A method for adjusting transmission power in satellite communication

Abstract

The present invention discloses a method for adjusting the transmission power in satellite communication, belonging to the technical field of satellite communication. First, the spectrum monitoring device on the satellite is used to scan the communication frequency band and adjacent frequency bands, construct multiple background noise power sequences and extract background noise features; then, corresponding sequences are formed according to the bit error rate and signal-to-noise ratio of the satellite transmission signal, and the bit error rate feature and signal-to-noise ratio feature are extracted; furthermore, the demand coefficient is obtained based on the gap between the latest bit error rate, signal-to-noise ratio and the target value; finally, multiple groups of background noise features, bit error rate features, signal-to-noise ratio features and demand coefficients are used as samples to be input into the trained fully connected layer - BP neural network to obtain the transmission power increment, which is added to the initial transmission power to determine the transmission power of the next signal, effectively improving the accuracy of transmission power adjustment.

Description

Technical Field

[0001] The present invention relates to the field of satellite communication technology, and particularly to a method for adjusting the transmission power in satellite communication. Background Art

[0002] With the increasing congestion of communication spectrum resources, different communication systems coexist within a limited frequency band, and the possibility of signal interference increases significantly. The modern communication network is highly dense and complex, making it extremely difficult to utilize spectrum resources. The growing demand for spectrum from various systems such as mobile communication, broadcasting, and satellite communication has made the problems of frequency congestion and interference more prominent. The traditional fixed-power transmission scheme is difficult to adapt to such a complex communication environment. Some existing methods for adjusting transmission power only adjust the transmission power based on the signal-to-noise ratio at the receiving end. Although the signal quality can be improved to a certain extent, this method does not consider the impact of the real-time spectrum of the frequency band used by satellite communication and its adjacent frequency bands on the current transmitted signal, so there is a problem of low accuracy in adjusting the transmission power. Summary of the Invention

[0003] Aiming at the above deficiencies in the prior art, the method for adjusting the transmission power in satellite communication provided by the present invention solves the problem of low accuracy in adjusting the transmission power existing in the prior art.

[0004] To achieve the above invention objective, the technical solution adopted by the present invention is: A method for adjusting the transmission power in satellite communication, comprising the following steps:

[0005] Scan the frequency band used by satellite communication and its adjacent frequency bands through the spectrum monitoring device on the satellite to construct multiple background noise power sequences;

[0006] Extract features from each background noise power sequence to obtain a set of background noise features;

[0007] According to the bit error rate and signal-to-noise ratio of each satellite transmitted signal, form a bit error rate sequence and a signal-to-noise ratio sequence, and extract features from the bit error rate sequence and the signal-to-noise ratio sequence respectively to obtain a set of bit error rate features and a set of signal-to-noise ratio features;

[0008] Obtain a demand coefficient according to the gap between the latest bit error rate and the target bit error rate, and the gap between the latest signal-to-noise ratio and the target signal-to-noise ratio;

[0009] Input multiple sets of background noise features, a set of bit error rate features, a set of signal-to-noise ratio features, and the demand coefficient as samples into the trained fully connected layer - BP neural network to obtain the transmission power increment;

[0010] Add the transmission power increment to the initial transmission power to obtain the transmission power of the next signal.

[0011] Furthermore, the process of constructing multiple background noise power sequences includes:

[0012] When the satellite is not transmitting a signal, a spectrum monitoring device on the satellite is used to scan the frequency band used for satellite communication to obtain a first background signal for each historical time period, calculate the background noise power, and construct a first background noise power sequence;

[0013] When the satellite is not transmitting a signal, a spectrum monitoring device on the satellite is used to scan a frequency band adjacent to the frequency band used for satellite communication to obtain a second background signal in each historical time period, calculate the background noise power, and construct a second background noise power sequence;

[0014] When the satellite is not transmitting a signal, the spectrum monitoring equipment on the satellite scans another adjacent frequency band used for satellite communication to obtain the third background signal in each time period, calculate the background noise power, and construct a third background noise power sequence.

[0015] Furthermore, the formula for calculating the background noise power is: , where P is the background noise power, x n is the nth signal value in the background signal of a time period, N is the length of the background signal of a time period, | | is the absolute value sign, and n is a positive integer.

[0016] Furthermore, a set of background noise features includes: a background noise power mean and a background noise power increment;

[0017] The mean background noise power is the mean of the background noise power sequence;

[0018] The background noise power increment is the difference between the latest background noise power and the earliest background noise power in the background noise power sequence.

