Rainfall forecasting method and device based on Beidou / GNSS water vapor

Abstract

The invention relates to the technical field of meteorological monitoring and forecasting, and discloses a Beidou / GNSS water vapor-based rainfall forecasting method and device, and the method comprises the steps: S1, laying a Beidou / GNSS monitoring station and a meteorological station; s2, receiving observation data from a Beidou / GNSS monitoring station, and synchronously collecting meteorological data of a meteorological station; the observation data comprises satellite observation values of pseudo range, carrier phase, Doppler frequency shift and signal intensity of a Beidou / GNSS satellite and ephemeris data, and the meteorological data comprises temperature, air pressure, humidity, wind speed, wind direction and precipitation; s3, calculating observation data by using a precise point positioning algorithm, and calculating troposphere atmospheric delay and the like; according to the method and the device, high-precision and real-time rainfall forecasting is realized based on Beidou / GNSS water vapor data and an artificial intelligence analysis model, and the timeliness and the accuracy of early warning are remarkably improved. By monitoring the dynamic change of the atmospheric precipitable water amount (PWV) in real time, the water vapor accumulation trend can be sensed in advance, and efficient early warning is provided for a rainfall event.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological monitoring and forecasting, and in particular to a rainfall forecasting method and device based on Beidou / GNSS water vapor. Background Art

[0002] In recent years, global warming has increased the frequency of extreme weather events, especially sudden heavy rainfall events. This has not only caused economic losses in the fields of transportation and agriculture, but also posed a serious threat to the safety of infrastructure. In areas with complex geographical environments such as mountains and rivers, heavy rainfall may trigger secondary disasters such as landslides, mud-rock flows, and river breaches, posing huge risks to residents' safety, infrastructure, and the ecological environment.

[0003] In recent years, the water vapor inversion technology of the global navigation satellite system has shown great application potential in the field of meteorological monitoring. This technology uses the delay information of GNSS signals when they pass through the atmospheric troposphere to invert the water vapor content in the atmosphere, thereby providing key forecast data for the occurrence of rainfall. GNSS water vapor inversion technology has the characteristics of high temporal and spatial resolution, all-weather operation, and relatively low deployment cost, and can achieve large-scale monitoring on the geographical coverage. Since the completion of the global networking of my country's Beidou-3 system, the GNSS water vapor monitoring technology with Beidou as the core has significant advantages in autonomous control, real-time and high precision, providing more reliable data support for meteorological disaster prevention.

[0004] Nevertheless, the current GNSS water vapor inversion method and existing rainfall forecast model still have shortcomings in practical applications, which are specifically manifested in the following aspects:

[0005] 1. Insufficient real-time performance: The existing Beidou / GNSS water vapor inversion system has delays in data collection, transmission and processing, which makes it difficult to meet the needs of real-time forecasting of rainfall events;

[0006] 2. Traditional models do not consider BeiDou / GNSS water vapor: Traditional models rely on statistical analysis methods and do not consider BeiDou water vapor data with high temporal and spatial resolution, resulting in low accuracy in responding to sudden rainfall;

[0007] 3. Lack of intelligent analysis: The data-driven artificial neural network has not been fully integrated, making it difficult to improve the accuracy of existing rainfall forecast models. Summary of the invention

[0008] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a rainfall forecasting method and device based on Beidou / GNSS water vapor. The method and device combine real-time water vapor monitoring and artificial intelligence analysis to efficiently capture the trend of water vapor changes, thereby achieving accurate early warning of rainfall events.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A rainfall forecasting method based on Beidou / GNSS water vapor comprises: S1, deploying Beidou / GNSS monitoring stations and meteorological stations; S2, receiving observation data from the Beidou / GNSS monitoring stations, and synchronously collecting meteorological data from the meteorological stations; the observation data comprises satellite observation values ​​of pseudorange, carrier phase, Doppler frequency shift and signal strength of Beidou / GNSS satellites, and ephemeris data, and the meteorological data comprises temperature, air pressure, humidity, wind speed, wind direction and precipitation; S3, solving the observation data by using a precise single point positioning algorithm, calculating the tropospheric atmospheric delay, separating the tropospheric dry delay part from the total atmospheric delay, and obtaining the tropospheric wet delay; S4, converting the tropospheric wet delay into Atmospheric precipitable water, perform time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend; S5, fuse the atmospheric precipitable water with the meteorological data as the input data of the artificial neural network; S6, establish a three-dimensional water vapor field in the rainfall forecast monitoring area, and use the three-dimensional water vapor field parameters as the input data of the artificial neural network; S7, construct and train the artificial neural network, input the input data into the trained artificial neural network, and obtain the short-term rainfall forecast result; S8, if the forecast result is an imminent rainfall event, generate rainfall information and push the rainfall information, the rainfall information includes the expected time, intensity and geographical range of the rainfall; S9, test, optimize and maintain the system.

