationship between multi-source data and azimuth angle through a state-action-reward mechanism, thereby providing both accuracy and speed. Based on the trained high-precision fusion model, the system completes multi-source data inference in real-time scenarios within seconds. Combined with the fast beam scanning capability of the RIS phased array, the system can directly output precise azimuth angle commands, achieving full-process automation and reducing satellite convergence time. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the overall process logic of a three-frequency satellite station multi-source data fusion and satellite alignment method provided in one embodiment of the present invention. Detailed Implementation

[0023] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0024] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for satellite alignment by multi-source data fusion of a three-frequency satellite station is provided, comprising: S100: Acquire signals from multiple frequency bands and set the transmit power ratio of each frequency band through a beamforming device; In one alternative embodiment, the beamforming device can be used for active antenna system (AAS) beamforming. By integrating multiple antenna elements and active amplifiers, the AAS dynamically adjusts the transmit power of each antenna element according to channel state information and user needs to optimize beam pointing and coverage. In another alternative embodiment, the beamforming device can also perform beam-selective transmit power control by precisely controlling the transmit power of the uplink beamforming resources by measuring the path loss associated with the downlink receive beamforming resources. In this embodiment of the invention, the beamforming device includes RIS phased array beamforming; Specifically, the satellite portable station activates signals in the S, Ku, and Ka bands, and each band uses RIS phased array beamforming. Then, the transmission power ratio of the three bands is set, with the specific transmission power ratio being: 35% for the S band, 35% for the Ku band, and 30% for the Ka band.

[0025] It should be noted that by using RIS phased array beamforming technology, the transmit power ratio of the S, Ku, and Ka bands can be precisely controlled to achieve coordinated optimization of multi-band signals, improve signal transmission efficiency and stability, enhance the reliability and coverage of satellite communication, and meet the high requirements of power grid emergency communication.

[0026] S200: Initialize the phase matrix of the beamforming device to obtain the initial beam pointing to the satellite, establish a communication link between the satellite and the portable station, and synchronize the timestamps and phase references of multiple frequency band signals; S300: Acquires communication data and environmental data of the portable station, preprocesses the communication data and environmental data, extracts features from the preprocessed communication data and environmental data, and obtains feature vectors; Specifically, it collects communication data and portable station environmental data in real time. The communication data includes signal strength, phase offset, and propagation delay in the S, Ku, and Ka frequency bands, while the portable station environmental data includes inertial navigation data, BeiDou positioning data, and environmental parameters.

[0027] S400: Integrate the feature vectors to obtain multidimensional feature vectors; S500: Analyzes multidimensional feature vectors using a fusion model to determine star alignment parameters; In one alternative embodiment, the fusion model can be a Transformer and Fully Connected Network (FCN) fusion model, which connects the output of the Transformer to the FCN, uses the output of the Transformer as the input of the FCN, and further processes it through the FCN to output the star alignment parameters. In another alternative embodiment, the fusion model can also be a fusion model of graph neural network (GNN) and recurrent neural network (RNN). The GNN is used to process multidimensional feature vectors to capture the relationship between different features. Then the output of the GNN is used as the input of the RNN to process the time series features and finally output the star correction parameters. In this embodiment of the invention, the fusion model includes a fusion model of convolutional neural network (CNN) and long short-term memory network (LSTM); Specifically, the CNN part: Input layer: [B,T,D]; Convolutional layer 1: 1D convolution (number of filters = 64, kernel size = 3, stride = 1, ReLU activation). Convolutional layer 2: 1D convolution (number of filters = 128, kernel size = 3, stride = 1, ReLU activation). Pooling layer: global average pooling, with an output shape of [B, 128].

[0028] LSTM part: The CNN output is flattened and then fed into the LSTM layer (hidden units = 256, bidirectional LSTM, Dropout = 0.3); the output shape is [B, 256].

[0029] Classification layer: Fully connected layer (output unit = number of disaster categories, Softmax activation).

