Short-wave communication highest available frequency prediction method and system based on long and short term prediction data fusion processing

Abstract

The invention discloses a short-wave communication highest available frequency prediction method based on long-term and short-term prediction data fusion processing. The method comprises the following steps: collecting F2 layer critical frequency in vertical detection data as long-term prediction data; identifying and marking an abnormal value in the original data, replacing the abnormal value by using long-term prediction data at the same moment in the current month, and taking data of an adjacent day after the abnormal value is replaced as short-term prediction data; performing data fusion on the long-term prediction data and the short-term prediction data to obtain continuous time sequence data; and inputting continuous time sequence data into the trained LSTM neural network model to obtain the highest available frequency of the short wave in the future time. According to the method, real-time fusion observation is estimated through Kalman filtering recursion, outliers are dynamically inhibited, and the optimal steady state of a linear system is kept; the data fusion technology is adopted to predict the highest available frequency, the characteristics of short-term prediction and long-term prediction are combined, and the highest available frequency of short-wave communication can be predicted more accurately in a long time.

Description

Technical Field

[0001] This application relates to the field of communication technology, and more specifically, to a method and system for predicting the highest available frequency for shortwave communication based on long-term and short-term prediction data fusion processing. Background Technology

[0002] In shortwave communication, the Maximum Usable Frequency (MUF) refers to the highest frequency of radio waves that can be reflected back to the ground by the ionosphere under specific ionospheric conditions and communication distances. If the radio wave frequency exceeds the MUF, it will penetrate the ionosphere and enter space, preventing skywave propagation. Radio waves with frequencies below the MUF may be reflected, but multipath effects and signal attenuation must be considered. The MUF value is influenced by various factors, including ionospheric electron density distribution, solar activity cycles, seasons, diurnal variations, and communication distance.

[0003] In existing technologies, using long-term forecast outliers to process short-term predicted MUFs results in a relatively large mean absolute error. This is because long-term forecast models utilize data such as monthly midpoints and critical frequencies of different ionospheric layers. While ionospheric models built by integrating these data can reasonably represent the average ionospheric state, their prediction accuracy is low when forecasting ionospheric changes at specific times. This is because the ionosphere undergoes significant changes in a short period, not only in its spatial distribution of latitude and longitude but also influenced by various factors such as diurnal variation, seasonality, and the solar cycle. These factors work together to cause drastic changes in the ionospheric state within a short time, making accurate prediction difficult using simple models. Therefore, using long-term forecast outliers to process short-term predicted MUFs results in a large mean absolute error.

[0004] With the deepening of ionospheric research and the increasing demands on shortwave communication quality from system applications, improving the prediction accuracy of the highest available frequency (MUF) of shortwave communication has become an urgent problem to be solved in the field of shortwave communication. Summary of the Invention

[0005] To address at least one deficiency or improvement need in the existing technology, this invention provides a method, system, device, and storage medium for predicting the highest available frequency (MUF) of shortwave communication based on long-term and short-term forecast data fusion processing. It uses Kalman filtering to replace outliers to ensure the continuity and accuracy of time series data in short-term forecasting, employs a data fusion method to fully utilize the advantages of long-term and short-term forecasting, and uses an LSTM model for long-term series data prediction. The organic combination of these multiple methods significantly improves the prediction accuracy of the MUF of shortwave communication.

[0006] To achieve the above objectives, according to a first aspect of the present invention, a method for predicting the highest available frequency for shortwave communication based on long-term and short-term prediction data fusion processing is provided, comprising the following steps: S101, collect the critical frequency of the F2 layer from the vertical sounding data as long-term prediction data; S201, identify and mark outliers in the original data, replace the outliers with long-term forecast data at the same time in the current month, and use the data of the next day after replacing the outliers as short-term forecast data. S301, The long-term forecast data and the short-term forecast data are fused to obtain continuous time series data; S401, The continuous time series data is input into the trained LSTM neural network model to obtain the highest available shortwave frequency for future times.

[0007] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S201 specifically includes: Outliers in the original data are identified and labeled. The N consecutive values ​​preceding the outlier are used as inputs for Kalman filter prediction. The outlier is replaced with the output value of the Kalman filter prediction to ensure the continuity and integrity of the time series data, where N is a positive integer.

