Ionized layer parameter prediction method and device based on sequence completion strategy

Abstract

The invention discloses an ionosphere parameter prediction method and device based on a sequence completion strategy, and relates to the technical field of ionosphere parameter prediction.The method comprises the steps that historical ionosphere sequence data are subjected to signal decomposition through multiple signal decomposition algorithms, and sub-components are obtained; performing comparative analysis on the basis of the sub-components and historical ionosphere sequence data to obtain loss components, determining error loss terms based on the sub-components, and performing signal completion according to the loss components and the error loss terms to obtain completion components; and inputting the complementation component and the sub-component as prediction components into a preset deep learning neural network to obtain a prediction result. According to the method, the problem that a traditional method neglects signal loss in a signal decomposition algorithm is solved, error signals generated in the decomposition process are converted into effective signals, and meanwhile the accuracy and effectiveness of a prediction result can be greatly improved by adopting a neural network processing method.

Description

Technical Field

[0001] The present application relates to the field of shortwave communication technology, and more specifically, to an ionospheric parameter prediction method and device based on a sequence completion strategy. Background Art

[0002] Shortwave communication mainly relies on ionospheric reflection to achieve non-relay long-distance communication. It has the characteristics of simple network establishment and strong survivability. In some special application scenarios such as deserts, oceans, and jungles, satellites, microwaves and other communication guarantee means cannot play a role. Shortwave communication can be used as a backup and emergency communication means to complete communication in the absence of communication infrastructure. Therefore, shortwave communication is widely used in military communications, civil aviation, shipping, meteorology, emergency rescue and other fields.

[0003] However, shortwave communication also has some shortcomings, which are reflected in the difficulty of frequency selection, small communication capacity, congestion caused by a large number of users, and narrow available frequency bands, which directly restrict the effectiveness of shortwave communication. The ionosphere is affected by many factors such as solar and geomagnetic activities, and its change law often presents a nonlinear change law that is difficult to characterize. In order to make more accurate predictions, a prediction method with stronger fitting ability is needed. The current shortwave communication ionosphere prediction method combining deep learning and decomposition has the following shortcomings:

[0004] 1. Lack of mining of ionospheric parameter variation characteristics: There are long-term and short-term variation patterns in the current ionospheric parameter changes, and the current neural network processing method lacks in-depth mining of the above different variation characteristics.

[0005] 2. Lack of explainability: Current ionospheric parameter prediction methods widely use deep learning methods as feature extraction methods, but the corresponding methods lack reliable explanations for how these methods improve the effectiveness of shortwave communication ionospheric parameter prediction methods.

[0006] 3. There is a lack of research on the signal loss problem that occurs in the decomposition method. The existing technical route of combining deep learning with the decomposition method to predict ionospheric parameters ignores the signal loss problem that exists in the decomposition method, which results in the decomposition method being unable to achieve the desired effect. Summary of the invention

[0007] In response to at least one defect or improvement need in the prior art, the present invention provides an ionospheric parameter prediction method and device based on a sequence completion strategy, which solves the defects of the prior art. By converting the error signal generated in the decomposition process into a valid signal and adopting a neural network processing method, the accuracy and effectiveness of the prediction result can be greatly improved. At the same time, the processing efficiency of the historical ionospheric sequence can be improved, providing strong support for shortwave communication frequency selection.

[0008] To achieve the above object, according to a first aspect of the present invention, a method for predicting ionospheric parameters based on a sequence completion strategy is provided, characterized in that it includes:

[0009] The historical ionospheric sequence data are decomposed using a variety of signal decomposition algorithms to obtain subcomponents;

[0010] Based on the comparative analysis of the sub-components and the historical ionospheric sequence data, a loss component is obtained, wherein the loss component includes a signal component lost after signal decomposition corresponding to each signal decomposition algorithm;

[0011] Based on the subcomponent, an error loss term is determined, and a signal is completed according to the loss component and the error loss term to obtain a completed component;

[0012] The completion component and the subcomponent are respectively input into a preset deep learning neural network as prediction components to obtain a prediction result.

[0013] Further, the comparative analysis based on the sub-component and the historical ionospheric sequence data to obtain the loss component includes:

[0014] The subcomponent is subtracted from the historical ionospheric sequence data to obtain a loss component.

[0015] Furthermore, the determining of the error loss term based on the sub-components includes: the error loss term is the component with the highest center frequency among the sub-components.