[0019] Furthermore, a set of bit error rate features includes: bit error rate mean and bit error rate delta;

[0020] The mean bit error rate is the mean of the bit error rate sequence composed of the bit error rates of satellite transmission signals at each historical moment;

[0021] The bit error rate increment is the difference between the latest bit error rate and the earliest bit error rate in the bit error rate sequence;

[0022] A set of SNR features includes: SNR mean and SNR delta;

[0023] The mean signal-to-noise ratio is the mean of the signal-to-noise ratio sequence composed of the signal-to-noise ratios of satellite transmission signals at each historical moment;

[0024] The SNR increment is the difference between the latest SNR and the earliest SNR in the SNR sequence.

[0025] Furthermore, the process of calculating the demand coefficient includes:

[0026] Judge whether the latest bit error rate is greater than the target bit error rate. If so, subtract the target bit error rate from the latest bit error rate to obtain the bit error rate difference, and take the ratio of the bit error rate difference to the target bit error rate as the first demand component. If not, assign 0 to the first demand component;

[0027] Judge whether the latest signal-to-noise ratio is less than the target signal-to-noise ratio. If so, subtract the latest signal-to-noise ratio from the target signal-to-noise ratio to obtain the signal-to-noise ratio difference, and take the ratio of the signal-to-noise ratio difference to the target signal-to-noise ratio as the second demand component; if not, assign 0 to the second demand component;

[0028] Add the first demand component and the second demand component to obtain the demand coefficient.

[0029] Furthermore, the fully connected layer - BP neural network includes: the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, the input layer, the first hidden layer, the second hidden layer, and the output layer;

[0030] The input end of the first fully connected layer is used to input the first group of background noise features;

[0031] The input end of the second fully connected layer is used to input the second group of background noise features;

[0032] The input end of the third fully connected layer is used to input the third group of background noise features;

[0033] The input end of the fourth fully connected layer is used to input a group of bit error rate features;

[0034] The input end of the fifth fully connected layer is used to input a group of signal-to-noise ratio features;

[0035] The input layer is used to input the outputs of the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, and the demand coefficient into the input end of the first hidden layer;

[0036] The output end of the first hidden layer is connected to the input end of the second hidden layer;

[0037] The input end of the output layer is connected to the output end of the second hidden layer, and its output end serves as the output end of the fully connected layer - BP neural network.

[0038] Further, the process of training the fully connected layer - BP neural network includes: a first training stage and a second training stage. The first training stage is used to train the fully connected layer - BP neural network according to the gap between the predicted transmit power increment and the maximum transmit power increment of the fully connected layer - BP neural network. The second training stage is used to retrain the fully connected layer - BP neural network after the first training stage to find the optimal weights and biases corresponding to the minimum transmit power increment.

[0039] Further, the process of the second training stage includes:

[0040] Input the samples of the nth training into the fully connected layer - BP neural network after the first training stage to obtain the transmit power increment of the nth training, where n is the number of training times in the second training stage;

[0041] When the transmit power increment of the nth training is greater than the maximum transmit power increment, take the maximum transmit power increment as the transmit power increment of the nth training, and add the transmit power increment of the nth training to the initial transmit power to obtain the transmit power of the nth training;

[0042] Transmit the satellite test signal using the transmit power of the nth training;

[0043] Statistically calculate the bit error rate and signal - to - noise ratio of the satellite test signal received by the receiving end, and calculate the second target value based on the objective function;

[0044] Update the weights and biases in the fully connected layer - BP neural network according to the second target value, increment n by 1, and loop the training process until the number of training times reaches the target number of training times;

[0045] In all trainings of the second training stage, mark the number of training times with a bit error rate less than the target bit error rate and a signal - to - noise ratio greater than the target bit error rate as the undetermined number of times;

[0046] Among all the undetermined numbers of times, mark the undetermined number of times corresponding to the minimum transmit power increment as the target number of times;

[0047] Take the weights and biases corresponding to the target number of times as the optimal parameters, and the training of the fully connected layer - BP neural network is completed.

[0048] Further, when the bit error rate is greater than the target bit error rate and the signal - to - noise ratio is less than the target signal - to - noise ratio, the objective function is: ; when the bit error rate is greater than the target bit error rate and the signal - to - noise ratio is greater than or equal to the target signal - to - noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal - to - noise ratio is less than the target signal - to - noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal - to - noise ratio is greater than or equal to the target signal - to - noise ratio, the objective function is: G = c, where Bact The bit error rate of the satellite test signal is B tar The target bit error rate is S act The signal-to-noise ratio of the satellite test signal is S tar The target signal-to-noise ratio is G, G is the second target value, and c is a constant between 0 and 1.

