A method and system for calculating a global average TEC model in a satellite-ground combined manner

Abstract

The application discloses a kind of global average TEC model calculation method and system of satellite-ground combination, belong to ionospheric key parameter monitoring technical field.It includes obtaining the occultation VTEC characteristic parameter calculated by occultation event;Obtain the VTEC characteristic parameter calculated by sparse GNSS site;The VTEC characteristic parameter calculated by occultation event is matched with the VTEC characteristic parameter calculated by sparse GNSS site, obtain the VTEC characteristic parameter calculated in each 30 minutes period before and after the reference time to be calculated;The VTEC characteristic parameter calculated in each 30 minutes period before and after the reference time to be calculated is input into the multilayer neural network model trained in advance, and the equivalent global average TEC of corresponding time is output and obtained;The application can reduce the error of single data source in combination with GNSS and occultation observation data, and the precision of ionospheric model can be significantly improved in combination.

Description

Technical Field

[0001] The present invention relates to a satellite-ground combined global average TEC model calculation method and system, belonging to the technical field of ionospheric key parameter monitoring. Background Art

[0002] Globally averaged TEC is a key parameter reflecting the distribution of electron density in the ionosphere. It offers significant advantages in describing global ionospheric variations, reducing noise and local biases, and supporting long-term research and prediction. Traditional methods calculate the global averaged TEC by taking an area-weighted average of the TEC values ​​over the grid cells of the Global Ionospheric Map (GIM). This requires ionospheric observations from hundreds of GNSS stations worldwide and a complex modeling process.

[0003] In addition, there are two methods to construct a global average TEC model based on GNSS observations: (1) based on sparse GNSS stations around the world; and (2) based on low-orbit satellite occultation observations.

[0004] Sparse GNSS stations are unevenly distributed across the surface. Stations are extremely scarce in areas such as oceans, deserts, high-altitude uninhabited areas, and areas with complex terrain, and their data is affected by communication links. Relying solely on data from sparse ground stations results in inaccurate ionospheric models. Occultation observations have limited coverage in high-latitude regions, and the number of satellites capable of occultation observations is also limited. Furthermore, occultation observations typically occur at orbital altitudes between 400 and 800 kilometers, limiting the ability to fully observe the ionosphere.

[0005] Affected by the above shortcomings, relying solely on sparse ground-based stations or occultation observations to invert global average TEC parameters will face the problem of low model accuracy due to poor observational representativeness. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a satellite-ground joint global average TEC model calculation method and system to solve the problem of low model accuracy caused by poor observation representativeness when only relying on sparse ground stations or occultation observations to invert global average TEC parameters;

[0007] In order to achieve the above objectives / solve the above technical problems, the present invention is implemented by adopting the following technical solutions:

[0008] A first aspect: a satellite-ground joint global average TEC model calculation method, the method comprising:

[0009] Obtain the occultation VTEC characteristic parameters calculated from the occultation event;

[0010] Obtain the site VTEC characteristic parameters calculated from sparse GNSS sites;

[0011] The VTEC characteristic parameters calculated from the occultation event are time-matched with the VTEC characteristic parameters calculated from the sparse GNSS stations to obtain the VTEC characteristic parameters calculated for 30 minutes before and after the reference time to be calculated.

[0012] The VTEC characteristic parameters calculated in the 30-minute periods before and after the reference time to be calculated are input into a pre-trained multi-layer neural network model, and the equivalent global average TEC at the corresponding time is output.

[0013] Optionally, the training of the multi-layer neural network model includes:

[0014] Collect the occultation VTEC characteristic parameters calculated from the occultation event;

[0015] Collect the site VTEC characteristic parameters calculated from sparse GNSS sites;

[0016] The VTEC characteristic parameters obtained from the occultation event and the VTEC characteristic parameters obtained from the sparse GNSS stations are time-matched to obtain a comprehensive dataset VTEC, which includes a training set and a validation set.

[0017] The training set and the validation set are input into the multi-layer neural network model for training to obtain a trained multi-layer neural network model.