[0030] Where B is the batch size, set to 64; T is the time step, set to 32, meaning that each training sample is 32 seconds; D is the feature dimension, a total of 18 dimensions, including the fluctuation value of the three-frequency signal, phase difference, time delay change, inertial navigation velocity / acceleration, BeiDou coordinate error, environmental temperature and humidity, etc.

[0031] It should be noted that the model extracts local features through CNN, processes the time series through LSTM, and finally outputs the disaster category through the classification layer. It effectively integrates multi-source data, accurately identifies power grid emergency scenarios, improves the adaptability and accuracy of satellite correction, and ensures the stability and reliability of the communication link.

[0032] S600: Based on the satellite alignment parameters, the phase matrix of the beamforming device is dynamically adjusted to optimize the beam pointing to the satellite and achieve satellite alignment.

[0033] It should be noted that by employing multi-band signal coordination and beamforming technology, the transmit power ratio is precisely set, and the communication link is optimized. Phase matrix initialization ensures the beam accurately points to the satellite, and preprocessing and feature extraction of communication and environmental data provide high-quality input for the fusion model. Multi-dimensional feature vector integration and fusion model analysis dynamically adjust the phase matrix of the beamforming device, achieving rapid and accurate satellite alignment, significantly shortening alignment time, improving the efficiency and reliability of power grid emergency communication, and meeting rapid response requirements.

[0034] In this embodiment of the invention, step S200 includes the following sub-steps A1-A3; In A1: The beam pointing angles corresponding to multiple frequency band signals are calculated based on satellite orbit parameters and the initial position of the portable station; In A2: Based on beamforming theory, an initial phase matrix is ​​generated; In A3: Based on the wavelength differences of multiple frequency band signals, frequency band compensation is performed on the initial phase matrix to obtain the initial beam pointing to the satellite.

[0035] Specifically, based on satellite orbital parameters including longitude, latitude, altitude and the initial position of the portable station, the beam pointing angles corresponding to the three-frequency signals (S / Ku / Ka) are calculated; Using the beamforming theory of phased array antennas, an initial phase matrix is ​​generated, and the phase value of each array element is represented as: in, For the spacing between array elements, For the coordinates of the array elements, For the first Frequency band wavelength; To address the wavelength differences between the S / Ku / Ka tri-frequency signals, frequency band compensation is applied to the phase matrix to ensure consistent beam pointing across the three frequencies. The wavelength differences are as follows: Compensation is expressed as: in, To use the Ku band wavelength as a reference, This is the initial phase matrix.

[0036] It should be noted that by accurately calculating the beam pointing angle and generating the initial phase matrix, combined with multi-band phase compensation, the beam pointing of the three-frequency signals is ensured to be consistent, which effectively improves the initial alignment accuracy and communication link stability of satellite communication, laying the foundation for efficient communication.

[0037] In this embodiment of the invention, step S300 includes the following sub-steps B1-B2; In B1: The communication data undergoes a first filtering process to separate signals into multiple frequency bands; In B2: The environmental data undergoes a second filtering process to remove noise data in the frequency band.

[0038] In one optional embodiment, the first filtering process can be wavelet transform filtering, which involves performing wavelet transform on the communication data, selecting a suitable wavelet basis, decomposing the signal into different frequency bands, and extracting the signals of the S, Ku, and Ka frequency bands through threshold processing or reconstruction. In another alternative embodiment, the first filtering process can also be adaptive filtering. An adaptive filter is designed, with communication data as input and known signal characteristics of the S, Ku, and Ka frequency bands as reference signals. The filter weights are adjusted through an adaptive algorithm to make the filter output as close as possible to the reference signal, thereby separating the signals of the three frequency bands. In this embodiment of the invention, the first filtering process includes bandpass filtering; Specifically, based on the center frequencies of the S / Ku / Ka bands (S = 2.5GHz, Ku = 12.5GHz, Ka = 30GHz) and bandwidths (S = 100MHz, Ku = 500MHz, Ka = 1GHz), three digital bandpass filters are designed with passband ranges of: S band: [2.25GHz, 2.75GHz]; Ku band: [12.25GHz, 12.75GHz]; Ka band: [29.5GHz, 30.5GHz]. The filter order is uniformly set to 8th order IIR (Infinite Impulse Response) structure, with a roll-off slope ≥60dB / decade, ensuring a stopband attenuation ≥50dB, thus separating the signals from the S, Ku, and Ka bands.