[0049] The beneficial effects of the present invention are as follows:

[0050] 1. The present invention constructs a background noise power sequence by scanning the satellite communication frequency band and adjacent frequency bands, and combines the features extracted from the bit error rate sequence and the signal-to-noise ratio sequence, as well as the demand coefficient obtained according to the gap between the bit error rate, signal-to-noise ratio and the target value, providing rich and comprehensive data samples for the fully connected layer - BP neural network. Compared with the traditional method that only relies on the signal-to-noise ratio at the receiving end, it can calculate the transmit power increment more accurately, thereby significantly improving the accuracy of transmit power adjustment and effectively ensuring the stability and accuracy of signal transmission.

[0051] 2. The present invention considers the influence of the real-time spectrum state of the frequency band used for satellite communication and its adjacent frequency bands on the signal, and at the same time obtains the demand coefficient according to the latest gap between the bit error rate and signal-to-noise ratio and the target value, which characterizes the actual demand of the current satellite communication system for transmit power adjustment, and improves the prediction accuracy of the fully connected layer - BP neural network. Description of the Drawings

[0052] Figure 1 is a flowchart of a method for adjusting transmit power in satellite communication;

[0053] Figure 2 is a schematic structural diagram of a fully connected layer - BP neural network. Detailed Embodiments

[0054] The following describes the detailed embodiments of the present invention to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the detailed embodiments. For those ordinary skilled in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are within the scope of protection.

[0055] As Figure 1 shown, a method for adjusting transmit power in satellite communication includes the following steps:

[0056] Scan the frequency band used for satellite communication and adjacent frequency bands through the spectrum monitoring device on the satellite to construct multiple background noise power sequences;

[0057] Extract features from each background noise power sequence to obtain a set of background noise features;

[0058] According to the bit error rate and signal-to-noise ratio of the signals transmitted by each satellite, a bit error rate sequence and a signal-to-noise ratio sequence are formed, and features are extracted from the bit error rate sequence and the signal-to-noise ratio sequence respectively to obtain a set of bit error rate features and a set of signal-to-noise ratio features;

[0059] According to the gap between the latest bit error rate and the target bit error rate, and the gap between the latest signal-to-noise ratio and the target signal-to-noise ratio, a demand coefficient is obtained;

[0060] Taking multiple groups of background noise features, a set of bit error rate features, a set of signal-to-noise ratio features and the demand coefficient as samples, input them into the trained fully connected layer - BP neural network to obtain the transmit power increment;

[0061] Add the transmit power increment to the initial transmit power to obtain the transmit power of the next signal.

[0062] In this embodiment, the initial transmit power is the average daily transmit power of the satellite.

[0063] In this embodiment, the process of constructing multiple background noise power sequences includes:

[0064] When the satellite does not transmit signals, scan the frequency band used for satellite communication through the spectrum monitoring device on the satellite to obtain the first background signals in each historical time period, calculate the background noise power, and construct the first background noise power sequence;

[0065] When the satellite does not transmit signals, scan an adjacent frequency band of the frequency band used for satellite communication through the spectrum monitoring device on the satellite to obtain the second background signals in each historical time period, calculate the background noise power, and construct the second background noise power sequence;

[0066] When the satellite does not transmit signals, scan another adjacent frequency band of the frequency band used for satellite communication through the spectrum monitoring device on the satellite to obtain the third background signals in each time period, calculate the background noise power, and construct the third background noise power sequence.

[0067] The spectrum monitoring device is also called a spectrum monitoring receiver or a spectrum analyzer.

[0068] The frequency bands commonly used for satellite communication, from low to high, include: L band, S band, C band, X band, Ku band, Ka band, V band, W band. If the satellite is going to use the S band to transmit signals, the adjacent frequency bands are the L band and the C band. If the satellite uses the L band to transmit signals, the adjacent frequency band is only the S band, and then one background noise power sequence is filled with 0.

[0069] In this embodiment, the formula for calculating the background noise power is: [[ID=3,6]]where P is the background noise power, xn is the nth signal value in the background signal for a time period, N is the length of the background signal for a time period, || is the absolute value symbol, and n is a positive integer.