[0011] In the present invention, preferably, the S1 includes: S11, deploying Beidou / GNSS monitoring stations and meteorological stations at selected sites in the target area; S12, installing Beidou / GNSS receivers and meteorological sensors at the selected sites; S13, configuring the data transmission network of the Beidou / GNSS monitoring stations and meteorological stations.

[0012] In the present invention, preferably, the formula for converting the tropospheric wet delay into atmospheric precipitable water in S4 is:

[0013] PWV=Π·ZWD,

[0014] Wherein, PWV is atmospheric precipitable water, Π is the conversion coefficient, and ZWD is the tropospheric wet delay; the calculation formula of the conversion coefficient is:

[0015]

[0016] Among them, Π is the conversion coefficient, ρ w is the density of liquid water, R w is the gas constant of water vapor, k 2 and k 3 is a constant, T m is the weighted average temperature.

[0017] In the present invention, preferably, the S6 includes: S61, discretizing the rainfall forecast monitoring area into a cell grid in three-dimensional space; S62, calculating the oblique path wet delay using the tropospheric wet delay, the formula is:

[0018]

[0019] Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value; S63, the oblique path water vapor content is calculated using the oblique path wet delay to calculate the water vapor density. The calculation formula for the oblique path water vapor content is:

[0020] SWV=Π·SWD=Π·∫N w ds,

[0021] Where SWV is the water vapor content of the oblique path, SWD is the wet delay of the oblique path, Π is the conversion coefficient, and N w is the wet refractive index and s is the signal path.

[0022] In the present invention, preferably, the process of calculating the water vapor density in S63 is: by dividing the rainfall forecast monitoring area into three-dimensional space cell networks, establishing a tomographic equation according to the intercept length of the GNSS signal propagation path, eliminating all invalid tomographic cell networks in the rainfall forecast monitoring area to reconstruct the tomographic equation, solving the tomographic equation through Kalman filtering, solving the wet refractive index, and converting it into water vapor density through the conversion coefficient.

[0023] In the present invention, preferably, the artificial neural network described in S7 adopts a BPNN model.

[0024] In the present invention, preferably, the S9 includes: S91, testing the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion and the accuracy of rainfall warning; S92, according to the test results, improving the accuracy of the precise single-point positioning algorithm solution, adjusting the structural parameters of the artificial neural network model and optimizing the data fusion algorithm; S93, regularly maintaining Beidou / GNSS monitoring stations, meteorological stations and data transmission networks, and periodically adjusting the artificial neural network model.

[0025] A rainfall forecasting device based on Beidou / GNSS water vapor comprises: a monitoring and acquisition module, used for receiving observation data from Beidou / GNSS monitoring stations, and synchronously collecting meteorological data from meteorological stations, wherein the observation data comprises satellite observation values ​​and ephemeris data, and the meteorological data comprises temperature, air pressure, humidity, wind speed, wind direction and precipitation; a delay calculation module, used for solving the observation data by using a precise single point positioning algorithm, calculating the tropospheric atmospheric delay, separating the dry delay part from the total atmospheric delay, and obtaining the tropospheric wet delay; a conversion module, used for converting the tropospheric wet delay into atmospheric precipitable water, and performing time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend; and a fusion module, used for The atmospheric precipitable water content is integrated with the meteorological data as the input data of the artificial neural network; the three-dimensional module is used to establish the three-dimensional water vapor field in the rainfall forecast monitoring area, and the three-dimensional water vapor field parameters are used as the input data of the artificial neural network; the analysis module is used to build and train the artificial neural network, and the input data is input into the trained artificial neural network to obtain the short-term rainfall forecast result. The artificial neural network adopts the BPNN model; the generation and push module is used to generate and push rainfall information if the forecast result is an imminent rainfall event. The rainfall information includes the expected time, intensity and geographical range of rainfall; the operation and maintenance module is used to test, optimize and maintain the system.

