Precision Troposphere Modeling Method, Device, Equipment and Medium

Abstract

The present invention discloses a precise tropospheric modeling method, device, equipment and medium, including: in the modeling data input link, the present invention introduces high-precision reference station data and a numerical weather prediction model with wide spatial coverage and strong stability. Through the tropospheric delay information fusion model, the advantages of these two types of information are fused to achieve complementary advantages, generating the basic tropospheric modeling data that combines the "high precision" of Beidou and the "wide spatial coverage and strong stability" of the numerical weather prediction model, that is, the improved grid-type tropospheric delay information, thus solving the problems of limited spatial expression ability and poor robustness of the modeling basic data in the prior art. At the same time, by introducing a coordinate normalization factor into the polynomial function, the numerical stability of the fitting process is controlled, and then the service area of the polynomial function model is expanded, effectively solving the problems of poor numerical stability and small applicable range in the existing modeling technology.

Description

Technical Field

[0001] The present invention relates to the technical field of tropospheric delay modeling in satellite navigation applications, and particularly to a precise tropospheric modeling method, device, equipment, and medium. Background Art

[0002] Tropospheric delay is one of the main error sources affecting the positioning, navigation, and timing accuracy of Beidou / Global Navigation Satellite System (GNSS). With the development of Beidou / GNSS positioning technology and its integrated application in the new generation of information technology industry, the service accuracy of the Beidou / GNSS augmentation system is being improved from the meter level to the decimeter / centimeter level, and its service mode is gradually expanding from ground-based (regional) to space-based (wide-area).

[0003] However, due to the limited spatial expression ability of the modeling input data, the weak robustness of the modeling process, and the small applicable range of the modeling expression method, the existing precise tropospheric modeling methods cannot meet the Beidou space-based augmentation service under the conditions of a large range (the entire Chinese region) and sparse and uneven distribution of Beidou / GNSS reference stations. Summary of the Invention

[0004] Based on this, it is necessary to provide a precise tropospheric modeling method, device, equipment, and medium to solve the problems such as the limited spatial expression ability of the modeling input data, the weak robustness of the modeling process, and the small applicable range of the modeling expression method.

[0005] A precise tropospheric modeling method, the method comprising:

[0006] Calculating grid-based tropospheric delay information of all reference stations within the service area based on a numerical weather prediction model, and inputting the grid-based tropospheric delay information into a tropospheric delay information fusion model to obtain the output improved grid-based tropospheric delay information; wherein, the tropospheric delay information fusion model is used to fuse the accuracy of reference station data and the wide spatial coverage and stability of the numerical weather prediction model;

[0007] Fitting the improved grid-based tropospheric delay information based on an improved polynomial function model with a coordinate normalization factor to obtain a precise tropospheric model.

[0008] In one embodiment, before calculating the grid-based tropospheric delay information of all reference stations within the service area based on the numerical weather prediction model, it further includes:

[0009] Calculating the first site-based tropospheric delay information of all reference stations within the service area based on the numerical weather prediction model, and calculating the second site-based tropospheric delay information of all reference stations within the service area based on the reference station data;

[0010] Taking the first site-based tropospheric delay information as the model input and the second site-based tropospheric delay information as the model output, a training dataset is constructed, and the generalized regression neural network model is trained based on the training dataset to obtain the tropospheric delay information fusion model.

[0011] In one embodiment, the training of the generalized regression neural network model based on the training dataset to obtain the tropospheric delay information fusion model includes:

[0012] Based on the performance function, the distribution density of the radial basis function in the generalized regression neural network model is adjusted to a preset value to obtain the tropospheric delay information fusion model; wherein, the performance function is determined based on the capacity of the first site-based tropospheric delay information and the error between the first site-based tropospheric delay information and the second site-based tropospheric delay information.