[0018] Optionally, the training of the multi-layer neural network model further includes:

[0019] The global average TEC is calculated hourly by area-weighted average in the grid product of the global ionosphere map obtained at 1-hour intervals. This is used as the benchmark data, and the accuracy of the model is adjusted by the benchmark data when training the model.

[0020] Optionally, the multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters, and then the normalized characteristic parameters are input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing, and finally reach the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

[0021] Optionally, the occultation VTEC characteristic parameters and site VTEC characteristic parameters include longitude, latitude, time and vertical total electron content VTEC.

[0022] Optionally, the matching time window is the hourly global mean TEC and the occultation and sparse GNSS site VTEC data within a 30-minute period before and after it.

[0023] Optionally, obtain the occultation VTEC characteristic parameters calculated from the occultation event, including:

[0024] The electron density profile product obtained by inverting a long series of occultation events is obtained, including the time, longitude, latitude, profile height and electron density sequence of the occultation, and the occultation VTEC characteristic parameters are calculated based on the profile height and electron density sequence.

[0025] Optionally, obtain the site VTEC characteristic parameters calculated for sparse GNSS sites, including:

[0026] Several sparsely distributed GNSS stations are selected around the world. The ionospheric puncture points are determined through the station receivers and observation satellites. The total electron content (TEC) on the slant path is calculated based on the GNSS observations and projected vertically to convert them into the VTEC characteristic parameters of the puncture point.

[0027] A second aspect: a satellite-ground combined global average TEC model calculation system, the system comprising:

[0028] A first data acquisition module is used to obtain occultation VTEC characteristic parameters calculated from the occultation event;

[0029] The second data acquisition module is used to obtain the site VTEC characteristic parameters calculated by the sparse GNSS site;

[0030] The data fusion module is used to time-match the VTEC characteristic parameters calculated from the occultation event with the VTEC characteristic parameters calculated from the sparse GNSS stations, and obtain the VTEC characteristic parameters calculated for each 30-minute period before and after the reference time to be calculated;

[0031] The data calculation module is used to input the VTEC characteristic parameters calculated in the 30-minute period before and after the reference time to be calculated into a pre-trained multi-layer neural network model, and output the equivalent global average TEC at the corresponding time.

[0032] Optionally, the multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters, and then the normalized characteristic parameters are input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing, and finally reach the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This paper combines sparse GNSS stations with occultation observations to construct a global average TEC model based on an artificial neural network. Occultation observations make up for the lack of GNSS station observations in areas such as oceans and deserts. GNSS station data can improve the observation accuracy in high-latitude areas. The combination of the two can also achieve more complete vertical coverage. Combining GNSS and occultation observation data can reduce the error of a single data source. The combined use can significantly improve the accuracy of the ionospheric model.

[0035] The present invention effectively compensates for the shortcomings of a single observation method, significantly improving the coverage, data accuracy and modeling capabilities of ionospheric monitoring, thereby improving the modeling accuracy of the global average TEC. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Shown is a flow chart of the method of the present invention;

[0037] Figure 2 Shown is a specific flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0039] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0040] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0041] like Figure 1 As shown, a satellite-ground joint global average TEC model calculation method is disclosed, the method comprising:

[0042] Obtain the occultation VTEC characteristic parameters calculated from the occultation event;

[0043] Obtain the site VTEC characteristic parameters calculated from sparse GNSS sites;

[0044] The VTEC characteristic parameters calculated from the occultation event are time-matched with the VTEC characteristic parameters calculated from the sparse GNSS stations to obtain the VTEC characteristic parameters calculated for 30 minutes before and after the reference time to be calculated.

[0045] The VTEC characteristic parameters calculated in the 30-minute periods before and after the reference time to be calculated are input into a pre-trained multi-layer neural network model, and the equivalent global average TEC at the corresponding time is output.