[0070] Co-channel interference: When multiple satellite communication systems or terrestrial communication devices operate on the same frequency, co-channel interference will occur. For example, in some popular frequency bands, there may be multiple communication systems of satellite operators operating simultaneously, and the signals between these systems interfere with each other, resulting in a decrease in the signal quality at the receiving end. To overcome co-channel interference, the transmitting end needs to increase the transmission power so that its own signal strength can still meet the demodulation requirements of the receiving end in the interference environment.

[0071] Adjacent-channel interference: Signals of adjacent frequencies interfere with the target signal due to reasons such as spectral leakage. For example, in the case of a tight allocation of satellite communication frequency bands, the interference of adjacent-channel signals cannot be ignored. To reduce the impact of adjacent-channel interference, the transmitting end needs to adjust the transmission power to ensure the communication quality.

[0072] In this embodiment, when a signal needs to be transmitted, the one-hour period before the current moment is equally divided into six 10-minute time periods. In each 10-minute period, the first background signal, the second background signal, and the third background signal are respectively obtained, and the background noise power is calculated to obtain the first background noise power sequence, the second background noise power sequence, and the third background noise power sequence with a length of 6.

[0073] In this embodiment, a set of background noise characteristics includes: the mean background noise power and the increment of the background noise power;

[0074] The mean background noise power is the mean of the background noise power sequence;

[0075] The increment of the background noise power is the difference between the latest background noise power and the earliest background noise power in the background noise power sequence.

[0076] The mean background noise power can represent the average situation of the background noise power sequence, and the increment of the background noise power reflects the change of the background noise power over a period of time. The power increment can sensitively capture this change trend, enabling the fully connected layer - BP neural network to timely understand the dynamic characteristics of the noise.

[0077] In this embodiment, a set of bit error rate characteristics includes: the mean bit error rate and the increment of the bit error rate;

[0078] The mean bit error rate is the mean of the bit error rate sequence composed of the bit error rates of the satellite transmitted signals at each historical moment;

[0079] The increment of the bit error rate is the difference between the latest bit error rate and the earliest bit error rate in the bit error rate sequence.

[0080] The average bit error rate reflects the average level of the bit error rate of satellite transmission signals in a historical period, and the bit error rate increment reflects the changing trend of the bit error rate over time, enabling the timely detection of fluctuations in the bit error rate.

[0081] A set of signal-to-noise ratio features includes: the average signal-to-noise ratio and the signal-to-noise ratio increment;

[0082] The average signal-to-noise ratio is the average of the signal-to-noise ratio sequence composed of the signal-to-noise ratios of satellite transmission signals at each historical moment;

[0083] The signal-to-noise ratio increment is the difference between the latest signal-to-noise ratio and the earliest signal-to-noise ratio in the signal-to-noise ratio sequence.

[0084] The average signal-to-noise ratio gives the average quality status of satellite transmission signals at historical moments, and the signal-to-noise ratio increment reflects the changing trend of the signal-to-noise ratio.

[0085] In this embodiment, the process of calculating the demand coefficient includes:

[0086] Determine whether the latest bit error rate is greater than the target bit error rate. If so, subtract the target bit error rate from the latest bit error rate to obtain the bit error rate difference, and take the ratio of the bit error rate difference to the target bit error rate as the first demand component. If not, assign 0 to the first demand component;

[0087] Determine whether the latest signal-to-noise ratio is less than the target signal-to-noise ratio. If so, subtract the latest signal-to-noise ratio from the target signal-to-noise ratio to obtain the signal-to-noise ratio difference, and take the ratio of the signal-to-noise ratio difference to the target signal-to-noise ratio as the second demand component; if not, assign 0 to the second demand component;

[0088] Add the first demand component and the second demand component to obtain the demand coefficient.

[0089] Voice communication has a relatively high tolerance for bit error rate, and the target bit error rate is set between 0.001 and 0.01. Data communication has extremely high requirements for accuracy, and the target bit error rate can be set between 0.000001 and 0.000000001. The target signal-to-noise ratio for ordinary voice calls is basically 15 - 25 dB to ensure clear voice signals, and the target signal-to-noise ratio for high-quality voice communication is 25 - 35 dB. For general data transmission (such as web browsing, simple file downloads, etc.): the target signal-to-noise ratio is usually 20 - 30 dB. For critical data transmission (such as financial transaction data, scientific research data transmission, etc.): the target signal-to-noise ratio often requires 25 - 40 dB or even higher.

