Ionized layer TEC prediction model integrating double mechanisms of self-attention and self-correlation

Abstract

The invention provides an ionosphere TEC prediction model integrating double mechanisms of self-attention and self-correlation, and belongs to the technical field of ionosphere delay prediction, and the model comprises a trend prediction module, a correlation analysis module and a sequence fusion module. Wherein the trend prediction module predicts the overall change trends of an ionized layer TEC sequence and space environment parameters by means of an autocorrelation operation mechanism; the correlation analysis module analyzes the mutual relation among the ionized layer TEC sequence, the space environment parameters and the longitude and latitude positions by using a self-attention mechanism; and the sequence fusion module performs further integration and optimization on the overall trend of the ionized layer TEC sequence by using the correlation information obtained by the correlation analysis module, and finally obtains an ionized layer TEC prediction result fusing the space environment and the latitude and longitude position information. According to the method, the spatial environment data and the longitude and latitude position information are fully utilized, and the accuracy of model prediction is greatly improved.

Description

Technical Field

[0001] The invention belongs to the technical field of ionospheric delay prediction, and in particular relates to an ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms. Background Art

[0002] Ionospheric delay is one of the main factors causing satellite positioning errors, especially affecting single point positioning (SPP). The magnitude of ionospheric delay can be approximately expressed by the vertical ionospheric total electron content (VTEC).

[0003] At present, the common classic ionospheric TEC models include: the broadcast ionosphere model (Klobuchar model) used in GPS, the NeQuick model used in the Galileo navigation system, and the global ionospheric TEC equivalent grid model. In addition, there are ionospheric TEC statistical prediction models constructed with the help of traditional mathematical statistical methods, such as the linear time series model (LinearTime Series, LTS), the auto-regressive moving average model (Auto-Regressive Moving Average, ARMA), etc. In recent years, with the rise of deep learning technology, many new ionospheric TEC prediction models based on deep learning technologies such as RNN, LSTM, GAN, CNN-LSTM, Transformer, etc. have also been widely proposed and applied.

[0004] The prediction accuracy of the traditional ionospheric TEC empirical model is low. This is because the traditional model has a small number of parameters and can only describe the changing trend of ionospheric TEC with a relatively simple equation, which cannot well reflect the specific laws of ionospheric TEC in different regions and time periods. In recent years, many ionospheric TEC deep learning models have increased the number of model parameters, and the accuracy has improved compared with traditional solutions. However, due to structural problems, the prediction effect of many models is unstable. They often have good prediction effects in mid-latitudes, but the prediction effect will drop significantly in low and high latitudes.

[0005] The global ionospheric TEC equivalent grid model is equivalently generated by the observation results of global stations, but the spatial resolution is low and cannot be specific to more detailed grid points. Many model algorithms predict the global ionospheric TEC equivalent grid. This type of model has two disadvantages. First, the prediction object is expanded to two dimensions, which significantly increases the complexity; second, it relies on the global ionospheric TEC equivalent grid map and cannot predict the ionospheric TEC at any latitude and longitude.

[0006] For deep neural network models, model training and prediction efficiency are crucial issues. In the past, many ionospheric TEC prediction models based on deep learning technology have been able to improve the prediction accuracy of ionospheric TEC by continuously increasing the number of parameters and model depth, but the consequence of doing so is that the training and prediction costs of the model have been significantly increased, especially many serial computing architectures based on recurrent neural networks. This type of model has many parameters, large amount of calculations, and low computing efficiency, and the computational cost and time cost are both high. Summary of the invention

[0007] The present invention uses deep learning technology to construct an ionospheric TEC prediction model that integrates the dual mechanisms of self-attention and self-correlation. In view of the ionospheric TEC prediction problem, the objectives of the present invention are mainly reflected in the following aspects:

[0008] 1. Improve prediction accuracy;

[0009] The present invention adopts deep learning technology to construct a model and reasonably increases the number of model parameters, so that the model can explore more complex ionospheric TEC variation laws and prevent the model from overfitting; the model cleverly integrates the dual mechanisms of self-attention and autocorrelation, and extracts the overall variation trend of ionospheric TEC with the help of autocorrelation operation. At the same time, the self-attention mechanism is used to analyze the relationship between ionospheric TEC, space environmental parameters and longitude and latitude positions, and deeply explores the influence of space environmental parameters and longitude and latitude positions on the variation law of ionospheric TEC, thereby improving the accuracy of ionospheric TEC prediction.

