A method for producing a large tropospheric delay dataset based on BeiDou continuously operating reference stations

Abstract

A method for producing a large tropospheric delay dataset based on Beidou continuously operating reference stations comprises: using original observation values of Beidou CORS stations to calculate the zenithal tropospheric total delay sequence of the CORS stations according to a precise single point positioning mode; using the zenithal tropospheric total delay sequence of the CORS stations, setting a missing day threshold, judging whether the tropospheric delay sequence has continuous missing data, and performing data segmentation; using the tropospheric delay sequence after the segmentation of the data segments, using a first-order difference method and a 3σ method to detect isolated tropospheric delay outliers; using the sequence after eliminating isolated tropospheric delay outliers, using wavelet decomposition, a threshold method, and a 3σ method to detect discrete tropospheric delay outliers; using the tropospheric delay sequence after eliminating isolated tropospheric delay outliers, using a deep neural network method to model and predict, and completing the missing data of the tropospheric delay sequence, and repeating the above steps until the production of the large tropospheric delay dataset of the entire CORS station network is completed.

Description

Technical Field

[0001] The present invention belongs to the field of Beidou / GNSS high-precision positioning, and in particular relates to a method for producing a large tropospheric delay data set based on a Beidou continuously operating reference station. Background Art

[0002] With the official launch of the BeiDou-3 global navigation satellite system in July 2020, China's Ministry of Natural Resources mandated that the national Continuously Operating Reference Station (CORS) network be upgraded and reconstructed by the end of 2022. Due to interruptions in the BeiDou CORS observation network and the lack of precise satellite products, the tropospheric delay series calculated using the CORS network inevitably contain gross errors and omissions. These tropospheric delay series containing gross errors and omissions negatively impact subsequent BeiDou / GNSS high-precision positioning. Therefore, as a key to BeiDou / GNSS high-precision positioning, the production of large tropospheric delay datasets based on BeiDou CORS is a key research topic.

[0003] There are currently two major technical difficulties in producing large tropospheric delay datasets.

[0004] The first is the processing of gross errors from a large number of station networks. Two common methods are used for processing Beidou / GNSS gross errors: hypothesis testing based on statistics and spectrum analysis. Hypothesis testing based on statistics requires that the sample data follow a specified distribution and can only remove one gross error at a time, thus requiring multiple iterative calculations and being inefficient. Spectral analysis based methods require that the sample data be evenly spaced in time. Tropospheric delay series based on the national Beidou CORS contain numerous gross error types and distributions, and the intervals between missing data are irregular. Furthermore, the number of Beidou CORS networks covering the entire country is enormous. Therefore, how to quickly and efficiently remove gross errors from such a large number of tropospheric delay series remains a current technical challenge.

[0005] The second challenge is completing sequences with large time spans. Currently, commonly used Beidou / GNSS data completion methods include spline interpolation and polynomial fitting. These methods are essentially linear fits, which are not very suitable for nonlinear patterns. Their applicability is also limited for completing sample sequences with large time spans. Tropospheric delay sequences based on the national Beidou CORS contain rich spatial information such as seasonal time, longitude / latitude / altitude, and altitude. The correlation between the tropospheric delay sequence and temporal and spatial information cannot be expressed using linear fitting. Furthermore, the nationwide Beidou CORS network spans a large time span. Therefore, how to quickly and accurately model the complex nonlinear relationships of tropospheric delays, and thereby complete massive tropospheric delay sequences, is the second current technical difficulty. Summary of the Invention

[0006] To address these two technical difficulties, this paper proposes a method for producing a large tropospheric delay dataset based on BeiDou continuously operating reference stations. This method can quickly, efficiently, and accurately remove gross errors from numerous tropospheric delay sequences and complete a large number of them, ultimately constructing a continuous and reliable large tropospheric delay dataset.

[0007] In order to achieve the above object, the technical solution of the present invention includes the following steps:

[0008] (1) Acquisition of tropospheric delay sequence: Using the original observation values of the BeiDou CORS station, the total zenith tropospheric delay sequence of the CORS station is obtained by solving the precise single point positioning mode.