[0008] Furthermore, the above-mentioned method for predicting the highest available frequency for shortwave communication also includes: The highest usable frequency for the nearby day is obtained by inverting ionospheric parameters using the critical frequency of the F2 layer obtained from vertical probes of the nearby day.

[0009] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S301 specifically includes: The long-term forecast data and the highest available frequency obtained from the vertical detection data are spliced ​​and fused to obtain continuous time series data.

[0010] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S401 further includes: Create an LSTM neural network model and configure the number of LSTM layers, the number of neurons, and the activation function. The continuous time series with the highest available frequency of the fused data is used as input, and the highest available frequency of the predicted day is used as output.

[0011] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, the training of the LSTM neural network model specifically includes: The data is divided into training set, validation set, and test set; The LSTM neural network model is trained using the training set data, and the network weights are adjusted using the backpropagation algorithm to minimize the prediction error. The trained LSTM neural network model is then validated and tested using the validation set and test set, respectively.

[0012] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, the workflow of the LSTM neural network model specifically includes: S4011, the LSTM neural network receives the input data at each time step and the hidden state and memory cell state of the previous time step; S4012, the input gate and forget gate of the LSTM neural network determine the information added to the memory cell and the information deleted from the memory cell; S4013, Update the state of the memory unit, update and integrate the information in the memory unit; S4014, the output gate determines the information read from the memory cell and uses it as the output information for the current time step.

[0013] According to a second aspect of the present invention, a shortwave communication maximum available frequency prediction system based on long- and short-term prediction data fusion processing is also provided, comprising: The long-term prediction data acquisition module is configured to collect the critical frequency of the F2 layer from the vertical probe data as long-term prediction data. The short-term forecast data acquisition module is configured to identify and mark outliers in the original data, replace the outliers with long-term forecast data at the same time in the current month, and use the data of the next day after replacing the outliers as short-term forecast data. A data fusion module is configured to fuse the long-term forecast data and the short-term forecast data to obtain continuous time series data. The data prediction module is configured to input the continuous time series data into a trained LSTM neural network model to obtain the highest available shortwave frequencies for future times.

[0014] According to a third aspect of the present invention, a shortwave communication maximum available frequency prediction device based on long-term and short-term prediction data fusion processing is also provided, which includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the steps of any of the methods described above.

[0015] According to a fourth aspect of the invention, a storage medium is also provided, which stores a computer program executable by an access authentication device, which performs the steps of any of the methods described above when the computer program is running.

[0016] According to a fifth aspect of the present invention, a computer program product is also provided, comprising a computer program / instructions, characterized in that, when executed by a processor, the computer program / instructions implement the steps of any of the methods described above.

[0017] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: (1) The present invention uses Kalman filter recursive estimation to fuse observations in real time, dynamically suppress outliers and maintain the optimal steady state of the linear system; LSTM uses a three-gating mechanism to accumulate long-term memory, automatically extract the deep patterns of the solar cycle, and has a robust gradient and wide generalization. It significantly reduces the mean absolute error on the measured set, taking into account both transient and long-term forecast accuracy.

[0018] (2) The present invention uses data fusion technology to predict the highest available frequency; it combines the characteristics of short-term and long-term prediction, and can more accurately predict the highest available frequency of shortwave communication over a longer period of time. Compared with short-term or long-term prediction alone, the data fusion method has a greater advantage in prediction accuracy.

