A method for completing radial missing data of PPI scanning of a three-dimensional laser wind measurement radar

Abstract

The application provides a kind of three-dimensional laser wind measurement radar PPI scanning radial missing data completion method.The method steps are:S1, set the pitch angle, azimuth angle interval parameters of three-dimensional laser wind measurement radar PPI scanning, obtain a complete scanning period data after 360 ° scanning of radar;S2, traverse each height layer radial data in scanning period data, if there is missing data in current height layer, then execute step S3, if there is no missing data, then continue to traverse the radial data of next height layer;S3, for the radial data of the same height layer with missing data, Fourier series or piecewise cosine function is used for data fitting, and the variation law curve of the radial data of the height layer is constructed;S4, using the variation law curve obtained by fitting, the missing data of current height layer is filled, and the missing data completion processing of the height layer is completed.The application is applied to the technical field of laser wind measurement radar data processing.

Description

Technical Field

[0001] This invention belongs to the field of laser wind radar data processing technology, specifically relating to a method for completing radial missing data in three-dimensional laser wind radar PPI (Plan Position Indicator) scans. It is applicable to the preprocessing and optimization of laser radar detection data in scenarios such as meteorological monitoring, aviation safety, and wind farm operation and maintenance. This method relies on an embedded computing platform to achieve efficient data repair, and can be integrated into the local processing unit or edge computing node of the laser radar. It features low resource consumption, high real-time performance, and strong robustness, meeting the data integrity assurance requirements of intelligent sensing terminals in next-generation information networks, and also providing crucial data support for the stable operation of related computer-aided equipment. Background Technology

[0002] Three-dimensional laser wind radar acquires radial wind speed information at different azimuth angles through PPI scanning, thereby enabling the observation and inversion of large-scale atmospheric wind fields. It plays a crucial role in various fields such as meteorological monitoring, aviation safety, and wind farm operation and maintenance. However, in actual operation, the radar beam is easily obstructed by obstacles such as buildings and terrain when scanning at low elevation angles. At the same time, the signal attenuates under adverse weather conditions such as precipitation and low visibility. Both of these situations can lead to missing radial wind speed data, resulting in a large number of blank spots in the radial wind speed matrix.

[0003] Such missing data severely disrupts the continuity of PPI data, thereby affecting the stability and reliability of wind field inversion. In applications with high data quality requirements, such as wind power assessment and atmospheric boundary layer analysis, high-quality, continuous radial wind speed data is crucial. Therefore, a high-precision and robust missing data completion method is urgently needed to improve the data availability of radar systems.

[0004] Existing methods for completing missing PPI scan data from lidar wind measuring radar mainly fall into two categories: one is a machine learning-based method that trains neural networks, interpolation networks, autoencoders, and other models to learn the spatial distribution characteristics of the wind field, thereby predicting the missing radial wind speed; the other is a method that relies on the inversion relationship of local observation points. This type of method is usually based on the physical projection relationship between the wind vector and the radial wind speed, and uses a small number of effective observations near the missing point to perform local inversion to estimate the radial wind speed value at the missing azimuth angle.

[0005] However, the aforementioned existing technologies have obvious drawbacks: for completion methods based on machine learning models, the model training and deployment costs are high. Deep learning models usually require GPU training and consume a lot of computing resources during the inference stage. Many LiDAR devices use embedded processing units, which are difficult to support such models. At the same time, the models are susceptible to noise and scene changes. Wind fields often have high non-uniformity and abrupt changes. Machine learning models learn distributions based on statistical laws, and their outputs may deviate from the real physical laws, making predictions unstable when the wind field changes abruptly.

[0006] Methods that rely on inversion relationships based on local observation points are sensitive to noisy data and errors are easily amplified. When the observation error is large or only a small number of observation points are relied upon, the error amplification effect is likely to occur. This problem is more serious when the angle difference involved in the inversion is small or the distribution is biased.

[0007] With the rapid development of next-generation information network infrastructure, the data quality of laser wind-measuring radar, as an intelligent meteorological sensing terminal, directly affects the reliability of upper-layer application systems. However, due to factors such as the complex field deployment environment, limited computing power of embedded hardware, and unstable communication links, raw radar data often suffers from partial missing data. This not only affects the accuracy of wind field inversion but may also trigger anomalies or interruptions in downstream computer-aided analysis systems. Therefore, there is an urgent need for a lightweight and robust data repair mechanism that can intelligently complete missing data locally without relying on cloud or high-performance computing resources, thereby ensuring the continuous and stable operation of the radar system and its associated auxiliary equipment (such as data acquisition controllers, edge servers, and remote monitoring terminals).