[0090] The present invention determines whether to calculate the first demand component by judging the magnitude relationship between the latest bit error rate and the target bit error rate. When the latest bit error rate is greater than the target bit error rate, the ratio of the bit error rate difference to the target bit error rate is used as the first demand component to quantify the degree of deviation of the bit error rate from the target value. If the bit error rate exceeds the target value by a large margin, the value of the first demand component is large, indicating an urgent need for adjusting the transmission power due to the bit error rate problem; if it does not exceed the target value, it is assigned a value of 0, indicating that the current bit error rate is within the acceptable range and there is no such power adjustment requirement.

[0091] For the signal-to-noise ratio, judge the magnitude between the latest value and the target value. If the latest signal-to-noise ratio is less than the target signal-to-noise ratio, subtract the latest signal-to-noise ratio from the target signal-to-noise ratio, and take the ratio of the signal-to-noise ratio difference to the target signal-to-noise ratio as the second demand component to reflect the deviation of the latest signal-to-noise ratio. If the target value is satisfied, it is assigned a value of 0, that is, there is no power adjustment requirement driven by this factor.

[0092] Adding the first demand component and the second demand component to obtain the demand coefficient realizes the comprehensive consideration of the adjustment demand of the satellite communication system for the transmission power from the perspectives of two key signal quality indicators, namely the bit error rate and the signal-to-noise ratio.

[0093] As Figure 2 shown, the fully connected layer - BP neural network includes: the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, the input layer, the first hidden layer, the second hidden layer, and the output layer;

[0094] The input end of the first fully connected layer is used to input the first set of background noise features;

[0095] The input end of the second fully connected layer is used to input the second set of background noise features;

[0096] The input end of the third fully connected layer is used to input the third set of background noise features;

[0097] The input end of the fourth fully connected layer is used to input a set of bit error rate features;

[0098] The input end of the fifth fully connected layer is used to input a set of signal-to-noise ratio features;

[0099] The input layer is used to input the outputs of the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, and the demand coefficient into the input end of the first hidden layer;

[0100] The output end of the first hidden layer is connected to the input end of the second hidden layer;

[0101] The input end of the output layer is connected to the output end of the second hidden layer, and its output end serves as the output end of the fully connected layer - BP neural network.

[0102] In this embodiment, the expressions of the fully connected layers are all: y = f(ω x1 x1 + b x1 ) + f(ω x2 x2 + b x2 ), where y is the output of the fully connected layer, x1 is the mean value, x2 is the increment, ω x1 is the weight of x1, b x1 is the bias of x1, ω x2 is the weight of x2, b x2 is the bias of x2, and f is the activation function.

[0103] As Figure 2 shown, there are 6 nodes in each of the input layer, the first hidden layer, and the second hidden layer in the fully connected layer - BP neural network.

[0104] The present invention performs feature fusion through each fully connected layer, and then inputs the first, second, third, fourth, and fifth fully connected layers and the demand coefficient into the BP neural network to achieve the improvement of the accuracy of predicting the transmit power increment through multi - dimensional feature combination.

[0105] In this embodiment, the fully connected layer - BP neural network can be trained using the existing gradient descent method. More preferably, the process of training the fully connected layer - BP neural network includes: a first training stage and a second training stage. The first training stage is used to train the fully connected layer - BP neural network according to the gap between the transmit power increment predicted by the fully connected layer - BP neural network and the maximum transmit power increment; the second training stage is used to retrain the fully connected layer - BP neural network after the first training stage to find the optimal weights and biases corresponding to the minimum transmit power increment.

[0106] The process of the first training stage includes:

[0107] Input the i - th sample into the fully connected layer - BP neural network to obtain the transmit power increment of the i - th training, where i is the number of training times in the first training stage;

[0108] Take the absolute value of the difference between the transmit power increment of the i - th training and the maximum transmit power increment as the first target value;

[0109] Update the weights and biases in the fully connected layer - BP neural network according to the first target value, increment i by 1, and loop the training process until the first target value is less than the target threshold, and the first - stage training of the fully connected layer - BP neural network is completed.

[0110] In this embodiment, the specific value of the target threshold can be set according to requirements. When higher training accuracy is required, the target threshold can be set to be close to 0, so that the predicted transmit power increment is extremely close to the maximum transmit power increment.