[0016] Furthermore, the preset deep learning neural network includes: a feature extraction part, a sequence extraction part and a sequence prediction part;

[0017] The feature extraction part includes a plurality of one-dimensional convolutional layers and a maximum pooling layer, which are used to extract the features of the input sequence from different angles; wherein the one-dimensional convolutional layer is used to extract the time-related features of the input sequence, and the maximum pooling layer is used to reduce the complexity of the input sequence;

[0018] The sequence extraction part includes a multi-layer bidirectional long short-term memory neural network and an attention mechanism layer, which is used to extract the time correlation features in the input sequence; wherein the long short-term memory network is used to extract the front-back correlation features of the input sequence, and the attention mechanism layer is used to strengthen the important sequence features;

[0019] The sequence prediction part includes multiple layers of fully connected layers, which are used to linearly combine the extracted feature information to obtain the prediction result of the input sequence.

[0020] Further, performing signal completion according to the loss component and the error loss term to obtain a completed component includes:

[0021] The loss component and the error loss term are added to obtain the completion component.

[0022] Further, the signal is completed according to the loss component and the error loss term to obtain the completed component, and further includes:

[0023] Configure a completion strategy; the completion strategy includes fitting based on a univariate time series model and fitting based on local linear correlation;

[0024] The univariate time series model fitting includes:

[0025] The input sequence is first fitted through the ARIMA model to obtain the fitted sequence;

[0026] Obtaining the correlation between the input sequence and the fitting sequence;

[0027] Eliminate the input sequence items whose correlation is lower than a preset threshold to obtain an intermediate sequence;

[0028] The intermediate sequence is input into a deep learning neural network to obtain a completion component.

[0029] According to the second aspect of the present invention, there is also provided an ionospheric parameter prediction device based on a sequence completion strategy, which includes: a decomposition unit, used to decompose the historical ionospheric sequence data using multiple signal decomposition algorithms to obtain sub-components; a loss analysis unit, used to compare and analyze the sub-components with the historical ionospheric sequence data to obtain loss components, wherein the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm; a completion unit, used to determine error loss terms based on the sub-components, and perform signal completion based on the loss components and the error loss terms to obtain completed components; a prediction unit, which inputs the completed components and sub-components as prediction components into a preset deep learning neural network to obtain prediction results.

[0030] According to a third aspect of the present invention, a computer-readable storage medium is further provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned ionospheric parameter prediction method based on the sequence completion strategy when running.

[0031] According to a fourth aspect of the present invention, there is also provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the ionospheric parameter prediction method based on the sequence completion strategy through the computer program.

[0032] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0033] (1) The present invention uses a variety of signal decomposition algorithms to decompose the historical ionospheric sequence data to obtain sub-components, and compares and analyzes the sub-components with the historical ionospheric sequence data to obtain loss components, which include signal components lost after signal decomposition by each signal decomposition algorithm; based on the sub-components, the error loss term is determined, and the signal is supplemented according to the loss component and the error loss term to obtain the supplement component; the supplement component and the sub-component are respectively input into the preset deep learning neural network as prediction components to obtain the prediction result. The problem of signal loss existing in the signal decomposition algorithm ignored by the traditional method is solved, and the error signal generated in the decomposition process is converted into the form of an effective signal. The neural network processing method can greatly improve the accuracy and effectiveness of the prediction result, and at the same time, it can improve the processing efficiency of the historical ionospheric sequence, providing strong support for the frequency selection of shortwave communication.

[0034] (2) The present invention also designs a convolutional neural network-bidirectional long short-term memory neural network and an attention mechanism function fusion deep learning model to predict the ionospheric sequence generated during the decomposition process, and improves the accuracy of the prediction through feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A schematic flow chart of an optional ionospheric parameter prediction method based on a sequence completion strategy provided in an embodiment of the present application;

[0037] Figure 2 A schematic diagram of a flow chart of another optional ionospheric parameter prediction method based on a sequence completion strategy provided in an embodiment of the present application;

[0038] Figure 3 A schematic diagram of the structure of an optional convolutional layer-long short-term memory neural network-attention mechanism neural network model provided in an embodiment of the present application;

[0039] Figure 4 A schematic diagram of the structure of an optional ionospheric parameter prediction device based on a sequence completion strategy provided in an embodiment of the present application;

[0040] Figure 5A schematic diagram of the structure of an optional electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0042] The terms "first", "second", "third", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0043] The ionosphere is part of the Earth's atmosphere and is located at an altitude of about 50 to 1000 kilometers from the Earth's surface. In this layer, due to the influence of solar radiation, atmospheric molecules are ionized to form a large number of free electrons and ions, thus forming a plasma state in the atmosphere. Since the composition and density distribution of the atmosphere at different altitudes are uneven, the frequency bands and spectral lines of the radiation sources that cause ionization due to solar radiation will also be different, resulting in uneven distribution of electron density at different altitudes in the ionosphere. This uneven distribution of electron density at different altitudes directly affects shortwave communication transmission.