[0026] The three-dimensional module includes: a division unit, which is used to discretize the rainfall forecast monitoring area into a cell network in three-dimensional space; and a slant path calculation unit, which is used to calculate the slant path wet delay using the tropospheric wet delay, and the formula is:

[0027]

[0028] Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value; the water vapor content calculation unit is used to calculate the oblique path water vapor content and water vapor density using the oblique path wet delay. The calculation formula of the oblique path water vapor content is:

[0029] SWV=Π·SWD=Π·∫N w ds,

[0030] Where SWV is the water vapor content of the oblique path, SWD is the wet delay of the oblique path, Π is the conversion coefficient, and N wis the wet refractive index, s is the signal path; the operation and maintenance module includes: a test unit, which is used to test the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion and the correctness of rainfall warning; an optimization unit, which is used to improve the accuracy of precise single-point positioning algorithm solution, adjust the structural parameters of the artificial neural network model and optimize the data fusion algorithm according to the test results; a delay unit, which is used to regularly maintain Beidou / GNSS monitoring stations, meteorological stations and data transmission networks, and periodically adjust the artificial neural network model.

[0031] A computer-readable storage medium comprises instructions. When the instructions are executed on a computer, the computer is enabled to execute the rainfall forecasting method based on Beidou / GNSS water vapor as described in any one of the above.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] Based on BeiDou / GNSS water vapor data and artificial intelligence analysis models, this method and device achieves high-precision, real-time rainfall forecasts, significantly improving the timeliness and accuracy of early warnings. By real-time monitoring of the dynamic changes in atmospheric precipitable water volume (PWV), it is possible to perceive the trend of water vapor accumulation in advance and provide efficient early warnings for rainfall events.

[0034] Compared with traditional rainfall forecasting methods, the present invention has the following three advantages:

[0035] 1. Strong real-time performance: The real-time collection of water vapor data by BeiDou CORS stations significantly shortens the warning lag time and ensures the timeliness of rainfall forecasts.

[0036] 2. High accuracy: By using multi-factor analysis models and deep learning technology, the warning accuracy of this method reaches more than 80%, effectively improving the accuracy of rainfall forecasts.

[0037] 3. Wide adaptability: This method is also applicable in complex terrain environments, providing support for meteorological disaster prevention along railways and in mountainous areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The present invention is a flowchart of a rainfall forecasting method based on BeiDou / GNSS water vapor according to an embodiment of the present invention.

[0039] Figure 2 This is a flow chart of S1 in a rainfall forecasting method based on Beidou / GNSS water vapor according to an embodiment of the present invention.

[0040] Figure 3 This is a flowchart of S6 in a rainfall forecasting method based on Beidou / GNSS water vapor according to an embodiment of the present invention.

[0041] Figure 4 Schematic diagram of the spatial distribution of GNSS signal lines and water vapor tomography areas in S6 in a rainfall forecasting method based on Beidou / GNSS water vapor according to an embodiment of the present invention.

[0042] Figure 5 This is a flowchart of S9 in a rainfall forecasting method based on Beidou / GNSS water vapor according to another embodiment of the present invention.

[0043] Figure 6 This is a schematic diagram of the artificial neural network model architecture of S7 in the rainfall forecasting method based on Beidou / GNSS water vapor in another embodiment of the present invention.

[0044] Figure 7 It is a structural schematic diagram of a rainfall forecasting device based on Beidou / GNSS water vapor according to another embodiment of the present invention.

[0045] Figure 8 It is a structural schematic diagram of a three-dimensional module in a rainfall forecasting device based on Beidou / GNSS water vapor according to another embodiment of the present invention.

[0046] Fig. 9 It is a structural schematic diagram of an operation and maintenance module in a rainfall forecasting device based on Beidou / GNSS water vapor according to another embodiment of the present invention.