[0039] In one optional embodiment, the second filtering process can be an extended Kalman filter (EKF) to establish a nonlinear dynamic model and an observation model for the portable station's environmental data. The dynamic model describes the changes in inertial navigation velocity and BeiDou coordinates, while the observation model describes the measurement data. The state estimate is recursively updated using EKF to remove abrupt changes in inertial navigation velocity and BeiDou coordinate drift noise. In another optional embodiment, the second filtering process can also be particle filtering, which initializes a set of particles, propagates the particles through a prediction step based on the dynamic model and observation model of the portable station environment data, and adjusts the particle weights through an update step to remove inertial navigation velocity mutations and BeiDou coordinate drift noise. In this embodiment of the invention, the second filtering process includes Kalman filtering; Specifically, a dual-channel Kalman filter architecture is adopted to independently filter the inertial navigation velocity and BeiDou coordinates, and to eliminate cross-interference through the coupling covariance matrix; The initial state vector and covariance matrix include inertial navigation velocity and BeiDou coordinates, where the inertial navigation velocity filter channel is represented as: State vector: State transition equation: Observation equation: in, This is the measured value of inertial navigation acceleration. For process noise, For measuring noise.

[0040] The BeiDou coordinate filtering channel is represented as follows: State vector: State transition equation: Observation equation: in, For BeiDou speed assistance quantity, This is process noise.

[0041] During the state update phase, the correlation between inertial navigation velocity and BeiDou coordinates is quantified by introducing a cross-channel covariance matrix, and the update is expressed as: in, The initial value is a zero matrix, and the coupling relationship between the standard guidance and Beidou data is learned step by step through iteration.

[0042] It should be noted that the carefully designed 8th-order IIR bandpass filter effectively separates the S, Ku, and Ka band signals, ensuring high selectivity and low interference, and providing a clean signal foundation for subsequent accurate processing. The dual-channel Kalman filter architecture independently processes the inertial navigation velocity and BeiDou coordinates, and eliminates cross-interference by coupling the covariance matrix, significantly improving data accuracy and reliability, and enhancing the system's stability and response capability in complex dynamic environments.

[0043] In this embodiment of the invention, after completing steps B1-B2, step S300 further includes steps B3-B4. In B3: different feature extraction methods are used to extract features from preprocessed communication data and environmental data; In B4: the feature vector includes the three-frequency signal strength fluctuation value, the three-frequency signal phase difference value, the three-frequency signal time delay change value, the inertial navigation velocity value, the inertial navigation acceleration value, the BeiDou coordinate accuracy value, the ambient temperature value, and the ambient humidity value.

[0044] Specifically, the extraction of the three-frequency signal strength fluctuation value includes performing sliding window statistics on the power values ​​of the S / Ku / Ka three-frequency signals, and calculating the fluctuation range of the signal strength as follows: in, For the first Within the first time window Instantaneous power value of the frequency band; If the fluctuation value is ≤5dB, it is marked as normal; otherwise, an alarm mechanism is triggered, that is, the gain of the RIS phased array is adjusted.

[0045] The extraction of phase difference values ​​for the three-frequency signals includes extracting the phase difference between the S-Ku and Ku-Ka band signals using a cross-correlation phase detection method: Short-time Fourier transform is performed on the separated S / Ku / Ka signals to extract the instantaneous phase, and the phase difference is calculated as follows: in, For the first Within a time window Instantaneous phase of the frequency band, For the first Within a time window Instantaneous phase of the frequency band, For the first Within a time window Instantaneous phase of the frequency band; The absolute value is then taken as the maximum value within the sliding window as the final phase difference.

[0046] If the phase difference is ≤π / 6, it is marked as well synchronized; otherwise, the phase compensation mechanism is triggered, that is, the RIS phase matrix is ​​adjusted.