[0044] Generally speaking, the ionosphere can be divided into D, E, F1, and F2 layers from low to high, taking the distribution of ionospheric electron density at different altitudes as the basic standard, and the corresponding electron density also shows a trend of change from low to high. At the same time, the ionospheric time prediction process aims to fit the change characteristics in the ionospheric parameter sequence. There are two methods for predicting ionospheric parameters: long-term prediction and short-term prediction. The long-term prediction method is used to predict the monthly statistical values ​​of statistical ionospheric parameters. The time scale covered and the predicted time range are large, and the prediction time granularity is large. Its expected goal is to provide a trend prediction for the long-term changes in the ionosphere, because its accuracy is limited because of the wide range and long time involved. The ionospheric short-term prediction method is also called the near-real-time prediction method. It covers a short time range and has a small prediction time granularity. The purpose is to predict the precise parameters of the ionosphere in the next few days or even hours.

[0045] In view of the above situation, the present invention is based on the ionospheric historical sequence, and through the signal decomposition method, the error phenomenon generated in the signal decomposition process is considered, and the error generated in the decomposition process is complemented by the sequence completion strategy. The long-term characteristics and various autocorrelation characteristics existing in the ionospheric sequence are considered, and a reasonable neural network model is designed to extract similar features existing in the ionospheric sequence. The present invention converts the error signal generated in the decomposition process into the form of an effective signal by evaluating and complementing the decomposed sub-signals, and uses a neural network processing method to obtain a prediction result with higher accuracy. This method can process historical ionospheric sequences more accurately and provide support for shortwave communication frequency selection.

[0046] According to one aspect of the embodiments of the present application, a method for predicting ionospheric parameters based on a sequence completion strategy is provided. Figure 1 The present invention describes an ionospheric parameter prediction method based on a sequence completion strategy provided in an embodiment of the present application.

[0047] Figure 1 is a flow chart of an optional ionospheric parameter prediction method based on a sequence completion strategy provided in an embodiment of the present application, such as Figure 1 As shown, the process of the method may include the following steps:

[0048] S102, decomposing the historical ionospheric sequence data by using a variety of signal decomposition algorithms to obtain sub-components.

[0049] S104, performing comparative analysis based on the sub-components and historical ionospheric sequence data to obtain loss components, wherein the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm.

[0050] S106, determining an error loss term based on the sub-component, and performing signal completion according to the loss component and the error loss term to obtain a completed component.

[0051] S108, inputting the completion component and the sub-component into a preset deep learning neural network as prediction components respectively to obtain a prediction result.

[0052] An optional ionospheric parameter prediction method based on a sequence completion strategy in an embodiment of the present application can be applied to short-term ionospheric prediction to provide refined ionospheric parameters within a few days.

[0053] Figure 2 A schematic flow chart of another optional ionospheric parameter prediction method based on a sequence completion strategy provided in an embodiment of the present application is as follows: Figure 2As shown, firstly, historical ionospheric sequence data, such as ionospheric foF2 parameters, are obtained, and different signal decomposition methods are used to decompose the obtained ionospheric sequence data to obtain different sub-signals.

[0054] Optionally, the signal decomposition algorithm includes but is not limited to the following methods: X11 decomposition algorithm, improved adaptive noise complete set empirical mode decomposition, EMD decomposition, wavelet transform-based method, Fourier transform-based method, decomposition method based on wavelet transform combined with Fourier transform, etc.

[0055] The decomposed sub-signals are processed, for example, each sub-signal is added and compared with the original sequence to obtain the difference between the sum of the decomposed sub-signals and the original signal, that is, the signal component lost after each signal decomposition algorithm performs signal decomposition is the loss component. Then, this difference is used as a valid signal to complete the signal to obtain the completed component. For example, the difference is superimposed with a component of the sub-signal generated by it to obtain the corresponding completed component.