[0008] Therefore, in view of the above-mentioned problems in the existing technology, the present invention proposes a method for completing radial missing data of three-dimensional laser wind radar PPI scanning, so as to solve the technical defects of existing completion methods such as high cost, poor stability and error amplification, and improve the accuracy and robustness of missing data completion. Summary of the Invention

[0009] The purpose of this invention is to provide a method for completing radial missing data in three-dimensional laser wind radar PPI scanning, so as to solve the problems of high cost, poor stability and easy amplification of errors in existing completion methods, and to achieve high-precision and high-robust radial missing data completion.

[0010] The technical solution adopted in this invention is a method for completing radial missing data in a three-dimensional laser wind radar PPI scan, which includes the following steps:

[0011] S1. Set the elevation angle and azimuth angle interval parameters of the three-dimensional laser wind radar PPI scan, and obtain the data of a complete scan cycle after the radar completes a 360° scan;

[0012] S2. Traverse the radial data of each height layer in the scanning cycle data. If there is missing data in the current height layer, execute step S3. If there is no missing data, continue to traverse the radial data of the next height layer.

[0013] S3. For radial data at the same height level with missing data, use Fourier series or piecewise cosine function to fit the data and construct the variation curve of the radial data at that height level.

[0014] S4. Using the variation curve obtained from the fitting, fill in the missing data of the current height layer to complete the missing data completion process for that height layer.

[0015] Further, in step S1, the complete scan cycle data consists of P data matrices distributed along the time dimension, sequentially labeled as follows: Each data matrix is ​​an m-row, n-column two-dimensional matrix, which together constitute a three-dimensional data set. The data matrix includes radial dimension, height dimension, time dimension, and matrix elements. The definitions of each dimension and element are as follows:

[0016] The radial dimension corresponds to the column direction of the matrix, forming n radial detection directions, identified by the column index j, where j is a positive integer from 1 to n;

[0017] The height dimension corresponds to the row direction of the matrix, containing m height detection layers, identified by the row index i, where i is a positive integer from 1 to m;

[0018] The time dimension corresponds to the radar continuously completing its operation with a period of T. p The data obtained from a complete 360° scan are arranged sequentially along the time dimension;

[0019] Matrix elements a ij This represents the valid observation data of the i-th height layer in the j-th radial direction. The data missing flag is set to -1000, which is used to mark the location where the radar did not collect valid data.

[0020] Furthermore, in step S3, when using Fourier series for data fitting, the Fourier series expression for the radial wind speed element in the i-th row of the data matrix is:

[0021] ,

[0022] In the formula, The fitted value of the radial wind speed element in the i-th row and j-th column of the data matrix; a i0 , a ik , b ikThe Fourier series coefficients corresponding to the data in the i-th row are obtained by fitting the effective radial wind speed data of that row and column. k denoted as the harmonic order of the Fourier series, and N is the truncation order of the series; θ j The azimuth angle of the radar beam corresponding to the j-th column of the data matrix.

[0023] Furthermore, in step S3, when using a piecewise cosine function for data fitting, the azimuth interval corresponding to the same height layer is divided into T consecutive segments. The fitted value of the element in the j-th column within the t-th segment is determined by the projection relationship between the segmented wind field and the radial direction, as shown in the formula:

[0024] ,

[0025] In the formula, This represents the fitted value of the radial wind speed element in the j-th column of the i-th row and t-th segment of the data matrix; i This is the row index of the data matrix, corresponding to the same range layer of the 3D laser wind measurement radar; t For segment numbers, t =1, 2, ..., T, where T is the total number of segments in the azimuth interval; u it , v it These are the eastward and northward wind field components within the i-th row and t-th segment, respectively; θ j The azimuth angle of the radar beam corresponding to the j-th column of the data matrix, and it belongs to the azimuth angle interval of the t-th segment; This is the fixed elevation angle for radar PPI scanning.

[0026] Furthermore, the key control parameters for piecewise fitting include the azimuth span. The data segments, data distribution, and local quality indicators; among them, the azimuth span. As the wind field size increases, its uniformity decreases, affecting accuracy. When decreasing, because In calculations, the data is in the denominator, which amplifies the error and requires balancing. The number of segments is 2 to n, where n is the number of columns in the data matrix. The data distribution can be left-skewed, right-skewed, or bilateral, with bilateral distribution providing balanced coverage of the overall characteristics of the segments. Local quality indicators should be selected as either large or small or the average.