[0046] In the specific implementation process of this embodiment, the training of the multi-layer neural network model includes:

[0047] Collect the occultation VTEC characteristic parameters calculated from the occultation event;

[0048] Collect the site VTEC characteristic parameters calculated from sparse GNSS sites;

[0049] The VTEC characteristic parameters obtained from the occultation event and the VTEC characteristic parameters obtained from the sparse GNSS stations are time-matched to obtain a comprehensive dataset VTEC, which includes a training set and a validation set.

[0050] The training set and the validation set are input into the multi-layer neural network model for training to obtain a trained multi-layer neural network model.

[0051] In this embodiment, the specific steps and training process of the method are described in detail:

[0052] Step 1: Obtain the electron density profile product obtained by inverting the long-term occultation event, including the time, longitude, latitude, profile height and electron density sequence of the occultation, and calculate the ionospheric vertical total electron content (VTEC) based on the profile height and electron density sequence.

[0053] Step 2: Select several sparsely distributed GNSS stations (generally no more than 20) evenly distributed around the globe. Ionospheric penetration points are determined using station receivers and observation satellites. TEC (STEC) along the slant path is calculated based on GNSS observations. This is projected vertically and converted into the VTEC (vertical total electron content) characteristic parameter of the penetration point. This creates a long-term dataset containing time, longitude and latitude of the penetration point, and VTEC.

[0054] Step 3: Obtain the 1-hourly gridded product of the Global Ionospheric Map (GIM) and calculate the hourly global mean TEC by area-weighted average as the benchmark data.

[0055] Step 4: Time-match the VTEC characteristic parameters calculated from the occultation event with the VTEC characteristic parameters calculated from the sparse GNSS stations. The matching time window is the hourly global average TEC and the occultation and station VTEC data within 30 minutes before and after.

[0056] Step 5: Construct a comprehensive dataset containing the VTEC characteristic parameters of the occultation observations (time, longitude, latitude, VTEC) and the VTEC characteristic parameters of the sparse sites (time, puncture point longitude, puncture point latitude, VTEC). The matched dataset is divided into a training set and a validation set.

[0057] Step 6: Construct a multi-layer neural network model. The input layer is the VTEC characteristic parameters (time, longitude, latitude, and VTEC) calculated from discrete occultation events during the different time periods described above, and the output layer is the global mean TEC at the corresponding time. The model includes multiple hidden layers to capture the complex nonlinear relationship between occultation observations and sparse site data.

[0058] Step 7: Input the training set and validation set into the multi-layer neural network model for training and save the model parameters.

[0059] Step 8: Calculate the equivalent global average TEC using the trained neural network model. Obtain the VTEC characteristic parameters (time, longitude, latitude, and VTEC) calculated for each 30-minute period before and after the reference time to be calculated. These parameters are input into the trained neural network model to obtain the equivalent global average TEC at the corresponding time.

[0060] The specific implementation principle is further explained for the occultation VTEC characteristic parameters calculated in step 1:

[0061] Occultation observation is a technology that uses the refraction and delay of satellite signals when they pass through the Earth's atmosphere to invert atmospheric and ionosphere parameters. The COSMIC (Constellation Observing System for Meteorology, Ionosphere and Climate) satellite system is one of the important platforms for occultation observation. It obtains electron density information of the ionosphere by receiving GPS signals and recording the changes in the signals when they pass through the ionosphere. The electron density profile product of the COSMIC satellite's occultation event provides the electron density distribution from the satellite to the Earth's surface. Electron density .in is the reference altitude The electron density at the vertical axis is calculated as follows: VTEC = ,in, and are the upper and lower limits of the integration, respectively, which usually correspond to the bottom and top altitudes of the ionosphere (e.g., 80 km to 1000 km). For a specific electron density profile, VTEC can be calculated using numerical integration methods (e.g., trapezoidal method or Simpson method). For example, assuming that the electron density line at discrete altitudes As mentioned above, VTEC can be approximately calculated by the following formula:

[0062] ;

[0063] Where N is the number of discrete height points; are adjacent height points; is the electron density at the corresponding height point.