[0056] Finally, the sequence data of each sub-signal and the sequence data of the complementary component are respectively input into the preset neural network to obtain the prediction results corresponding to each sequence data. Finally, the prediction results are added together to obtain the final sequence prediction result and error analysis is performed.

[0057] The present invention converts the error signal generated in the decomposition process into a valid signal through the above steps S102 to S108, and adopts a neural network processing method to greatly improve the accuracy and effectiveness of the prediction result, while improving the processing efficiency of the historical ionospheric sequence, providing strong support for shortwave communication frequency selection.

[0058] The specific implementation of the present invention is further described in detail below in conjunction with the drawings and examples.

[0059] Specifically, the historical ionospheric sequence data are decomposed using a variety of signal decomposition algorithms to obtain sub-components corresponding to each decomposition algorithm, for example, sub-components of different intrinsic mode functions IMF are obtained through EMD decomposition.

[0060] Preferably, in this embodiment, the signal decomposition algorithm is an EMD algorithm. Compared with other signal decomposition algorithms, the EMD algorithm has the characteristics of noise resistance and no need for prior information in time-frequency processing.

[0061] Specifically, the subcomponents are compared and analyzed with the historical sequence data to obtain the loss component of each subcomponent compared with the historical sequence data. The loss component is the result of the difference between the subcomponent and the historical sequence data, and the loss component refers to the signal component lost after each decomposition using the decomposition algorithm.

[0062] Preferably, in this embodiment, different eigenmode functions IMf are obtained by using EMD decomposition N0 -IMf N The loss component is IMf N0 -IMf N The result of subtracting from historical ionospheric series data.

[0063] Specifically, determining an error loss term of a subcomponent, the error loss term being the center frequency of the component with the highest frequency contribution in the subcomponent or the center frequency of the main frequency component;

[0064] Preferably, in this embodiment, the error loss term is IMf N0 -IMf N The IMF center frequency with the highest center frequency.

[0065] Specifically, based on the completion strategy, which includes three types of completion strategies: model-based fitting, local linear combination-based fitting, and time convolution module TCN-based fitting, completion is performed to obtain the completion components.

[0066] This embodiment provides two specific strategies for completion, including:

[0067] S4.1. Add the loss component and the error loss term to obtain the completion component;

[0068] S4.2, configure the completion strategy; the completion strategy includes fitting based on a univariate time series model and fitting based on local linear correlation;

[0069] Specifically, the univariate time series model fitting includes:

[0070] The input sequence is first fitted through the ARIMA model to obtain the fitted sequence;

[0071] Get the correlation between the input sequence and the fitted sequence; within a certain range, the ones with high correlation are retained in the input sequence;

[0072] Eliminate the input sequence items whose correlation is lower than the preset threshold to obtain the intermediate sequence;

[0073] The intermediate sequence is input into the deep learning neural network to obtain the completion component.

[0074] Specifically, the historical sequence data is truncated into several subsequences. The autocorrelation function and partial correlation function of the input sequence are obtained, and the lag step number is calculated; the calculation method can be calculated based on SPSS software. For the extracted subsequences, the correlation function of each subsequence is calculated respectively to obtain the fitness of each subsequence, and the fitness condition is configured, that is, the subsequence with the maximum lag step number that satisfies the fitness. That is, the maximum lag step number corresponding to the optimal fitness of each subsequence is found. The input sequence is completed according to the lag step number of the subsequence to obtain the completed component. Since the subsequence with the maximum lag step number corresponds to the fitness, the tail of the subsequence corresponding to the maximum lag step number is completed to the subsequence length according to a fixed length.

[0075] Furthermore, the deep learning neural network model structure in this embodiment includes: a feature extraction part, a sequence extraction part, and a sequence prediction part. Figure 3 A schematic structural diagram of an optional convolutional layer-long short-term memory neural network-attention mechanism neural network model provided for an embodiment of the present application.

[0076] The feature extraction part consists of multiple one-dimensional convolutional layers and maximum pooling layers, which are used to extract the features of the input sequence from different angles. Among them, the one-dimensional convolutional layer is used to extract the time-related features of the input sequence, and the maximum pooling layer is used to reduce the complexity of the input sequence. The size of the convolution kernel and the pooling step are determined by the distribution of the input sequence and the task itself.