[0027] Furthermore, in step S1, the elevation angle is a low elevation angle positive value, and the azimuth interval is set according to the actual observation requirements, so that the radar can cover 0°-360° all-round detection.

[0028] Furthermore, in step S2, each height layer is traversed in order of radial distance, and radial data loss is detected layer by layer. The presence of missing data is determined by identifying the missing data markers in the data matrix.

[0029] Furthermore, the harmonic order of the Fourier series k Adjust according to the complexity of the radial data curve.

[0030] Furthermore, when using piecewise cosine function fitting, the local wind field is assumed to have uniformity and continuity, and the wind field inversion of the corresponding segment is completed by relying on at least two valid observations.

[0031] Further, the specific steps of step S4 are as follows: substitute the azimuth angle corresponding to the missing data into the fitted change curve, calculate the compensation result of the missing data, and then fill the compensation result into the corresponding position of the data matrix to complete the missing data completion.

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

[0033] 1. Low computational cost and strong adaptability: This invention uses Fourier series or piecewise cosine function for data fitting, without relying on GPU for model training. The calculation process is simple and consumes less computing resources. It can be directly embedded into the embedded processing unit of LiDAR, solving the problems of high training and deployment costs and incompatibility with embedded devices in existing machine learning model completion methods, and significantly improving the feasibility of the method in practical radar systems.

[0034] 2. High stability and strong anti-interference capability: This invention fits radial data at the same altitude layer, using a large amount of effective azimuth data at this layer to support inference for missing points. It constructs the variation law based on the approximate cosine characteristics of the radial wind speed-azimuth curve, without relying on the assumptions of temporal continuity and strictly uniform wind field. Compared to machine learning models, this invention is not affected by the limitations of statistical laws. Even under abrupt wind field changes and noise interference, it can maintain the stability of the compensation results, avoiding deviations from the true physical laws. Compared to traditional local observation point inversion methods, this invention reduces the dependence on a small number of observation points through a global fitting or optimized piecewise fitting strategy, reducing error amplification effects and improving the reliability of the compensation results.

[0035] 3. High Completion Accuracy and Excellent Data Integrity: This invention accurately captures the complex variation patterns of the radial wind speed-azimuth curve through Fourier series fitting, making it particularly suitable for scenarios with uneven wind fields and complex curve shapes. Furthermore, by using piecewise cosine function fitting and optimizing key parameters, it can adapt to different data distributions and missing data situations, achieving high-precision fitting in scenarios with both sufficient and sparse observation data coverage. The flexible selection of these two fitting methods maximizes the accuracy of missing data completion, ensuring that the completed PPI scan data achieves 0°-360° omnidirectional coverage and continuous distribution, significantly improving data integrity. This provides high-quality data support for subsequent wind field inversion, thereby enhancing the applicability and reliability of 3D laser wind radar in practical scenarios such as meteorological monitoring, aviation safety, and wind farm operation and maintenance.

[0036] 4. In practical engineering deployments, the method of this invention can be embedded as a firmware module and integrated into the main control board of the lidar or its supporting data acquisition auxiliary equipment. After the radar completes a PPI scan, the embedded processor immediately calls this completion algorithm to screen and repair the original matrix, ensuring that the data stream output to the external system is complete and free of anomalies. This mechanism not only improves the data availability of a single-machine system but also provides standardized, high-quality data input for application scenarios such as multi-radar networking and remote centralized monitoring, effectively supporting the collaborative operation and intelligent maintenance of distributed meteorological sensing nodes in next-generation information networks. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method of the present invention;

[0038] Figure 2 A schematic diagram of modeling the PPI scanning data matrix of a 3D laser wind radar;

[0039] Figure 3 A graph showing the relationship between azimuth angle and radial wind speed at the same distance layer;

[0040] Figure 4 This is an example of comparing Fourier series fitting before and after. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0042] In the three-dimensional laser wind radar data matrix modeling of this invention, the projection relationship from the real wind vector to the radial wind elements within the data matrix is ​​characterized by the following formula:

[0043] ,

[0044] The parameters are defined as follows:

[0045] 1. :correspond Figure 2 As shown OK The first column of the data matrix line (number) (height layer), the first Column (number) Radial wind element (radial direction), in m / s;

[0046] 2. : respectively the first The actual wind components at different altitudes, including east-west, north-south, and vertical winds, are all in m / s.