[0064] The time, longitude, and latitude in the vertical total electron content (VTEC) are determined as follows:

[0065] Time: The time of the occultation event can be determined from the timestamp of the satellite signal;

[0066] Longitude and latitude: The longitude and latitude of the projected point on the Earth's surface when the occultation event occurs can be calculated based on the geometric relationship between the satellite orbit parameters and the occultation event.

[0067] The specific implementation principle is further explained for the site VTEC characteristic parameters calculated in step 1:

[0068] The sparse GNSS station calculation is based on GNSS observation data, but it selects 20 representative stations around the world to represent the global stations. It still inverts the ionospheric delay by processing pseudorange and carrier phase observations, and then calculates VTEC.

[0069] GNSS observation data:

[0070] GNSS systems (such as GPS, GLONASS, and Galileo) transmit dual-frequency signals (e.g., L1 and L2 bands), allowing receivers to measure pseudorange and carrier phase. These observations contain information about ionospheric delay, which can be used to calculate TEC (total electron content).

[0071] STEC (Slant Total Electron Content) calculation:

[0072] (1) Pseudorange observation equation:

[0073] ;

[0074] in, is the pseudorange observation value from the receiver to satellite i, ( , , ) is the position coordinate of satellite i, ( , , ) is the position coordinate of the receiver, c is the speed of light, is the receiver clock error. is the clock error of satellite i. is the ionospheric delay, is the tropospheric delay, is the observation noise.

[0075] (2) Carrier phase observation equation:

[0076] ;in, is the carrier phase observation from the receiver to satellite i. is the carrier wavelength. is the whole-week ambiguity.

[0077] (3) Dual-frequency observation equation: In order to eliminate the influence of ionospheric delay, dual-frequency observation values ​​can be used. Assuming that the L1 and L2 frequency bands are used, the dual-frequency observation equation can be expressed as:

[0078] ;

[0079] ;

[0080] in, are the pseudorange observation values ​​of L1 and L2 frequency bands respectively. are the carrier phase observation values ​​of the L1 and L2 frequency bands respectively; the ionospheric delay I can be calculated from the dual-frequency observation values:

[0081] ;

[0082] in, and are the frequencies of the L1 and L2 bands respectively, are the wavelengths of the L1 and L2 frequency bands respectively.

[0083] VTEC (Vertical Total Electron Content) calculation:

[0084] STEC is the total electron content along the oblique path, while VTEC is the total electron content along the vertical path. VTEC can be calculated by projecting STEC onto the vertical direction.

[0085] VTEC=STEC·cos(elevation); where elevation is the elevation angle of the satellite, that is, the angle between the satellite signal and the ground plane.

[0086] Regarding step 8, global average TEC calculation, the specific implementation principle is further explained:

[0087] Obtain CODE GIM products. CODE GIM products are global ionospheric VTEC (vertical total electron content) grid data generated based on global GNSS observation data. They are usually provided in the form of a 1°×1° longitude and latitude grid with a time resolution of 1 hour.

[0088] Area-weighted average calculation of grid data. In order to calculate the global average TEC, it is necessary to perform area-weighted average processing on the GIM grid data. The specific steps are as follows:

[0089] (1) Calculation of grid area:

[0090] For each 1°×1° grid cell, its area A can be approximately calculated by the following formula:

[0091] ; where R is the average radius of the Earth (about 6371 km). is the latitude of the i-th grid row. is the factor that converts degrees to radians.

[0092] (2) Area weighted average formula:

[0093] The global mean TEC can be calculated by taking the area-weighted average of the VTEC values ​​of all grid cells:

[0094] ;

[0095] Where N and M are the number of grid cells in latitude and longitude respectively (usually 180 and 360). is the VTEC value of the grid cell in row i and column j.

[0096] In this embodiment, the multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters. The normalized characteristic parameters are then input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing before reaching the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

[0097] In this embodiment, the multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters. The normalized characteristic parameters are then input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing before reaching the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

[0098] The method of the present invention is described below with a specific embodiment: an embodiment of calculating the global mean TEC using the TEC product and CODE GIM product obtained by occultation observations and GNSS station observations respectively;

[0099] Download the COSMIC satellite's discrete electron density profiles for 11 years (one solar cycle) from 2010 to 2020. Calculate the characteristic parameters of the occultation events, including time, longitude, latitude, and VTEC. Select 20 GNSS stations evenly distributed around the world.