[0077] The sequence extraction part consists of a multi-layer bidirectional long short-term memory neural network and an attention mechanism layer, which is used to extract the time correlation features in the input sequence; among them, the long short-term memory network is used to extract the front-end correlation features of the input sequence, and the attention mechanism layer is used to strengthen important sequence features.

[0078] The sequence prediction part consists of multiple layers of fully connected layers, which are used to linearly combine the extracted feature information to obtain the prediction results of the input sequence.

[0079] This embodiment provides two prediction system models constructed based on sequence completion strategies for prediction.

[0080] Embodiment 1:

[0081] This embodiment provides a completion strategy, which is based on model fitting. First, the historical sequence data is intercepted according to a certain rule. For example, according to the actual situation, the number of subsequences intercepted is 10, and the subsequence length is kept as 100, and the correlation function of each subsequence is calculated. By setting different step numbers to calculate the correlation function, all the correlation function values ​​of each subsequence can be calculated, and fitting is performed according to the correlation function value. According to the goodness of fit (the greater the goodness of fit, the closer the correlation function is to 1), the useless historical sequence information is eliminated.

[0082] Specifically, on the basis of obtaining the correlation function, a subsequence with a lower correlation function value is extracted, and after the subsequence with a low value is completed, it is merged with the previous subsequence to form a new historical sequence as the input sequence for predicting ionospheric parameters.

[0083] The ionospheric parameter prediction model is constructed using the completed data. First, the characteristic part of the historical sequence is extracted to extract the time-related features of the input sequence. The kernel size of the one-dimensional convolution and the step size of the pooling layer are determined by the distribution size of the input sequence. Then comes the sequence extraction layer, which uses a bidirectional convolutional neural network to extract the front-back correlation features of the input sequence. Among them, the attention mechanism layer is used to strengthen important sequence features. Finally, the extracted feature information is linearly combined through multiple layers of fully connected layers to obtain the prediction results of the input sequence.

[0084] Embodiment 2:

[0085] This embodiment provides a second completion strategy, that is, to obtain a completion sequence based on local linear fitting. For the extracted subsequences, the correlation function of each subsequence is calculated respectively to obtain the degree of fit of each subsequence, and the subsequence whose degree of fit satisfies the condition, that is, the maximum number of lag steps that satisfies the degree of fit is configured. That is, the maximum number of lag steps corresponding to the optimal degree of fit of each subsequence is found. Specifically, the following steps are included:

[0086] Set the parameter k, k is the number of lag steps, and calculate the optimal fit of each subsequence.

[0087] Given the best fit, it is expressed as an autoregressive model as follows:

[0088] st=ψ(st-1,…,st-k),k=1,2,…,K

[0089] Among them, K is the number of lag steps, st is the input sequence, and ψ is the autoregressive model.

[0090] The optimal fit k is obtained, where k is the optimal number of lag steps. For example, if the number of lag steps is 4, the subsequences with a lag step of 4 are overlapped and sampled, and the input sequence is completed according to the number of lag steps.

[0091] In summary, the present invention has the following characteristics:

[0092] The present invention proposes an innovative signal decomposition method based on completion to assist deep learning ionospheric prediction method, which solves the problem of signal loss existing in the signal decomposition algorithm that is ignored by traditional methods.

[0093] By evaluating and completing the decomposed sub-signals, the error signals generated during the decomposition process are converted into effective signals, and a neural network processing method is used to obtain a more accurate prediction result. This method can more accurately process historical ionospheric sequences and provide support for shortwave communication frequency selection.

[0094] The present invention also designs a convolutional neural network-bidirectional long short-term memory neural network and an attention mechanism function fusion deep learning model to predict the ionospheric sequence generated during the decomposition process, and improves the accuracy of the prediction by feature extraction. According to another aspect of the embodiment of the present application, an ionospheric parameter prediction device for implementing the above-mentioned ionospheric parameter prediction method based on sequence completion strategy is also provided.

[0095] Figure 4 is a schematic structural diagram of an optional ionospheric parameter prediction device based on a sequence completion strategy according to an embodiment of the present application, such as Figure 4 As shown, the device may include:

[0096] A decomposition unit 402 is used to decompose the historical ionospheric sequence data using a variety of signal decomposition algorithms to obtain sub-components;

[0097] A loss analysis unit 404 is used to compare and analyze the sub-components with the historical ionospheric sequence data to obtain loss components, where the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm;

[0098] A completion unit 406 is used to determine an error loss term based on the sub-component, and perform signal completion according to the loss component and the error loss term to obtain a completed component;

[0099] The prediction unit 408 inputs the completion component and the sub-component into a preset deep learning neural network as prediction components to obtain a prediction result.