[0010] 2. Improve spatial resolution;

[0011] The present invention completely gets rid of the dependence on the global ionospheric TEC equivalent grid map by introducing specific latitude and longitude position information, and can predict the ionospheric TEC for any specified latitude and longitude position. The present invention uses the TEC data of sites distributed at different latitude and longitude positions around the world for model training, so that the model can learn the influence of latitude and longitude positions on the law of ionospheric TEC changes, rather than being limited to fixed grid points, thereby improving the spatial resolution.

[0012] 3. Reduce the amount of calculation and improve the calculation efficiency;

[0013] The present invention uses autocorrelation operation to predict the overall change trend of ionospheric TEC. Different from the serial operation of the general recurrent neural network prediction model, autocorrelation performs parallel operation in the sequence as a whole. This process requires fewer parameters and has higher operation efficiency, which can give full play to the parallel operation acceleration characteristics of GPU. In addition to the autocorrelation operation, the self-attention mechanism and the subsequent sequence fusion process also adopt a parallel operation structure. Therefore, this model will take up less running memory and achieve more accurate prediction results in a shorter operation time.

[0014] The technical solution of the present invention is: an ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms, the model comprising: a trend prediction module, a correlation analysis module and a sequence fusion module; wherein,

[0015] The trend prediction module uses the autocorrelation operation mechanism to predict the overall change trend of the ionospheric TEC sequence and the space environment parameters; including: the trend prediction module receives the matrix composed of historical ionospheric TEC and space environment parameters As input, the ionospheric TEC value and the space environment parameters are analyzed independently by autocorrelation operation, and their prediction trend matrix is ​​output. ;

[0016] The correlation analysis module uses the self-attention mechanism to analyze the relationship between the ionospheric TEC sequence, space environment parameters and longitude and latitude positions; including: the correlation analysis module receives the matrix composed of historical ionospheric TEC sequence, space environment parameters and longitude and latitude As input, the self-attention mechanism is used to mine the relationship between the ionospheric TEC value, various spatial environmental parameters and longitude and latitude, and the matrix containing their correlation at each moment is output. ;

[0017] The sequence fusion module uses the correlation information obtained by the correlation analysis module to further integrate and optimize the overall trend of the ionospheric TEC sequence, and finally obtains the ionospheric TEC prediction result that integrates the space environment and longitude and latitude position information; including: the sequence fusion module simultaneously receives the output of the trend prediction module Output of the correlation analysis module , using the masked self-attention method, and The information contained is reasonably integrated together, and finally the predicted time series vector of ionospheric TEC is output , length is L.

[0018] The present invention has the following beneficial effects:

[0019] 1. The present invention adopts deep learning technology to construct a model, and innovatively integrates the dual mechanisms of self-attention and autocorrelation. On the one hand, the autocorrelation operation is used to efficiently predict the overall change trend of ionospheric TEC. On the other hand, the self-attention mechanism is used to deeply explore the mutual relationship between ionospheric TEC, space environment parameters and longitude and latitude positions, and finally the overall trend of ionospheric TEC is further integrated and optimized with the help of the correlation information between the three. In this way, the space environment data and longitude and latitude position information are fully utilized, and the accuracy of model prediction is greatly improved.

[0020] 2. The present invention uses the longitude and latitude position information as one of the model input data, and uses the self-attention mechanism to analyze the influence of the longitude and latitude position on the ionospheric TEC variation law, so that the model can make corresponding ionospheric TEC predictions for any specified longitude and latitude position, thereby getting rid of the dependence on the global ionospheric TEC equivalent grid map and improving the spatial resolution of the model prediction.

[0021] 3. The present invention completely abandons the previous scheme of using recurrent neural network serial operation to predict ionospheric TEC, and instead uses autocorrelation operation to perform parallel operation on the basis of the entire sequence, which greatly accelerates the prediction process of the overall change trend of ionospheric TEC. Similarly, subsequent correlation analysis and sequence fusion operations also all adopt parallel operation structure, which reduces the number of parameters required for the entire model and improves the operation efficiency. It is also more conducive to accelerating the operation with the help of GPU and other devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is the overall structural framework diagram of the model of the present invention;

[0023] Figure 2 This is the structural details of the trend prediction module;

[0024] Figure 3 This is the structural details of the correlation analysis module;

[0025] Figure 4 Figure 2 shows the structural details of the sequence fusion module. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme 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 each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above-mentioned purpose, the present invention adopts the following technical scheme.