[0019] (3) The data fusion technology of the present invention can accurately predict the highest available frequency of shortwave communication in the next few days under limited information conditions in unfamiliar areas and in situations where short-term forecast data is relatively scarce, thereby improving the shortwave communication guarantee capability. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a method for predicting the highest available frequency of shortwave communication based on long-term and short-term prediction data fusion processing, provided for an embodiment of this application; Figure 2 This is a schematic diagram of the LSTM neural network model structure provided in the embodiments of this application; Figure 3 A schematic diagram comparing short-term prediction and fusion prediction 12h MUF provided in the embodiments of this application; Figure 4 A schematic diagram comparing the 12-hour MUF error of short-term prediction and fusion prediction provided in the embodiments of this application; Figure 5 A schematic diagram comparing short-term and long-term 12h MUF predictions provided in this application embodiment; Figure 6 A schematic diagram comparing the 12h MUF error of short-term and long-term predictions provided in the embodiments of this application; Figure 7 A schematic diagram comparing fusion prediction and long-term prediction 12h MUF provided in the embodiments of this application; Figure 8 A schematic diagram showing the comparison of fusion prediction and long-term prediction 12h MUF error provided in the embodiments of this application; Figure 9 A schematic diagram comparing the long-term forecast and the short-term forecast 1-day MUF provided in the embodiments of this application; Figure 10 A schematic diagram comparing the 1-day MUF error of long-term forecast and short-term forecast provided in the embodiments of this application; Figure 11 This is a schematic diagram comparing long-term prediction and fusion prediction 1-day MUF provided in an embodiment of this application; Figure 12 A schematic diagram comparing the 1-day MUF error of long-term prediction and fusion prediction provided in the embodiments of this application; Figure 13 This is a schematic diagram comparing the short-term prediction and the fusion prediction 1-day MUF provided in the embodiments of this application; Figure 14 This is a schematic diagram comparing the 1-day MUF error of short-term prediction and fusion prediction provided in the embodiments of this application; Figure 15 A schematic diagram comparing the long-term prediction and the fusion prediction 3-day MUF provided in the embodiments of this application; Figure 16 A schematic diagram showing the comparison of 3-day MUF errors between long-term prediction and fusion prediction provided in the embodiments of this application; Figure 17 A schematic diagram comparing the 3-day MUF of long-term forecast and short-term forecast provided in the embodiments of this application; Figure 18 A schematic diagram comparing the 3-day MUF error of long-term and short-term forecasts provided in the embodiments of this application; Figure 19 This is a schematic diagram comparing the short-term prediction and the fusion prediction 3-day MUF provided in the embodiments of this application; Figure 20 A schematic diagram showing the comparison of 3-day MUF errors between short-term prediction and fusion prediction provided in the embodiments of this application; Figure 21 This is a schematic diagram illustrating the comparison between the fused 2-day data prediction and the actual MUF provided in an embodiment of this application. Figure 22 This is a schematic diagram showing the comparison between the predicted error of fused 2-day data and the actual MUF, provided in an embodiment of this application. Figure 23A schematic diagram of a shortwave communication maximum available frequency prediction system based on data fusion is provided for an embodiment of this application; Detailed Implementation To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0023] According to a first aspect of the present invention, a method for predicting the highest available frequency for shortwave communication based on long-term and short-term prediction data fusion processing is provided, as shown in the appendix. Figure 1 As shown, it includes the following steps: S101, the critical frequency of the F2 layer in the vertical sounding data is collected as long-term prediction data. It should be noted that using long-term prediction outliers to process short-term predicted MUF has a relatively large mean absolute error. This is because long-term prediction models use data such as monthly mid-terms and critical frequencies of various ionospheric layers. Although the ionospheric model established by integrating these data can reasonably represent the average ionospheric state, its prediction accuracy is not high when predicting ionospheric changes at specific times. This is because the ionosphere undergoes significant changes in a short period, which are not only reflected in its spatial distribution of latitude and longitude but also affected by various factors such as daytime, seasons, and solar cycles. These factors work together to cause drastic changes in the ionospheric state in a short period, making accurate prediction difficult using simple models. Therefore, using long-term prediction outliers to process short-term predicted MUF has a large mean absolute error. Furthermore, the data collected in this invention comes from the vertical sounding station established by the National Space Science Data Center in Zuoling Town, Wuhan, and the critical frequency of the F2 layer in the vertical sounding data from March 1st to 4th, 2024, is selected as the calculation data.