[0013] In one embodiment, the tropospheric delay information fusion model includes an input layer, a pattern layer, a summation layer, and an output layer. Inputting the grid-based tropospheric delay information into the tropospheric delay information fusion model to obtain the output improved grid-based tropospheric delay information includes:

[0014] Each neuron in the input layer passes the input input vector to all neurons in the pattern layer through a linear function; wherein, the number of neurons in the input layer is equal to the dimension of the grid-based tropospheric delay information, and the input vector is any vector in the grid-based tropospheric delay information;

[0015] Each neuron in the pattern layer passes the input vector to the summation layer through a radial basis function; wherein, the number of neurons in the pattern layer is equal to the number of the grid-based tropospheric delay information, and the radial basis function indicates the exponential form of the square of the Euclidean distance between the input vector and the grid-based tropospheric delay information;

[0016] The first summation neuron in the summation layer calculates the output sum of the pattern layer and passes it to the output layer, and the second summation neuron in the summation layer calculates the weighted sum of the pattern layer and passes it to the output layer; wherein, the output sum is obtained by arithmetically summing the outputs of all neurons in the pattern layer, and the weighted sum is obtained by weighted summing the outputs of all neurons in the pattern layer;

[0017] The output layer calculates the ratio of the weighted sum to the output sum and takes the obtained ratio as the improved grid-based tropospheric delay information.

[0018] In one embodiment, the improved polynomial function model based on the coordinate normalization factor fits the improved grid tropospheric delay information to obtain a precise tropospheric model, including:

[0019] Based on the least squares method, use the improved polynomial function model composed of the longitude difference and latitude difference between the grid point and the regional center point, the coordinate normalization factor, and the polynomial coefficients to fit the improved grid tropospheric delay information to obtain a precise tropospheric model.

[0020] In one embodiment, the obtaining of the output improved grid tropospheric delay information includes:

[0021] Obtain the output improved grid tropospheric delay information reduced to the surface, based on the elevation reduction method of the empirical tropospheric model, strip the elevation component of the improved grid tropospheric delay information reduced to the surface, and project it onto the geoid to obtain the improved grid tropospheric delay information reduced to the geoid height.

[0022] The fitting of the improved grid tropospheric delay information includes: fitting the improved grid tropospheric delay information reduced to the geoid height.

[0023] In one embodiment, the elevation reduction method based on the empirical tropospheric model strips the elevation component of the improved grid tropospheric delay information reduced to the surface and projects it onto the geoid to obtain the improved grid tropospheric delay information reduced to the geoid height, including:

[0024] Calculate the first tropospheric delay information at the geoid height and the second tropospheric delay information at the grid point height based on the empirical tropospheric model, calculate the difference between the first tropospheric delay information and the second tropospheric delay information, and use the obtained difference as the elevation component;

[0025] Calculate the sum of the improved grid tropospheric delay information reduced to the surface and the elevation component, and use the obtained sum as the improved grid tropospheric delay information reduced to the geoid height.

[0026] A precise tropospheric modeling device, the device includes:

[0027] A modeling data input module, configured to calculate the grid tropospheric delay information of all reference stations within the service area coverage based on the numerical weather prediction model, input the grid tropospheric delay information into the tropospheric delay information fusion model, and obtain the output improved grid tropospheric delay information; wherein, the tropospheric delay information fusion model is used to fuse the accuracy of the numerical weather prediction model and the wide spatial coverage and stability of the reference station data.

[0028] A modeling and expression module, configured to fit improved grid tropospheric delay information based on an improved polynomial function model carrying a coordinate normalization factor, so as to obtain a precise tropospheric model.

[0029] A computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to execute the steps of the above-mentioned precise tropospheric modeling method.

[0030] A precise tropospheric modeling device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to execute the steps of the above-mentioned precise tropospheric modeling method.