[0047] Extracting the time delay variation values ​​of the three-frequency signals includes calculating the time delay variation of the three-frequency signals using the cross-correlation function method, and performing discrete cross-correlation calculations on the S / Ku / Ka signals: in, For the first Sampling sequence of frequency band signal, For time delay.

[0048] Find the time delay corresponding to the peak of cross-correlation, and calculate the time delay difference between adjacent time points: in, For the first The time delay corresponding to the peak cross-correlation within each time window For the first The time delay corresponding to the peak cross-correlation within each time window; If the time delay variation is ≤50ns, it is marked as stable; otherwise, clock calibration is triggered.

[0049] Inertial navigation velocity and acceleration value extraction includes velocity extraction, which extracts the magnitude from the inertial navigation velocity vector output by the Kalman filter: in, Let the magnitude of the inertial navigation velocity vector be . These are the components of the inertial navigation velocity vector in the x, y, and z directions.

[0050] like If it is marked as normal, otherwise an abnormal motion status alarm will be triggered.

[0051] Accelerometer extraction involves directly reading the output value of the inertial navigation accelerometer, and then extracting the modulus value after low-pass filtering. in, Let the magnitude of the inertial navigation acceleration vector be . These are the components of the inertial navigation acceleration vector in the x, y, and z directions.

[0052] like If it is marked as normal, otherwise a device stability check will be triggered.

[0053] BeiDou coordinate accuracy extraction includes positioning error calculation. The BeiDou coordinate error calculated through differential positioning is expressed as follows: in, These are the error components of BeiDou coordinates in the x, y, and z directions, respectively.

[0054] If Errorp≤2m, mark it as high precision; otherwise, trigger BeiDou signal reacquisition.

[0055] Ambient temperature and humidity value extraction includes temperature extraction, which involves reading the output voltage value of the environmental sensor and converting it into a temperature value using a linear fitting formula. in, This is the ambient temperature value. This refers to the voltage value output by the environmental sensor. is a coefficient.

[0056] like Mark as normal; otherwise, trigger heat dissipation control.

[0057] Humidity extraction involves reading the output from the humidity sensor and converting it into a humidity value using a calibration formula. in, This represents the ambient humidity value. is a coefficient.

[0058] like Mark as normal; otherwise, trigger a device sealing check.

[0059] It should be noted that if the signal strength fluctuations in the S / Ku / Ka bands exceed 5dB, it indicates that the link may experience power reduction due to atmospheric attenuation, rain attenuation, or interference. This necessitates triggering gain adjustment or a RIS phased array gain compensation mechanism to prevent communication interruption. Synchronous monitoring of signal strength fluctuations across the three bands can help determine the type of interference source, providing a basis for multi-band resource allocation strategies.

[0060] If the phase difference between the S-Ku and Ku-Ka bands exceeds π / 6, it may lead to demodulation errors or carrier recovery failures in multi-band signals. Dynamic compensation using the RIS phase matrix is ​​necessary to ensure coherent superposition of multi-band signals at the receiver. Phase differences often reflect multipath effects or ionospheric disturbances, providing prior information for subsequent channel coding error correction.

[0061] If the time delay variation exceeds ±50ns, it may cause frame synchronization deviation or sampling clock jitter, which needs to be corrected by the BeiDou timing module or RIS clock calibration mechanism to ensure the timing consistency of multi-band signals. The time delay variation is related to the moving speed of the portable station, and can be combined with inertial navigation data to help calculate the Doppler frequency shift caused by relative motion.

[0062] Inertial navigation speed (0-10m / s) and acceleration (0-3m / s²) are used to determine whether the portable station is stationary, moving, or vibrating, and dynamically adjust the RIS beam pointing strategy; speed / acceleration outside the range may indicate equipment collision or loose installation, triggering an alarm and entering a safe mode.

[0063] A coordinate accuracy of ±2m is crucial for ensuring the RIS beam points to the satellite. Decreased accuracy can cause the beam to deviate from the satellite, requiring recovery through differential positioning or assisted navigation. Combining BeiDou coordinates with inertial navigation velocity allows for the calculation of the portable station's three-dimensional trajectory and optimization of the beam tracking algorithm.