[0024] S201, identify and label outliers in the original data, replace the outliers with long-term forecast data from the same time of the current month, and use the data from the nearest day after the outlier replacement as short-term forecast data. It should be noted that this step is used to process short-term forecast data, specifically by identifying and labeling outliers in the original data. For detected outliers, long-term forecast data from the same time of the current month (i.e., March 2024) is used as a replacement value; after replacing the outlier, the continuity and integrity of the time series data are ensured. Furthermore, for detected outliers, the five consecutive values ​​preceding the outlier are used as input to the Kalman filter prediction, and the resulting output value is used as a replacement value; after replacing the outlier, the continuity and integrity of the time series data are ensured. Kalman filtering is a recursive estimation method that can update the filter value in real time based solely on newly observed data without relying on a large amount of known data, maintaining optimal results. It is particularly suitable for linear dynamic systems, combining the system's predicted values ​​and actual observed values, and providing the optimal estimate of the system state through an optimization process. When dealing with data containing outliers, Kalman filtering can reduce the impact of outliers on the estimation results by adjusting its noise covariance matrix, thereby maintaining the stability of the prediction.

[0025] S301, the long-term forecast data and the short-term forecast data are fused to obtain continuous time series data. It should be noted that this step is used to achieve the fusion of long-term and short-term forecast data. Due to electromagnetic environmental limitations, there is limited early-stage data for short-term predictions of the highest available frequency for ionospheric shortwave communication, and the accuracy of the predicted longitude decreases over time, affecting the communication quality of shortwave communication. To compensate for the lack of data, long-term forecast data is used as a supplement; that is, long-term forecast data and recent short-term forecast data are fused to increase the length of known data, thereby improving the accuracy of frequency prediction and addressing time issues. The basic principle of data fusion is to fully utilize the long-term forecast model based on ITU-R P.533 and short-term predictions based on vertical sounding data, and then use a neural network to provide a highly accurate high-availability frequency for shortwave communication over a relatively long period. This invention uses the ITU-R P.533 model as a background and comprehensively utilizes the LSTM neural network algorithm for data processing and fusion, thereby achieving the fusion of the long-term forecast model and short-term prediction data based on vertical sounding data.

[0026] S401, the continuous time series data is input into the trained LSTM neural network model to obtain the highest available shortwave frequencies in the future. It should be noted that the construction and training of the LSTM neural network model includes: creating an LSTM neural network model that takes the F2 layer critical frequency time series processed by Kalman filtering as input; dividing the data into training, validation, and test sets; training the LSTM model and adjusting the network weights. The prediction generation process specifically involves: using the trained LSTM model for prediction, inputting the processed continuous time series into the LSTM neural network, and outputting the highest available shortwave frequencies over a longer future period.

[0027] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S201 specifically includes: Outliers in the original data are identified and marked. The N consecutive values ​​preceding the outlier are used as input to a Kalman filter prediction. The output value of the Kalman filter prediction is then used to replace the outlier, ensuring the continuity and integrity of the time series data. Here, N is a positive integer, such as 5, 6, 8, or other suitable values. It should be noted that the Kalman filter prediction process of this invention requires identifying and marking outliers in the original data; for each detected outlier, the preceding, for example, five values ​​are used as input to the Kalman filter prediction, and the resulting output value is used as a replacement value; after replacing the outlier, the continuity and integrity of the time series data are ensured.

[0028] Furthermore, the above-mentioned method for predicting the highest available frequency for shortwave communication also includes: The highest usable frequency for shortwave communication on the nearest day is obtained by inverting ionospheric parameters using the F2 layer critical frequencies obtained from vertical sounding on the nearest day. It should be noted that because ionospheric sounding equipment can only provide vertical sounding data one day in advance, a data fusion technique will be used to provide a more accurate 5-day maximum usable frequency for shortwave communication. This involves fusing long-term forecast data with vertical sounding data to obtain the maximum usable frequency for shortwave communication on the next day. Specifically, the highest usable frequencies for 3rd and 4th days are obtained by inverting ionospheric parameters using the F2 layer critical frequencies obtained from vertical sounding on the 3rd and 4th days.

[0029] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S301 specifically includes: The highest available frequency obtained from the long-term forecast data and the vertical sensing data is then concatenated and fused to obtain continuous time-series data. It should be noted that, based on the aforementioned highest available frequency, the long-term forecast data and the highest available frequency obtained from the vertical sensing data are concatenated and fused to obtain continuous time-series data.