[0027] The present invention proposes an ionospheric TEC prediction model that integrates the dual mechanisms of self-attention and autocorrelation. The model takes 9 pre-processed data of historical 24-hour ionospheric TEC, geomagnetic index (Dst), ap index, F10.7 index, sunspot number (SSN), X-ray index (X-ray), D-region absorption prediction (D-RAP), latitude and longitude as input, and outputs the ionospheric TEC prediction sequence for the next 24 hours.

[0028] The model consists of three modules: Trend Prediction Module, Relation Analysis Module and Series Merging Module. The Trend Prediction Module uses the autocorrelation operation mechanism to predict the overall change trend of ionospheric TEC and space environment parameters; the Relation Analysis Module uses the self-attention mechanism to analyze the relationship between ionospheric TEC, space environment parameters and longitude and latitude positions; the Series Merging Module uses the correlation information obtained from the Relation Analysis Module to further integrate and optimize the overall trend of ionospheric TEC, and finally obtains the ionospheric TEC prediction result that integrates the space environment and longitude and latitude position information.

[0029] 1. Input Data

[0030] The model needs to input 9 data: ionospheric TEC, geomagnetic index (Dst), ap index, F10.7 index, sunspot number (SSN), X-ray index (X-ray), D-region absorption prediction data (D-RAP), latitude and longitude.

[0031] (1) Data description

[0032] Ionospheric TEC: The average value of ionospheric vertical TEC (VTEC) at the satellite penetration point above the station or in the specified area, with a time resolution of 30 minutes. The data within one day are formed into a time series of length 48.

[0033] Space environment parameters: geomagnetic index (Dst), ap index, F10.7 index, sunspot number (SSN), X-ray index (X-ray), D-region absorption prediction (D-RAP), a total of 6 items. The temporal and spatial resolutions of the original data are not uniform, and they need to be processed by difference or mean respectively to make their temporal and spatial resolutions consistent with the ionospheric TEC data. The five data items of geomagnetic index, ap index, F10.7 index, sunspot number, and X-ray index are one-dimensional time series, while the D-region absorption prediction data are given in the form of a global grid, and specific values ​​need to be assigned according to different longitudes and latitudes. The specific temporal and spatial resolutions of the space environment parameters are shown in Table 1.

[0034] Latitude and longitude: The latitude and longitude of the site or the center of a specified area, indicating the geographical location of the area.

[0035] Table 1 Spatial and temporal resolution of spatial environmental parameters

[0036] (2) Data preprocessing

[0037] The units of the seven data items, ionospheric TEC, geomagnetic index, ap index, F10.7 index, sunspot number, X-ray index, and D-zone absorption prediction data, are different, and there may be large numerical differences between them, which will affect the model inference. They need to be standardized separately so that the mean of each data item is 0 and the variance is 1. The standardization formula is formula (1):

[0038] (1)

[0039] in, represents the mean, represents the standard deviation, is the standardized value, The data to be processed.

[0040] The longitude and latitude are normalized, scaling the longitude range [-180, 180] and the latitude range [-90, 90] to the range [-1, 1]. The normalization formula is formula (2):

[0041] (2)

[0042] in, is the normalized value, For the data to be processed, is the maximum value, As the minimum value, the 9 pre-processed time series data are merged into a matrix , Each row represents a data, a total of 9 rows; each column represents the data of a moment, a total of Column. Matrix As the final input data format of this algorithm model, this algorithm recommends setting the first row as the time series data of ionospheric TEC, the last two rows as the longitude and latitude data, and the middle 6 rows as the time series data of various space environment parameters, so as to facilitate the design of subsequent masks.

[0043] 2. Model Structure

[0044] This model is generally divided into three modules: Trend Prediction Module, Relation Analysis Module and Series Merging Module. Figure 1 It shows the interconnection between the three major modules of this model.

[0045] The trend prediction module receives a matrix consisting of historical ionospheric TEC values ​​and space environment parameters As input, the ionospheric TEC value and the space environment parameters are analyzed independently by autocorrelation operation, and their approximate predicted trends are output. .