[0100] It should be noted that the decomposition unit 402 in this embodiment can be used to execute the above step S102, the loss analysis unit 404 in this embodiment can be used to execute the above step S104, the completion unit 406 in this embodiment can be used to execute the above step S106, and the prediction unit 408 in this embodiment can be used to execute the above step S108.

[0101] Through the above modules, the present invention uses a variety of signal decomposition algorithms to decompose the historical ionosphere sequence data to obtain sub-components, and compares and analyzes the sub-components with the historical ionosphere sequence data to obtain loss components, which include signal components lost after signal decomposition by each signal decomposition algorithm; based on the sub-components, the error loss term is determined, and the signal is supplemented according to the loss component and the error loss term to obtain the supplement component; the supplement component and the sub-component are respectively input into the preset deep learning neural network as prediction components to obtain the prediction result. The problem of signal loss existing in the signal decomposition algorithm ignored by the traditional method is solved, and the error signal generated in the decomposition process is converted into the form of an effective signal. The neural network processing method can greatly improve the accuracy and effectiveness of the prediction results, and at the same time, it can improve the processing efficiency of the historical ionosphere sequence, providing strong support for the frequency selection of shortwave communication.

[0102] It should be noted here that the examples and scenarios implemented by the above-mentioned modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiments. It should be noted that the above-mentioned modules as part of the device can run in a hardware environment and can be implemented by software or hardware, wherein the hardware environment includes a network environment.

[0103] According to another aspect of the embodiment of the present application, a storage medium is further provided. Optionally, in this embodiment, the storage medium can be used to execute the program code of any of the above-mentioned ionospheric parameter prediction methods based on sequence completion strategy in the embodiment of the present application.

[0104] Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps:

[0105] S1, the historical ionospheric sequence data is decomposed using a variety of signal decomposition algorithms to obtain subcomponents;

[0106] S2, performing comparative analysis based on the sub-components and the historical ionospheric sequence data to obtain loss components, wherein the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm;

[0107] S3, determining an error loss term based on the sub-component, and performing signal completion according to the loss component and the error loss term to obtain a completed component;

[0108] S4, inputting the completion component and the sub-component into a preset deep learning neural network as prediction components respectively to obtain a prediction result.

[0109] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, which will not be described in detail in this embodiment.

[0110] Among them, computer-readable storage media may include, but are not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives and magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0111] According to another aspect of an embodiment of the present application, an electronic device for implementing the above-mentioned ionospheric parameter prediction method based on sequence completion strategy is also provided. The electronic device may be a server, a terminal, or a combination thereof.

[0112] Figure 5 is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present application, such as Figure 5 As shown, it includes a processor 502, a communication interface 504, a memory 506 and a communication bus 508, wherein the processor 502, the communication interface 504, and the memory 506 communicate with each other through the communication bus 508, wherein:

[0113] Memory 506, used for storing computer programs;

[0114] The processor 502 is used to execute the computer program stored in the memory 506 to implement the following steps:

[0115] S1, the historical ionospheric sequence data is decomposed using a variety of signal decomposition algorithms to obtain subcomponents;

[0116] S2, performing comparative analysis based on the sub-components and the historical ionospheric sequence data to obtain loss components, wherein the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm;

[0117] S3, determining an error loss term based on the sub-component, and performing signal completion according to the loss component and the error loss term to obtain a completed component;

[0118] S4, inputting the completion component and the sub-component into a preset deep learning neural network as prediction components respectively to obtain a prediction result.

[0119] Optionally, the communication bus may be a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The communication interface is used for communication between the electronic device and other devices.

[0120] The memory may include RAM, or may include non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0121] As an example, the memory 506 may include, but is not limited to, the receiving unit 402, the first verification unit 404, the second verification unit 406, and the adding unit 408 in the ionospheric parameter prediction device based on the sequence completion strategy. In addition, other module units in the ionospheric parameter prediction device based on the sequence completion strategy may also be included but are not limited to, which will not be repeated in this example.

