Wind field data correction method of spaceborne lidar based on ground-based lidar

Abstract

The present invention relates to a satellite-borne laser radar wind field data correction method based on a ground-based laser wind measuring radar, which belongs to the field of radio, and includes: calculating horizontal wind speed components U1 and V1, as well as horizontal wind speed components U2 and V2 at different altitudes according to collected data; determining an optimal nonlinear regression correction model; generating U and V wind fields with high spatial resolution; constructing a power spectrum correction model to generate power spectrum signals of high spatial resolution grids U and V after correction; generating a satellite-borne laser wind measuring radar antenna direction weighting function to obtain a weight coefficient corresponding to each grid point in the high spatial resolution grid; constructing a satellite-borne laser radar wind field data correction model, and finally obtaining corrected satellite-borne laser wind measuring radar wind field data. The present invention realizes the accurate simulation of the power spectrum of U and V components of satellite-borne laser wind measuring radar with low spatial resolution, and realizes the accurate and efficient correction of U and V components of satellite-borne laser wind measuring radar.

Description

Technical Field

[0001] The invention relates to the field of radio, and in particular to a method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind measuring radar. Background Art

[0002] The atmospheric wind field is a core element of atmospheric dynamics. It directly affects processes such as energy distribution, humidity transfer and cloud formation in the atmosphere, which in turn determines the evolution of weather systems and the occurrence of weather phenomena. Secondly, wind field measurement is crucial in climate research, which helps to understand the mechanism of global climate change and provide a basis for climate prediction. In addition, wind field measurement also has important applications in aviation and navigation. Accurate wind field data can help optimize route selection, reduce fuel consumption, and ensure the safety of flight and navigation. In summary, atmospheric wind field measurement is not only a basic tool for weather forecasting and climate research, but also plays an important role in aviation, navigation, environmental protection and other fields, and has a far-reaching impact on ensuring social security and promoting sustainable development.

[0003] At present, the measurement of atmospheric wind field mainly includes the following methods: (1) ground meteorological station measurement, which can accurately obtain the atmospheric wind field near the ground. However, its spatial coverage is limited, mainly concentrated on land, with less data in oceans and remote areas, and can only measure the wind field near the surface, and cannot provide high-altitude wind speed and direction information; (2) sounding balloon measurement, which can provide wind field data at multiple heights in the vertical direction, but the measurement period and spatial coverage are limited, and the operation is complicated and costly, making it unsuitable for large-scale long-term monitoring; (3) ground-based Doppler radar, ground-based lidar and airborne radar remote sensing inversion, which can provide wind fields at different heights in a large range, but the detection capability for mountainous areas or offshore areas is very limited, and it cannot provide global long-term wind field information; (4) satellite-borne laser wind radar, which can realize global wind field monitoring, but its data accuracy is not high; (5) numerical weather rapid update forecast model inversion, which can provide global wind field data at different heights for continuous time, but the simulated data accuracy is often not high, especially in areas with complex terrain.

[0004] Based on the above analysis, it is difficult to obtain long-term, large-scale and accurate wind field information by relying solely on a single wind field information detection method. In addition, the correction of conventional satellite-borne laser radar wind field data often starts from the data end, without considering the impact of satellite platforms, environmental factors on echo signals, and the reliability of other reference wind field data. Therefore, the reliability and universality of this correction effect are often not high. Therefore, how to start from the perspective of satellite-borne laser wind radar power spectrum signal generation, comprehensively utilize ground-based and satellite-borne multi-source wind field observation data, and carry out the correction of satellite-borne laser wind radar power spectrum signal and wind field data is a problem that needs to be considered at present. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar, thereby solving the deficiencies of the prior art.

[0006] The object of the present invention is achieved by the following technical solution: a satellite-borne laser radar wind field data correction method based on a ground-based laser wind radar, the correction method comprising:

[0007] Step 1: Calculate the horizontal wind speed components U1 and V1, and the horizontal wind speed components U2 and V2 at different altitudes based on the collected data;

[0008] Step 2: Use the obtained U1 and V1 as dependent variable data sets, and U2 and V2 as independent variable data sets to establish a nonlinear regression correction model, and construct an evaluation index RUV based on the calculated correlation coefficients RU and RV, and determine the best nonlinear regression correction model based on the evaluation index;

[0009] Step 3: Correct the U2 and V2 components through the optimal nonlinear regression correction model, and establish the power spectrum of horizontal wind speed components at different altitudes based on the correction. Start the weather rapid update forecast mode according to the collected data, and set the network model of the CFD calculation area to generate U and V wind fields with high spatial resolution.