[0031] The present invention provides a precise tropospheric modeling method, device, equipment and medium. In the modeling data input link, the present invention introduces high-precision reference station data and a numerical weather prediction mode with wide spatial coverage and strong stability. Through the tropospheric delay information fusion model, the advantages of these two types of information are fused to achieve complementary advantages, generating tropospheric modeling basic data that combines the "high precision" of Beidou and the "wide spatial coverage and strong stability" of the numerical weather prediction mode, that is, improved grid tropospheric delay information, thus solving the problems of limited spatial expression ability and poor robustness of the modeling basic data in the prior art. At the same time, by introducing a coordinate normalization factor into the polynomial function, the numerical stability of the fitting process is controlled, thereby expanding the service area of the polynomial function model, effectively solving the problems of poor numerical stability and small application range in the existing modeling technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Among them:

[0034] Figure 1 It is a schematic flow chart of the precise tropospheric modeling method in the first embodiment;

[0035] Figure 2 It is a schematic flow chart of the precise tropospheric modeling method in the second embodiment;

[0036] Figure 3 It is a schematic structural diagram of a generalized regression neural network model;

[0037] Figure 4 It is an error distribution diagram of the delay information before and after fusion;

[0038] Figure 5 The test results between the fitting error and the polynomial degree when the coordinate normalization factor k takes different values;

[0039] Figure 6 The experimental comparison diagram of the present invention compared with the prior art;

[0040] Figure 7 The structural schematic diagram of the precise tropospheric modeling device in one embodiment;

[0041] Figure 8 The structural block diagram of the precise tropospheric modeling device in one embodiment. Specific embodiments

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0043] As Figure 1 shown, Figure 1 is the flow schematic diagram of the precise tropospheric modeling method in the first embodiment. The steps provided by the precise tropospheric modeling method in this embodiment include:

[0044] Step 102, calculate the grid-type tropospheric delay information of all reference stations within the coverage service area based on the numerical weather prediction model, and input the grid-type tropospheric delay information into the tropospheric delay information fusion model to obtain the output improved grid-type tropospheric delay information.

[0045] In this embodiment, it is also necessary to pre-train a tropospheric delay information fusion model, which is used to fuse the accuracy of reference station data and the wide spatial coverage and stability of the numerical weather prediction model. Refer to Figure 2 , the training process of this model includes:

[0046] (1), calculate the first station-type tropospheric delay information of all reference stations within the coverage service area based on the numerical weather prediction model, and calculate the second station-type tropospheric delay information of all reference stations within the coverage service area based on the reference station data.

[0047] Among them, the numerical weather prediction model refers to a method that, based on the actual situation of the atmosphere, under certain initial and boundary conditions, performs numerical calculations using a large computer to solve the fluid mechanics and thermodynamics equations describing the weather evolution process, and predicts the atmospheric motion state and weather phenomena in a future period of time.

[0048] The reference station data refers to all the measured satellite data at the reference station. It can be understood that the reference station data has higher accuracy, while the data under the numerical weather prediction model has better wide - space coverage and stability.

[0049] The troposphere is the atmosphere layer below 50 km from the ground, including the troposphere and the stratosphere. When GPS passes through the troposphere and the stratosphere, its propagation speed will change, and the propagation path will bend, resulting in additional path delay, which is called tropospheric delay. The site - type tropospheric delay information is the tropospheric delay data measured at all base stations within the service area. Among them, the calculation method of the site - type tropospheric delay information belongs to the prior art and will not be elaborated here.

[0050] (2) Construct a training data set with the first site - type tropospheric delay information as the model input and the second site - type tropospheric delay information as the model output, and train the generalized regression neural network model based on the training data set to obtain the tropospheric delay information fusion model.

[0051] See Figure 3 , the generalized regression neural network model to be trained includes a total of four - layer structures, namely the input layer, the pattern layer, the summation layer, and the output layer. The function of each layer is as follows:

[0052] Each neuron in the input layer passes the input vector to all neurons in the pattern layer through a linear function.

[0053] Among them, the number of neurons in the input layer is equal to the dimension of the input tropospheric delay information, and the input vector is any one vector in the input tropospheric delay information. In this embodiment, the number of neurons in the input layer is 4, corresponding to the longitude, latitude, elevation, and numerical weather prediction model tropospheric delay information at the selected site, denoted as x1, x2, L, x4. The dimension of the output layer data is 1, that is, the fused tropospheric delay information. Suppose there are m input samples, and the i - th input sample is denoted as The corresponding i - th output set is denoted as Y i =[y i .

[0054] Each neuron in the pattern layer passes the input vector to the summation layer through a radial basis function.