[0009] (2) Segmentation of tropospheric delay sequence: Using the total zenith tropospheric delay sequence of the CORS station, a missing day threshold is set to determine whether there is a continuous missing in the tropospheric delay sequence and perform data segmentation.

[0010] (3) Detection of isolated tropospheric delay anomalies: Using the tropospheric delay sequence after segmentation of the data segments, the first-order difference method and the 3σ method are used to detect isolated tropospheric delay anomalies.

[0011] (4) Detection of discrete tropospheric delay anomalies: Using a sequence that eliminates isolated tropospheric delay anomalies, discrete tropospheric delay anomalies are detected using wavelet decomposition, threshold method, and 3σ method.

[0012] (5) Completion of missing data of tropospheric delay series: Using the tropospheric delay series that has been free of isolated and discrete outliers, deep neural network methods are used for modeling and prediction to complete the missing data of the tropospheric delay series.

[0013] (6) Repeat the above steps until the continuous and reliable large dataset of tropospheric delays of the entire CORS station network is completed.

[0014] The present invention has numerous beneficial effects.

[0015] First, isolated and discrete tropospheric delay anomalies are detected in steps. This method is highly efficient in detecting anomalies for the numerous Beidou CORS stations across China. Second, a deep neural network approach is used to complete missing tropospheric delay data. This method effectively reduces modeling errors in tropospheric delay sequences with large time spans and complex spatiotemporal correlations, thereby minimizing prediction errors in the completed tropospheric delays.

[0016] In summary, this invention addresses numerous issues related to BeiDou CORS tropospheric delay gross error processing and sequence completion, proposing a comprehensive and efficient method for producing a large BeiDou / GNSS tropospheric delay dataset. This approach holds great promise for the construction and upgrade of a nationwide BeiDou CORS network, which spans a large number of stations and a significant timeframe. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments are briefly introduced below.

[0018] Figure 1 This is a block diagram of the principles of producing a large tropospheric delay dataset based on Beidou continuously operating reference stations, provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0019] Figure 1 The following is a block diagram showing the principle of producing a large tropospheric delay data set based on the Beidou continuously operating reference station. Figure 1 The method for producing a large tropospheric delay dataset based on Beidou continuously operating reference stations is described in detail.

[0020] 1. Acquisition of tropospheric delay series

[0021] The original observation values of the BeiDou / GNSS Continuously Operating Reference Station (CORS) are obtained, and the total zenith tropospheric delay sequence of the CORS station is obtained according to the precise point positioning mode.

[0022] 2. Segmentation of tropospheric delay series

[0023] Determine whether there is continuity loss in the tropospheric delay sequence and set a threshold for the number of missing days. If the threshold is exceeded, the tropospheric delay sequence is segmented.

[0024] 3. Detection of isolated tropospheric delay anomalies

[0025] There are two main types of tropospheric delay series anomalies: isolated (appearing alone) and discrete (appearing in succession). Different methods are used to detect these two types of tropospheric delay series anomalies.

[0026] For isolated outliers: use the first-order difference method and the 3σ method for detection.

[0027] The specific implementation steps are:

[0028] First, the tropospheric delay series ZTD i (i=1,2,...,n) performs a difference to obtain the first-order difference sequence:

[0029] ΔZTDi =ZTD i+1 -ZTD i (i=1,2,...,n)

[0030] Then, the standard deviation σ of the first-order difference sequence is calculated using the Bessel formula:

[0031]

[0032] The confidence level is selected as 99.7%, and the corresponding critical value is 3. If the difference ΔZTD i Not satisfied with:

[0033] |ΔZTD i |<3σ,i=1,2,...,n

[0034] Then determine the corresponding ZTD i+1 It is an isolated gross error.

[0035] 4. Detection of discrete tropospheric delay anomalies

[0036] For discrete outliers: wavelet decomposition, threshold method and 3σ method are used for detection.