[0064] Temperature (0-50℃) and humidity (0-100%) directly affect antenna material deformation, sensor drift, and RF component performance. RIS phased array calibration parameters need to be adjusted in real time based on environmental parameters. Extreme temperature and humidity may cause equipment aging or condensation, triggering heat dissipation / dehumidification control in advance to avoid communication interruptions.

[0065] In this embodiment of the invention, step S400 includes the following sub-steps C1-C2; In C1: Normalize the different eigenvectors separately to obtain the normalized eigenvectors; In C2: The normalized feature vector is divided into weight coefficients, and the normalized feature vectors are concatenated according to the weight coefficients to obtain a multidimensional feature vector.

[0066] In this embodiment of the invention, a multi-dimensional feature fusion method based on weighted splicing is used to integrate feature vectors; Specifically, the strength, phase, and delay of the three-frequency signals are normalized to the [0,1] interval using Min-Max; the inertial navigation features, BeiDou features, and environmental features are normalized using Z-Score.

[0067] Weight coefficients are assigned based on feature type; For example, the signal feature weighting coefficient is 0.45, which accounts for the largest proportion and reflects the quality of the communication link; The inertial navigation and BeiDou feature weighting coefficient is 0.35 to ensure spatial positioning accuracy; The environmental feature weighting coefficient is 0.2 to help compensate for external disturbances.

[0068] The normalized feature vectors are weighted and concatenated according to their weights to form the final multidimensional feature vector.

[0069] It should be noted that by using Min-Max normalization and Z-Score standardization, different feature dimensions are unified, and multi-dimensional feature vectors are formed by weighted concatenation, which enhances the consistency of feature expression, improves the quality of model input, and significantly improves the accuracy and efficiency of star correction.

[0070] In this embodiment of the invention, step S500 includes the following sub-steps D1-D3; In D1: Training data for multiple power grid emergency scenarios are acquired, and multi-dimensional feature vectors are preprocessed, including feature encoding, data standardization, and data partitioning. In D2: The fusion model is trained using the preprocessed multidimensional feature vectors. The objective function is to minimize the time required for satellite alignment. The objective function combines the error between the adjustment amount of the alignment parameters and the actual optimal alignment parameters, and introduces physical constraints and system constraints, including physical range constraints of the alignment parameters, signal quality constraints, and upper limit constraints of the adjustment time. In D3: The threshold of the fusion model is dynamically adjusted according to the type of power grid emergency scenario and environmental parameters, and the satellite correction parameters are output.

[0071] Specifically, the preprocessing of the multidimensional feature vectors includes encoding the multidimensional feature vectors into time series tensors with shape [B,T,D], performing Z-Score standardization on numerical features, and performing One-Hot encoding on the classification labels, dividing them into training, validation, and test sets in an 8:1:1 ratio.

[0072] Training the fusion model involves using cross-entropy loss as the loss function for multi-class classification tasks; the optimizer is the Adam optimizer with an initial learning rate of 0.001 and weight decay of 1e-5.

[0073] The input time series tensor is processed by a CNN to extract local features, and then an LSTM is used to model the temporal dependencies, ultimately outputting the disaster classification probability. The loss gradient is calculated, and the CNN convolutional kernel weights, LSTM gating parameters, and fully connected layer weights are updated using a chain rule. The Adam optimizer is used to adjust the learning rate, decaying by 50% every 50 epochs, and an early stopping strategy is applied: training terminates if the validation set loss does not decrease for three consecutive epochs.

[0074] In each training round, data is read in batches, forward-backward propagation is performed, and model parameters are updated; every 5 rounds, classification accuracy is evaluated on the validation set, and the best model parameters are recorded.

[0075] Validation metrics include classification accuracy and F1 score.

[0076] If the validation set accuracy is below 85%, increase the number of CNN filters (e.g., 128 → 256) or adjust the number of LSTM hidden units (256 → 512). If overfitting occurs, increase the Dropout rate (0.3→0.5) or introduce L2 regularization.

[0077] After training, the CNN-LSTM model weights are saved in ONNX format for embedded deployment. The final model performance is evaluated on the test set, requiring a classification accuracy of ≥90% and an F1 score of ≥0.88.