[0047] In the attached figure: 1. Monitoring and acquisition module; 2. Delay calculation module; 3. Conversion module; 4. Fusion module; 5. Three-dimensional module; 501. Division unit; 502. Oblique path calculation unit; 503. Water vapor content calculation unit; 6. Analysis module; 7. Generation and push module; 8. Operation and maintenance module; 801. Test unit; 802. Optimization unit; 803. Delay unit. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a component centered. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a component centered. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a component centered. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0051] See also Figure 1 A preferred embodiment of the present invention provides a rainfall forecasting method based on Beidou / GNSS water vapor, comprising:

[0052] S1, deploy Beidou / GNSS monitoring stations and meteorological stations.

[0053] Specifically, Figure 2 As shown, S1 includes:

[0054] S11, select sites in the target area to deploy BeiDou / GNSS monitoring stations and meteorological stations.

[0055] Choose appropriate geographical locations in the target area to deploy BeiDou / GNSS monitoring stations and weather stations. When selecting a site, consider terrain features, climate conditions, and communication conditions to ensure that the station can continuously and stably collect data and has a good observation field of view.

[0056] S12, install BeiDou / GNSS receivers and meteorological sensors at selected sites.

[0057] High-precision Beidou / GNSS receivers are installed at selected monitoring sites to collect satellite observation data in real time. At the same time, meteorological sensors are installed to monitor meteorological factors such as temperature, air pressure, humidity and wind speed, providing multi-dimensional data support for rainfall forecasts.

[0058] S13, configure the data transmission network of Beidou / GNSS monitoring stations and meteorological stations.

[0059] After the BeiDou / GNSS monitoring station and weather station are installed, configure the data transmission network to ensure that the monitoring data is transmitted to the data center in real time through the NTRIP protocol.

[0060] S2, receives observation data from BeiDou / GNSS monitoring stations and simultaneously collects meteorological data from meteorological stations.

[0061] Receive observation data (binary messages in RTCM format) from BeiDou / GNSS monitoring stations in real time through the NTRIP protocol. Specifically, the observation data may include satellite observation values ​​of pseudorange, carrier phase, Doppler shift and signal strength of BeiDou / GNSS satellite systems, as well as ephemeris data. Synchronously collect data from meteorological stations. Specifically, meteorological data may include temperature, air pressure, humidity, wind speed, wind direction and precipitation parameters, and upload them to the cloud server in real time through the MODBUS protocol.

[0062] S3, using the precise single point positioning algorithm to solve the observation data, calculate the tropospheric atmospheric delay, separate the dry delay part from the total atmospheric delay, and obtain the tropospheric wet delay.

[0063] The precise point positioning algorithm (PPP) first uses the data provided by IGS (The International GNSS Service) to calculate the satellite orbit parameters and clock errors, and then uses the phase and pseudo-range observations of a single point for non-differential positioning processing.

[0064] PPP technology is the core technology used to obtain high-precision positioning data in weather forecasting. Error correction is performed through SSR's real-time precise ephemeris and clock recovery technology. The observation equation is:

[0065] P i =ρ+c·(dt-dT)+I i +T+ε p ...... (1),

[0066] Φ i =ρ+c·(dt-dT)-I i +T+λ·N+ε φ ......(2),

[0067] Among them, P i is the pseudorange observation value, Φ i is the carrier phase observation value, ρ is the geometric distance, c is the speed of light in vacuum, dt and dT are the clock differences between the receiver and the satellite respectively, I i Ionospheric delay, T is tropospheric delay, λ is carrier wavelength, N is integer ambiguity, ε p and ε φ are the observation noise of pseudorange and carrier phase respectively.

[0068] The tropospheric delay T is used as the parameter to be estimated in the observation equation and is recursively solved using the Kalman filter method to obtain the tropospheric atmospheric delay result required for subsequent calculations.

[0069] The PPP algorithm can be used to calculate the total atmospheric delay ZTD at the epoch of the observation station; the tropospheric dry delay ZHD is also called the static delay, and its variation characteristics are relatively stable. It can be calculated using empirical models such as Saastamoinen based on the station location, water vapor pressure, absolute temperature, surface atmospheric pressure and other parameters. The calculation results of various empirical models are relatively different; while the tropospheric wet delay ZWD represents the satellite signal propagation delay caused by the water vapor content in the atmosphere. Its variation characteristics are relatively active and difficult to estimate with empirical models. Relatively accurate ZWD parameters can be obtained by solving the total atmospheric delay and deducting the dry delay part. That is:

[0070] ZWD=ZTD-ZHD......(3),

[0071] S4, convert the tropospheric wet delay into atmospheric precipitable water, and perform time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend.

