Low-altitude three-dimensional wind field inversion method and system based on single wind measurement laser radar

Abstract

The invention discloses a low-altitude three-dimensional wind field inversion method and system based on a single wind measurement laser radar, relates to the technical field of radar detection, and solves the technical problems that a traditional wind measurement laser radar cannot directly obtain a velocity vector, data missing of a key area is caused by a blocking effect, and detection nodes are not uniform. The method specifically comprises the following steps: S1, executing a multi-elevation-angle body scanning detection mode and a wind profile detection mode by using a wind measurement laser radar, and obtaining radial wind speed data of a target area; s2, quality control is carried out on the radial wind speed, isolated abnormal points are eliminated, and isolated data missing points are filled; s3, carrying out blocking correction on a single scanning surface by using a smooth spline interpolation method; s4, after the radial wind speed is corrected based on the wind profile, an initial background three-dimensional wind field is obtained; and high-precision wind field inversion of a wind field in a three-dimensional space can be realized by using the radial wind speed of multi-layer body scanning of a single wind measurement laser radar.

Description

Technical Field

[0001] The present invention belongs to the field of radar detection, and in particular relates to a low-altitude three-dimensional wind field inversion method and system based on a single wind measuring laser radar. Background Art

[0002] Three-dimensional wind field inversion with high temporal and spatial resolution plays a crucial role in supporting the safe operation and efficient development of the low-altitude economy. Its importance permeates the entire process of low-altitude economic planning and construction, production operations, and safety supervision. As a remote sensing device, wind lidar offers advantages in long detection range and high temporal and spatial resolution, making it widely used for atmospheric wind field detection.

[0003] In the process of low-altitude atmospheric wind field detection, traditional wind measurement lidar has the following problems: 1) Limited by the principle of the wind lidar instrument itself, it can only directly measure the radial wind speed within the line of sight and cannot directly obtain the velocity vector; 2) During radar detection, the beam is easily blocked by low-altitude obstacles such as buildings, terrain, and debris. This blocking effect can lead to data loss in critical areas. 3) The detection nodes of LiDAR are uneven. As the distance increases, the measurement points become sparse and cannot meet the actual application requirements. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a low-altitude three-dimensional wind field inversion method and system based on a single wind measuring lidar, which is used to solve the technical problems raised in the above-mentioned background technology. The present invention solves the above-mentioned problems by performing high-precision inversion of the three-dimensional wind field based on the volume scanning detection results of a single wind measuring lidar.

[0005] To achieve the above objectives, a first aspect of the present invention provides a low-altitude three-dimensional wind field inversion method based on a single wind measuring lidar, comprising: Collect historical data of low-altitude three-dimensional wind field inversion from a wind laser radar, and construct an elevation step and azimuth step prediction model for a single wind laser radar in a multi-elevation volume scan detection mode based on the historical data; the elevation step and azimuth step prediction model outputs a predicted adjustment value; the predicted adjustment value is less than a preset value; Based on a single wind measurement lidar in a multi-elevation volume scanning detection mode, detection is started and, based on the predicted adjustment value, the current scanning plane is adjusted to the next target scanning plane, and detection is continued until the entire target area is detected, thereby obtaining radial wind speed data of the target area, wherein the radial wind speed data of the target area includes radial wind speed data of several scanning planes; preprocessing the acquired radial wind speed data of the target area based on the spatial continuity of the wind field to obtain preprocessed radial wind speed data; Based on the wind profile detection mode, the wind profile of the target area is obtained. Based on the wind profile, the vertical wind speed data of all radar measurement points are initialized, and the preprocessed radial wind speed data is corrected to obtain the corrected radial wind speed data. The horizontal wind vector is inverted using the corrected radial wind speed data to obtain the background three-dimensional wind field.