[0078] The training objective is to minimize the time required for star alignment, and the objective function is defined as follows: in, For model parameters, The input is a multidimensional feature vector. For satellite alignment time; The satellite alignment time is expressed as: in, To correct the parameter adjustment amount, This represents the maximum adjustment rate of the RIS phased array. These are the correction parameters for the model prediction. These are the actual optimal correction parameters. These are the weighting coefficients.

[0079] The training process includes physical and system constraints. Physical range constraints of correction parameters: in This represents the physical limit for phase adjustment of RIS phased arrays.

[0080] Signal quality constraints: in, This is the minimum signal-to-noise ratio threshold for the communication link.

[0081] Adjusting the time limit constraint: in, This is the maximum allowed satellite alignment time by the system.

[0082] Transform it into a differentiable form and update the model parameters through backpropagation. The Adam optimizer is used, employing projective gradient descent. If the predicted parameters exceed the physical range, they are truncated to the boundary value. The signal quality constraint and the upper limit constraint of the adjustment time are transformed into penalty terms added to the objective function, expressed as: in, The weighting coefficient of the penalty term, As expected of the input data, To adjust the upper limit of the time.

[0083] The threshold is dynamically adjusted based on the type of power grid emergency scenario and environmental parameters.

[0084] It should be noted that by jointly optimizing the adjustment time and parameter error, the model can maintain high accuracy while being quickly corrected, avoiding the problem of excessive adjustment time caused by traditional single error minimization. The introduction of physical and system constraints ensures the feasibility of the model output and prevents communication interruption due to out-of-bounds prediction parameters or insufficient signal quality.

[0085] In this embodiment of the invention, step S600 includes the following sub-steps E1-E5; In E1: Calculate the adjustment value of the phase matrix based on the star correction parameters; In E2: The corresponding phase adjustment amount is calculated for each frequency band signal, and the phase adjustment matrix is ​​obtained by weighted fusion; In E3: The calibration frequency is dynamically adjusted according to the moving speed of the portable station; In E4: The phase adjustment matrix is ​​superimposed on the current phase matrix to obtain the updated phase matrix; In E5: The phase value is converted into a bias voltage by a field-programmable gate array controller, which drives each element of the reconfigurable smart surface phased array to adjust the beam pointing in real time and achieve star correction.

[0086] Specifically, based on the star correction parameters, the real-time adjustment value of the RIS phased array phase matrix is ​​expressed as follows: in, , For the adjusted azimuth and elevation angles, For the current azimuth and elevation angles, , The coordinates of the array element.

[0087] If the transmission power needs to be adjusted Through the gain coefficient, using The signal amplitude is compensated, but the phase matrix itself only adjusts the phase, so there is no need to directly modify the phase value; the power is only adjusted at the subsequent transmitting end.

[0088] For the S / Ku / Ka bands, the phase adjustment amount is calculated separately, and the final phase matrix is ​​weighted and fused to achieve multi-band beam alignment: in, These are the weights for the S, Ku, and Ka frequency bands, respectively. , , Prioritize ensuring the stability of Ka-band links. For the final phase matrix, These are the adjustment amounts for the S, Ku, and Ka frequency bands, respectively. Will With the current phase matrix Superimpose and update the RIS phased array: The phase value is modulo 2π to ensure physical feasibility.

[0089] In vibration scenarios, the relationship between the correction frequency and the moving speed is a threshold-triggered dynamic adjustment. The specific rules include that if the moving speed of the portable station v > 5 m / s is detected, the correction frequency is immediately increased from the traditional 30 milliseconds / time to 2 milliseconds / time. High-speed movement (>5m / s) may cause satellite signals to attenuate rapidly. Shortening the correction cycle can compensate for beam offset in time and avoid communication interruption. In vibration scenarios, equipment is prone to sudden high-speed movement due to vibration or ground displacement, so link stability must be prioritized.

[0090] Will The phase values ​​in the array are mapped to the RIS hardware control registers, and the phase adjustment value for each array element is... For a specific bias voltage .