[0100] The ionospheric puncture point is determined by the station receiver and the observation satellite, and the TEC on the slant path (i.e., STEC) is calculated based on the GNSS observations. The STEC is then projected vertically and converted into the VTEC characteristic parameters of the puncture point.

[0101] Repeat step (2) to obtain the VTEC characteristic parameters of each station in turn, forming a long-term data set containing time, puncture point longitude, puncture point latitude, TEC and other characteristics, and further obtain the VTEC characteristic parameters.

[0102] The global ionospheric map (GIM) grid product of 1 hour from the Center for Orbit Determination in Europe (CODE) is obtained, and the global average TEC is calculated by area-weighted average.

[0103] Step (4) is repeated to calculate the global average TEC of each hour in the above long time series (2010-2020) from the global ionospheric map (GIM) grid product.

[0104] The VTEC characteristic parameters calculated from the occultation event are time-matched with the VTEC characteristic parameters calculated from the sparse GNSS site, and the matching time window is the occultation and site VTEC data in the 30-minute period before and after the global average TEC of each hour.

[0105] An integrated data set is constructed, containing the VTEC characteristic parameters (time, longitude, latitude, VTEC) of the occultation observation and the VTEC characteristic parameters (time, piercing point longitude, piercing point latitude, VTEC) of the sparse site. The data set is divided into a training set and a validation set. (Here, the data set ratio is set to 8:2).

[0106] A back propagation neural network (BPNN) model containing 3 layers of hidden layers is constructed. The input layer is the VTEC characteristic parameters (time, longitude, latitude, VTEC) calculated in the above different time periods, and the output layer is the global average TEC at the corresponding time.

[0107] The training set and validation set data are input into the BPNN, the model is trained to convergence, and the model parameters are saved.

[0108] The trained neural network model is used to calculate the equivalent global average TEC. The VTEC characteristic parameters (time, longitude, latitude, VTEC) calculated from the discrete occultation events of the COSMIC satellite in the 30-minute period before and after the reference time to be calculated are input into the trained neural network model, and the equivalent global average TEC at the corresponding time is obtained.

[0109] Embodiment 2: A star-ground combined global average TEC model calculation system, the system comprising:

[0110] A first data acquisition module for acquiring occultation VTEC characteristic parameters calculated from occultation events;

[0111] A second data acquisition module for acquiring site VTEC characteristic parameters calculated from sparse GNSS sites;

[0112] A data fusion module for time-matching the VTEC characteristic parameters calculated from the occultation event with the VTEC characteristic parameters calculated from the sparse GNSS site to obtain the VTEC characteristic parameters calculated in the 30-minute period before and after the reference time to be calculated.

[0113] The data calculation module is used to input the VTEC characteristic parameters calculated in the 30-minute period before and after the reference time to be calculated into a pre-trained multi-layer neural network model, and output the equivalent global average TEC at the corresponding time.

[0114] In this embodiment, the multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters. The normalized characteristic parameters are then input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing before reaching the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

[0115] This paper combines sparse GNSS stations with occultation observations to construct a global average TEC model based on an artificial neural network. Occultation observations compensate for the limited GNSS observations in areas such as oceans and deserts. GNSS station data can improve observation accuracy in high-latitude regions, and the combination of the two can also achieve more complete vertical coverage. Combining GNSS and occultation observation data can reduce the errors of a single data source, and their combined use can significantly improve the accuracy of ionospheric models.

[0116] This method effectively makes up for the shortcomings of a single observation method, significantly improving the coverage, data accuracy and modeling capabilities of ionospheric monitoring, thereby improving the modeling accuracy of the global average TEC.