[0010] Step 4: Input the generated high spatial resolution U and V wind fields and the collected data into the lidar simulator to obtain data set 1, use the power spectra of U and V components at different altitudes as data set 2, form data set 3 based on data set 1 and data set 2, construct a power spectrum correction model based on data set 2 and data set 3, and generate the corrected high spatial resolution grid U and V power spectrum signals;

[0011] Step 5: Calculate the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, simulate the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna, generate the direction weighting function of the satellite-borne laser wind radar antenna, and finally obtain the weight coefficient corresponding to each grid point in the high spatial resolution grid;

[0012] Step six, generate the power spectrum signals of the U and V components of the low spatial resolution spaceborne laser wind radar according to the obtained weight coefficients and the power spectrum signals of the high spatial resolution grids U and V, and finally obtain the simulated U and V components of the low spatial resolution spaceborne laser wind radar as the input data set, and use the U and V components actually observed by the spaceborne laser wind radar as the output data set. Use the input data set and the output data set as the training set to construct a satellite-borne laser radar wind field data correction model, and finally obtain the corrected satellite-borne laser wind radar wind field data.

[0013] The step one comprises:

[0014] A1. Collect horizontal wind speed and direction data at different heights on ground-based wind towers, collect ground-based laser wind radar observation parameters and DBS scanning mode detection data, collect GFS data, global observation data and cloud parameter information data, collect space-borne laser wind radar observation parameters and inverted wind field data;

[0015] A2. According to the horizontal wind speed WS and wind direction WD data collected at different heights on the ground-based wind tower, WS and WD are converted from the polar coordinate system to the rectangular coordinate system, and U1 and V1 are calculated according to the decomposition formula of the horizontal wind speed components U1 and V1;

[0016] According to the radial velocity data collected in the DBS scanning mode detection data of the ground-based laser wind measuring radar, combined with the beam elevation angle information in the DBS scanning mode, the horizontal wind speed components U2 and V2 are projected onto the radial beams in the four directions of due east, due south, due west and due north, respectively, and a group of equations between the radial velocities in the four directions and U2 and V2 are established. The U2 and V2 components at different altitude layers are obtained by solving the group of equations.

[0017] The second step comprises:

[0018] B1, taking the obtained U1 and V1 components as the dependent variable data set, according to the altitude layer where the dependent variable data set is located, finding the laser wind radar observation data layer closest to these altitude layers, and extracting the U2 and V2 component data in these laser wind radar observation data layers as the independent variable data set;

[0019] B2. Input the dependent variable data set and the independent variable data set into the nonlinear regression correction model library together, establish multiple nonlinear regression correction models between U2 and V2 as independent variable data sets and corresponding U1 and V1 as dependent variable data sets, calculate the correlation coefficients RU and RV of U2 as independent variable and U1 as dependent variable, V2 as independent variable and V1 as dependent variable in each linear regression correction model, construct the evaluation index of joint correlation coefficient RUV, take the maximum amplitude of joint correlation coefficient as the evaluation standard, and determine the best nonlinear regression correction model.

[0020] The step 3 of correcting the U and V components by using the optimal nonlinear regression correction model and establishing the power spectra of the U and V components at different altitudes after correction includes:

[0021] C1. Based on the determined optimal nonlinear regression correction model, the horizontal wind speed components U2 and V2 at different height layers observed by the laser wind radar are input to correct the U2 and V2 components, and the corrected U2_c and V2_c components are obtained;

[0022] C2. According to the power spectrum signal of the DBS scanning mode of the collected ground-based laser wind radar, the least square error is set as the error function, and the power spectrum model is fitted using the nonlinear least square method;

[0023] C3. The corrected U2_c and V2_c components and the ground-based laser wind radar observation parameters are input as input data into the fitted power spectrum model to establish the power spectrum of the horizontal wind speed U2_c and V2_c components at different height layers based on the corrected data.