[0030] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, step S401 further includes: Create an LSTM neural network model, configuring the number of LSTM layers, neurons, and activation functions. Take the continuous time series of the highest available frequency from the fused data as input, and predict the highest available frequency for each day as output. It's important to note that this LSTM neural network model takes the time series of the highest available frequency from the processed fused data as input. Design the network structure based on the characteristics of the data, including the number of LSTM layers, neurons, and activation functions.

[0031] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, the training of the LSTM neural network model specifically includes: The data is divided into training, validation, and test sets. The LSTM neural network model is trained using the training set data, and the network weights are adjusted using the backpropagation algorithm to minimize prediction error. The trained LSTM neural network model is then validated and tested using the validation and test sets, respectively. It should be noted that during the training of the LSTM neural network model, the data is divided into training, validation, and (optionally) test sets. The LSTM model is trained using the training set data, and the network weights are adjusted using the backpropagation algorithm to minimize prediction error. The trained LSTM model is then used for prediction; the processed continuous time series data is input into the LSTM neural network, and the highest available frequency for March 5th is output.

[0032] Furthermore, LSTM uses its gating mechanism to control the flow of information, enabling the network to learn long-term dependencies in time series data. However, LSTM neural networks may be affected to some extent when dealing with data containing outliers, as outliers can disrupt the inherent patterns and regularities of the data. Kalman filtering shows better mean absolute error (MAO) in short-term predictions compared to LSTM neural networks when handling outliers, demonstrating the advantage of Kalman filtering in outlier handling. Although LSTM neural networks are very powerful in processing time series data, Kalman filtering, with its optimization and recursive characteristics, may better maintain prediction stability when dealing with outliers. Therefore, it can be concluded that Kalman filtering demonstrates better performance in short-term predictions when handling outliers.

[0033] The advantages of LSTM neural networks in predicting the highest available frequency (MUF) for shortwave communication are mainly reflected in their long-term memory capability, time series modeling ability, ability to solve the vanishing and exploding gradient problems, adaptability and generalization ability, and practical results. LSTM neural networks, through their unique gating mechanisms (including input gate, forget gate, and output gate), can effectively process and remember key information in long sequences. This characteristic allows LSTM to consider data changes over longer periods when dealing with ionospheric maximum available frequency (MUF) prediction, thus making more accurate predictions. In ionospheric MUF prediction, changes in time series data are crucial to the accuracy of the prediction results. LSTM can effectively utilize this information, learning more long-term and complex dependencies during training, thereby improving prediction accuracy. LSTM neural networks have strong adaptability and generalization ability, achieving good performance on different datasets and tasks. In ionospheric MUF prediction, LSTM can adapt to different ionospheric conditions and change patterns, thus making more accurate predictions. This paper verifies the effectiveness and accuracy of LSTM in ionospheric prediction through actual data processing and model training.

[0034] LSTM manages the flow of information through three main gate structures. The input gate decides whether to allow new information to be added to the current state. This selective information addition mechanism allows LSTM to focus on information beneficial to the task while ignoring irrelevant details. The forget gate filters information, determining what to retain and what to forget based on the degree of forgetting of previous information. This selective forgetting allows LSTM to remember key information for a long time, avoiding interference from other unnecessary information. The output gate is responsible for outputting information, determining what information to output in the next step, selectively outputting information through a gating mechanism. Through the control mechanism of these three gates, LSTM can effectively improve the dependency problem in long sequence processing. Simultaneously, due to the gating mechanism, LSTM can better manage gradients during training, thereby alleviating the gradient vanishing problem and making the model more stable and efficient when training long sequences. Its structure is shown in the attached figure. Figure 2 As shown.

[0035] Furthermore, in the above-mentioned method for predicting the highest available frequency for shortwave communication, the workflow of the LSTM neural network model specifically includes: S4011, the LSTM neural network receives the input data at each time step and the hidden state and memory cell state of the previous time step; it should be noted that at each time step, the LSTM receives the input data, the hidden state and memory cell state of the previous time step for use.

[0036] S4012, the input gate and forget gate of the LSTM neural network determine the information added to the memory unit and the information deleted from the memory unit; it should be noted that the input gate and forget gate determine which new information should be added to the memory unit and which old information should be deleted from the memory unit.