[0055] Among them, the number of neurons in the pattern layer is equal to the number of tropospheric delay information, that is, the above-mentioned m. The radial basis function indicates the exponential form of the square of the Euclidean distance between the input vector and the tropospheric delay information. Its mathematical expression is:

[0056]

[0057] In the formula, p i is the output of neuron i in the pattern layer, X is all the first-site tropospheric delay information, and X i is the i-th first-site tropospheric delay information. σ is a hyperparameter of the model, which needs to be set in advance or can be determined by an optimization algorithm.

[0058] The first summation neuron in the summation layer calculates the sum of the outputs of the pattern layer and passes it to the output layer. The second summation neuron in the summation layer calculates the weighted sum of the pattern layer and passes it to the output layer.

[0059] Among them, the sum of outputs is obtained by arithmetically summing the outputs of all neurons in the pattern layer. Its data expression is:

[0060]

[0061] The weighted sum is obtained by weighted summing the outputs of all neurons in the pattern layer. Its mathematical expression is:

[0062]

[0063] The output layer calculates the ratio of the weighted sum to the sum of outputs. The output of neuron j corresponds to the j-th element of the estimated result , and its mathematical expression is:

[0064]

[0065] In this application, j = 1.

[0066] In the above model, σ in the pattern layer is the only undetermined parameter of the model, which represents the distribution density spread of the radial basis function. Spread has a relatively large impact on the performance and accuracy of the generalized regression neural network model. Theoretically speaking, the smaller the spread, the more accurate the approximation of the function, but the approximation process will be less smooth; the larger the spread, the smoother the approximation process, but the approximation error will be relatively large.

[0067] In the actual training process, first determine the performance function MAE, which is expressed as:

[0068]

[0069] Among them, Q is the number of first-site tropospheric delay information, and ei is the error between the i-th first site tropospheric delay information and the i-th second site tropospheric delay information.

[0070] Based on this performance function MA, using the first site tropospheric delay information as the model input and the second site tropospheric delay information as the model output, a training dataset is constructed. The distribution density spread of the radial basis function in the generalized regression neural network model is adjusted pairwise to a preset value until both the performance and accuracy of the model meet the preset requirements, thereby obtaining a trained tropospheric delay information fusion model.

[0071] See Figure 2 , and then the grid tropospheric delay information of all reference stations within the service coverage area can be calculated based on the numerical weather prediction model. Among them, this grid data refers to the internal data stored in a grid structure in the computer. Then, the grid tropospheric delay information (equivalent to the first site tropospheric delay information input during training) and the coordinates (longitude, latitude, altitude) of each grid point are input into the tropospheric delay information fusion model, and the improved grid tropospheric delay information output by the model is obtained. This improved grid tropospheric delay information not only integrates the accuracy of the reference station data but also integrates the wide spatial coverage and stability of the numerical weather prediction model.

[0072] Meanwhile, see Figure 4 , Figure 4 is the error distribution diagram of the delay information before and after fusion. The tropospheric delay error distribution before fusion is Figure 4 Regions 1 and 2 in. It can be seen that there is a systematic deviation in the error distribution before fusion (the peak is offset). The tropospheric delay error distribution after fusion is Regions 1 and 3, without systematic deviation (the peak is exactly at 0), which indicates that by fusing the reference station data, the systematic deviation of the original data can be eliminated, and the accuracy of the tropospheric modeling basic dataset (all improved grid tropospheric delay information) is improved.

[0073] Step 104, based on the improved polynomial function model with coordinate reduction factors, fit the improved grid tropospheric delay information to obtain a precise tropospheric model.

[0074] In a specific embodiment, see Figure 2 , and the complexity of the modeling basic dataset is also reduced through the elevation reduction method, which specifically includes:

[0075] (1). Obtain the improved grid tropospheric delay information reduced to the surface, and based on the elevation reduction method of the empirical tropospheric model, strip out the elevation component of the improved grid tropospheric delay information reduced to the surface and project it onto the geoid to obtain the improved grid tropospheric delay information at the height of the geoid.