[0046] The correlation analysis module receives the matrix composed of historical ionospheric TEC values, space environment parameters and longitude and latitude. As input, the self-attention mechanism is used to mine the relationship between the ionospheric TEC value, various spatial environmental parameters and longitude and latitude, and the matrix containing their correlation at each moment is output. .

[0047] The sequence fusion module also receives the output of the trend prediction module The output of the correlation analysis module , using a special masked self-attention method, and The information contained is reasonably integrated together, and finally the predicted time series vector of ionospheric TEC is output , the predicted length is L.

[0048] (1) Trend prediction module structure details

[0049] The main function of the trend prediction module is to analyze the changing trends of the ionospheric TEC values ​​and various space environment parameters over time. Its core algorithms are auto-correlation and series decomposition. At the same time, the feed-forward neural network is used to add variable parameters to increase the complexity of the model. Figure 2 shown.

[0050] The trend prediction module is composed of N layers in series. The internal structure of each layer is the same. The specific process is as follows: input data , first through the autocorrelation operation unit, then through the sequence decomposition unit to obtain the trend component (Trend) and the period component (Season), the period component passes through the feedforward neural network and then undergoes a sequence decomposition, the trend components of the two sequence decompositions are weighted summed, and then the trend component and the period component enter the next layer respectively, and the data passes through the autocorrelation operation unit and the feedforward neural network and then has a skip connection to add. The trend component output by the last layer and periodic component , and the mean of the original input data Add together to get the final output ,Right now:

[0051] (3)

[0052] ①Autocorrelation operation

[0053] Input Data First, three different linear layer mappings are performed to form three data matrices Q, K, and V respectively, and then the fast Fourier transform (FFT) and inverse Fourier transform IFFT are used to solve Autocorrelation function after mapping :

[0054] (4)

[0055] After that, find out The time shift corresponding to the largest value of k in the front :

[0056] (5)

[0057] Among them, arg TopK is to find the top k largest values; the weights corresponding to these k time-shifted vectors are calculated through the Softmax function :

[0058] (6)

[0059] remember Indicates that the time series data V is cyclically shifted The output of the autocorrelation operation module can be expressed as:

[0060] (7)

[0061] It should be noted that during the entire process, the rows in the three data matrices Q, K, and V are independent of each other and are calculated in parallel, and the amount of time shift required for each row in the V matrix is ​​not necessarily the same.

[0062] ②Sequence decomposition

[0063] The sequence decomposition module smoothes the input data using the sliding average method , thus obtaining the trend component and periodic component , the sliding average process can be completed through the average pooling operation (AvgPool). In order to keep the time series length unchanged, padding (Padding) is required on both sides of the time series. The process is shown in formula (8):

[0064] (8)

[0065] (2) Correlation analysis module structure details

[0066] The main function of the correlation analysis module is to explore the mutual relationship between the ionospheric TEC, various space environment parameters and longitude and latitude positions. Its core algorithm is the self-attention mechanism, and the feedforward neural network is also used to increase the complexity. The specific structure is shown in Figure 3 .

[0067] The correlation analysis module is composed of M layers in series. Each layer has the same structure but different specific parameters. The specific execution process of each layer is as follows: input the data matrix consisting of ionospheric TEC value, space environment parameters and longitude and latitude , First, it passes through the masked self-attention unit (Masked Self-Attention), performs skip connection and layer normalization (Add& LayerNorm), then passes through the feedforward neural network (FeedForward), and finally performs skip connection and layer normalization before outputting to the next layer. The output of the last layer is the relevant information matrix of each data The overall process of each layer in the correlation analysis module is as follows:

[0068] (9)

[0069] in, Represents the input data of each layer in the correlation analysis module, express Through the masked self-attention unit and the intermediate variable after layer normalization, Represents the output data of each layer.

[0070] ① Masked self-attention unit

[0071] Input Data First, three different linear layer mappings are performed to form three data matrices Q, K, and V respectively. Then, the attention matrix A is calculated using Q and K, and then the mask Mask(A) is added. The Softmax function is used to calculate the weight corresponding to each row of Mask(A), and finally multiplied by the V matrix. The process is as shown in formula (10):

[0072] (10)

[0073] in, is the scaling factor, which is equal to the time series length. The self-attention unit adopts a multi-head structure, and each head executes the above process independently. The process is the same, but the parameters are different, which increases the complexity of the model and enables the model to learn more complex structures.