[0037] S4013, Update the state of the memory unit, update and integrate the information in the memory unit; it should be noted that updating the state of the memory unit integrates new information into the memory unit and deletes information that is no longer needed.

[0038] S4014, the output gate determines the information read from the memory cell and uses it as the output information for the current time step. It should be noted that the output gate determines which information to read from the memory cell and use it as the output information for the current time step.

[0039] Based on the above technical solutions of the present invention, the algorithm performance was analyzed.

[0040] First, a performance analysis and comparison predicting the next 12 hours is provided; see the appendix for details. Figure 3-8 As shown. Among them, the 12-hour mean absolute error (MAE) of short-term prediction is 0.89873 MHz; the 12-hour mean absolute error (MAE) of long-term prediction is 1.5984 MHz; and the 12-hour mean absolute error (MAE) of fused prediction is 1.3112 MHz.

[0041] Secondly, a performance analysis and comparison for the next day is provided in the appendix. Figure 9-14 As shown in the figure, the mean absolute error (MAE) for long-term prediction over one day is 1.2054 MHz; the mean absolute error (MAE) for short-term prediction over one day is 0.86156 MHz; and the mean absolute error (MAE) for fused prediction over one day is 1.1137 MHz. The prediction results show that, within a short timeframe, the short-term predictions are better, with a smaller mean absolute error, demonstrating that the LSTM neural network can achieve better accuracy under short-term prediction conditions. The long-term predictions have larger errors, consistent with the conclusion that real-time predictions have larger deviations and lower accuracy. The proposed data fusion-based prediction technique for the highest available frequency of shortwave communication shows better prediction accuracy within a short timeframe than the data predicted by the long-term prediction model.

[0042] Secondly, a performance analysis and comparison for the next three days is provided in the appendix. Figure 15-20 As shown. Among them, the 3-day mean absolute error (MAE) of short-term prediction is 1.2888 MHz; the 3-day mean absolute error (MAE) of fusion prediction is 0.97945 MHz; and the 3-day mean absolute error (MAE) of long-term prediction is 1.0978 MHz.

[0043] After in-depth analysis of the various prediction results, we can draw the following conclusions: 1. Performance Analysis of Short-Term Forecasts: Short-term forecasts (typically for forecasts within the next 12 hours) exhibited relatively low mean absolute error (MAE), particularly within 12-hour forecasts, where the MAE was 0.89873 MHz. However, when the forecast time span increased to 1 day or 3 days, the MAE of short-term forecasts increased significantly, reaching 0.86156 MHz and 1.2888 MHz, respectively. This indicates that while LSTM neural networks perform well in short-term forecasts, their forecasting performance is not ideal for longer time spans, with higher errors, reflecting the limitations of short-term forecasting models in dealing with long-term trend changes.

[0044] 2. Analysis of the effectiveness of long-term forecasts: In the 12-hour and 1-day forecasts, the MAEs were 1.5984 MHz and 1.2054 MHz respectively. Although slightly higher than the short-term forecasts, the error fluctuations were small, showing a relatively stable effect. When the forecast time span was further increased to 3 days, the MAE was 1.0978 MHz, still maintaining a certain degree of stability. This indicates that the long-term forecast model has certain advantages in handling long-term trend changes.

[0045] 3. Analysis of the effectiveness of data fusion prediction: The MAE of data fusion prediction within 12 hours is 1.3112 MHz, slightly higher than that of short-term prediction. However, the advantages of data fusion prediction become increasingly apparent when predicting longer future time periods. In 1-day and 3-day predictions, the MAEs are 1.1137 MHz and 0.97945 MHz, respectively, both lower than those of short-term and long-term predictions. This indicates that by integrating data from different sources or types, data fusion technology can more accurately predict the highest available frequency for shortwave communication over longer periods.

[0046] Therefore, while short-term forecasting has a low prediction error in the short term, it is less effective in the long term; long-term forecasting maintains a relatively stable effect over a longer period; and data fusion forecasting exhibits even lower prediction errors over a longer period, especially when predicting the highest available shortwave communication frequencies for the next three days, where its advantages are particularly evident. Thus, data fusion-based technology can predict the highest available shortwave communication frequencies for a relatively long period, providing strong support for related applications.