[0037] The specific implementation steps are:

[0038] First, we select the appropriate wavelet basis function and decomposition order based on the regularity and smoothness of the tropospheric delay sequence and the complexity of the tropospheric delay characteristics. After comprehensive consideration, we choose the db8 wavelet basis function and 8-layer decomposition order. The original tropospheric delay sequence can be decomposed into:

[0039] ZTD i =a8 i +(d8 i +d7 i +d6 i +d5 i +d4 i +d3 i +d2 i +d1 i )

[0040] Among them, a8 i is the low-frequency layer, and d1, d2, …, d8 are the high-frequency layers.

[0041] Then, the trend term of the tropospheric delay time series (a8 i ), subtracted from the original tropospheric delay sequence, to obtain the residual term ZTD i -a8 i . Use the threshold method to test the residual term. If the residual does not meet

[0042] |ZTDi -a8 i | <threshold

[0043] , then determine the corresponding ZTD i is a discrete gross error, where threshold is a given threshold.

[0044] Finally, the noise terms of the tropospheric delay time series (layers d1, d2, …, d8) are extracted. According to the rules of signal spectrum decomposition, the low-frequency terms (layers d4–d8) still contain some tropospheric delay series trend terms, but contain fewer noise terms and cannot be used for tropospheric delay anomaly detection. The high-frequency terms (layer d1) contain too low a level of noise and cannot be used for tropospheric delay anomaly detection. Therefore, layers d2 and d3 are selected for tropospheric delay series anomaly detection.

[0045] In the d2 and d3 layers, the 3σ method is combined to detect abnormal values of the tropospheric delay series. If d2 i or d3 i Not satisfied with:

[0046]

[0047] Then determine the corresponding ZTD i It is a discrete gross error.

[0048] 5. Completion of missing data of tropospheric delay series

[0049] First, the results of step 4 (after removing isolated and discrete outliers) are used as the training set for deep neural network (DNN) modeling. The input parameters are time information (Year, DoY, HoD), and the output parameter is the tropospheric delay (ZTD). The deep neural network model is as follows:

[0050] ZTD=f DNN (Year,DoY,HoD)

[0051] Secondly, the deep neural network model is trained by setting parameters such as the number of layers, number of nodes, learning rate, loss function, etc. The model is trained using the training set data, and finally a trained tropospheric delay deep neural network model is obtained.

[0052] Next, use the training residual Calculate the noise error σ DNN :

[0053]

[0054] Where ZTD i is the tropospheric delay sequence obtained in step 4 after removing isolated and discrete outliers, The tropospheric delay series obtained for deep neural network modeling is

[0055] Then, using the trained deep neural network model, the relevant parameters of the time to be completed are input to obtain the tropospheric delay of the missing data segment.

[0056] Finally, the predicted missing tropospheric delay is added with the noise ε DNN , we get the tropospheric delay completion result. The formula is as follows:

[0057]

[0058] Where, is the tropospheric delay sequence obtained in step 5, ε DNN The mean is 0 and the variance is σ DNN of white noise.

[0059] Steps 1 to 5 complete the tropospheric delay time series for one CORS station. Repeat steps 1 to 5 for other CORS stations to complete the continuous and reliable tropospheric delay large dataset for the entire CORS station network.

[0060] In some embodiments, a computer is further provided. The computer includes a processor and a memory. The memory is configured to store non-transitory computer-readable instructions (e.g., one or more computer program modules). The processor is configured to execute the non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by the processor, one or more steps of the method for producing a large tropospheric delay dataset based on a Beidou continuously operating reference station are performed. The memory and the processor may be interconnected via a bus system and / or other forms of connection mechanisms.

[0061] For example, the processor may be a central processing unit (CPU), a graphics processing unit (GPU), or other processing units with data processing capabilities and / or program execution capabilities. For example, the central processing unit (CPU) may be an X86 or ARM architecture. The processor may be a general-purpose processor or a special-purpose processor, and may control other components in the computer to perform desired functions.

[0062] For example, the memory may include any combination of one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, an erasable programmable read-only memory (EPROM), a compact disc read-only memory (CD-ROM), a USB memory, a flash memory, etc. One or more computer program modules may be stored on the computer-readable storage medium, and the processor may execute the one or more computer program modules to implement various functions of the computer.