[0122] The above-mentioned processor can be a general-purpose processor, which can include but not be limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processing), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0123] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0124] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0125] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0126] In the several embodiments provided in the present application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0127] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0128] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0129] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially 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, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, disk or optical disk and other media that can store program code.

[0130] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0131] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0132] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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.

[0133] It will be easily understood by those skilled in the art that the above description is only 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 in the protection scope of the present invention.

Claims

1. A method for predicting ionospheric parameters based on sequence completion strategy, characterized in that: include: The historical ionospheric sequence data are decomposed using a variety of signal decomposition algorithms to obtain subcomponents; Based on the comparative analysis of the sub-components and the historical ionospheric sequence data, a loss component is obtained, wherein the loss component includes a signal component lost after signal decomposition corresponding to each signal decomposition algorithm; Based on the subcomponent, an error loss term is determined, and a signal is completed according to the loss component and the error loss term to obtain a completed component; The completion component and the subcomponent are respectively input into a preset deep learning neural network as prediction components to obtain a prediction result.

2. The ionospheric parameter prediction method based on sequence completion strategy according to claim 1, characterized in that: The step of performing comparative analysis based on the sub-component and the historical ionospheric sequence data to obtain the loss component includes: The subcomponent is subtracted from the historical ionospheric sequence data to obtain a loss component.

3. The ionospheric parameter prediction method based on sequence completion strategy according to claim 1, characterized in that: The determining of the error loss term based on the sub-components includes: the error loss term is the component with the highest center frequency among the sub-components.

4. The ionospheric parameter prediction method based on sequence completion strategy according to claim 1, characterized in that: The preset deep learning neural network includes: a feature extraction part, a sequence extraction part and a sequence prediction part; The feature extraction part includes a plurality of one-dimensional convolutional layers and a maximum pooling layer, which are used to extract the features of the input sequence from different angles; wherein the one-dimensional convolutional layer is used to extract the time-related features of the input sequence, and the maximum pooling layer is used to reduce the complexity of the input sequence; The sequence extraction part includes a multi-layer bidirectional long short-term memory neural network and an attention mechanism layer, which is used to extract the time correlation features in the input sequence; wherein the long short-term memory network is used to extract the front-back correlation features of the input sequence, and the attention mechanism layer is used to strengthen the important sequence features; The sequence prediction part includes multiple layers of fully connected layers, which are used to linearly combine the extracted feature information to obtain the prediction result of the input sequence.

5. The ionospheric parameter prediction method based on sequence completion strategy according to claim 1, characterized in that: The signal is completed according to the loss component and the error loss term to obtain the completed component, including: The loss component and the error loss term are added to obtain the completion component.

6. The ionospheric parameter prediction method based on sequence completion strategy according to claim 5, characterized in that: The signal is completed according to the loss component and the error loss term to obtain the completed component, and further includes: Configure a completion strategy; the completion strategy includes fitting based on a univariate time series model and fitting based on local linear correlation; The univariate time series model fitting includes: The input sequence is first fitted through the ARIMA model to obtain the fitted sequence; Obtaining the correlation between the input sequence and the fitting sequence; Eliminate the input sequence items whose correlation is lower than a preset threshold to obtain an intermediate sequence; The intermediate sequence is input into a deep learning neural network to obtain a completion component.

7. The ionospheric parameter prediction method based on sequence completion strategy according to claim 6, characterized in that: The local linear correlation-based fitting includes: Obtaining the autocorrelation function and partial correlation function of an input sequence, and calculating the corresponding lag steps; the input sequence is a non-overlapping subsequence; Get the fitness of each subsequence and configure the fitness condition, that is, the subsequence with the maximum number of lag steps that satisfies the fitness; The input sequence is completed according to the lag steps of the subsequence to obtain a completed component.

8. An ionospheric parameter prediction device based on a sequence completion strategy, characterized in that: include: A decomposition unit, used for decomposing the historical ionospheric sequence data by using a variety of signal decomposition algorithms to obtain sub-components; A loss analysis unit, configured to perform comparative analysis based on the subcomponents and the historical ionospheric sequence data to obtain loss components, wherein the loss components include signal components lost after signal decomposition corresponding to each signal decomposition algorithm; A completion unit, configured to determine an error loss term based on the sub-component, and perform signal completion according to the loss component and the error loss term to obtain a completed component; The prediction unit inputs the completion component and the sub-component into a preset deep learning neural network as prediction components to obtain a prediction result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images