[0047] 3. : No. Radar beam azimuth angle corresponding to the radial direction (values ​​range from 0° to 360°);

[0048] 4. : Fixed elevation angle for radar PPI scanning (in this invention, it is a low elevation angle positive value);

[0049] This formula forms the basis for the correlation between radial wind elements in the data matrix and the actual wind field, providing a theoretical basis for data completion and wind field inversion in this invention.

[0050] in, Figure 2 Multiple data matrices distributed along the time dimension are presented. Each matrix is ​​a two-dimensional matrix with m rows and n columns, forming a three-dimensional data set. The definitions of the radial (azimuth) dimension, the height dimension, and the time dimension, as well as the rules for identifying missing data, are clearly defined.

[0051] The radial wind speed data characteristics of a single row in the data matrix of this invention can be obtained through... Figure 3 The attached figure shows the relationship between azimuth and radial wind speed at the same distance layer of a 3D laser wind radar. It intuitively illustrates the radial wind speed data characteristics and technical logic upon which the data completion method of this invention relies: Compared with other data processing methods, this invention selects radial wind speed data from different azimuths in a single row of the data matrix. The curve characteristics are approximately cosine, and only a small number of missing points in a single curve need to be completed each time. The reasoning for missing points can be supported by a large amount of existing valid data, and the selection of this processing dimension has clear technical feasibility. However, when the radar performs PPI scanning, the non-uniformity of the wind field and measurement errors will make the curve shape more complex. Such complex curves cannot directly fit the projection formula from the real wind to the radial wind, and it is difficult to directly complete the missing values ​​through the projection relationship. Therefore, in order to complete the missing values ​​of the elements in a single row of the data matrix corresponding to this complex curve, it is necessary to use Fourier series or piecewise functions to fit the curve and complete the missing values ​​by constructing its change law.

[0052] For the complex radial wind speed-azimuth curve in a single row of the data matrix of this invention, this patent uses Fourier series or piecewise functions to fit it, so as to accurately construct the variation law of the curve and provide a basic model for filling in missing data.

[0053] Example 1: Missing Data Completion Based on Fourier Series Fitting

[0054] 1. Set PPI scan parameters and acquire raw scan data.

[0055] The present invention takes the PPI scan data collected between 9:42:26 and 9:45:15 on December 25, 2025 as the specific implementation object.

[0056] The radar PPI scanning core parameters were set as follows: scanning elevation angle of 6°, radial scanning interval of 3° (covering 0°-360° omnidirectional coverage, corresponding to 120 azimuth sampling points, i.e., data matrix n=120 columns); radial range coverage of 180m~11580m, corresponding to 96 altitude layers (each altitude layer matched with a different radial distance, i.e., data matrix m=96 rows); the radar was controlled to perform PPI scanning under the above parameters to acquire raw scanning data from 9:42:26 to 9:45:15 on December 25, 2025; missing radial wind speed (RWS) values ​​in the raw data are marked with "-1000", indicating localized missing radial data, such as... Figure 4 The area marked by the arrow on the left. Figure 4 The example shows a comparison before and after Fourier series fitting. The left side shows the result before imputation (including missing regions), and the right side shows the result after imputation, demonstrating the continuity and completeness of the data after imputation.

[0057] 2. Traverse each height layer and determine the data missing status.

[0058] Following the radial distance order, the 96 height layers corresponding to the PPI scan data (i.e., data matrix m=96 rows) are traversed, with each layer corresponding to 120 azimuth sampling points (i.e., data matrix n=120 columns). The radial data missing situation is detected layer by layer: if there are missing RWS values ​​in the 120 azimuth points of the current height layer, the subsequent imputation step is performed to perform imputation; if there are no missing values, the next layer is traversed. In this embodiment, all 96 height layers have missing data to varying degrees, and imputation processing is performed on all of them. After imputation, the missing rate of the entire data matrix (96 rows × 120 columns) is reduced to 0%.

[0059] 3. Perform Fourier series fitting on the radial data of the target height layer.

