Multi-feature fusion ionospheric TEC modeling method and system based on gold satellite data

Abstract

The application discloses a multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data, which acquires multi-dimensional channel space features of a to-be-predicted point and inputs the multi-dimensional channel space features into a trained TEC model to output a TEC model value. The application realizes the spatio-temporal matching of GOLD radiation data and MIT TEC data through an automatic process, constructs an input vector fusing periodic features, extracts local feature correlation ability by using a convolutional neural network (CNN), simultaneously introduces a spatial attention mechanism, realizes large-range, high-precision and rapid monitoring of TEC, and provides a new scheme for GNSS ionospheric error correction.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of ionospheric modeling, and particularly relates to a multi-feature fusion ionospheric TEC modeling method and system based on GOLD satellite data. BACKGROUND

[0002] Ionospheric total electron content (TEC) is a key parameter affecting the positioning accuracy of global navigation satellite systems (GNSS), and its accurate prediction is of great significance to navigation and positioning, communication support and other fields. Traditional TEC monitoring methods mostly rely on ground-based observation data or physical model derivation, which has the problems of limited observation coverage, high model complexity and poor real-time performance.

[0003] Traditional TEC measurement methods (such as ground-based GNSS and ionospheric vertical sounders) mostly rely on electromagnetic wave penetration delay inversion mode, but they are point observations and lack large-scale imaging observation technology.

[0004] The ionospheric photometer carried by the GOLD (Global-scale Observations of the Limb and Disk) satellite can observe far ultraviolet band (OI 135.6nm) airglow radiation, which is produced by ionospheric oxygen atoms excited by solar radiation, and has strong correlation with solar activity intensity, ionospheric electron density and ionospheric TEC, providing a good data source for TEC monitoring. However, there is no good solution for TEC inversion based on GOLD data at present. SUMMARY

[0005] In order to overcome the above-mentioned deficiencies of the prior art, the present application provides a multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data, which realizes the spatio-temporal matching of GOLD radiation data and MIT TEC data through an automated process, constructs an input vector that fuses periodic features, uses a convolutional neural network (CNN) to extract local feature correlation capabilities, and introduces a spatial attention mechanism to realize large-scale, high-precision and rapid monitoring of TEC, providing a new solution for GNSS ionospheric error correction.

[0006] According to an aspect of the present application, a multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data is provided, comprising:

[0007] Obtain the multi-dimensional features of the point to be predicted and input them into the trained TEC model to output the TEC model value; the training of the TEC model comprises:

[0008] Obtain the GOLD data and preprocess it;

[0009] Obtain the MIT TEC data and perform spatio-temporal matching with the preprocessed GOLD data;

[0010] Perform feature conversion on the spatio-temporally matched data to form multi-channel spatial features and multi-dimensional temporal features;

[0011] Construct a data set using spatial features, temporal features, and target TECs;

[0012] Construct a TEC model using a spatio-temporal dual-branch convolutional neural network architecture, wherein the temporal branch is used to process multi-dimensional temporal features, and the spatial branch is used to process multi-channel spatial features. The processing results of the temporal branch and the spatial branch are fused through a fusion layer, and the output is the TEC model value;

[0013] Perform model training on the constructed data set and output the trained TEC model.

[0014] As a further technical solution, the GOLD data is obtained, including:

[0015] Collect GOLD L1C level NI1 data, covering target years, target time periods, and specified time pairs;

[0016] Search for channel files according to the annual accumulation day range and time pairs, and match the data of the same time in the northern and southern hemispheres.

[0017] As a further technical solution, the preprocessing includes:

[0018] Extract the radiation value and wavelength variable from the NI1 data, and match the radiation value closest to the set wavelength;

[0019] Grid the discrete radiation data at regular grid intervals, calculate the average radiation value in each grid, and generate regular grid data;

[0020] Fill the radiation values in the abnormal data area with null values.