[0076] Specifically, based on the empirical troposphere model, calculate the first tropospheric delay information at the geoid height and the second tropospheric delay information at the grid point height, calculate the difference between the first tropospheric delay information and the second tropospheric delay information, and use the obtained difference as the elevation component; calculate the sum of the improved grid-type tropospheric delay information reduced to the surface and the elevation component, and use the obtained sum as the improved grid-type tropospheric delay information reduced to the geoid height. This process is expressed as:

[0077]

[0078] In the above formula, represents the first tropospheric delay information, represents the second tropospheric delay information; represents the elevation component; represents the improved grid-type tropospheric delay information reduced to the surface, represents the improved grid-type tropospheric delay information reduced to the geoid height. Through the above process, the elevation component of the tropospheric delay information is stripped, thereby reducing the subsequent fitting complexity.

[0079] (2) Fit the improved grid-type tropospheric delay information reduced to the geoid height.

[0080] That is, based on the improved polynomial function model, fit the improved grid-type tropospheric delay information reduced to the geoid height after dimensionality reduction to obtain a precise troposphere model. The specific fitting process is as follows:

[0081] Based on the least squares method, use the improved polynomial function model composed of the longitude difference and latitude difference between the grid point and the regional center point, the coordinate reduction factor, and the polynomial coefficients to fit the improved grid-type tropospheric delay information. Its mathematical expression is:

[0082]

[0083] In the above formula, n is the maximum degree of the polynomial describing the horizontal distribution of tropospheric information; k is the coordinate reduction factor, which needs to be determined through testing; dL is the longitude difference between the grid point and the regional center point, and dB is the latitude difference between the grid point and the regional center point; a ij are the polynomial coefficients, which are estimated by the least squares method.

[0084] See Figure 5 , Figure 5 is the test result between the fitting error and the polynomial degree when taking different coordinate reduction factors k. Through Figure 5It can be seen that when k = 1 (i.e., non-normalized coordinates), the degree of the polynomial fitting function is maximally increased to 8 times to reach its maximum fitting ability. At this time, the fitting error is relatively large, which is 11 mm. When k takes a smaller value, the fitting error of the polynomial when reaching the maximum fitting ability also decreases accordingly, indicating that introducing an appropriate coordinate normalization factor can improve the fitting ability of the polynomial function, thereby demonstrating the effectiveness of the improved polynomial function model proposed by the present invention. When k = 0.1, the polynomial reaches its maximum fitting ability at the 15th degree, and its fitting error is 7.0 mm. However, when k is further decreased from 0.1 to 0.01, the fitting residual starts to increase when the degree of the polynomial exceeds 13. Therefore, k = 0.1 is the optimal value.

[0085] After the fitting of the model coefficients is completed, a precise troposphere model can be obtained. In actual operation, the coefficients of the improved polynomial function model are also improved through real-time coding and broadcast to users to realize the application of the precise troposphere model.

[0086] Finally, reference can be made to Figure 6 , Figure 6 which is the experimental comparison diagram of the present invention compared with the prior art. The model error of using the tropospheric delay information of a single reference station usually reaches its fitting limit at a relatively low degree, and the accuracy when reaching the fitting limit is relatively low. This is mainly because using the tropospheric delay information of a single reference station as the basic data for modeling, its distribution is relatively sparse and uneven, and the spatial expression ability is limited. Compared with the model established using the tropospheric delay information of a single reference station, the model established using the tropospheric delay information of a single numerical weather prediction model significantly improves the spatial expression ability of the basic data for modeling. Therefore, the modeling accuracy can be further improved by increasing the degree of the polynomial. However, it has an obvious systematic bias, indicating that there is a systematic bias in the tropospheric delay information of a single numerical weather prediction model. The model that fuses the tropospheric delay information of the reference station and the numerical weather prediction model combines the advantages of the non-bias of the tropospheric information system of the reference station and the strong spatial expression ability of the tropospheric information of the numerical weather prediction model, and avoids their respective deficiencies, achieving the optimal modeling effect. At the same time, in the present invention, by introducing a coordinate normalization factor into the polynomial function, the numerical stability of the fitting process is controlled, thereby expanding the service area of the polynomial function model, and effectively solving the problems of poor numerical stability and small applicable range in the existing modeling technology.