[0006] Another aspect of the present application provides a low-altitude three-dimensional wind field inversion system based on a single wind measuring lidar, comprising: A prediction model construction and training unit, used for constructing and training elevation step and azimuth step prediction models, comprising a data preprocessing module, a model architecture construction module, a training optimization module, a confidence assessment module, and a feedback iteration module; An elevation step and azimuth step prediction unit, used to deploy elevation step and azimuth step prediction models and calculate prediction adjustment values; The detection control module is used to control a single wind laser radar to detect in the wind profile detection mode and to perform adjustment detection in the multi-elevation volume scan detection mode according to the predicted adjustment value; A radial wind speed quality control module is used to perform quality control on the radial wind speed, eliminate isolated abnormal points and fill in isolated data missing points; A three-dimensional wind field inversion module for radar measurement points is used to perform blockage correction on a single scanning surface using a smoothing spline interpolation method, initialize the vertical wind speed of all radar measurement points based on the vertical wind speed obtained in the wind profile detection mode, correct the radial wind speed, obtain the initial background three-dimensional wind field, and use a three-dimensional variational method to invert the three-dimensional wind speed of each radar measurement point based on the initial background wind field; The target area three-dimensional wind field interpolation module is used to interpolate the radar measurement point data into the three-dimensional space grid of the target area using the inverse distance weighted method to obtain the background three-dimensional wind field.

[0007] Compared with the prior art, the present invention has the following beneficial effects: This application constructs a single wind measuring lidar based on historical data, and uses an elevation step and azimuth step prediction model in a multi-elevation volume scanning detection mode to effectively improve the efficiency and accuracy of data acquisition by a single wind measuring lidar. The continuity of the wind field is then used to calculate the pre-processed radial wind speed data. Subsequently, based on the wind profile collected by the single wind measuring lidar, the vertical wind speed data of all radar measuring points are initialized, and the pre-processed radial wind speed data is corrected to obtain the corrected radial wind speed data. The horizontal wind vector is inverted using the corrected radial wind speed data to obtain the background three-dimensional wind field, thereby achieving high-precision wind field inversion of the wind field in three-dimensional space. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0009] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Schematic diagram of a wind laser radar detection mode in an embodiment of the present invention; Figure 3 This is a rendering of quality control using the low-altitude three-dimensional wind field inversion method and system based on a single wind laser radar provided by an embodiment of the present invention; Among them, (a) is the original radial wind speed effect diagram of a single PPI scan, (b) is the effect diagram of filling isolated points and missing points through quality control, and (c) is the radial wind speed effect diagram after blocking correction; Figure 4 Schematic diagram of a three-dimensional wind field slice in an embodiment of the present invention; Figure 5 A schematic diagram is constructed for the system of the present invention; Figure 6 Schematic diagram of the construction of the prediction model and training unit in the present invention. DETAILED DESCRIPTION

[0010] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0011] See also Figure 1 - Figure 6 On the one hand, this application provides a low-altitude three-dimensional wind field inversion system based on a single wind measuring lidar, comprising: A prediction model construction and training unit 10 is used to construct and train the elevation step and azimuth step prediction models. The prediction model construction and training unit includes a data preprocessing module 11, a model architecture construction module 12, a training optimization module 13, a confidence assessment module 14, and a feedback iteration module 15; The elevation step and azimuth step prediction unit 20 is used to deploy elevation step and azimuth step prediction models and calculate prediction adjustment values; The detection control module 30 is used to control a single wind laser radar to detect in the wind profile detection mode and to perform adjustment detection in the multi-elevation volume scan detection mode according to the predicted adjustment value; A radial wind speed quality control module 40 is used to perform quality control on the radial wind speed, remove isolated abnormal points and fill in isolated data missing points; A radar point 3D wind field inversion module 50 is configured to perform blockage correction on a single scanning surface using a smoothing spline interpolation method, initialize the vertical wind speed of all radar points based on the vertical wind speed obtained in the wind profile detection mode, correct the radial wind speed, obtain an initial background 3D wind field, and invert the 3D wind speed of each radar point using a 3D variational method based on the initial background wind field. The target area three-dimensional wind field interpolation module 60 is used to interpolate the radar measurement point data into the three-dimensional space grid of the target area using the inverse distance weighted method to obtain the background three-dimensional wind field.

[0012] In an exemplary embodiment, the data preprocessing module 11 receives data input from the detection control module 30, including radial wind speed data in the multi-elevation volume scanning detection mode, the elevation and azimuth angles of the scanning surface, and the scanning surface time interval; and receives input data from the radar measurement point three-dimensional wind field inversion module 50, including the low-altitude wind field gradient (horizontal wind speed shear, vertical wind speed vertical change) obtained by inverting the first N scanning surfaces, and radar equipment status parameters; the above data is cleaned by the data preprocessing module 11; and also includes data collected by the radial wind speed quality control module 40, data feedback from the target area three-dimensional wind field interpolation module 60, and so on.