[0072] The tropospheric wet delay is converted into atmospheric precipitable water vapor (PWV) through the conversion model. The formula is:

[0073] PWV=Π·ZWD......(4),

[0074] Among them, Π is the conversion coefficient, which depends on the temperature and atmospheric pressure and can be approximately calculated as follows:

[0075]

[0076] Among them, ρ w is the density of liquid water, R w is the gas constant of water vapor, k 2 and k 3 is a constant, T m is the weighted average temperature.

[0077] The converted PWV values ​​are analyzed in time series to detect the trend of water vapor accumulation. The real-time PWV increment and PWV growth rate are updated with a 72-hour time window to analyze the trend of water vapor accumulation. The dynamic change of PWV is a precursor to rainfall, which can effectively reflect the accumulation of water vapor in the atmosphere and provide a basis for rainfall forecast.

[0078] S5, the atmospheric precipitable water is fused with meteorological data as input data of artificial neural network.

[0079] PWV was integrated with meteorological data (temperature, air pressure, humidity, wind speed, wind direction and rainfall), and the data were standardized to ensure that different factors had relatively consistent dimensions and proportions as input data for the artificial neural network.

[0080] S6, establishing a three-dimensional water vapor field in the rainfall forecast monitoring area, and using the three-dimensional water vapor field parameters as input data of the artificial neural network.

[0081] Specifically, Figure 3 As shown, S6 includes:

[0082] S61, discretize the rainfall forecast monitoring area into a cell grid in three-dimensional space.

[0083] like Figure 4 As shown in the figure, the rainfall forecast monitoring area is discretized into a cell grid in three-dimensional space. In the horizontal direction, it is evenly divided according to the longitude and latitude of the survey area. In the vertical direction, the height from the ground to the top of the troposphere (set to 11km) is divided into 12 height intervals using a non-uniform division method based on the characteristics of water vapor content changing with height.

[0084] S62, calculating the slant path wet delay using the tropospheric wet delay.

[0085] The tropospheric wet delay ZWD calculated by multiple Beidou / GNSS monitoring stations in the rainfall forecast monitoring area is projected onto the oblique path direction of the GNSS signal line through a mapping projection function to obtain the oblique path wet delay SWD. The formula is:

[0086] Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value.

[0087] S63, using the oblique path wet delay to calculate the oblique path water vapor content, and calculate the water vapor density.

[0088] The slant path water vapor content SWV refers to the height of liquid water per unit area converted from the water vapor contained in the water vapor column in the propagation direction of the GNSS signal. The calculation formula for the slant path water vapor content is:

[0089] SWV=Π·SWD=Π·∫N w ds......(7),

[0090] Π represents the conversion coefficient, which can be calculated by formula (8), N wRepresents wet refractive index; s represents the signal path. The process of calculating water vapor density is as follows: through the three-dimensional space cell grid divided by the monitoring area, the tomographic equation is established according to the intercept length of the GNSS signal propagation path, all invalid tomographic cells in the monitoring area are eliminated to reconstruct the tomographic equation, and the tomographic equation is solved by Kalman filtering. The wet refractive index is calculated and converted into water vapor density through the conversion coefficient to complete the establishment of the three-dimensional water vapor field. The three-dimensional water vapor field parameters (latitude and longitude, elevation grid data) are used as the input data of the artificial neural network.

[0091] S7, constructing and training an artificial neural network, inputting the input data into the trained artificial neural network to obtain a short-term rainfall forecast result.

[0092] The input data includes PWV parameters, meteorological data (meteorological factors), and three-dimensional water vapor field parameters (latitude and longitude, elevation grid data). The standardized PWV parameters, meteorological data, and three-dimensional water vapor field parameters (latitude and longitude, elevation grid data) are substituted into the artificial neural network with geographic space parameters and time domain parameters as input layer training data. Preferably, this embodiment adopts a BPNN model (BP neural network), and the BPNN model is used as an example for explanation below. The CV cross-validation method is used to confirm the optimal number of hidden layer nodes based on the training data. The nonlinear mapping, parallel processing, and adaptive learning characteristics of the neural network are used to compare the training results with the actual recorded rainfall conditions, calculate the model rainfall forecast hit rate, adjust the data weights of PWV, meteorological data, and three-dimensional water vapor field parameters, and iteratively learn to improve the event response relationship. A three-month time span is used as the model calculation update cycle, and the accuracy of model training is improved by updating new data. The artificial neural network model architecture is as follows: Figure 6 shown.