[0047] Applying the above solution of this invention to a practical case, location A (32.8953N, 115.8137E) and location B (28.11305N, 112.8144E) need to implement shortwave communication throughout the day on March 5, 2024, UTC. A fixed-frequency shortwave communication link needs to be established between the two locations for all-weather shortwave communication. A rhombic antenna is used. To ensure the reliability of shortwave communication, the operating frequency needs to be changed every hour. The ionospheric sounding equipment arrives at the designated area one day in advance and can provide ionospheric vertical sounding data for March 4. Because the ionospheric sounding equipment can only provide vertical sounding data one day in advance, in order to provide a more accurate maximum available frequency for shortwave communication on the 5th, data fusion technology will be used. By fusing long-term forecast data with vertical sounding data, the maximum available frequency for shortwave communication in the future day will be obtained. The specific steps are as follows: First, predict the long-term forecast data for March using ICEPAC software; second, perform ionospheric parameter inversion on the F2 layer critical frequencies of vertical sounding on March 3rd and 4th to obtain the highest available frequencies for March 3rd and 4th; third, concatenate and fuse the obtained long-term forecast data with the highest available frequencies obtained from vertical sounding data; fourth, create an LSTM neural network model that takes the time series of the highest available frequencies from the processed fused data as input. Design the network structure based on the characteristics of the data, including the number of LSTM layers, neurons, and activation functions. Divide the data into a training set, a validation set, and (optionally) a test set. Train the LSTM model using the training set data, and adjust the network weights using the backpropagation algorithm to minimize the prediction error; fifth, use the trained LSTM model for prediction, inputting the processed continuous time series into the LSTM neural network to output the highest available frequency for March 5th.

[0048] See the attached forecast results. Figure 21-22 Comparative analysis revealed that fusing long-term and short-term one-day forecasts resulted in a lower error, with a mean absolute error (MAE) of 0.88259 MHz, demonstrating good performance. Firstly, this is somewhat accidental, as the ionosphere is influenced by various factors such as solar radiation, atmospheric circulation, and geomagnetic environment, exhibiting complex time-varying dispersion characteristics. This characteristic means that the structure and properties of the ionosphere change significantly with time, location, and season, consequently affecting the highest usable frequency for shortwave communication. However, this also illustrates that data fusion forecasting can offer significant advantages even with limited original data.

[0049] It is evident that in predicting the highest available frequency for shortwave communication, short-term prediction demonstrates a significant advantage over a shorter period, accurately predicting frequencies within the near future. However, the effectiveness of short-term prediction decreases as the prediction range expands to longer timeframes. In contrast, long-term prediction exhibits more stable performance in predicting the highest available frequency for shortwave communication. To fully leverage the advantages of both prediction methods, this invention employs data fusion technology to predict the highest available frequency. This method combines the characteristics of both short-term and long-term prediction, enabling more accurate prediction of the highest available frequency for shortwave communication over a longer period. Compared to either short-term or long-term prediction alone, the data fusion-based method offers greater accuracy. Furthermore, the data fusion technology of this invention, even in unfamiliar regions with relatively scarce short-term prediction data, can accurately predict the highest available frequency for shortwave communication within the next few days under limited information conditions, thereby enhancing shortwave communication support capabilities.

[0050] According to a second aspect of the present invention, a system for predicting the highest available frequency for shortwave communication based on long- and short-term prediction data fusion processing is also provided, as shown in the appendix. Figure 23 As shown, it includes: The long-term prediction data acquisition module is configured to collect the critical frequency of the F2 layer from the vertical probe data as long-term prediction data. The short-term forecast data acquisition module is configured to identify and mark outliers in the original data, replace the outliers with long-term forecast data at the same time in the current month, and use the data of the next day after replacing the outliers as short-term forecast data. A data fusion module is configured to fuse the long-term forecast data and the short-term forecast data to obtain continuous time series data. The data prediction module is configured to input the continuous time series data into a trained LSTM neural network model to obtain the highest available shortwave frequencies for future times.

[0051] It should be noted that the shortwave communication maximum available frequency prediction system based on long-term and short-term prediction data fusion processing of the present invention is implemented in accordance with the steps in the aforementioned shortwave communication maximum available frequency prediction method based on data fusion, and will not be repeated here.