[0013] The model architecture building module 12 is used to design the model and input the model into the training optimization module 13; The training optimization module 13 is used to receive the cleaned data output by the data preprocessing module 11 for training the model output by the model architecture construction module 12; and output the trained model to the confidence evaluation module 14 for performance evaluation; The confidence assessment module 14 is used to receive the model data input by the training optimization module 13, and compare and verify the predicted data generated based on the model data input by the training optimization module 13 with the data output by the data preprocessing module 11; if the model performance meets the standards, the qualified model data is fed back to the elevation step and azimuth step prediction unit 20 for deployment; if the model performance does not meet the standards, the evaluation structure is transmitted to the feedback iteration module 15 for triggering the optimization process; The feedback iteration module 15 is used to receive the evaluation structure of the model performance that does not meet the standards output by the confidence assessment module 14, and perform mechanism operation based on the results of the confidence assessment module 14. If the model performance does not meet the standards (such as the error exceeds the threshold or the stability is insufficient), the iterative mechanism is triggered as follows: feedback to the data preprocessing module to optimize the data cleaning or division method, feedback to the model architecture construction module to adjust the sub-network structure or the number of neurons, and feedback to the training optimization module to correct the loss function or hyperparameters.

[0014] See also Figure 1 - Figure 6 Another embodiment of the present invention provides a low-altitude three-dimensional wind field inversion method based on a single wind laser radar, comprising: S1. The wind laser radar performs multi-angle volume scanning detection mode and wind profile detection mode. Based on a single wind laser radar in multi-angle volume scanning detection mode, it starts detection of the target area and obtains radial wind speed data of the target area. It collects the front area including the current scanning area after the detection is turned on. The complete observation data corresponding to each scanning surface includes radial wind speed data, elevation angle and azimuth angle of the corresponding scanning surface, and the complete observation data also includes enhanced input data including the front The low-altitude wind field gradient obtained by inverting each scanning surface, the time interval between the current scanning surface and the previous scanning surface, and the radar equipment status parameters. The low-altitude wind field gradient includes horizontal wind speed shear and vertical wind speed vertical change, and then the predicted adjustment value is obtained through the elevation step and azimuth step prediction model calculation; among them, the multi-elevation volume scanning detection mode is composed of a plane position indication scanning probe with a preset elevation sequence. The elevation sequence is from 0° to 90°, the elevation step is 5°, and the azimuth step is 2°; the wind profile detection mode is a multi-beam detection mode with an elevation angle of 70°; the schematic diagram of the wind measurement lidar detection mode is shown in Figure 2 shown.

[0015] In an exemplary embodiment, the elevation step and azimuth step prediction model construction method includes: Collect historical data of low-altitude three-dimensional wind field inversion from wind laser radars, and construct elevation step and azimuth step prediction models for a single wind laser radar in multi-elevation volume scanning detection mode based on the historical data. The elevation step and azimuth step prediction models output predicted adjustment values. If the predicted adjustment values ​​are less than preset values, the predicted adjustment values ​​include the elevation step value and azimuth step value from the current scanning plane to the next target scanning plane. Specifically, the construction method of the elevation step and azimuth step prediction models includes: data cleaning, data partitioning, model architecture, model training and optimization, and model verification and iteration. The elevation and azimuth stepping model architecture includes a core network, a feature input layer, a regional submodule, and a fusion and output layer using a long short-term memory (LSTM) network. In the regional submodule, the hidden layer is set to near the ground and near the altitude, and the two parallel subnetworks use different numbers of neurons. The elevation step and azimuth step prediction models use the minimization of the inversion error of the radial wind speed data of the next target scanning surface as the loss function.

[0016] S2. Perform quality control on radial wind speed, eliminate isolated outliers and fill in missing values, specifically including: Based on the continuity of the wind field, this quality control method realizes the identification, elimination and filling of abnormal wind speed and missing values ​​through a sliding window, combined with sub-interval statistics and threshold judgment. Its core idea is to use spatial continuity to preprocess the acquired radial wind speed data of the target area to obtain preprocessed radial wind speed data. The preprocessing method includes the identification, elimination and filling of missing values ​​of abnormal wind speed in the target area; and the blocking correction of each scanning surface in the target area to fill in the missing points in the blocking area of ​​the corresponding scanning surface.