[0091] Voltage conversion is expressed as: in, This is the full-scale bias voltage of the PIN diode.

[0092] via FPGA controller The electromagnetic wave reflection phase is adjusted in real time and sent to each unit of the RIS phased array; the update period is Tupdate=2 milliseconds to ensure that the beam pointing follows the satellite position change in real time.

[0093] It should be noted that traditional methods rely on mechanical adjustments or inefficient algorithms, requiring 10 minutes for satellite alignment. This invention directly outputs correction parameters through a trained model, avoiding trial-and-error parameter tuning; it quickly calculates the phase adjustment amount using analytical formulas, eliminating the need for iterative optimization; and it achieves real-time updates of the phase matrix through FPGA hardware acceleration, reducing the update cycle to 2 milliseconds per update, and shortening the satellite alignment time from the traditional 10 minutes to within 3 minutes.

[0094] The above is a schematic scheme of a three-frequency satellite station multi-source data fusion and satellite alignment correction method according to this embodiment. It should be noted that the technical solution of this three-frequency satellite station multi-source data fusion and satellite alignment correction system belongs to the same concept as the technical solution of the aforementioned three-frequency satellite station multi-source data fusion and satellite alignment correction method. Details not described in detail in this embodiment can be found in the description of the aforementioned three-frequency satellite station multi-source data fusion and satellite alignment correction method.

[0095] The tri-frequency satellite station multi-source data fusion and satellite alignment system in this embodiment includes: The ratio setting module is used to acquire signals from multiple frequency bands and set the transmit power ratio of each frequency band through a beamforming device; An initialization module is used to initialize the phase matrix of the beamforming device, obtain an initial beam pointing to the satellite, establish a communication link between the satellite and the portable station, and synchronize the timestamps and phase references of multiple frequency band signals. The feature extraction module is used to acquire communication data and environmental data of the portable station, preprocess the communication data and environmental data, and extract features from the preprocessed communication data and environmental data to obtain feature vectors. An integration module is used to integrate the feature vectors to obtain multidimensional feature vectors; The parameter determination module is used to analyze the multidimensional feature vector using the fusion model to determine the star correction parameters; The adjustment module is used to dynamically adjust the phase matrix of the beamforming device according to the satellite alignment parameters, optimize the beam pointing to the satellite, and achieve satellite alignment.

[0096] This embodiment also provides a computer device suitable for satellite calibration using multi-source data fusion at a three-frequency satellite station, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a three-frequency satellite station multi-source data fusion and satellite alignment method as proposed in the above embodiments.

[0097] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a three-frequency satellite station multi-source data fusion and satellite alignment method as proposed in the above embodiment.

[0098] The storage medium proposed in this embodiment and the method for realizing multi-source data fusion and satellite calibration of three-frequency satellite stations proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0099] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for satellite alignment and calibration using multi-source data fusion from a three-frequency satellite station, characterized in that, include: Multiple frequency band signals are acquired, and the transmit power ratio of each frequency band is set through a beamforming device; The phase matrix of the beamforming device is initialized to obtain the initial beam pointing to the satellite, a communication link is established between the satellite and the portable station, and the timestamps and phase references of multiple frequency band signals are synchronized. Acquire communication data and environmental data of the portable station, preprocess the communication data and environmental data, extract features from the preprocessed communication data and environmental data to obtain feature vectors; The feature vectors are integrated to obtain a multidimensional feature vector; The multidimensional feature vectors are analyzed using a fusion model to determine the star alignment parameters; Based on the satellite alignment parameters, the phase matrix of the beamforming device is dynamically adjusted to optimize the beam pointing towards the satellite and achieve satellite alignment.

2. The method for satellite alignment and calibration of a three-frequency satellite station using multi-source data fusion as described in claim 1, characterized in that, Initializing the phase matrix of the beamforming device includes: The beam pointing angles corresponding to multiple frequency band signals are calculated based on satellite orbit parameters and the initial position of the portable station; Based on beamforming theory, an initial phase matrix is ​​generated; Based on the wavelength differences of signals in multiple frequency bands, frequency band compensation is performed on the initial phase matrix to obtain the initial beam pointing towards the satellite.