[0021] As a further technical solution, MIT TEC data is obtained and spatio-temporally matched with the preprocessed GOLD data, including:

[0022] Load the MIT TEC data, and extract the corresponding TEC values according to the target time and spatial range;

[0023] Filter the MIT TEC data within the time threshold range based on the GOLD data time;

[0024] Filter the MIT TEC data within the spatial threshold range, take the TEC value closest to the radiation value as the matching label, and generate "radiation value-TEC" paired data.

[0025] As a further technical solution, feature conversion is performed on the spatio-temporally matched data, including:

[0026] Temporal features: convert hours and accumulated days into sine and cosine expressions, respectively;

[0027] Spatial features: convert longitude and latitude into sine and cosine expressions, respectively, and obtain geomagnetic parameters;

[0028] Physical features: logarithmically transform the set wavelength radiation value, calculate the local time according to the longitude and latitude, and then add a day and night marker.

[0029] As a further technical solution, the multi-channel spatial features and multi-dimensional temporal features formed include:

[0030] The multi-channel spatial features include: longitude sine, longitude cosine, latitude sine, latitude cosine, geomagnetic parameters, radiation value, radiation value logarithm, and day and night marker;

[0031] The multi-dimensional temporal features include: hour sine, hour cosine, accumulated day sine, and accumulated day cosine.

[0032] As a further technical solution, in the TEC model constructed by the spatio-temporal dual-branch convolutional neural network architecture:

[0033] The spatial branch includes multiple convolutional blocks, each of which uses LeakyReLU activation, is flattened after convolution processing, and is connected using two fully connected layers, which use LeakyReLU activation;

[0034] The temporal branch processes multi-dimensional temporal features by two fully connected layers and uses LeakyReLU activation.

[0035] According to an aspect of the present application, a multi-feature fusion ionospheric TEC modeling system based on GOLD satellite data is provided, comprising:

[0036] A first main module for obtaining multi-dimensional channel spatial features of a point to be predicted;

[0037] A second main module for inputting the multi-dimensional channel spatial features of the point to be predicted into a trained TEC model to output a TEC model value; the training of the TEC model includes:

[0038] Obtain GOLD data and pre-process it;

[0039] Obtain MIT TEC data and perform spatio-temporal matching with the pre-processed GOLD data;

[0040] Perform feature conversion on the spatio-temporally matched data to form multi-channel spatial features and multi-dimensional temporal features;

[0041] Construct a data set with spatial features, temporal features, and target TEC;

[0042] The TEC model is constructed by using a space-time dual-branch convolutional neural network architecture, wherein the time branch is used for processing multi-dimensional time features, the space branch is used for processing multi-channel space features, and the processing results of the time branch and the space branch are fused through a fusion layer and then output TEC model values;

[0043] Model training is performed on the constructed data set, and a trained TEC model is output.

[0044] According to an aspect of the present application, a multi-feature fusion ionospheric TEC modeling device based on GOLD satellite data is provided, comprising a memory and a processor, the memory storing program instructions executed by the processor, and the processor calling the program instructions to execute the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data.

[0045] According to an aspect of the present application, a non-transitory computer readable storage medium is provided, which stores computer instructions for executing the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data.

[0046] Compared with the prior art, the present application has the following advantages:

[0047] 1. Data source innovation:

[0048] For the first time, the 135.6nm radiation intensity of the GOLD satellite is used as the core data source for ionospheric TEC inversion, and the ionospheric TEC is inverted.

[0049] 2. Feature engineering innovation.

[0050] Periodic feature encoding. For the periodicity of time (local time), space (longitude and latitude), and season (yearly cumulative day), the positive and negative sine vectors are converted to avoid the boundary mutation problem of direct numerical coding, and to better conform to the space-time variation law of the earth itself.

[0051] Physical feature optimization. The logarithmic transformation of the GOLD 135.6nm radiation value is performed to reduce the skewness of the data distribution and make the distribution more balanced, which is conducive to the data training to obtain a more accurate model; a new day-night marker coding is added to adapt to the ionospheric day-night difference and show the influence of solar activity behind the ionosphere.