[0060] The target height layer was selected as the height layer corresponding to a radial length of 3780m (the 31st row in the data matrix m=96 rows). Among the 120 azimuth points in the original data of this layer, the RWS corresponding to azimuth angles of 180° and 198° were missing values ​​(marked as -1000). The two missing values ​​in the target height layer were removed, and the remaining 118 valid RWS data and corresponding azimuth angle data were extracted. The Fourier series fitting method was used, and the 5th harmonic was used to fit the valid data to obtain the Fourier series coefficients of this height layer as [0.31003, 1.68119, 6.02630, -0.69150, -0.47996, 0.03079, -0.34441, -0.02039, -0.19894, -0.24030, 0.13762].

[0061] 4. Complete the missing data based on the fitting results and verify the effect.

[0062] The missing azimuth angles (180°, 198°) of the target height layer were substituted into the fitted Fourier series model to calculate the corresponding RWS compensation results. These compensation results were then filled into the corresponding columns of the target height layer in the data matrix (96 rows × 120 columns) (i.e., the columns corresponding to 180° and 198°), completing the compensation for that layer. This logic was followed to fill in the missing data for the remaining height layers. After compensation, the missing rate of the entire data matrix (96 rows × 120 columns) was reduced to 0%, achieving omnidirectional coverage from 0° to 360° with continuous distribution. The radial wind speed distribution of the PPI scan after compensation is shown below. Figure 4 As shown on the right, the left side of the figure is the result before the imputation (the missing areas marked by arrows correspond to blank columns in the data matrix), and the right side of the figure is the result after the imputation. After imputation, the 96-row × 120-column data matrix achieves 0°-360° full coverage and continuous distribution.

[0063] Example 2: Missing Data Completion Based on Piecewise Cosine Function Fitting

[0064] 1. Set PPI scan parameters and acquire raw scan data.

[0065] Similar to Example 1, the scanning elevation angle was set to 6°, the radial scanning interval was 3°, and the radial distance coverage range was 180m to 11580m. The original scanning data for the same time period were obtained, and the data matrix was 96 rows × 120 columns. Missing values ​​were marked as "-1000".

[0066] 2. Traverse each height layer and determine the data missing status.

[0067] Consistent with Example 1, 96 height layers are traversed and the status of missing data is determined. All height layers need to be filled in.

[0068] 3. Perform piecewise cosine function fitting on the radial data of the target height layer.

[0069] Select the same target height layer as in Example 1 (row 31), which has 2 missing values; divide the azimuth angle interval (0°-360°) corresponding to this height layer into 12 continuous segments, with each segment having an azimuth angle span of... =30°; For each segment, effective observation data are extracted, and based on the assumption of local wind field homogeneity, a piecewise cosine function fitting formula is used to determine the eastward and northward wind field components of each segment. u t , v t Among them, the segmented data volume is controlled so that each segment contains at least 3 valid observations, the data distribution adopts a two-sided distribution to evenly cover the overall features of the segment, and the mean value is selected for the local quality index to improve the similarity between the fitting and the observation.

[0070] 4. Complete the missing data based on the fitting results and verify the effect.

[0071] The missing azimuth angles (180°, 198°) were substituted into the fitting curves of the corresponding segments to calculate the compensation results of the missing data, which were then filled into the corresponding positions of the data matrix. After compensation, the data continuity of the target height layer was good, and the missing rate of the entire data matrix was reduced to 0%.

[0072] Results and Analysis

[0073] The completion effects of Examples 1 and 2 above were compared, and the evaluation indicators included completion accuracy (mean square error of the actual data), computation time, and stability (completion deviation in areas of sudden wind field changes). The results are shown in the table below:

[0074]

[0075] Comparison conclusion:

[0076] 1. In terms of completion accuracy: The Fourier series fitting method in Example 2 has the smallest MSE and the highest completion accuracy; the piecewise cosine function fitting method in Example 1 has the second highest accuracy.

[0077] 2. Calculation time: The calculation time of Example 1 and Example 2 is short, both within 50ms.

[0078] 3. In terms of stability: The completion deviations of Examples 1 and 2 in the wind field abrupt change region are small, at 0.51 m / s and 0.63 m / s respectively, indicating good stability.

[0079] In summary, the two fitting methods of this invention are superior to existing technologies in terms of completion accuracy, computational efficiency, and stability. They can effectively solve the technical defects of existing methods and are more suitable for the completion of radial missing data in 3D laser wind radar PPI scanning.