[0024] The step 3 of starting the weather rapid update forecast mode according to the collected data and setting the network model of the CFD calculation area to generate high spatial resolution U and V wind fields includes:

[0025] D1. Use the obtained GFS data, global real-time observation data and cloud parameter information data as the input of the weather rapid update forecast model RAP, set the corresponding parameters and schemes, and start RAP to generate wind speed and wind direction grid data of different altitude layers with medium spatial resolution M1;

[0026] D2. Based on the generated medium spatial resolution wind speed and wind direction grid data, generate the wind speed and wind direction of the CFD boundary wind field according to the CFD boundary field range, set the network model of the CFD calculation area, and set the output spatial resolution to a high level. Run the CFD mode to generate a high spatial resolution horizontal wind speed component wind field, represented by U and V.

[0027] The fourth step comprises:

[0028] E1. Input the generated high spatial resolution horizontal wind speed component U and V wind field, as well as the collected spaceborne laser wind radar observation parameters and GFS data into the laser radar simulator, and run the laser radar simulator to generate the U and V power spectrum signals of each grid point under the high spatial resolution grid as data set 1;

[0029] E2, taking the U and V component power spectra of different altitude layers after the ground-based laser wind radar correction as data set 2, calculating the distance between all grid points in data set 2 and data set 1, and selecting the data that matches the space of data set 2 from data set 1 to form data 3 based on the shortest distance as the selection criterion;

[0030] E3. Use dataset 2 as the output dataset and dataset 3 as the input dataset to build a power spectrum correction model based on random forest and generate the corrected power spectrum signals of high spatial resolution grids U and V.

[0031] The step 5 of calculating the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, simulating the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna, and generating the satellite-borne laser wind radar antenna direction weighting function includes:

[0032] F1. Based on the generated corrected high-resolution grid and combined with the low-spatial-resolution grid of the spaceborne laser wind radar, the number of high-spatial-resolution grids included in each low-spatial-resolution grid is counted;

[0033] F2. Calculate the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, input the observation parameters of the satellite-borne laser wind radar into the satellite-borne laser wind radar antenna simulator, and simulate the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna;

[0034] F3. According to the spatial distribution diagram of antenna radiation energy, the angle information of each point in the antenna side lobe deviating from the center point of the main lobe and the amplitude below the level of the center point of the main lobe are obtained, and these angle information and amplitude information are input into the Taylor weighting function to generate the direction weighting function of the space-borne laser wind radar antenna.

[0035] The weight coefficient corresponding to each grid point in the high spatial resolution grid finally obtained in step 5 includes:

[0036] G1. Based on the calculated distance △R, combined with the operating altitude of the satellite-borne laser wind radar platform and the longitude and latitude of the sub-satellite point, and taking into account the curvature of the earth, the zenith angle of each grid point in the high spatial resolution grid is calculated based on the space-to-ground geometric relationship between the generated corrected high spatial resolution grid and the satellite-borne laser wind radar platform;

[0037] G2. Input the zenith angle information into the established satellite-borne laser wind radar antenna direction weighting function to obtain the weight coefficient corresponding to each grid point in the high spatial resolution grid.

[0038] The step six comprises:

[0039] H1, multiply the weight coefficient corresponding to each grid point in the obtained high spatial resolution grid by the power spectrum signals of the U and V components on the generated high spatial resolution grid points, carry out weighted synthesis of U and V in the Doppler frequency domain on the low spatial resolution grid, and generate the power spectrum signals of the U and V components of the low spatial resolution spaceborne laser wind radar;

[0040] The echo power and noise coefficient received by the space-borne laser wind radar are added to the power spectrum signal to generate a power spectrum signal containing the U and V components of the real noise signal. The U and V components of the simulated low spatial resolution space-borne laser wind radar are obtained by using the Fourier transform spectrum estimation method.

[0041] H2, the obtained simulated low spatial resolution spaceborne laser wind radar U and V components are combined into a pair of data sets, and used as the input data set, and the actual observed U and V components of the spaceborne laser wind radar are combined into a pair of data sets, and used as the output data set;

[0042] The input data set and the output data set are used as training sets to construct a correction model for spaceborne lidar wind field data based on the adaptive boosting model. After the training is completed, the wind field data observed by the lidar is input again to obtain the corrected spaceborne lidar wind field data.