[0017] S21. Methods for eliminating isolated outliers include: Measuring point with effective wind speed value As the center, a 5×3 sliding window window1 is constructed, where 5 and 3 are the radial and tangential window sliding numbers respectively. Assume that the wind speed measurement interval of the radar is [-60, 60], which is divided into 30 wind speed subintervals with a wind speed interval of 4m / s. -60m / s and 60m / s are the minimum and maximum radial wind speeds that the radar can detect. The frequency of occurrence of the wind speed value of window1 in each subinterval is counted. When the measuring point Radial wind speed When the frequency of the interval is less than 4, it is identified as an isolated point, removed, and set as a missing value. All measuring points are traversed to complete the removal of all isolated points in the detection area.

[0018] S22. Methods for filling missing values ​​include: Missing points with missing values A 7×5 sliding window window2 is constructed as the center, where 7 and 5 are the radial and tangential sliding numbers of the window respectively. When the proportion of valid wind speed values ​​in window2 is greater than 50%, the window has the ability to fill missing values. The frequency of occurrence of wind speed values ​​in window2 in each subinterval is counted. When the maximum interval frequency is greater than 5, the wind speed of the corresponding subinterval is used to fill the missing value. All missing measurement points are traversed to complete the processing of missing values ​​that can be filled in the detection area.

[0019] After repeating S21 and S22 operations three times, all abnormal wind speeds are identified, eliminated, and missing values ​​are filled.

[0020] S3. Use the smoothing spline interpolation method to achieve blockage correction of a single scan surface. The methods include: Extraction range gate All radial wind speeds at and the corresponding azimuth ,in =1, 2, ..., ,in is the total number of azimuth angles; The sequence of missing wind speed values ​​caused by obstruction is , the corresponding azimuth sequence ,in =1, 2, ..., ; The sequence of effective wind speed values ​​that are not blocked is , the corresponding azimuth sequence ,in =1, 2, ..., ; When L / N is less than the threshold of 20%, the condition of blocking correction is met, and the smoothing spline interpolation method is used to fill the missing points in the blocking area. The sequence based on the effective wind speed value is , the corresponding azimuth sequence , according to the goal of smoothing spline fitting, the goal is to find a function , so that the following objective function is minimized: {f}_{(x)}=min\left ( \left [ {{\sum{}}^{NL}_{k=1}{({y}_{k}-f({{\phi}_{k}}^{}))}^{2}+\lambda \int\left ( \left ( {} \right ) \right )\left ( {{f}^{''}\left ( {\phi} \right )} \right )^{2}d\phi} \right ] \right ) ; in, is the fitting error of the data point, that is, the residual sum of squares, represents the smoothness penalty term of the curve, where The second derivative of the function measures the curvature of the curve by adjusting the smoothing parameter Controls the trade-off between fitting error and smoothness, The larger the value, the smoother the curve. Take 0.2.

[0021] Furthermore, the sequence of missing wind speed values Fill in .

[0022] S4, based on the wind profile correction radial wind speed, calculate the background three-dimensional wind field. The vertical wind speed obtained based on the wind profile detection mode is initialized to the vertical wind speed of all radar measurement points. , and correct the radial wind speed, and further use the radial wind speed to invert the horizontal wind vector , and The east-west wind and the north-south wind are respectively used to obtain the background wind field . Specifically including: The wind profile detection mode obtains the wind profile using the vertical wind speed Initialize the vertical wind speed of all radar measurement points, using the vertical wind speed , correct the radial wind speed at each measuring point , remove the radial projection of the vertical velocity. The specific method is as follows: ; Based on the corrected radial wind speed , invert the two-dimensional wind field of each PPI scanning surface , each measuring point As the initial background wind field.

[0023] S5. Use the three-dimensional variational method to invert the three-dimensional wind speed at each node: Specific methods include: Based on the initial background wind field, the three-dimensional wind field components are calculated using the variational method and obtained by optimizing the objective function. , the objective function The expression is as follows: ; in, {J}_{b}=\sum _{i,j} {{W}_{b}}\left [ {{({u}_{i,j}-{u}^{b}_{ij})}^{2}+{({v}_{i,j}-{v}^{b}_{ij})}^{2}+{({w}_{i,j}-{w}^{b}_{ij})}^{2}} \right ] ; ; ; in, is the background velocity term, is the radial velocity term, is a smoothing term, determined by the divergence, vorticity and Laplace operator, , and are the corresponding weight coefficients, which are 1, 2, and 10 respectively.