[0063] In some embodiments, a computer-readable storage medium is further provided, which is used to store non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, one or more steps in the above-mentioned method for producing a large tropospheric delay data set based on the Beidou continuously operating reference station can be implemented. That is, when the method for producing a large tropospheric delay data set based on the Beidou continuously operating reference station provided in the embodiment of the present application is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium. For relevant instructions on the storage medium, please refer to the corresponding description of the memory in the computer above, which will not be repeated here.

Claims

1. A method for producing a large tropospheric delay dataset based on BeiDou continuously operating reference stations, characterized in that: include: Step 1: Using the original observations of the BeiDou CORS station, the total zenith tropospheric delay sequence of the CORS station is obtained according to the precise point positioning mode. Step 2: Using the total zenith tropospheric delay sequence of the CORS station, we set a missing day threshold to determine whether there is a continuous missing in the tropospheric delay sequence and perform data segmentation. Step 3: Use the tropospheric delay sequence after segmentation of data segments and use the first-order difference method and 3 Method to detect isolated tropospheric delay anomalies, first-order difference method and 3 Methods Methods for detecting isolated tropospheric delay outliers include: First-order difference method: tropospheric delay series Perform a difference to obtain the first-order difference sequence: ; 3 Method: Use Bessel's formula to calculate the standard deviation of the first-order difference sequence , select the confidence level as 99.7%, the corresponding critical value is 3, if the difference Dissatisfied , then determine the corresponding It is an isolated gross error; Step 4: Using the sequence that removes isolated tropospheric delay outliers, the wavelet decomposition, threshold method and 3 When using the wavelet decomposition method to detect discrete tropospheric delay anomalies, the db8 wavelet basis function and 8-layer decomposition order are selected, and the tropospheric delay sequence after removing isolated outliers in step 3 is decomposed into: in, is the low-frequency layer, , ,…, is the high frequency layer; Extracting the Noise Term of Tropospheric Delay Time Series Using Wavelet Decomposition , ,…, Layer, select and The layer is used for detecting anomalies of tropospheric delay series. and Layer, combined The method detects abnormal values of the tropospheric delay series. or Not satisfied with: Then determine the corresponding is a discrete gross error, where N is the number of tropospheric delay sequences; Step 5: Using the tropospheric delay series from which isolated and discrete outliers have been removed, deep neural network methods are used for modeling and prediction to complete the missing data of the tropospheric delay series and create a large tropospheric delay dataset.

2. The method according to claim 1, characterized in that The deep neural network model is: Where, the input parameter is time information: , which are year, accumulated days per year, and number of hours in a day, respectively.

3. The method according to claim 2, characterized in that Methods to fill in missing data of tropospheric delay series include: The tropospheric delay sequence obtained in step 4 after eliminating isolated and discrete outliers is , the tropospheric delay sequence obtained by deep neural network modeling in step 5 is , calculate the training residual sequence at the corresponding moment , using the training residual sequence , calculate the error in noise : Where, n is the number of training samples; Then, use the trained deep neural network model and input the time parameters that need to be completed: 、 、 , get the missing tropospheric delay, add the noise to the predicted missing tropospheric delay , that is, the tropospheric delay completion result is obtained; Where, is the tropospheric delay sequence obtained in step 5, The mean is 0 and the variance is of white noise.

4. A computer, characterized in that: include: processor; a memory comprising one or more computer program modules; The one or more computer program modules are stored in the memory and configured to be executed by the processor, and the one or more computer program modules include instructions for implementing the method for producing a large tropospheric delay data set based on the Beidou continuously operating reference station as described in any one of claims 1 to 3.

5. A computer-readable storage medium for storing non-transitory computer-readable instructions, characterized in that: When the non-transitory computer-readable instructions are executed by a computer, the method for producing a large tropospheric delay data set based on the Beidou continuously operating reference station as described in any one of claims 1 to 3 can be implemented.

Images