[0043] The present invention has the following advantages: a satellite-borne laser radar wind field data correction method based on a ground-based laser wind radar, based on wind speed and direction data of different height layers of a ground-based wind tower and radial velocity data of a DBS scanning mode of a ground-based laser wind radar, using the deviation of the two data, respectively constructing optimal nonlinear correction models for the deviation of U and V wind components, thereby realizing accurate correction of the U and V wind components of the ground-based laser wind radar; based on the power spectrum data of the ground-based laser wind radar, a nonlinear least squares method is used to construct a power spectrum model, and the corrected U and V wind components of the ground-based laser wind radar are input to realize correction of the U and V power spectrum of the ground-based laser wind radar; based on GFS, global real-time observation data and cloud parameter data, a RAP model and a CFD model are combined to simulate U and V wind fields with high spatial resolution; based on the high spatial resolution U and V wind field, taking into account the observation parameters of spaceborne laser radar, using advanced laser radar signal simulator, the simulation of high spatial resolution U and V power spectra is realized; using the corrected ground-based laser radar U and V power spectra as the true value, the random forest model is used to realize the correction of the simulated high spatial resolution U and V power spectra; based on the observation parameters of spaceborne laser wind measuring radar, using the spaceborne laser radar antenna simulator, the weight coefficient of each grid point in the high-resolution grid point can be accurately calculated; considering the noise of the real environment, the strategy of frequency domain weighting and power spectrum spatial synthesis is adopted to realize the accurate simulation of the power spectrum of U and V components of low spatial resolution spaceborne laser wind measuring radar, and based on the actual observation data of U and V of spaceborne laser wind measuring radar, the adaptive lifting model is adopted to realize the accurate and efficient correction of U and V components of spaceborne laser wind measuring radar. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the process of the present invention;

[0045] Figure 2 This is a flow chart for implementing the correction of U and V power spectra of the ground-based laser wind radar of the present invention;

[0046] Figure 3 This is a flow chart for realizing high spatial resolution U and V wind fields of the present invention;

[0047] Figure 4 It is a flow chart for realizing high spatial resolution U and V power spectrum correction of the present invention;

[0048] Figure 5 A flowchart for calculating the weight coefficient of a high spatial resolution grid point according to the present invention;

[0049] Figure 6 The present invention is a flowchart for implementing the correction of the U and V components of the spaceborne laser wind radar. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided below in conjunction with the drawings is not intended to limit the scope of protection of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application. The present invention is further described below in conjunction with the drawings.

[0051] like Figure 1 As shown, the present invention specifically relates to a method for correcting wind field data of a satellite-borne laser radar based on the echo signal of a ground-based laser wind radar, which specifically includes the following contents:

[0052] Step 1: Collect horizontal wind speed and direction data at different heights on the ground-based wind tower; collect ground-based laser wind radar observation parameters and wind profiler beam swing (DBS) scanning mode detection data; collect Global Forecast System (GFS) data, global observation data and cloud parameter information data; collect space-borne Aeolus radar observation parameters and inverted wind field data;

[0053] Among them, the observation parameters of ground-based laser wind radar mainly refer to data such as wavelength, pulse energy, observation elevation angle, observation azimuth, temporal and spatial resolution, and noise level;

[0054] The DBS scanning mode of the ground-based laser wind radar refers to the way in which the beam quickly scans in the four directions of due east, due south, due west and due north at a high elevation angle;

[0055] The detection data of DBS scanning mode of ground-based laser wind radar mainly refers to power spectrum, radial velocity, spectrum width, spectrum intensity and signal-to-noise ratio.