[0087] In one embodiment, as Figure 7 shown, a precise troposphere modeling device is proposed, and the device includes:

[0088] A modeling data input module 702 is configured to calculate grid - type tropospheric delay information of all reference stations within a service coverage area based on a numerical weather prediction model, input the grid - type tropospheric delay information into a tropospheric delay information fusion model, and obtain the output improved grid - type tropospheric delay information. Among them, the tropospheric delay information fusion model is used to fuse the accuracy of the numerical weather prediction model and the wide - space coverage and stability of reference station data.

[0089] A modeling expression module 704 is configured to fit the improved grid - type tropospheric delay information based on an improved polynomial function model carrying a coordinate normalization factor to obtain a precise tropospheric model.

[0090] Figure 8 The internal structure diagram of a precise tropospheric modeling device in an embodiment is shown. As Figure 8 shown, the precise tropospheric modeling device includes a processor, a memory, and a network interface connected through a system bus. Among them, the memory includes a non - volatile storage medium and an internal memory. The non - volatile storage medium of the precise tropospheric modeling device stores an operating system and can also store a computer program. When the computer program is executed by the processor, the processor can implement the precise tropospheric modeling method. The internal memory can also store a computer program. When the computer program is executed by the processor, the processor can execute the precise tropospheric modeling method. Those skilled in the art can understand that Figure 8 the structure shown in

[0091] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the precise tropospheric modeling device to which the solution of the present application is applied. The specific precise tropospheric modeling device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0092] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented: calculating grid tropospheric delay information of all reference stations within a service area based on a numerical weather prediction model, inputting the grid tropospheric delay information into a tropospheric delay information fusion model to obtain improved grid tropospheric delay information as output; fitting the improved grid tropospheric delay information based on an improved polynomial function model carrying a coordinate reduction factor to obtain a precise tropospheric model.

[0093] It should be noted that the above precise tropospheric modeling method, device, equipment, and computer-readable storage medium belong to a general inventive concept, and the content in the embodiments of the precise tropospheric modeling method, device, equipment, and computer-readable storage medium can be mutually applicable.

[0094] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0095] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0096] The above embodiments only illustrate several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A precise troposphere modeling method, characterized in that, The method includes: Calculating grid tropospheric delay information of all reference stations within the service coverage area based on a numerical weather prediction model, and inputting the grid tropospheric delay information into a tropospheric delay information fusion model to obtain the output improved grid tropospheric delay information; wherein, the tropospheric delay information fusion model is used to fuse the accuracy of reference station data and the wide spatial coverage and stability of the numerical weather prediction model; Fitting the improved grid tropospheric delay information based on an improved polynomial function model with a coordinate normalization factor to obtain a precise tropospheric model; Wherein, before calculating the grid tropospheric delay information of all reference stations within the service coverage area based on the numerical weather prediction model, it further includes: Calculating the first site-based tropospheric delay information of all reference stations within the service coverage area based on the numerical weather prediction model, and calculating the second site-based tropospheric delay information of all reference stations within the service coverage area based on reference station data; Constructing a training dataset with the first site-based tropospheric delay information as the model input and the second site-based tropospheric delay information as the model output, and training a generalized regression neural network model based on the training dataset to obtain the tropospheric delay information fusion model; Wherein, training the generalized regression neural network model based on the training dataset to obtain the tropospheric delay information fusion model includes: Adjusting the distribution density of the radial basis function in the generalized regression neural network model to a preset value according to a performance function to obtain the tropospheric delay information fusion model; wherein, the performance function is determined based on the capacity of the first site-based tropospheric delay information and the error between the first site-based tropospheric delay information and the second site-based tropospheric delay information.