[0052] 3. Model structure innovation.

[0053] A space-time dual-branch CNN structure with space attention is proposed. Convolution is used to extract the local correlation of input features (latitude, longitude, time, 135.6nm radiation value, day-night label, geomagnetic element). The spatial branch extracts grid features through multiple convolution blocks and embeds a space attention mechanism (dynamically generates a region weight map to enhance key region features). The time branch processes periodic time features through a fully connected layer. The fusion layer splices the dual-branch features and outputs TEC model values consistent with the input grid dimension through a fully connected network. BRIEF DESCRIPTION OF DRAWINGS

[0054] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings used in the embodiments or prior art description will be briefly described as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0055] Figure 1 The flowchart of the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data provided by the embodiments of the present application is shown.

[0056] Figure 2 The training flowchart of the TEC model provided by the embodiments of the present application is shown.

[0057] Figure 3 The GOLD 135.6nm radiation value-MIT TEC comparison chart provided by the embodiments of the present application is shown.

[0058] Figure 4 The training and verification loss diagram provided by the embodiments of the present application is shown.

[0059] Figure 5 The TEC model value and real value scatter diagram provided by the embodiments of the present application is shown.

[0060] Figure 6 The residual distribution histogram provided by the embodiments of the present application is shown.

[0061] Figure 7 The residual space distribution diagram provided by the embodiments of the present application is shown.

[0062] Figure 8 The TEC real value and model value space distribution comparison chart provided by the embodiments of the present application is shown. DETAILED DESCRIPTION

[0063] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application. In addition, the technical features in each of the embodiments or in a single embodiment provided by the present application can be combined with each other at will to form new technical solutions, and such combination is not restricted by the order of steps and / or structure mode, but should be based on the fact that it can be realized by those of ordinary skill in the art. When the combination of technical solutions appears to be contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist, and is not within the scope of protection required by the present application.

[0064] The embodiments of the present application provide a multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data, as shown in Figure 1 The method comprises the following steps of: first, acquiring multi-dimensional channel space features of a to-be-predicted point; then, inputting the acquired multi-dimensional features of the to-be-predicted point into a trained TEC model, and outputting a TEC model value.

[0065] As shown in Figure 2 The training of the TEC model comprises the following steps of:

[0066] S1, data acquisition and preparation.

[0067] GOLD data: collect GOLD L1C level NI1 data (containing CHA / CHB two channels), cover target year, target period and specified time pair.

[0068] MIT TEC data acquisition: acquire global TEC grid products published by MIT as TEC label data.

[0069] Automatic generation of file list: search CHA / CHB files according to the range of annual accumulated days and time pairs, and match south and north hemisphere data at the same time.

[0070] S2, GOLD data preprocessing.

[0071] Radiation value extraction: read the NI1 file, extract the radiation value and wavelength variable, match the radiation value closest to the wavelength of 135.6 nm, and filter invalid missing values. It should be noted that the GOLD radiation value and TEC value data are both point-shaped, and cannot be completely matched. Therefore, the GOLD data and the TEC data at the closest geographical position and time are matched together to facilitate subsequent neural network training.

[0072] Griding: Discrete radiation data is gridded with regular grid spacing, and the average radiation value in each grid is calculated to generate regular grid data.

[0073] Region Mask: Fill the radiation values of the abnormal data region with null values to exclude the interference data in this region.

[0074] S3, MIT TEC data loading and space-time matching.

[0075] MIT TEC data loading: Load MIT TEC files and extract corresponding TEC values according to target time and spatial range.

[0076] Space-time matching: Select MIT data within a time threshold (e.g., 10 minutes) based on GOLD data time; select MIT data within a threshold range (e.g., 0.1°) in space, take the nearest TEC value as the matching label, and generate "radiation value-TEC" paired data.

[0077] After pairing the radiation value and TEC, the radiation value is one of the features participating in the training, and the TEC is the label for supervised training. The purpose of pairing is to calculate the loss function in the neural network, and the TEC value is used as the label and the TEC value predicted by the radiation value is subtracted.