[0056] GFS data include atmospheric temperature, atmospheric pressure, humidity, wind speed and direction, ground temperature, sea surface temperature, etc.;

[0057] Global observation data include real-time observation data such as satellite remote sensing, radar, ground weather stations, aircraft observations and radiosonde data;

[0058] Cloud parameter information data mainly refers to data sets such as cloud-water mixing ratio, rain-water mixing ratio, liquid water path, cloud top height, cloud bottom height, and total cloud amount;

[0059] Observation parameters of spaceborne laser wind radar, including the platform's operating altitude, operating cycle, laser wavelength, vertical resolution at different altitudes, pulse energy, pulse repetition frequency, receiver bandwidth and beam width, etc.;

[0060] Step 2: If Figure 2 As shown, according to the horizontal wind speed WS and wind direction WD data at different heights on the ground-based wind tower collected in step 1, WS and WD are converted from the polar coordinate system to the rectangular coordinate system, and according to the decomposition formula of the horizontal wind speed components U1 and V1, U1 and V1 are calculated. According to the radial velocity data in the DBS scanning mode detection data of the ground-based laser wind measurement radar collected in step 1, combined with the beam elevation angle information in the DBS scanning mode, the horizontal wind speed components U2 and V2 are projected onto the radial beams in the four directions of due east, due south, due west and due north, respectively, and a set of equations between the radial velocities in the four directions and U2 and V2 is established. The U2 and V2 components at different heights can be obtained by solving this set of equations.

[0061] Among them, the decomposition formulas of horizontal wind speed components U1 and V1 are: U1=-WS´sin(WD), V1=-WS´cos(WD);

[0062] Step 3: Take the U1 and V1 components of different height layers on the ground-based wind tower obtained in step 2 as the dependent variable data set, find the laser wind radar observation data layer closest to these height layers according to the height layer where the dependent variable data set is located, and extract the U2 and V2 component data in these laser wind radar observation data layers as the independent variable data set. Input the dependent variable data set and the independent variable data set into the nonlinear regression correction model library together, and establish multiple nonlinear regression correction models between U2 and V2 as independent variable data sets and the corresponding U1 and V1 as dependent variable data sets. Then, calculate the correlation coefficients RU and RV of U2 as an independent variable and U1 as a dependent variable, and V2 as an independent variable and V1 as a dependent variable in each nonlinear regression correction model. After that, construct the evaluation index of the joint correlation coefficient RUV, and determine the best nonlinear regression correction model with the maximum amplitude of the joint correlation coefficient as the evaluation standard.

[0063] Among them, the nonlinear regression correction model library mainly includes exponential model, power law model, logarithmic model, polynomial regression and exponential power model;

[0064] The evaluation index calculation of the joint correlation coefficient RUV includes ;

[0065] Step 4: Based on the optimal nonlinear regression correction model determined in step 3, the two components U2 and V2 of the horizontal wind speed of the laser wind radar at different heights calculated in step 2 are input to realize the correction of the U2 and V2 components and obtain the corrected U2_c and V2_c components. According to the power spectrum signal of the DBS scanning mode of the ground-based laser wind radar collected in step 1, the error function is defined as the least squares error, and the nonlinear least squares method is used to realize the fitting of the power spectrum model. Afterwards, the corrected U2_c and V2_c components and the observation parameters of the ground-based laser wind radar are used as input data and input into the fitted power spectrum model to realize the establishment of the power spectrum based on the corrected horizontal wind speed components U2_c and V2_c at different heights.

[0066] Step 5: Figure 3 As shown in the figure, the GFS data, global real-time observation data and cloud parameter information data obtained in step 1 are used as the input of the rapid weather forecast model (Rapid Refresh, RAP), the grid spacing is set to 3 kilometers, the time step is 360 seconds, the number of domains is 2, the microphysics scheme is selected as Thompson, the longwave and shortwave radiation schemes are selected as the Rapid Radiative Transfer Model for GCMs (RRTMG), the boundary layer parameterization scheme is selected as the boundary layer scheme (Mellor-Yamada-Janjic, MYJ), the surface physics parameterization scheme is selected as MYJ, and the surface physics parameterization scheme is selected as the Noah Land Surface Model with Multi-Parameterization Options (Noah-MP). Start the RAP mode to generate wind speed and wind direction grid data for different altitude layers with a medium spatial resolution of 3 kilometers. Based on the generated 3 km spatial resolution wind speed and wind direction grid data, the wind speed and wind direction of the CFD boundary wind field are generated according to the boundary field range of the computational fluid dynamics model (CFD). Set the network model of the CFD calculation area, set the output to 30 meters spatial resolution, run the CFD mode, and generate a 30-meter high spatial resolution horizontal wind speed component wind field, represented by U and V.