[0093] The real-time calculated ZWD result of the Beidou / GNSS monitoring station is converted into PWV by formula (4), and converted into SWD by formula (6). Then, a coefficient matrix is ​​established according to the three-dimensional space cell network by formula (7), and the wet refractive index is estimated by Kalman filtering iteration, which is converted into the water vapor density of the cell network through the conversion coefficient.

[0094] The PWV results, synchronously collected meteorological station data, and three-dimensional water vapor field data are substituted into the trained BPNN model as the input layer to generate the output layer results of short-term rainfall forecast in the plane grid area.

[0095] S8: If the forecast result is that a rainfall event is about to occur, rainfall information is generated and pushed.

[0096] If the artificial neural network predicts that a rainfall event is about to occur, rainfall information is generated. The rainfall information (warning information) can include the expected time, intensity and geographical range of rainfall for reference by relevant departments and to take corresponding measures. The rainfall information is pushed to the emergency response department and other relevant units in real time through the network platform to ensure the timeliness of the warning transmission.

[0097] S9, test, optimize and maintain the system.

[0098] Specifically, Figure 5 As shown, S9 may include:

[0099] S91 tests the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion, and the accuracy of rainfall warning.

[0100] After the system was deployed, a series of tests were conducted to verify its accuracy and stability. The tests included the stability of data transmission, the accuracy of delay calculation, the accuracy of PWV conversion, and the accuracy of rainfall warning. The system's response speed and prediction effect under sudden weather conditions were verified by simulating rainfall events.

[0101] S92, based on the test results, improve the accuracy of the precise single-point positioning algorithm, adjust the structural parameters of the artificial neural network model, and optimize the data fusion algorithm.

[0102] According to the test results, the performance of each module of the system is continuously optimized, including improving the accuracy of PPP solution, adjusting the structural parameters of artificial neural network models (such as BPNN models), and optimizing data fusion algorithms to improve the prediction accuracy and response speed of the overall system.

[0103] S83, regularly maintain BeiDou / GNSS monitoring stations, meteorological stations and data transmission networks, and periodically adjust artificial neural network models (such as BPNN models).

[0104] After the system is put into use, the monitoring equipment and data transmission network are regularly maintained to ensure the normal operation of the equipment and the continuous transmission of data. In addition, according to climate change and the environmental characteristics of the monitoring area, the artificial neural network model (such as the BPNN model) is periodically adjusted to ensure that it always has the prediction ability to adapt to different seasons and meteorological conditions.

[0105] like Figure 7 As shown, an embodiment of the present invention further provides a rainfall forecasting device based on Beidou / GNSS water vapor, comprising:

[0106] The monitoring and collection module 1 is used to receive observation data from the Beidou / GNSS monitoring station and synchronously collect meteorological data from the meteorological station. The observation data includes satellite observation values ​​and ephemeris data. The meteorological data includes temperature, air pressure, humidity, wind speed, wind direction and precipitation.

[0107] The delay calculation module 2 is used to solve the observation data using the precise single point positioning algorithm, calculate the tropospheric atmospheric delay, separate the dry delay part from the total atmospheric delay, and obtain the tropospheric wet delay.

[0108] The conversion module 3 is used to convert the tropospheric wet delay into atmospheric precipitable water, and perform time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend.

[0109] The fusion module 4 is used to fuse the atmospheric precipitable water with the meteorological data as input data of the artificial neural network.

[0110] The three-dimensional module 5 is used to establish the three-dimensional water vapor field in the rainfall forecast monitoring area, and use the three-dimensional water vapor field parameters as the input data of the artificial neural network. Figure 8 As shown, the three-dimensional module 5 includes:

[0111] The division unit 501 is used to discretize the rainfall forecast monitoring area into cell grids in three-dimensional space.

[0112] The slant path calculation unit 502 is used to calculate the slant path wet delay using the tropospheric wet delay, and the formula is:

[0113]

[0114] Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value.