[0078] S4, feature construction and dataset division.

[0079] Feature construction: Convert the matched data into features: (1) Time features: convert hours and year accumulation days into sine and cosine representations; (2) Spatial features: convert latitude and longitude into radians, express them as sine and cosine, and obtain magnetic declination, magnetic inclination, and magnetic field strength components. These geomagnetic parameters use the geomagnetic data of the corresponding year provided by the International Geomagnetic Reference Field (IGRF); (3) Physical features: logarithmic transformation of 135.6 nm radiation value (but input original value and logarithmic value as features), calculation of local time according to latitude and longitude and time, and addition of day and night markers.

[0080] Dataset division: filter samples with high missing proportion, fill feature missing values with channel mean, fill target missing values with overall mean; generate 8-channel spatial features (radiation value, radiation value logarithm, latitude and longitude sine and cosine, geomagnetic parameters, day and night markers) and 4-dimensional time features (hour / year accumulation day sine and cosine) according to the constructed features; divide the dataset into training set and validation set in the form of (spatial feature, time feature, target TEC) tuple according to 8:2.

[0081] S5, construction of space-time dual-branch CNN model.

[0082] (1) Spatial branch: composed of multiple convolutional blocks, each block uses LeakyReLU activation, and after convolution processing, it is flattened and connected using two fully connected layers, which use LeakyReLU activation;

[0083] (2) Temporal branch: 4-dimensional temporal features are processed by two fully connected layers, and LeakyReLU activation is used.

[0084] (3) Fusion layer: fuse the output features of the spatial branch and the temporal branch to generate the final model value. Use fully connected neural network connection and output.

[0085] S6, model training.

[0086] The training set and the validation set are used to train and validate the model.

[0087] S7, model evaluation and use.

[0088] Evaluation indicators: calculate regression indicators (MSE, RMSE, MAE, R²) and residual statistics (mean, median, standard deviation, positive and negative proportion, etc.), visualize loss curves, model values and true values, residual distribution histogram, residual spatial distribution, etc.

[0089] Model deployment: save model weights, including model parameters, feature normalizer, target variable normalizer; when using, input the 12-dimensional features of the predicted point (including 8-dimensional channel space features and 4-dimensional time features), after normalization, input the model, and output the TEC model value.

[0090] As an embodiment, taking 2023 data as an example, according to the data acquisition and processing, model construction and training process, the steps of the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data provided by the application are explained, and the specific process is as follows:

[0091] 1. Experimental data configuration.

[0092] Determine the data year and range: 2023, annual accumulation day 281-290;

[0093] Read GOLD NI1 data, MIT TEC data, and set the spatial range: longitude [-90°, -20°], latitude [-40°, 40°]. The comparison of GOLD 135.6nm radiation value and MIT TEC is shown in FIG. Figure 3

[0094] 2. Match the GOLD data file.

[0095] ​Search CHA / CHB file according to 18 groups of time per day, merge the data files of the northern and southern hemispheres, grid the 135.6 nm radiation value according to a 1.0 degree grid, apply the regional mask to the longitude [-90°, -75°] and latitude [30°, 40°], and exclude abnormal values.

[0096] 3. Match the GOLD and MIT data.

[0097] Search and match the MIT TEC data according to a 10-minute time tolerance, a 0.1-degree spatial tolerance, and the GOLD data, and save the matched data.

[0098] 4. Feature engineering and dataset construction.

[0099] Construct 12-dimensional features (yearly day sine, yearly day cosine, hour sine, hour cosine, geographic longitude sine, geographic longitude cosine, geographic latitude sine, geographic latitude cosine, 135.6 nm radiation value and its logarithmic value, geomagnetic parameters (magnetic declination, magnetic inclination, magnetic field strength component), day and night markers) for the matched GOLD and MIT TEC data; divide the training set and the test set in an 8:2 ratio, and save the dataset.