[0067] Step 6: Figure 4As shown in the figure, the high spatial resolution U and V wind fields generated in step 5, as well as the satellite-borne laser wind radar observation parameters and GFS data collected in step 1 are input into the laser radar simulator (LidarSim). Running the simulator can generate the U and V power spectrum signals of each grid point under the high spatial resolution grid (dataset 1). Calculate the distance between the U and V component power spectra (dataset 2) of different altitude layers corrected by the ground-based laser wind radar generated in step 4 and all the grid points in data set 1. Using the shortest distance as the selection criterion, select the data that matches the space of data set 2 from data set 1 to form data set 3; use data set 2 as the output data set and data set 3 as the input data set to construct a power spectrum correction model based on random forest to generate the corrected high spatial resolution grid U and V power spectrum signals.

[0068] Step 7: Figure 5 As shown, based on the corrected high spatial resolution grid generated in step 6, combined with the low spatial resolution grid of the satellite-borne laser wind radar, the number of high spatial resolution grids included in each low spatial resolution network is counted. Afterwards, the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid is calculated. The satellite-borne laser wind radar observation parameters are input into the satellite-borne laser radar antenna simulator to simulate the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna. According to the spatial distribution diagram of the antenna radiation energy, the angle information of each point in the antenna side lobe deviating from the center point of the main lobe and the amplitude below the level of the main lobe center point are obtained. Thereafter, these angle information and amplitude information are input into the Taylor weighting function to generate the satellite-borne laser wind radar antenna direction weighting function.

[0069] Step 8: Based on the distance △R calculated in step 7, combined with the operating altitude of the satellite-borne laser wind radar platform and the longitude and latitude of the sub-satellite point, and taking into account the curvature of the earth, the zenith angle of each grid point in the high spatial resolution grid is calculated based on the satellite-ground spatial geometric relationship between the corrected high-resolution grid generated in step 6 and the satellite-borne laser wind radar platform. The zenith angle information is input into the satellite-borne laser wind radar antenna direction weighting function established in step 7 to obtain the weight coefficient corresponding to each grid point in the high spatial resolution grid.

[0070] Step 9: Figure 6As shown in FIG. 8 , the weight coefficient corresponding to each grid point in the high spatial resolution grid obtained in step 8 is multiplied by the power spectrum signals of the U and V components on the high spatial resolution grid points generated in step 6, and then the weighted synthesis of U and V in the Doppler frequency domain is carried out on the low spatial resolution grid, and finally the power spectrum signals of the U and V components of the low spatial resolution spaceborne laser wind radar can be generated. After that, the echo power, noise coefficient, etc. received by the spaceborne laser wind radar are added to the power spectrum signal to generate the power spectrum signals of the U and V components containing the real noise signal, and then the Fourier transform spectrum estimation method is used to obtain the simulated U and V components of the low spatial resolution spaceborne laser wind radar.

[0071] Step 10: Use the simulated low spatial resolution spaceborne laser wind radar U component obtained in step 9 as the input data set; use the U component actually observed by the spaceborne laser wind radar as the output data set; based on the adaptive lifting model, use the input and output data sets as training sets to construct a satellite-borne laser radar wind field data correction model based on the adaptive lifting model; after the model training is completed, input the wind field data observed by the spaceborne laser wind radar again to obtain the corrected satellite-borne laser wind radar wind field data.