[0117] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A satellite-ground joint global average TEC model calculation method, characterized in that: The method comprises: Obtain the occultation VTEC characteristic parameters calculated from the occultation event; Obtain the site VTEC characteristic parameters calculated from sparse GNSS sites; The VTEC characteristic parameters calculated from the occultation event are time-matched with the VTEC characteristic parameters calculated from the sparse GNSS stations to obtain the VTEC characteristic parameters calculated for 30 minutes before and after the reference time to be calculated. The VTEC characteristic parameters calculated in the 30-minute period before and after the reference time to be calculated are input into a pre-trained multi-layer neural network model, and the equivalent global average TEC at the corresponding time is output; The multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters. The normalized characteristic parameters are then input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing before reaching the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

2. The satellite-ground combined global average TEC model calculation method according to claim 1, characterized in that: The training of the multi-layer neural network model includes: Collect the occultation VTEC characteristic parameters calculated from the occultation event; Collect the site VTEC characteristic parameters calculated from sparse GNSS sites; The VTEC characteristic parameters obtained from the occultation event and the VTEC characteristic parameters obtained from the sparse GNSS stations are time-matched to obtain a comprehensive dataset VTEC, which includes a training set and a validation set. The training set and the validation set are input into the multi-layer neural network model for training to obtain a trained multi-layer neural network model.

3. The satellite-ground combined global average TEC model calculation method according to claim 2, characterized in that: The training of the multi-layer neural network model also includes: The global average TEC is calculated hourly by area-weighted average in the grid product of the global ionosphere map obtained at 1-hour intervals. This is used as the benchmark data, and the accuracy of the model is adjusted by the benchmark data when training the model.

4. The satellite-ground combined global average TEC model calculation method according to claim 1, characterized in that: The occultation VTEC characteristic parameters and site VTEC characteristic parameters include longitude, latitude, time and vertical total electron content VTEC.

5. The satellite-ground combined global average TEC model calculation method according to claim 1, characterized in that: The matching time window is the hourly global mean TEC and the occultation and sparse GNSS site VTEC data within a 30-minute period before and after it.

6. The satellite-ground combined global average TEC model calculation method according to claim 1, characterized in that: Obtain the occultation VTEC characteristic parameters calculated from the occultation event, including: The electron density profile product obtained by inverting a long series of occultation events is obtained, including the time, longitude, latitude, profile height and electron density sequence of the occultation, and the occultation VTEC characteristic parameters are calculated based on the profile height and electron density sequence.

7. The satellite-ground combined global average TEC model calculation method according to claim 1, characterized in that: Obtain the site VTEC characteristic parameters calculated for sparse GNSS sites, including: Several sparsely distributed GNSS stations are selected around the world. The ionospheric puncture points are determined through the station receivers and observation satellites. The total electron content (TEC) on the slant path is calculated based on the GNSS observations and projected vertically to convert them into the VTEC characteristic parameters of the puncture point.

8. A satellite-ground joint global average TEC model calculation system, characterized by: The system comprises: A first data acquisition module is used to obtain occultation VTEC characteristic parameters calculated from the occultation event; The second data acquisition module is used to obtain the site VTEC characteristic parameters calculated by the sparse GNSS site; The data fusion module is used to time-match the VTEC characteristic parameters calculated from the occultation event with the VTEC characteristic parameters calculated from the sparse GNSS stations, and obtain the VTEC characteristic parameters calculated for each 30-minute period before and after the reference time to be calculated; The data calculation module is used to input the VTEC characteristic parameters calculated in the 30-minute period before and after the reference time to be calculated into a pre-trained multi-layer neural network model, and output the equivalent global average TEC at the corresponding time; The multi-layer neural network model includes an input layer, a hidden layer, and an output layer. The input layer is used to clean and normalize the VTEC characteristic parameters. The normalized characteristic parameters are then input and sequentially passed through the neurons of each hidden layer for weighted summation and activation function processing before reaching the output layer. The output layer uses a linear activation function to directly output the weighted summation result to obtain the predicted global average TEC value. The multi-layer neural network model uses a loss function to calculate the square difference between the predicted value and the actual value, and minimizes the error by adjusting the model parameters.

Images