[0111] The process of the second training stage includes:

[0112] Input the samples of the nth training into the fully connected layer - BP neural network after the first training stage (i.e., use the weights and parameters of the first training stage as the starting values of the second training stage) to obtain the transmit power increment of the nth training, where n is the number of training times in the second training stage;

[0113] When the transmit power increment of the nth training is greater than the maximum transmit power increment, take the maximum transmit power increment as the transmit power increment of the nth training, and add the transmit power increment of the nth training to the initial transmit power to obtain the transmit power of the nth training;

[0114] Transmit the satellite test signal using the transmit power of the nth training;

[0115] Statistically calculate the bit error rate and signal - to - noise ratio of the satellite test signal received at the receiving end, and calculate the second target value based on the objective function;

[0116] Update the weights and biases in the fully connected layer - BP neural network according to the second target value, increment n by 1, and loop the training process until the number of training times reaches the target number of training times;

[0117] In all the trainings of the second training stage, mark the number of training times with a bit error rate less than the target bit error rate and a signal - to - noise ratio greater than the target bit error rate as the pending number of times;

[0118] Among all the pending numbers of times, mark the pending number of times corresponding to the minimum transmit power increment as the target number of times;

[0119] Take the weights and biases corresponding to the target number of times as the optimal parameters, and the training of the fully connected layer - BP neural network is completed.

[0120] In the present invention, the in - sample data includes: multiple groups of background noise characteristics, one group of bit error rate characteristics, one group of signal - to - noise ratio characteristics, and a demand coefficient.

[0121] The first training stage trains the network with the gap between the predicted transmit power increment and the maximum transmit power increment as the first target value, which can guide the network to adjust the parameters in the direction of the maximum transmit power increment, enabling the network to have the weights and biases for predicting the maximum transmit power increment. However, the maximum transmit power increment is not the optimal choice, so the second training stage is set up to find the weights and biases that meet the requirements.

[0122] In the second training stage, when the predicted transmit power increment is greater than the maximum transmit power increment, the maximum transmit power increment is taken as the actual transmit power increment, which avoids the transmit power exceeding the maximum value allowed by the system, ensures the safe operation of satellite communication equipment, prevents damage to the equipment caused by excessive transmit power, and also avoids unnecessary interference to other communication systems.

[0123] By transmitting satellite test signals and counting the bit error rate and signal-to-noise ratio at the receiving end to calculate the second target value, and then updating the weights and biases of the network based on this target value. The training method based on the actual communication effect enables the network to learn the real relationship between the transmit power and the communication quality, rather than just the theoretical numerical relationship. Thus, in actual applications, the network can more accurately adjust the transmit power according to the feedback of the communication quality, improving the reliability and stability of communication.

[0124] Since the satellite is powered by batteries and the adjustment of the satellite's transmit power cannot be increased without limit, therefore, the number of times corresponding to the minimum transmit power increment is selected as the target number of times among the pending number of times, so as to find the training result that makes the transmit power adjustment most reasonable (with the smallest increment) on the premise of meeting the communication quality requirements. The corresponding weights and biases are used as the optimal parameters. In subsequent applications, the neural network can not only ensure the communication quality but also adjust the transmit power in an energy-saving manner.

[0125] In this embodiment, when the bit error rate is greater than the target bit error rate and the signal-to-noise ratio is less than the target signal-to-noise ratio, the objective function is: ; when the bit error rate is greater than the target bit error rate and the signal-to-noise ratio is greater than or equal to the target signal-to-noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal-to-noise ratio is less than the target signal-to-noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal-to-noise ratio is greater than or equal to the target signal-to-noise ratio, the objective function is: G = c, where B act is the bit error rate of the satellite test signal, B tar is the target bit error rate, S act is the signal-to-noise ratio of the satellite test signal, S tar is the target signal-to-noise ratio, G is the second target value, and c is a constant between 0 and 1.

[0126] The present invention takes the ratio of the bit error rate to the target bit error rate when the bit error rate is greater than the target bit error rate, and takes the ratio of the target signal-to-noise ratio to the signal-to-noise ratio when the signal-to-noise ratio is less than the target signal-to-noise ratio, reflecting the situation where the signal-to-noise ratio is less than the target signal-to-noise ratio. This makes the target value of the objective function larger when the bit error rate is greater and the signal-to-noise ratio is smaller, so that the decrease amplitude of the weight and bias changes with the signal-to-noise ratio and the bit error rate. When both the bit error rate and the signal-to-noise ratio meet the requirements, a constant between 0 and 1 is assigned to the target value to facilitate the update of the weight and bias. For example, c is 0.1. A smaller value of c results in a more detailed search for weights and biases, requiring more training times.

[0127] In this embodiment, in the second stage, the target number of training times is set at 500 times.