[0072] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used for various other combinations, modifications and improvements, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar, characterized in that: The correction method comprises: Step 1: Calculate the horizontal wind speed components U1 and V1, as well as the horizontal wind speed components U2 and V2 at different altitudes based on the collected data; specifically including: A1. Collect horizontal wind speed and direction data at different heights on ground-based wind towers, collect ground-based laser wind radar observation parameters and DBS scanning mode detection data, collect GFS data, global observation data and cloud parameter information data, collect space-borne laser wind radar observation parameters and inverted wind field data; A2. According to the horizontal wind speed WS and wind direction WD data collected at different heights on the ground-based wind tower, WS and WD are converted from the polar coordinate system to the rectangular coordinate system, and U1 and V1 are calculated according to the decomposition formula of the horizontal wind speed components U1 and V1; According to the radial velocity data collected from the DBS scanning mode detection data of the ground-based laser wind measuring radar, combined with the beam elevation angle information in the DBS scanning mode, the horizontal wind speed components U2 and V2 are projected onto the radial beams in the four directions of due east, due south, due west and due north, respectively, and a set of equations between the radial velocities in the four directions and U2 and V2 is established. The U2 and V2 components at different altitudes are obtained by solving the set of equations. Step 2: Use the obtained U1 and V1 as dependent variable data sets, and U2 and V2 as independent variable data sets to establish a nonlinear regression correction model, and construct the evaluation index RUV based on the calculated correlation coefficients RU and RV, and determine the best nonlinear regression correction model based on the evaluation index; specifically, it includes: B1, taking the obtained U1 and V1 components as the dependent variable data set, according to the altitude layer where the dependent variable data set is located, finding the laser wind radar observation data layer closest to these altitude layers, and extracting the U2 and V2 component data in these laser wind radar observation data layers as the independent variable data set; B2. Input the dependent variable data set and the independent variable data set into the nonlinear regression correction model library, respectively establish multiple nonlinear regression correction models between U2 and V2 as independent variable data sets and corresponding U1 and V1 as dependent variable data sets, respectively calculate the correlation coefficients RU and RV of U2 as independent variable and U1 as dependent variable, V2 as independent variable and V1 as dependent variable in each linear regression correction model, construct the evaluation index of joint correlation coefficient RUV, take the maximum amplitude of joint correlation coefficient as the evaluation standard, and determine the best nonlinear regression correction model; Step 3: Correct the U2 and V2 components through the optimal nonlinear regression correction model, and establish the power spectrum of horizontal wind speed components at different altitudes based on the correction. Start the weather rapid update forecast mode according to the collected data, and set the network model of the CFD calculation area to generate U and V wind fields with high spatial resolution. The step 3 of correcting the U and V components by using the optimal nonlinear regression correction model and establishing the power spectra of the U and V components at different altitudes after correction includes: C1. Based on the determined optimal nonlinear regression correction model, the horizontal wind speed components U2 and V2 at different height layers observed by the laser wind radar are input to correct the U2 and V2 components, and the corrected U2_c and V2_c components are obtained; C2. According to the power spectrum signal of the DBS scanning mode of the collected ground-based laser wind radar, the least square error is set as the error function, and the power spectrum model is fitted using the nonlinear least square method; C3, the corrected U2_c and V2_c components and the ground-based laser wind radar observation parameters are used as input data and input into the fitted power spectrum model to establish the power spectrum of the horizontal wind speed U2_c and V2_c components at different height layers after correction; Step 4: Input the generated high spatial resolution U and V wind fields and the collected data into the lidar simulator to obtain data set 1, use the power spectra of U and V components at different altitudes as data set 2, form data set 3 based on data set 1 and data set 2, construct a power spectrum correction model based on data set 2 and data set 3, and generate the corrected high spatial resolution grid U and V power spectrum signals; Step 5: Calculate the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, simulate the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna, generate the direction weighting function of the satellite-borne laser wind radar antenna, and finally obtain the weight coefficient corresponding to each grid point in the high spatial resolution grid; Step six, generate the power spectrum signals of the U and V components of the low spatial resolution spaceborne laser wind radar according to the obtained weight coefficients and the power spectrum signals of the high spatial resolution grids U and V, and finally obtain the simulated U and V components of the low spatial resolution spaceborne laser wind radar as the input data set, and use the U and V components actually observed by the spaceborne laser wind radar as the output data set. Use the input data set and the output data set as the training set to construct a satellite-borne laser radar wind field data correction model, and finally obtain the corrected satellite-borne laser wind radar wind field data.