[0115] The water vapor content calculation unit 503 is used to calculate the water vapor content of the oblique path and the water vapor density by using the oblique path wet delay. The calculation formula of the oblique path water vapor content is:

[0116] SWV=Π·SWD=Π·∫N w ds,

[0117] Where SWV is the water vapor content of the oblique path, SWD is the wet delay of the oblique path, Π is the conversion coefficient, and N w is the wet refractive index and s is the signal path.

[0118] The analysis module 6 is used to construct and train an artificial neural network, input the input data into the trained artificial neural network, and obtain the short-term rainfall forecast result. The artificial neural network adopts a BPNN model.

[0119] The generating and pushing module 7 is used to generate and push rainfall information if the forecast result is that a rainfall event is about to occur, and the rainfall information includes the time, intensity and geographical range of the expected rainfall.

[0120] The operation and maintenance module 8 is used to test, optimize and maintain the system. Fig. 9 As shown, the operation and maintenance module 8 includes: a test unit 801, which is used to test the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion and the accuracy of rainfall warning; an optimization unit 802, which is used to improve the accuracy of precise single-point positioning algorithm solution, adjust the structural parameters of the artificial neural network model and optimize the data fusion algorithm according to the test results; a delay unit 803, which is used to regularly maintain Beidou / GNSS monitoring stations, meteorological stations and data transmission networks, and periodically adjust the artificial neural network model.

[0121] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned Beidou / GNSS water vapor-based rainfall forecasting method embodiment is implemented, and the same technical effect can be achieved. Among them, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0122] The above description is a detailed description of the preferred feasible embodiments of the present invention, but the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modified changes completed under the technical spirit suggested by the present invention should fall within the patent scope covered by the present invention.

Claims

1. A rainfall forecasting method based on Beidou / GNSS water vapor, characterized in that: include: S1, deploy Beidou / GNSS monitoring stations and meteorological stations; S2, receives observation data from BeiDou / GNSS monitoring stations and simultaneously collects meteorological data from weather stations; The observation data include satellite observation values ​​of pseudorange, carrier phase, Doppler shift and signal strength of Beidou / GNSS satellites, and ephemeris data, and the meteorological data include temperature, air pressure, humidity, wind speed, wind direction and precipitation; S3, using the precise point positioning algorithm to solve the observation data, calculate the tropospheric atmospheric delay, separate the tropospheric dry delay from the total atmospheric delay, and obtain the tropospheric wet delay; S4, converting the tropospheric wet delay into atmospheric precipitable water, and performing time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend; S5, fusion of atmospheric precipitable water and meteorological data as input data of artificial neural network; S6, establishing a three-dimensional water vapor field in the rainfall forecast monitoring area, and using the three-dimensional water vapor field parameters as input data of the artificial neural network; S7, constructing and training an artificial neural network, inputting input data into the trained artificial neural network, and obtaining a short-term rainfall forecast result; S8, if the forecast result is that a rainfall event is about to occur, rainfall information is generated and pushed, the rainfall information including the expected time, intensity and geographical range of the rainfall; S9, test, optimize and maintain the system.

2. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 1, characterized in that: The S1 includes: S11, select sites in the target area to deploy BeiDou / GNSS monitoring stations and meteorological stations; S12, install BeiDou / GNSS receivers and meteorological sensors at selected sites; S13, configure the data transmission network of Beidou / GNSS monitoring stations and meteorological stations.

3. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 1, characterized in that: The formula for converting the tropospheric wet delay to atmospheric precipitable water as described in S4 is: PWV=Π·ZWD, Where PWV is atmospheric precipitable water, Π is the conversion factor, and ZWD is the tropospheric wet delay; The calculation formula of the conversion coefficient is: Among them, Π is the conversion coefficient, ρ w is the density of liquid water, R w is the gas constant of water vapor, k2 and k3 are constants, T m is the weighted average temperature.

4. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 1, characterized in that: The S6 includes: S61, discretizing the rainfall forecast monitoring area into a cell grid in three-dimensional space; S62, calculate the slant path wet delay using the tropospheric wet delay, the formula is: Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value; S63, using the oblique path wet delay to calculate the oblique path water vapor content, calculate the water vapor density, the oblique path water vapor content calculation formula is: SWV=Π·SWD=Π·∫N w ds, Where SWV is the water vapor content of the oblique path, SWD is the wet delay of the oblique path, Π is the conversion coefficient, and N w is the wet refractive index and s is the signal path.

5. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 4 is characterized in that: The process of calculating water vapor density described in S63 is: by dividing the rainfall forecast monitoring area into three-dimensional space cell grids, establishing a tomographic equation according to the intercept length of the GNSS signal propagation path, eliminating all invalid tomographic cell grids in the rainfall forecast monitoring area to reconstruct the tomographic equation, solving the tomographic equation through Kalman filtering, solving the wet refractive index, and converting it into water vapor density through the conversion coefficient.

6. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 1, characterized in that: The artificial neural network described in S7 adopts the BPNN model.

7. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 1, characterized in that: The S9 includes: S91, tests the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion, and the accuracy of rainfall warning; S92, based on the test results, improve the accuracy of the precise point positioning algorithm, adjust the structural parameters of the artificial neural network model, and optimize the data fusion algorithm; S93, regularly maintain BeiDou / GNSS monitoring stations, meteorological stations and data transmission networks, and periodically adjust the artificial neural network model.

8. A rainfall forecasting device based on Beidou / GNSS water vapor, characterized in that: include: The monitoring and collection module is used to receive observation data from the Beidou / GNSS monitoring station and simultaneously collect meteorological data from the meteorological station. The observation data includes satellite observation values ​​and ephemeris data. The meteorological data includes temperature, air pressure, humidity, wind speed, wind direction and precipitation. The delay calculation module is used to solve the observation data using the precise point positioning algorithm, calculate the tropospheric atmospheric delay, separate the dry delay part from the total atmospheric delay, and obtain the tropospheric wet delay; A conversion module is used to convert the tropospheric wet delay into atmospheric precipitable water, and perform time series analysis on the atmospheric precipitable water to detect the water vapor accumulation trend; A fusion module is used to fuse atmospheric precipitable water with meteorological data as input data for artificial neural networks; The three-dimensional module is used to establish the three-dimensional water vapor field in the rainfall forecast monitoring area and use the three-dimensional water vapor field parameters as the input data of the artificial neural network; The analysis module is used to construct and train an artificial neural network, input the input data into the trained artificial neural network, and obtain the short-term rainfall forecast result, wherein the artificial neural network adopts a BPNN model; A generation and push module, for generating and pushing rainfall information if the forecast result is an upcoming rainfall event, wherein the rainfall information includes the time, intensity and geographical range of the expected rainfall; The operation and maintenance module is used to test, optimize and maintain the system.

9. The rainfall forecasting method based on Beidou / GNSS water vapor according to claim 8, characterized in that: The three-dimensional module includes: Division unit is used to discretize the rainfall forecast monitoring area into a cell network in three-dimensional space; The slant path calculation unit is used to calculate the slant path wet delay using the tropospheric wet delay. The formula is: Where SWD is the slant path wet delay, mf w is the wet mapping function, ZWD is the tropospheric wet delay, mf g is the gradient mapping function, e is the altitude angle of the GNSS signal, α is the azimuth information of the GNSS signal, is the horizontal moisture gradient term in the north-south direction, is the wet horizontal gradient term in the east-west direction, and ε is the unmodeled residual value; The water vapor content calculation unit is used to calculate the water vapor content of the oblique path and the water vapor density by using the oblique path wet delay. The calculation formula of the oblique path water vapor content is: SWV=Π·SWD=Π·∫N w ds, Where SWV is the water vapor content of the oblique path, SWD is the wet delay of the oblique path, Π is the conversion coefficient, and N w is the wet refractive index, s is the signal path; The operation and maintenance module includes: The test unit is used to test the stability of data transmission, the accuracy of delay calculation, the accuracy of atmospheric precipitable water conversion, and the accuracy of rainfall warning; The optimization unit is used to improve the accuracy of the precise point positioning algorithm, adjust the structural parameters of the artificial neural network model, and optimize the data fusion algorithm according to the test results; The delay unit is used to regularly maintain Beidou / GNSS monitoring stations, weather stations and data transmission networks, and to periodically adjust the artificial neural network model.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes instructions, and when the instructions are executed on a computer, the computer is enabled to execute the rainfall forecasting method based on Beidou / GNSS water vapor as described in any one of claims 1 to 7.

Images