[0128] In this embodiment, in the first training phase, the formula for updating weights and biases is: , , where ω i+1 is the weight of the i+1th training, b i+1 is the bias of the i+1th training, ω i is the weight of the i-th training, b i is the bias of the i-th training, G i is the first target value of the i-th training, and μ is the step size.

[0129] In the second training phase, the formula for updating weights and biases is: , , where ω n+1 is the weight of the n+1th training, b n+1 is the bias of the n+1th training, ω n is the weight of the nth training, b n is the bias of the nth training, G n is the second target value of the nth training.

[0130] In this embodiment, the empirical value of the step length μ is 0.001, and the specific value can be adjusted as needed during the training process.

[0131] This method constructs a background noise power sequence by scanning the satellite communication frequency band and adjacent frequency bands. This method combines features extracted from bit error rate and signal-to-noise ratio sequences with demand coefficients derived from the differences between the bit error rate and signal-to-noise ratio and target values. This provides a rich and comprehensive data sample for the fully connected layer-based BP neural network. Compared to traditional methods that rely solely on the receiving end's signal-to-noise ratio, this method can more accurately calculate the transmit power increment, significantly improving the precision of transmit power regulation and effectively ensuring the stability and accuracy of signal transmission.

[0132] The present invention takes into account the influence of the real-time spectrum state of the frequency band used for satellite communication and its adjacent frequency bands on the signal. At the same time, according to the latest bit error rate and the gap between the signal-to-noise ratio and the target value, a demand coefficient is obtained to characterize the actual demand for transmit power adjustment in the current satellite communication system, thereby improving the prediction accuracy of the fully connected layer-BP neural network.

[0133] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for adjusting transmission power in satellite communication, characterized in that, Including the following steps: Scanning the frequency bands used for satellite communication and adjacent frequency bands through the spectrum monitoring device on the satellite to construct multiple background noise power sequences; Extracting features from each background noise power sequence to obtain a set of background noise features; According to the bit error rate and signal-to-noise ratio of each satellite transmission signal, constructing a bit error rate sequence and a signal-to-noise ratio sequence, and respectively extracting features from the bit error rate sequence and the signal-to-noise ratio sequence to obtain a set of bit error rate features and a set of signal-to-noise ratio features; Obtaining a demand coefficient according to the gap between the latest bit error rate and the target bit error rate, and the gap between the latest signal-to-noise ratio and the target signal-to-noise ratio; Taking multiple sets of background noise features, a set of bit error rate features, a set of signal-to-noise ratio features, and the demand coefficient as samples and inputting them into the trained fully connected layer - BP neural network to obtain the transmit power increment; Adding the transmit power increment to the initial transmit power to obtain the transmit power of the next signal; The process of training the fully connected layer - BP neural network includes: a first training stage and a second training stage. The first training stage is used to train the fully connected layer - BP neural network according to the gap between the transmit power increment predicted by the fully connected layer - BP neural network and the maximum transmit power increment; the second training stage is used to retrain the fully connected layer - BP neural network after the first training stage to find the optimal weights and biases corresponding to the minimum transmit power increment; The process of the second training stage includes: Inputting the samples of the nth training into the fully connected layer - BP neural network after the first training stage to obtain the transmit power increment of the nth training, where n is the number of training times in the second training stage; When the transmit power increment of the nth training is greater than the maximum transmit power increment, taking the maximum transmit power increment as the transmit power increment of the nth training, and adding the transmit power increment of the nth training to the initial transmit power to obtain the transmit power of the nth training; Transmitting a satellite test signal using the transmit power of the nth training; Statistically analyzing the bit error rate and signal-to-noise ratio of the satellite test signal received at the receiving end, and calculating a second target value based on the objective function; Updating the weights and biases in the fully connected layer - BP neural network according to the second target value, incrementing n by 1, and looping the training process until the number of training times reaches the target number of training times; In all the trainings in the second training stage, marking the number of training times when the bit error rate is less than the target bit error rate and the signal-to-noise ratio is greater than the target bit error rate as the pending number of times; Among all the pending numbers of times, marking the pending number of times corresponding to the minimum transmit power increment as the target number of times; Taking the weights and biases corresponding to the target number of times as the optimal parameters, and the training of the fully connected layer - BP neural network is completed.