2. The method according to claim 1, characterized in that, The tropospheric delay information fusion model includes an input layer, a pattern layer, a summation layer, and an output layer. Inputting the grid tropospheric delay information into the tropospheric delay information fusion model to obtain the output improved grid tropospheric delay information includes: Each neuron in the input layer passes the input input vector to all neurons in the pattern layer through a linear function; wherein, the number of neurons in the input layer is equal to the dimension of the grid tropospheric delay information, and the input vector is any one vector in the grid tropospheric delay information; Each neuron in the pattern layer passes the input vector to the summation layer through a radial basis function; wherein, the number of neurons in the pattern layer is equal to the number of the grid tropospheric delay information, and the radial basis function indicates the exponential form of the square of the Euclidean distance between the input vector and the grid tropospheric delay information; The first summation neuron in the summation layer calculates the output sum of the pattern layer and passes it to the output layer, and the second summation neuron in the summation layer calculates the weighted sum of the pattern layer and passes it to the output layer; wherein, the output sum is obtained by arithmetically summing the outputs of all neurons in the pattern layer, and the weighted sum is obtained by weighted summing the outputs of all neurons in the pattern layer; The output layer calculates the ratio of the weighted sum to the output sum, and uses the obtained ratio as the improved grid tropospheric delay information.

3. The method according to claim 1, characterized in that, Based on the improved polynomial function model with a coordinate normalization factor, fitting the improved grid tropospheric delay information to obtain a precise tropospheric model, including: Based on the least squares method, using the improved polynomial function model composed of the longitude difference and latitude difference between the grid point and the regional center point, the coordinate normalization factor, and the polynomial coefficients to fit the improved grid tropospheric delay information to obtain a precise tropospheric model.

4. The method according to claim 1, characterized in that, The obtaining of the output improved grid tropospheric delay information includes: Obtaining the improved grid tropospheric delay information reduced to the surface, based on the elevation reduction method of the empirical tropospheric model, stripping the elevation component of the improved grid tropospheric delay information reduced to the surface, and projecting it onto the geoid to obtain the improved grid tropospheric delay information at the height of the geoid. The fitting of the improved grid tropospheric delay information includes: fitting the improved grid tropospheric delay information at the height of the geoid.

5. The method according to claim 4, characterized in that, The elevation reduction method based on the empirical tropospheric model, stripping the elevation component of the improved grid tropospheric delay information reduced to the surface, and projecting it onto the geoid to obtain the improved grid tropospheric delay information at the height of the geoid, includes: Calculating the first tropospheric delay information at the height of the geoid and the second tropospheric delay information at the grid point height based on the empirical tropospheric model, calculating the difference between the first tropospheric delay information and the second tropospheric delay information, and using the obtained difference as the elevation component. Calculating the sum of the improved grid tropospheric delay information reduced to the surface and the elevation component, and using the obtained sum as the improved grid tropospheric delay information at the height of the geoid.

6. A precise troposphere modeling device, characterized in that, The device includes: A modeling data input module, configured to calculate the grid tropospheric delay information of all reference stations within the coverage service area based on the numerical weather prediction model, input the grid tropospheric delay information into the tropospheric delay information fusion model, and obtain the output improved grid tropospheric delay information; wherein, the tropospheric delay information fusion model is used to fuse the accuracy of the numerical weather prediction model and the wide spatial coverage and stability of the reference station data. A modeling expression module, configured to fit the improved grid tropospheric delay information based on the improved polynomial function model with a coordinate normalization factor to obtain a precise tropospheric model. The device is further configured to: Calculate the first site-based tropospheric delay information of all reference stations within the coverage service area based on the numerical weather prediction model, and calculate the second site-based tropospheric delay information of all reference stations within the coverage service area based on the reference station data. Construct a training data set with the first site-based tropospheric delay information as the model input and the second site-based tropospheric delay information as the model output, and train the generalized regression neural network model based on the training data set to obtain the tropospheric delay information fusion model. The device is further configured to: adjust the distribution density of the radial basis function in the generalized regression neural network model to a preset value based on a performance function, so as to obtain the tropospheric delay information fusion model; wherein, the performance function is determined based on the capacity of the first site-type tropospheric delay information and the error between the first site-type tropospheric delay information and the second site-type tropospheric delay information.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 5.

8. A precise troposphere modeling device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 5.

Images