[0100] 5. CNN model construction and training.

[0101] The convolutional neural network (CNN) model is constructed, including:

[0102] (1) Spatial branch: composed of multiple convolutional blocks, each block uses LeakyReLU activation, is flattened after convolution processing, is connected using two fully connected layers, and uses LeakyReLU activation;

[0103] (2) Time branch: processes 4-dimensional time features using two fully connected layers, and uses LeakyReLU activation;

[0104] (3) Fusion layer: fuses the output features of the spatial branch and the time branch to generate the final model value. A fully connected neural network is used for connection and output.

[0105] 6. Model training.

[0106] The training set and the validation set are used to train the model, and the training and validation loss is as shown in Figure 4 .

[0107] The training to validation loss no longer decreases, triggering early stopping, and obtaining the best validation loss.

[0108] 7. Model evaluation and testing.

[0109] The evaluation results include the RMSE, MAE, R², residual mean, and positive residual proportion of the test set. The residual distribution histogram is as shown inFigure 6 As shown in FIG. 6, the residual space distribution diagram is as shown in FIG. 7. Figure 7 As shown in FIG. 7.

[0110] The model is saved, including model weights, scaler_X, scaler_y, etc.

[0111] TEC test example: input the to-be-predicted point information, calculate the 12-dimensional features of the to-be-predicted point information through feature conversion, input the features into the model after normalization by the scaler_X, output the normalized TEC value, and obtain the final TEC model value through the scaler_y. As shown in FIG. 8, the scatter diagram of the TEC real value and the model value is as shown in FIG. 9. Figure 5 As shown in FIG. 9. Figure 8 As shown in FIG. 9.

[0112] The implementation basis of each embodiment of the present application is realized through programmed processing of a device with processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present application are packaged into various modules. Based on this actual situation, on the basis of the above-mentioned embodiments, the embodiment of the present application provides a multi-feature fusion ionospheric TEC modeling system based on GOLD satellite data, which is used to execute one of the above-mentioned multi-feature fusion ionospheric TEC modeling methods based on GOLD satellite data.

[0113] The system comprises: a first main module configured to acquire multi-dimensional channel space features of a to-be-predicted point; and a second main module configured to input the multi-dimensional channel space features of the to-be-predicted point into a trained TEC model and output a TEC model value. The training of the TEC model comprises: acquiring GOLD data and performing preprocessing; acquiring MIT TEC data and performing spatiotemporal matching with the preprocessed GOLD data; performing feature conversion on the spatiotemporally matched data to form multi-channel space features and multi-dimensional time features; constructing a data set with the space features, the time features and target TEC; constructing a TEC model with a spatiotemporal double-branch convolutional neural network architecture, wherein the time branch is configured to process the multi-dimensional time features, the space branch is configured to process the multi-channel space features, and the processing results of the time branch and the space branch are fused through a fusion layer and output a TEC model value; and performing model training on the constructed data set and outputting a trained TEC model.

[0114] The embodiment of the application provides a multi-feature fusion ionospheric TEC modeling system based on GOLD satellite data, faces problems in traditional TEC measurement methods and existing TEC inversion based on GOLD data, adopts the plurality of modules, realizes space-time matching of GOLD radiation data and MIT TEC data through an automatic process, constructs an input vector fusing periodic characteristics, extracts local feature correlation ability by using a convolutional neural network (CNN), simultaneously introduces a space attention mechanism, realizes large-range, high-precision and rapid monitoring of TEC, and provides a new scheme for GNSS ionospheric error correction.

[0115] It should be noted that the system embodiments provided by the application are used to implement the methods in the method embodiments, and are also used to implement the methods in other method embodiments provided by the application, the difference is only that corresponding function modules are arranged, the principle is basically the same as that of the above-mentioned system embodiments provided by the application, as long as the person skilled in the art improves the modules in the above-mentioned system embodiments on the basis of the above-mentioned system embodiments, refers to the specific technical solutions in other method embodiments, obtains corresponding technical means by combining technical features, and the technical solutions formed by the technical means, on the premise of ensuring the practicability of the technical solutions, the corresponding system embodiments are obtained, and the methods in other method embodiments are implemented.