2. The method for adjusting the transmission power in satellite communication according to claim 1, wherein The process of constructing multiple background noise power sequences includes: When the satellite does not transmit signals, scanning the frequency bands used for satellite communication through the spectrum monitoring device on the satellite to obtain the first background signals in each historical time period, calculating the background noise power, and constructing the first background noise power sequence; When the satellite does not transmit signals, a neighboring frequency band of the frequency band used for satellite communication is scanned by the spectrum monitoring device on the satellite to obtain the second background signals in each historical time period, calculate the background noise power, and construct the second background noise power sequence; When the satellite does not transmit signals, another neighboring frequency band of the frequency band used for satellite communication is scanned by the spectrum monitoring device on the satellite to obtain the third background signals in each time period, calculate the background noise power, and construct the third background noise power sequence.

3. The method for adjusting the transmission power in satellite communication according to claim 2, wherein, The formula for calculating the background noise power is as follows: , where P is the background noise power, and x n is the nth signal value in the background signal for a period of time, N is the length of the background signal for a period of time, | | is the absolute value symbol, and n is a positive integer.

4. The method for adjusting the transmission power in satellite communication according to claim 1, wherein A set of background noise characteristics includes: the mean background noise power and the background noise power increment; The mean background noise power is the mean of the background noise power sequence; The background noise power increment is the difference between the latest background noise power and the earliest background noise power in the background noise power sequence.

5. The method for adjusting the transmission power in satellite communication according to claim 1, wherein, A set of bit error rate characteristics includes: the mean bit error rate and the bit error rate increment; The mean bit error rate is the mean of the bit error rate sequence composed of the bit error rates of the satellite transmitted signals at each historical moment; The bit error rate increment is the difference between the latest bit error rate and the earliest bit error rate in the bit error rate sequence; A set of signal-to-noise ratio characteristics includes: the mean signal-to-noise ratio and the signal-to-noise ratio increment; The mean signal-to-noise ratio is the mean of the signal-to-noise ratio sequence composed of the signal-to-noise ratios of the satellite transmitted signals at each historical moment; The signal-to-noise ratio increment is the difference between the latest signal-to-noise ratio and the earliest signal-to-noise ratio in the signal-to-noise ratio sequence.

6. The method for adjusting the transmission power in satellite communication according to claim 1, characterized in that, The process of calculating the demand coefficient includes: Judge whether the latest bit error rate is greater than the target bit error rate. If so, subtract the target bit error rate from the latest bit error rate to obtain the bit error rate difference, and take the ratio of the bit error rate difference to the target bit error rate as the first demand component. If not, assign 0 to the first demand component; Judge whether the latest signal-to-noise ratio is less than the target signal-to-noise ratio. If so, subtract the latest signal-to-noise ratio from the target signal-to-noise ratio to obtain the signal-to-noise ratio difference, and take the ratio of the signal-to-noise ratio difference to the target signal-to-noise ratio as the second demand component; If not, assign 0 to the second demand component; Add the first demand component and the second demand component to obtain the demand coefficient.

7. The method for adjusting the transmission power in satellite communication according to claim 1, wherein The fully connected layer - BP neural network includes: the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, the input layer, the first hidden layer, the second hidden layer, and the output layer; The input end of the first fully connected layer is used to input the first set of background noise characteristics; The input end of the second fully connected layer is used to input the second set of background noise characteristics; The input end of the third fully connected layer is used to input the third set of background noise characteristics; The input end of the fourth fully connected layer is used to input a set of bit error rate characteristics; The input end of the fifth fully connected layer is used to input a set of signal-to-noise ratio characteristics; The input layer is used to input the outputs of the first fully connected layer, the second fully connected layer, the third fully connected layer, the fourth fully connected layer, the fifth fully connected layer, and the demand coefficient into the input end of the first hidden layer; The output end of the first hidden layer is connected to the input end of the second hidden layer; The input end of the output layer is connected to the output end of the second hidden layer, and its output end serves as the output end of the fully connected layer - BP neural network.

8. The method for adjusting the transmission power in satellite communication according to claim 1, wherein When the bit error rate is greater than the target bit error rate and the signal-to-noise ratio is less than the target signal-to-noise ratio, the objective function is: ; when the bit error rate is greater than the target bit error rate and the signal-to-noise ratio is greater than or equal to the target signal-to-noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal-to-noise ratio is less than the target signal-to-noise ratio, the objective function is: ; when the bit error rate is less than or equal to the target bit error rate and the signal-to-noise ratio is greater than or equal to the target signal-to-noise ratio, the objective function is G = c, where B act is the bit error rate of the satellite test signal, B tar is the target bit error rate, S act is the signal-to-noise ratio of the satellite test signal, S tar is the target signal-to-noise ratio, G is the second target value, and c is a constant between 0 and 1.

Images