3. A method for satellite alignment and calibration of multi-source data fusion from a three-frequency satellite station as described in claim 1 or 2, characterized in that, Preprocessing the communication data and environmental data includes: The communication data undergoes a first filtering process to separate signals into multiple frequency bands; The environmental data undergoes a second filtering process to remove noise data in the frequency band.

4. The method for satellite alignment and calibration of a three-frequency satellite station using multi-source data fusion as described in claim 3, characterized in that... Feature extraction from preprocessed communication and environmental data includes: Different feature extraction methods are used to extract features from preprocessed communication data and environmental data; The feature vector includes the intensity fluctuation value of the three-frequency signal, the phase difference value of the three-frequency signal, the time delay change value of the three-frequency signal, the inertial navigation velocity value, the inertial navigation acceleration value, the BeiDou coordinate accuracy value, the ambient temperature value, and the ambient humidity value.

5. The method for satellite alignment and fusion of multi-source data from a three-frequency satellite station as described in claim 4, characterized in that, Integrating the feature vectors includes: Normalize the different eigenvectors separately to obtain the normalized eigenvectors; The normalized feature vectors are divided into weight coefficients, and the normalized feature vectors are concatenated according to the weight coefficients to obtain multidimensional feature vectors.

6. The method for satellite alignment and calibration of a three-frequency satellite station using multi-source data fusion as described in claim 5, characterized in that, Determining the satellite alignment parameters includes: Training data from multiple power grid emergency scenarios were acquired, and multidimensional feature vectors were preprocessed, including feature encoding, data standardization, and data partitioning. The fusion model is trained using the preprocessed multidimensional feature vectors. The objective function is to minimize the time required for satellite alignment. The objective function combines the error between the adjustment amount of the alignment parameters and the actual optimal alignment parameters, and introduces physical constraints and system constraints, including physical range constraints of the alignment parameters, signal quality constraints, and upper limit constraints of the adjustment time. The threshold of the fusion model is dynamically adjusted according to the type of power grid emergency scenario and environmental parameters, and the satellite correction parameters are output.

7. The method for satellite alignment and fusion of multi-source data from a three-frequency satellite station as described in claim 6, characterized in that, Dynamically adjusting the phase matrix of the beamforming device includes: Calculate the adjustment value of the phase matrix based on the star correction parameters; Calculate the corresponding phase adjustment amount for each frequency band signal, and obtain the phase adjustment matrix through weighted fusion; The calibration frequency is dynamically adjusted according to the moving speed of the portable station; The phase adjustment matrix is ​​superimposed on the current phase matrix to obtain the updated phase matrix; The phase value is converted into a bias voltage by a field-programmable gate array controller, which drives each element of the reconfigurable intelligent surface phased array to adjust the beam pointing in real time and achieve star correction.

8. A tri-frequency satellite station multi-source data fusion and satellite alignment system, employing the tri-frequency satellite station multi-source data fusion and satellite alignment method as described in any one of claims 1-7, characterized in that, include: The ratio setting module is used to acquire signals from multiple frequency bands and set the transmit power ratio of each frequency band through a beamforming device; An initialization module is used to initialize the phase matrix of the beamforming device, obtain an initial beam pointing to the satellite, establish a communication link between the satellite and the portable station, and synchronize the timestamps and phase references of multiple frequency band signals. The feature extraction module is used to acquire communication data and environmental data of the portable station, preprocess the communication data and environmental data, and extract features from the preprocessed communication data and environmental data to obtain feature vectors. An integration module is used to integrate the feature vectors to obtain multidimensional feature vectors; The parameter determination module is used to analyze the multidimensional feature vector using the fusion model to determine the star correction parameters; The adjustment module is used to dynamically adjust the phase matrix of the beamforming device according to the satellite alignment parameters, optimize the beam pointing to the satellite, and achieve satellite alignment.

9. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the three-frequency satellite station multi-source data fusion and satellite alignment method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of a three-frequency satellite station multi-source data fusion and satellite alignment method according to any one of claims 1 to 7.

Images