[0116] Based on the same inventive concept as the foregoing embodiments, the embodiment of the application also provides a multi-feature fusion ionospheric TEC modeling device based on GOLD satellite data, comprising a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data.

[0117] In the embodiment of the application, the memory can be a non-volatile memory such as a hard disk (HDD) or a solid-state drive (SSD), and can also be a volatile memory such as a random-access memory (RAM). The memory can be any other medium capable of carrying or storing desired program codes in the form of instructions or data structures and capable of being accessed by a computer, but is not limited to this. The memory in the embodiment of the application can also be a circuit or other any device capable of realizing a storage function, used for storing program instructions and / or data.

[0118] In an embodiment of the present application, the processor can be a general purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the present application. The general purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as hardware processor execution, or executed by a combination of hardware and software modules in the processor.

[0119] Based on the same inventive concept as the foregoing embodiments, the embodiments of the present application also provide a non-transitory computer readable storage medium storing computer instructions, which make the computer execute the disclosed method for modeling ionospheric TEC based on GOLD satellite data, the steps being as follows:

[0120] Obtain the multi-dimensional features of the to-be-predicted point and input the trained TEC model to output the TEC model value; the training of the TEC model comprises:

[0121] Obtain the GOLD data and perform preprocessing;

[0122] Obtain the MIT TEC data and perform spatio-temporal matching with the preprocessed GOLD data;

[0123] Perform feature conversion on the spatio-temporally matched data to form multi-channel spatial features and multi-dimensional temporal features;

[0124] Construct a data set with the spatial features, the temporal features and the target TEC;

[0125] Construct the TEC model in a spatio-temporal dual-branch convolutional neural network architecture, wherein the temporal branch is used to process the multi-dimensional temporal features, the spatial branch is used to process the multi-channel spatial features, and the processing results of the temporal branch and the spatial branch are fused through a fusion layer to output the TEC model value;

[0126] Perform model training on the constructed data set to output the trained TEC model.

[0127] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0128] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks or in conjunction with the flowcharts. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in one or more flowcharts and / or blocks.

[0129] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks or in conjunction with the flowcharts. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in one or more flowcharts and / or blocks.

[0130] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks or in conjunction with the flowcharts. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in one or more flowcharts and / or blocks.

[0131] In summary of the above embodiments, the application proposes a multi-feature fusion CNN ionospheric TEC modeling method based on GOLD satellite data, realizes spatio-temporal matching of GOLD radiation data and MIT TEC data through an automatic process, constructs an input vector fusing periodic features, uses CNN to extract local feature correlation ability, simultaneously introduces a spatial attention mechanism, realizes large-scale, high-precision and rapid monitoring of TEC, and provides a new scheme for GNSS ionospheric error correction.

[0132] The terms "including", "containing", "having" and "including" and any variations thereof in the specification and in the claims and the above-described accompanying drawings are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or apparatus that includes a list of steps or units, without being limited to clearly listed steps or units, but can include other steps or units that are not clearly listed or inherent to the process, method, product or apparatus.

[0133] It should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present application.

Claims

1. A multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data, characterized in that, include: Obtain the multidimensional features of the point to be predicted and input them into the trained TEC model, then output the TEC model value. The training of the TEC model includes: Acquire GOLD data and perform preprocessing; Acquire MIT TEC data and perform spatiotemporal matching with preprocessed GOLD data; The spatiotemporally matched data undergoes feature transformation to form multi-channel spatial features and multi-dimensional temporal features. This feature transformation includes: temporal features: converting hours and annual days into sine and cosine expressions respectively; spatial features: converting latitude and longitude into sine and cosine expressions respectively, and obtaining geomagnetic parameters; physical features: performing a logarithmic transformation on the set wavelength radiation value, calculating local time based on latitude, longitude, and time, and then adding day and night markers. A dataset is constructed using spatial features, temporal features, and target TECs. The TEC model is constructed using a spatiotemporal dual-branch convolutional neural network architecture, where the temporal branch is used to process multi-dimensional temporal features and the spatial branch is used to process multi-channel spatial features. The processing results of the temporal and spatial branches are fused by a fusion layer to output the TEC model value. The model is trained on the constructed dataset, and the trained TEC model is output.