[0074] ②Mask design

[0075] A suitable mask needs to be added to the self-attention unit. The three assumptions on which the mask is constructed are: the variation pattern of the ionospheric TEC value is affected by the longitude and latitude position and various spatial environmental parameters; the various spatial environmental parameters influence each other and are affected by the longitude and latitude position at the same time, but are not affected by the ionospheric TEC value; the longitude and latitude are not affected by any factors because they are fixed values.

[0076] To ensure that the above three assumptions can be realized, it is necessary to set a suitable mask to mask certain values ​​in the attention matrix A. The attention matrix A row in the represents the influence of other rows of data on this row of data, such as represents the influence of the j-th row data on the i-th row data. Then, if the j-th row data has no influence on the i-th row data itself, then Mask it with a mask, usually marking it as negative infinity, so After the Softmax function, the corresponding weight is 0.

[0077] According to the previous data processing suggestions, assuming that the input data The first row is the time series data of ionospheric TEC, the last two rows are the longitude and latitude data, and the middle 6 rows are the time series data of various space environment parameters. Then we can first design a mask matrix of the same size as the attention matrix A. , according to the design requirements of the mask matrix, the mask matrix The specific values ​​are as shown in formula (11):

[0078] (11)

[0079] According to the matrix The value of attention matrix Process it in the way of formula (12):

[0080] (12)

[0081] (3) Structural details of the sequence fusion module

[0082] The main task of the sequence fusion module is to integrate the trend information predicted by the trend prediction module The relevant information between the data mined by the correlation analysis module The TEC prediction sequence is finally output. , L is the target prediction length. The core structure of the sequence fusion module is the self-attention mechanism, which uses a linear fully connected layer to fuse various data, and adds skip connections to keep the mean of the prediction sequence stable. Figure 4 The specific structure of the sequence fusion module is shown.

[0083] Trend information predicted by the trend prediction module Cascade the latitude and longitude data to form , thus the relevant information output by the correlation analysis module The dimensions are the same, At the same time, it is used as the input of Q and K to fuse the self-attention unit, and Then it is used as K input to the fused self-attention unit. The output of the fused self-attention unit is recorded as , remove The two rows of longitude and latitude data in , and then with Add together to get :

[0084] (13)

[0085] Then, it is normalized by layer and transformed into The weighted sum of each row of :

[0086] (14)

[0087] Initially entered trend information In the above example, the time series corresponding to the ionospheric TEC is recorded as ,Will and Add and smooth the time series to get the final ionospheric TEC prediction sequence :

[0088] (15)

[0089] Two skip connection operations are performed in the process of the sequence fusion module. The main purpose is to maintain the time series mean, because the fusion of the self-attention unit may cause a large change in the mean of the time series, and the layer normalization operation forces the mean of the time series to be set to 0.

[0090] ① Fusion of self-attention units

[0091] The fused self-attention unit is basically the same as the masked self-attention unit in the correlation analysis module, and the added mask is also the same. The difference is that the generation of the V data matrix of the fused self-attention unit does not require linear mapping.

[0092] Since the masks added by the two modules are the same, the output of the fused self-attention unit Chinese: The time series corresponding to the ionospheric TEC is given by The time series corresponding to the space environment parameters is obtained by weighting the remaining 8 rows of data except the sequence corresponding to the ionosphere TEC; the longitude and latitude data remain unchanged. At this time, the longitude and latitude information has been fused into various data, so the longitude and latitude data will not be fused in the subsequent linear weighted fusion.

[0093] ② Timing smoothing

[0094] The time series smoothing operation is mainly to eliminate high-frequency noise and make the output prediction data smoother. Practice has shown that not adding time series smoothing will cause the prediction sequence to have more "burrs".

[0095] 3. Training and prediction

[0096] This algorithm model is a deep neural network model, so it needs to be trained with a large amount of data before the ionospheric TEC time series can be predicted.

[0097] The key points when training the model are as follows:

[0098] (1) Data selection and processing

[0099] In order to learn the influence of longitude and latitude on ionospheric TEC, the data selection needs to involve data from different longitude and latitude regions around the world as much as possible. If the observation site area data is used as the training set, the selected sites should be distributed as evenly as possible.