[0024] S6. Using an inverse distance weighted method, interpolate the radar point data into a three-dimensional spatial grid, where the three-dimensional spatial grid is a three-dimensional spatial grid of the target area. The specific method includes: Convert the coordinates of the radar measurement points into Cartesian coordinates , the coordinates of the interpolation points are also converted to Cartesian coordinates , find the distance from the point to be interpolated The wind speed value of the point to be interpolated at the nearest 9 radar measurement points , can be expressed as the weighted sum of the neighborhood point values, that is: ; in, Wind speed value at the nearby point, is the weight function; ; in, For nearby points To the point to be interpolated distance, is the distance index, The larger the value, the more obvious the near point effect. Take 2, is the smoothing factor, take 0.2, such as Figure 4 The schematic diagram of the three-dimensional wind field slice is shown, where the colors represent the wind speed values.

[0025] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A low-altitude three-dimensional wind field inversion method based on a single wind laser radar is characterized by: include, Collect historical data of low-altitude three-dimensional wind field inversion from a wind laser radar, and construct an elevation step and azimuth step prediction model for a single wind laser radar in a multi-elevation volume scan detection mode based on the historical data; the elevation step and azimuth step prediction model outputs a predicted adjustment value; the predicted adjustment value is less than a preset value; Based on a single wind measurement lidar in a multi-elevation volume scanning detection mode, detection is started and, based on the predicted adjustment value, the current scanning plane is adjusted to the next target scanning plane, and detection is continued until the entire target area is detected, thereby obtaining radial wind speed data of the target area, wherein the radial wind speed data of the target area includes radial wind speed data of several scanning planes; preprocessing the acquired radial wind speed data of the target area based on the spatial continuity of the wind field to obtain preprocessed radial wind speed data; Based on the wind profile detection mode, the wind profile of the target area is obtained. Based on the wind profile, the vertical wind speed data of all radar measurement points are initialized, and the preprocessed radial wind speed data is corrected to obtain the corrected radial wind speed data. The horizontal wind vector is inverted using the corrected radial wind speed data to obtain the background three-dimensional wind field.

2. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 1 is characterized in that: Based on a single wind measurement lidar in multi-elevation volume scanning detection mode, the target area is opened for detection; Collect complete observation data corresponding to the first N scanning surfaces including the current scanning surface after detection is turned on, and obtain the predicted adjustment value through calculation of the elevation step and azimuth step prediction models; The complete observation data includes radial wind speed data, the elevation angle and azimuth angle of the corresponding scanning surface, and the predicted adjustment value includes the elevation angle step value and the azimuth step value from the current scanning surface to the next target scanning surface.

3. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 2 is characterized in that: The complete observation data also includes enhanced input data including the low-altitude wind field gradient obtained by inverting the first N scanning surfaces, the time interval between the current scanning surface and the previous scanning surface, and the radar equipment status parameters. The low-altitude wind field gradient includes horizontal wind speed shear and vertical wind speed vertical change.

4. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 1 is characterized in that: The method for constructing the elevation step and azimuth step prediction model includes: data cleaning, data partitioning, model architecture, model training and optimization, and model verification and iteration; The elevation step and azimuth step model architecture includes a core network, a feature input layer, a sub-regional sub-module, and a fusion and output layer using a long short-term memory network (LSTM). In the sub-regional sub-module, the hidden layer is set to near the ground and near the altitude, and the two parallel sub-networks use different numbers of neurons. The elevation step and azimuth step prediction model uses minimization of the inversion error of radial wind speed data of the next target scanning surface as a loss function.

5. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 1 is characterized in that: The method for preprocessing the radial wind speed data of the target area obtained based on the spatial continuity of the wind field includes identifying, eliminating and filling missing values ​​of abnormal wind speeds in the target area; correcting the obstruction of each scanning surface in the target area, and filling missing points in the obstruction area of ​​the corresponding scanning surface.

6. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 5 is characterized in that: The method of correcting the obstruction of each scanning surface in the target area and filling the missing points in the obstruction area of ​​the corresponding scanning surface includes: Extraction range gate All radial wind speeds at and the corresponding azimuth ,in =1, 2, ..., ,in is the total number of azimuth angles; The sequence of missing wind speed values ​​caused by obstruction is , the corresponding azimuth sequence ,in =1, 2, ..., ; The sequence of effective wind speed values ​​that are not blocked is , the corresponding azimuth sequence ,in =1, 2, ..., ; when When the value is less than the threshold value P3, the condition of blocking correction is met, and the smoothing spline interpolation method is used to fill the missing points in the blocking area. The sequence based on the effective wind speed value is , the corresponding azimuth sequence , obtain the objective function according to the smoothing spline fitting : {f}_{(x)}=min\left ( \left [ {{\sum{}}^{N-L}_{k=1}{({y}_{k}-f({{\phi}_{k}}^{}))}^{2}+\lambda \int\left ( \left ( {} \right ) \right )\left ( {{f}^{''}\left ( {\phi} \right )} \right )^{2}d\phi} \right ] \right ) ; in, is the fitting error of the data point, represents the smoothness penalty term of the curve, where The second derivative of a function measures the curvature of a curve, is the smoothing parameter.

7. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 1 is characterized in that: The method for obtaining the corrected radial wind speed data includes: The wind profile detection mode obtains the wind profile using the vertical wind speed Initialize the vertical wind speed of all radar measurement points; Using vertical wind speed , correct the radial wind speed at each measuring point , remove the radial projection of the vertical velocity and obtain the corrected radial wind speed : ； Based on the corrected radial wind speed , invert the horizontal wind vector of each PPI scanning surface , and The east-west wind and the north-south wind are respectively As the initial background wind field.

8. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 7 is characterized in that: The method of obtaining the background three-dimensional wind field includes: The three-dimensional wind speed at each radar point is inverted using the three-dimensional variational method. , use the variational method to calculate the three-dimensional wind field components, and get the three-dimensional wind speed by optimizing the objective function. The objective function The expression is as follows: ； in: {J}_{b}=\sum _{i,j} {{W}_{b}}\left [ {{({u}_{i,j}-{u}^{b}_{ij})}^{2}+{({v}_{i,j}-{v}^{b}_{ij})}^{2}+{({w}_{i,j}-{w}^{b}_{ij})}^{2}} \right ] ; ； ； in, is the background velocity term, is the radial velocity term, is a smoothing term, determined by the divergence, vorticity and Laplace operator, , and is the corresponding weight coefficient; The inverse distance weighted method is used to interpolate the radar measurement point data into the three-dimensional spatial grid of the target area to obtain the background three-dimensional wind field.

9. The low-altitude three-dimensional wind field inversion method based on a single wind laser radar according to claim 8, characterized in that: The method of interpolating the radar measurement point data into the three-dimensional space grid of the target area using the inverse distance weighted method includes: Convert the coordinates of the radar measurement points into Cartesian coordinates , the coordinates of the interpolation point are converted to Cartesian coordinates ; Find the point to be interpolated The wind speed value of the nearest Q radar adjacent measurement points to be interpolated It is expressed as the weighted sum of the neighborhood point values, that is: ； in, Wind speed value at the nearby point, is the weight function; ； in, For nearby points To the point to be interpolated distance, is the distance index, is the smoothing factor.

10. A low-altitude three-dimensional wind field inversion system based on a single wind laser radar is characterized by: include: A prediction model construction and training unit, used for constructing and training elevation step and azimuth step prediction models, comprising a data preprocessing module, a model architecture construction module, a training optimization module, a confidence assessment module, and a feedback iteration module; An elevation step and azimuth step prediction unit, used to deploy elevation step and azimuth step prediction models and calculate prediction adjustment values; The detection control module is used to control a single wind laser radar to detect in the wind profile detection mode and to perform adjustment detection in the multi-elevation volume scan detection mode according to the predicted adjustment value; A radial wind speed quality control module is used to perform quality control on the radial wind speed, eliminate isolated abnormal points and fill in isolated data missing points; A three-dimensional wind field inversion module for radar measurement points is used to perform blockage correction on a single scanning surface using a smoothing spline interpolation method, initialize the vertical wind speed of all radar measurement points based on the vertical wind speed obtained in the wind profile detection mode, correct the radial wind speed, obtain the initial background three-dimensional wind field, and use a three-dimensional variational method to invert the three-dimensional wind speed of each radar measurement point based on the initial background wind field; The target area three-dimensional wind field interpolation module is used to interpolate the radar measurement point data into the three-dimensional space grid of the target area using the inverse distance weighted method to obtain the background three-dimensional wind field.

Images