2. The multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data according to claim 1, characterized in that, Obtain GOLD data, including: Collect GOLD L1C-level NI1 data, covering target years, target time periods, and specified time pairs; Based on the year-to-date range and time, search channel files to match data from the Northern and Southern Hemispheres at the same time.

3. The multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data according to claim 2, characterized in that, The preprocessing includes: Extract the radiation value and wavelength variable from the NI1 data and match the radiation value closest to 135.6 nm; The discrete radiation data is gridded according to a regular grid spacing, and the average radiation value in each grid is calculated to generate regular grid data. Fill the radiation values ​​in the abnormal data area with empty values.

4. The multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data according to claim 1, characterized in that, Acquire MIT TEC data and perform spatiotemporal matching with preprocessed GOLD data, including: Load MIT TEC data and extract the corresponding TEC values ​​based on the target time and spatial range; Using the time frame of GOLD data as a benchmark, filter MIT TEC data within a time threshold range; Filter the MIT TEC data within the spatial threshold range, take the TEC value closest to the radiation value as the matching label, and generate "radiation value-TEC" paired data.

5. The multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data according to claim 1, characterized in that, The resulting multi-channel spatial features and multi-dimensional temporal features include: Multi-channel spatial characteristics include: longitude sine, longitude cosine, latitude sine, latitude cosine, geomagnetic parameters, radiance, logarithm of radiance, and day / night markers; Multidimensional time features include: hourly sine, hourly cosine, yearly sine, and yearly cosine.

6. The multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data according to claim 1, characterized in that, In the TEC model built with a spatiotemporal dual-branch convolutional neural network architecture: The spatial branch includes multiple convolutional blocks, each activated by LeakyReLU. After convolution, the blocks are flattened and connected by two fully connected layers, which are activated by LeakyReLU. The time branch is processed by two fully connected layers to handle multidimensional time features and activated using LeakyReLU.

7. A multi-feature fusion ionospheric TEC modeling system based on GOLD satellite data, characterized in that, include: The first main module is used to obtain the multi-dimensional channel spatial features of the point to be predicted. The second main module is used to input the multi-dimensional channel spatial features of the point to be predicted into the trained TEC model and output the TEC model value; the training of the TEC model includes: Acquire GOLD data and perform preprocessing; Acquire MIT TEC data and perform spatiotemporal matching with preprocessed GOLD data; The spatiotemporally matched data undergoes feature transformation to form multi-channel spatial features and multi-dimensional temporal features. This feature transformation includes: temporal features: converting hours and annual days into sine and cosine expressions respectively; spatial features: converting latitude and longitude into sine and cosine expressions respectively, and obtaining geomagnetic parameters; physical features: performing a logarithmic transformation on the set wavelength radiation value, calculating local time based on latitude, longitude, and time, and then adding day and night markers. A dataset is constructed using spatial features, temporal features, and target TECs. The TEC model is constructed using a spatiotemporal dual-branch convolutional neural network architecture, where the temporal branch is used to process multi-dimensional temporal features and the spatial branch is used to process multi-channel spatial features. The processing results of the temporal and spatial branches are fused by a fusion layer to output the TEC model value. The model is trained on the constructed dataset, and the trained TEC model is output.

8. A multi-feature fusion ionospheric TEC modeling device based on GOLD satellite data, characterized in that, The system includes a memory and a processor, wherein the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to execute the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the multi-feature fusion ionospheric TEC modeling method based on GOLD satellite data as described in any one of claims 1 to 6.

Images