[0052] Overall, compared with the prior art, the above technical solutions conceived in this invention can achieve the following beneficial effects: This invention uses Kalman filtering recursively to estimate real-time fused observations, dynamically suppresses outliers, and maintains the optimal steady state of the linear system; LSTM uses a three-gating mechanism to accumulate long-term memory, automatically extracts deep patterns of the solar cycle, has robust gradients and wide generalization, significantly reduces the mean absolute error on the measured set, and takes into account both transient and long-term forecast accuracy.

[0053] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0054] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0055] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0056] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0057] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0058] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0060] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0061] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0062] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the highest available frequency for shortwave communication based on long-term and short-term forecast data fusion processing, characterized in that, Includes the following steps: S101, collect the critical frequency of the F2 layer from the vertical sounding data as long-term prediction data; S201, identify and mark outliers in the original data, replace the outliers with long-term forecast data at the same time in the current month, and use the data of the next day after replacing the outliers as short-term forecast data. S301, The long-term forecast data and the short-term forecast data are fused to obtain continuous time series data; S401, The continuous time series data is input into the trained LSTM neural network model to obtain the highest available shortwave frequency for future times.

2. The method for predicting the highest available frequency for shortwave communication as described in claim 1, characterized in that, Step S201 specifically includes: Outliers in the original data are identified and labeled. The N consecutive values ​​preceding the outlier are used as inputs for Kalman filter prediction. The outlier is replaced with the output value of the Kalman filter prediction to ensure the continuity and integrity of the time series data, where N is a positive integer.

3. The method for predicting the highest available frequency for shortwave communication as described in claim 2, characterized in that, Also includes: The highest usable frequency for the nearby day is obtained by inverting ionospheric parameters using the critical frequency of the F2 layer obtained from vertical probes of the nearby day.

4. The method for predicting the highest available frequency for shortwave communication as described in claim 1, characterized in that, Step S301 specifically includes: The long-term forecast data and the highest available frequency obtained from the vertical detection data are spliced ​​and fused to obtain continuous time series data.

5. The method for predicting the highest available frequency for shortwave communication as described in claim 1, characterized in that, Step S401 also includes: Create an LSTM neural network model and configure the number of LSTM layers, the number of neurons, and the activation function. The continuous time series with the highest available frequency of the fused data is used as input, and the highest available frequency of the predicted day is used as output.

6. The method for predicting the highest available frequency for shortwave communication as described in claim 5, characterized in that, The training of the LSTM neural network model specifically includes: The data is divided into training set, validation set, and test set; The LSTM neural network model is trained using the training set data, and the network weights are adjusted using the backpropagation algorithm to minimize the prediction error. The trained LSTM neural network model is then validated and tested using the validation set and test set, respectively.

7. The method for predicting the highest available frequency for shortwave communication as described in claim 6, characterized in that, The workflow of the LSTM neural network model specifically includes: S4011, the LSTM neural network receives the input data at each time step and the hidden state and memory cell state of the previous time step; S4012, the input gate and forget gate of the LSTM neural network determine the information added to the memory cell and the information deleted from the memory cell; S4013, Update the state of the memory unit, update and integrate the information in the memory unit; S4014, the output gate determines the information read from the memory cell and uses it as the output information for the current time step.

8. A system for predicting the highest available frequency for shortwave communication based on the fusion processing of long-term and short-term forecast data, characterized in that, include: The long-term prediction data acquisition module is configured to collect the critical frequency of the F2 layer from the vertical probe data as long-term prediction data. The short-term forecast data acquisition module is configured to identify and mark outliers in the original data, replace the outliers with long-term forecast data at the same time in the current month, and use the data of the next day after replacing the outliers as short-term forecast data. A data fusion module is configured to fuse the long-term forecast data and the short-term forecast data to obtain continuous time series data. The data prediction module is configured to input the continuous time series data into a trained LSTM neural network model to obtain the highest available shortwave frequencies for future times.

9. A device for predicting the highest available frequency for shortwave communication based on long-term and short-term forecast data fusion processing, characterized in that, The method includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images