[0080] Finally, it should be emphasized that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan, characterized in that, The method includes the following steps: S1. Set the elevation angle and azimuth angle interval parameters of the three-dimensional laser wind radar PPI scan, and obtain the data of a complete scan cycle after the radar completes a 360° scan; S2. Traverse the radial data of each height layer in the scanning cycle data. If there is missing data in the current height layer, proceed to step S3. If there is no missing data, continue to traverse the radial data of the next height layer. S3. For radial data at the same height level with missing data, use Fourier series or piecewise cosine function to fit the data and construct the variation curve of the radial data at that height level. S4. Using the fitted change curve, fill in the missing data of the current height layer to complete the missing data completion process for this height layer. In step S3, when using Fourier series for data fitting, the Fourier series expression for the radial wind speed element in the i-th row of the data matrix is: ， In the formula, The fitted value of the radial wind speed element in the i-th row and j-th column of the data matrix; a i0 , a ik , b ik The Fourier series coefficients corresponding to the data in the i-th row are obtained by fitting the effective radial wind speed data of that row and column. k denoted as the harmonic order of the Fourier series, and N is the truncation order of the series; θ j The radar beam azimuth angle corresponding to the j-th column of the data matrix; When using a piecewise cosine function for data fitting, the azimuth interval corresponding to the same height layer is divided into T consecutive segments. The fitted value of the element in the j-th column within the t-th segment is determined by the projection relationship between the segmented wind field and the radial direction, as shown in the formula: , In the formula, This represents the fitted value of the radial wind speed element in the j-th column of the i-th row and t-th segment of the data matrix; i is the row index of the data matrix, corresponding to the same range layer of the three-dimensional laser wind measurement radar. t For segment numbers, t =1, 2, ..., T, where T is the total number of segments in the azimuth interval; u it , v it These are the eastward and northward wind field components within the i-th row and t-th segment, respectively; θ j The azimuth angle of the radar beam corresponding to the j-th column of the data matrix, and it belongs to the azimuth angle interval of the t-th segment; A fixed elevation angle for radar PPI scanning; Key control parameters for piecewise fitting include azimuth span. The data segments, data distribution, and local quality indicators; among them, the azimuth span. As the wind field size increases, its uniformity decreases, affecting accuracy. When decreasing, because In calculations, the data is in the denominator, which amplifies the error and requires balancing. The number of segments is 2 to n, where n is the number of columns in the data matrix. The data distribution can be left-skewed, right-skewed, or bilateral, with bilateral distribution providing balanced coverage of the overall characteristics of the segments. Local quality indicators should be selected as either large or small or the average.

2. The method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan according to claim 1, characterized in that, In step S1, the complete scan cycle data consists of P data matrices distributed along the time dimension, sequentially labeled as follows: Each data matrix is ​​an m-row, n-column two-dimensional matrix, which together constitute a three-dimensional data set. The data matrix includes radial dimension, height dimension, time dimension, and matrix elements. The definitions of each dimension and element are as follows: The radial dimension corresponds to the column direction of the matrix, forming n radial detection directions, identified by the column index j, where j is a positive integer from 1 to n; The height dimension corresponds to the row direction of the matrix, containing m height detection layers, identified by the row index i, where i is a positive integer from 1 to m; The time dimension corresponds to the radar continuously completing its operation with a period of T. p The data obtained from a complete 360° scan are arranged sequentially along the time dimension; Matrix elements a ij This represents the valid observation data of the i-th height layer in the j-th radial direction. The data missing flag is set to -1000, which is used to mark the location where the radar did not collect valid data.

3. The method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan according to claim 1, characterized in that, In step S1, the elevation angle is a low elevation angle with a positive value, and the azimuth interval is set according to the actual observation requirements so that the radar can cover 0°-360° all-round detection.

4. The method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan according to claim 1, characterized in that, In step S2, each height layer is traversed in order of radial distance, and radial data missingness is detected layer by layer. The presence of missing data is determined by identifying the missing data marker in the data matrix.

5. The method for completing radial missing data in a three-dimensional laser wind radar PPI scan according to claim 1, characterized in that, Harmonic order of Fourier series k Adjust according to the complexity of the radial data curve.

6. The method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan according to claim 1, characterized in that, When using piecewise cosine function fitting, the local wind field is assumed to have uniformity and continuity, and the wind field inversion of the corresponding segment is completed by relying on at least two valid observations.

7. The method for completing radial missing data in a three-dimensional laser wind-measuring radar PPI scan according to claim 1, characterized in that, The specific steps of step S4 are as follows: Substitute the azimuth angle corresponding to the missing data into the fitted change curve, calculate the compensation result of the missing data, and then fill the corresponding position of the data matrix with the compensation result to complete the missing data completion.

Images