[0100] Due to the different resolutions of spatial environmental data, data preprocessing requires difference or mean processing of spatial environmental data. The resolution of many spatial environmental parameters is 1 day or 1 hour, so the overall resolution should not be too low, preferably within the range of 10 to 30 minutes.

[0101] (2) Model training parameters

[0102] When setting model hyperparameters, it is best to have more layers in the trend prediction module than in the correlation analysis module. After testing, the effect is better when the number of layers in the trend prediction module is 6 and the number of layers in the correlation analysis module is 4. It is recommended to use the Adam algorithm for model optimization, and the learning rate is set to , an exponentially decaying dynamic learning rate can be used to gradually reduce the learning rate during training. After testing, the model can be trained smoothly and quickly on the NVIDIA RTX3080 graphics card (10G video memory), and the hardware requirements are not high. The specific parameters of model training are shown in Table 2.

[0103] Table 2 Model training parameter details

Claims

1. An ionospheric TEC prediction model integrating self-attention and autocorrelation mechanisms, characterized by: The model includes: a trend prediction module, a correlation analysis module and a sequence fusion module; wherein, The trend prediction module uses the autocorrelation operation mechanism to predict the overall change trend of the ionospheric TEC sequence and the space environment parameters; including: the trend prediction module receives the matrix composed of historical ionospheric TEC and space environment parameters As input, the ionospheric TEC value and the space environment parameters are analyzed independently by autocorrelation operation, and their prediction trend matrix is ​​output. ; The correlation analysis module uses the self-attention mechanism to analyze the relationship between the ionospheric TEC sequence, space environment parameters and longitude and latitude positions; including: the correlation analysis module receives the matrix composed of historical ionospheric TEC sequence, space environment parameters and longitude and latitude As input, the self-attention mechanism is used to mine the relationship between the ionospheric TEC value, various spatial environmental parameters and longitude and latitude, and the matrix containing their correlation at each moment is output. ; The sequence fusion module uses the correlation information obtained by the correlation analysis module to further integrate and optimize the overall trend of the ionospheric TEC sequence, and finally obtains the ionospheric TEC prediction result that integrates the space environment and longitude and latitude position information; including: the sequence fusion module simultaneously receives the output of the trend prediction module The output of the correlation analysis module , using the masked self-attention method, and The information contained is reasonably integrated together, and finally the predicted time series vector of ionospheric TEC is output .

2. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 1, characterized in that: The model takes 9 pre-processed data including historical 24-hour ionospheric TEC, geomagnetic index Dst, ap index, F10.7 index, sunspot number SSN, X-ray index X-ray, D-region absorption prediction D-RAP, latitude and longitude as input, and outputs the ionospheric TEC prediction sequence for the next 24 hours.

3. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 2, characterized in that: Ionospheric TEC is: the average value of the vertical TEC of the ionosphere at the satellite penetration point above the station or in the specified area, with a time resolution of 30 minutes, and the data within one day are formed into a time series of length 48; The space environment parameters include: geomagnetic index Dst, ap index, F10.7 index, sunspot number SSN, X-ray index X-ray, D-region absorption prediction D-RAP; the space environment parameters are processed by difference or mean respectively to make their temporal and spatial resolution consistent with the ionospheric TEC series. The geomagnetic index, ap index, F10.7 index, sunspot number, and X-ray index are all one-dimensional time series, while the D-region absorption prediction data are given in the form of a global grid, and specific values ​​are assigned according to different longitudes and latitudes. Latitude and longitude are: the latitude and longitude of the site or the center of a specified area, indicating the geographical location of the area.

4. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 3, characterized in that: The ionospheric TEC, geomagnetic index, ap index, F10.7 index, sunspot number, X-ray index, and D-region absorption prediction data are standardized so that the mean of each data item is 0 and the variance is 1. The standardized formula is formula (1): (1) in, represents the mean, represents standard deviation; is the standardized value, The data to be processed; The longitude and latitude are normalized, scaling the longitude range [-180, 180] and the latitude range [-90, 90] to the range [-1, 1]. The normalization formula is formula (2): (2) in, is the normalized value, For the data to be processed, is the maximum value, As the minimum value, the preprocessed time series data is merged into a matrix , Each row represents a data, a total of 9 rows; each column represents the data of a moment, a total of Column; Matrix It will be used as the final input data format for this algorithm model.

5. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 4, characterized in that: The trend prediction module is composed of N layers in series, and the internal structure of each layer is the same. The specific process is as follows: input data , first through the autocorrelation operation unit, and then through the sequence decomposition unit to obtain the trend component Trend and the periodic component Season, the periodic component is decomposed again after passing through the feedforward neural network, the trend components of the two sequence decompositions are weighted summed, and then the trend component and the periodic component enter the next layer respectively. After the data passes through the autocorrelation operation unit and the feedforward neural network, there is a skip connection addition, and the trend component output by the last layer and periodic component , and the mean of the original input data Add together to get the final output ,Right now: (3)。 6. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 5, characterized in that: The autocorrelation operation is as follows: Input Data First, three different linear layer mappings are performed to form three data matrices Q, K, and V respectively, and then the fast Fourier transform FFT and inverse Fourier transform IFFT are used to solve Autocorrelation function after mapping : (4) Afterwards, find The time shift corresponding to the largest value of k in the front : (5) Among them, arg TopK is to find the top k largest values; the weights corresponding to these k time-shifted vectors are calculated through the Softmax function : (6) remember Indicates that the time series data V is cyclically shifted The output of the autocorrelation operation module is expressed as follows: (7) During the whole process, the rows in the three data matrices Q, K, and V are independent of each other and are calculated in parallel, and the amount of time shift required for each row in the V matrix is ​​not necessarily the same.

7. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 6, characterized in that: The sequence decomposition is as follows: Smoothing input data using the sliding average method , thus obtaining the trend component and periodic component , the sliding average process can be completed through the average pooling operation AvgPool. In order to keep the time series length unchanged, it is necessary to fill Padding on both sides of the time series. The process is shown in formula (8): (8)。 8. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 7, characterized in that: The correlation analysis module is composed of M layers in series. Each layer has the same structure but different specific parameters. The specific execution process of each layer is as follows: input the data matrix consisting of ionospheric TEC value, space environment parameters and longitude and latitude , First, the data is passed through the masked self-attention unit Masked Self-Attention, and skip connection and layer normalization Add & LayerNorm are performed. Then, it is passed through the feedforward neural network FeedForward, and finally skip connection and layer normalization are performed before output to the next layer. The output of the last layer is the relevant information matrix of each data. , the overall process is formula (9): (9) in, Represents the input data of each layer in the correlation analysis module, express Through the masked self-attention unit and the intermediate variable after layer normalization, Represents the output data of each layer.

9. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 8, characterized in that: The masked self-attention unit is specifically: Input Data First, three different linear layer mappings are performed to form three data matrices Q, K, and V respectively. Then, the attention matrix A is calculated using Q and K, and then the mask Mask(A) is added. The Softmax function is used to calculate the weight corresponding to each row of Mask(A), and finally multiplied by the V matrix. The process is as shown in formula (10): (10) in, is the scaling factor, which is equal to the time series length; Add a suitable mask to the self-attention unit and design a mask matrix of the same size as the attention matrix A. , according to the design requirements of the mask matrix, the mask matrix The specific values ​​are as shown in formula (11): (11) According to the matrix The value of attention matrix Process it in the way of formula (12): (12)。 10. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 9, characterized in that: Trend information predicted by the trend prediction module Cascade the latitude and longitude data to form , thus the relevant information output by the correlation analysis module The dimensions are the same, At the same time, it is used as the input of Q and K to fuse the self-attention unit, and Then K is used as the input of the fused self-attention unit, and the output of the fused self-attention unit is recorded as , remove The two rows of longitude and latitude data in , and then with Add together to get : (13) Then, it is normalized by layer and transformed into The weighted sum of each row of : (14) Initially entered trend information In the above example, the time series corresponding to the ionospheric TEC is recorded as ,Will and Add and smooth the time series to get the final ionospheric TEC prediction sequence : (15)。 11. The ionospheric TEC prediction model integrating self-attention and autocorrelation dual mechanisms according to claim 10, characterized in that: The fused self-attention unit is the same as the masked self-attention unit in the correlation analysis module, and the added mask is the same, wherein the generation of the V data matrix of the fused self-attention unit does not require linear mapping.

Images