2. The method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar according to claim 1, characterized in that: The step 3 of starting the weather rapid update forecast mode according to the collected data and setting the network model of the CFD calculation area to generate high spatial resolution U and V wind fields includes: D1. Use the obtained GFS data, global real-time observation data and cloud parameter information data as the input of the weather rapid update forecast model RAP, set the corresponding parameters and schemes, and start RAP to generate wind speed and wind direction grid data of different altitude layers with medium spatial resolution M1; D2. Based on the generated medium spatial resolution wind speed and wind direction grid data, generate the wind speed and wind direction of the CFD boundary wind field according to the CFD boundary field range, set the network model of the CFD calculation area, and set the output spatial resolution to a high level. Run the CFD mode to generate a high spatial resolution horizontal wind speed component wind field, represented by U and V.

3. The method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar according to claim 1, characterized in that: The fourth step comprises: E1. Input the generated high spatial resolution horizontal wind speed component U and V wind field, as well as the collected spaceborne laser wind radar observation parameters and GFS data into the laser radar simulator, and run the laser radar simulator to generate the U and V power spectrum signals of each grid point under the high spatial resolution grid as data set 1; E2, taking the U and V component power spectra of different altitude layers after the ground-based laser wind radar correction as data set 2, calculating the distance between all grid points in data set 2 and data set 1, and selecting the data that matches the space of data set 2 from data set 1 to form data 3 based on the shortest distance as the selection criterion; E3. Use dataset 2 as the output dataset and dataset 3 as the input dataset to build a power spectrum correction model based on random forest and generate the corrected power spectrum signals of high spatial resolution grids U and V.

4. The method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar according to claim 1, characterized in that: The step 5 of calculating the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, simulating the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna, and generating the satellite-borne laser wind radar antenna direction weighting function includes: F1. Based on the generated corrected high-resolution grid and combined with the low-spatial-resolution grid of the spaceborne laser wind radar, the number of high-spatial-resolution grids included in each low-spatial-resolution grid is counted; F2. Calculate the distance △R between each high spatial resolution grid and the center point of the low spatial resolution grid, input the observation parameters of the satellite-borne laser wind radar into the satellite-borne laser wind radar antenna simulator, and simulate the spatial distribution diagram of the radiation energy of the satellite-borne laser wind radar antenna; F3. According to the spatial distribution diagram of antenna radiation energy, the angle information of each point in the antenna side lobe deviating from the center point of the main lobe and the amplitude below the level of the center point of the main lobe are obtained, and these angle information and amplitude information are input into the Taylor weighting function to generate the direction weighting function of the space-borne laser wind radar antenna.

5. The method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar according to claim 4, characterized in that: The weight coefficient corresponding to each grid point in the high spatial resolution grid finally obtained in step 5 includes: G1. Based on the calculated distance △R, combined with the operating altitude of the satellite-borne laser wind radar platform and the longitude and latitude of the sub-satellite point, and taking into account the curvature of the earth, the zenith angle of each grid point in the high spatial resolution grid is calculated based on the space-to-ground geometric relationship between the generated corrected high spatial resolution grid and the satellite-borne laser wind radar platform; G2. Input the zenith angle information into the established satellite-borne laser wind radar antenna direction weighting function to obtain the weight coefficient corresponding to each grid point in the high spatial resolution grid.

6. The method for correcting wind field data of a satellite-borne laser radar based on a ground-based laser wind radar according to claim 1, characterized in that: The step six comprises: H1, multiply the weight coefficient corresponding to each grid point in the obtained high spatial resolution grid by the power spectrum signals of the U and V components on the generated high spatial resolution grid points, carry out weighted synthesis of U and V in the Doppler frequency domain on the low spatial resolution grid, and generate the power spectrum signals of the U and V components of the low spatial resolution spaceborne laser wind radar; The echo power and noise coefficient received by the space-borne laser wind radar are added to the power spectrum signal to generate a power spectrum signal containing the U and V components of the real noise signal. The U and V components of the simulated low spatial resolution space-borne laser wind radar are obtained by using the Fourier transform spectrum estimation method. H2, the obtained simulated low spatial resolution spaceborne laser wind radar U and V components are combined into a pair of data sets, and used as the input data set, and the actual observed U and V components of the spaceborne laser wind radar are combined into a pair of data sets, and used as the output data set; The input data set and the output data set are used as training sets to construct a correction model for spaceborne lidar wind field data based on the adaptive boosting model. After the training is completed, the wind field data observed by the lidar is input again to obtain the corrected spaceborne lidar wind field data.

Images