A data fusion method for low-altitude multi-source wind fields

By constructing a three-dimensional background field and assimilating and fusing multi-source data, the problem of insufficient coverage and resolution in low-altitude wind field information acquisition was solved, the collaborative utilization of heterogeneous data sources was realized, and the accuracy and reliability of wind field data were improved.

CN123116187APending Publication Date: 2026-11-17JINAN LOW ALTITUDE ECONOMIC DEVELOPMENT GROUP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611628638.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing methods for acquiring low-level wind field information, the three types of data sources—surface wind observation, profile wind observation, and numerical model forecast—are used independently, making it difficult to simultaneously meet the requirements for coverage and spatial resolution. Furthermore, the differences in format, benchmark, and quality characteristics between different types of data make it difficult to directly coordinate their use. Existing assimilation methods have failed to fully coordinate background field and observation information.

Method used

By acquiring surface wind observation data, profile wind observation data, and numerical model forecast data, a three-dimensional background field is constructed and multi-source data assimilation and fusion are performed. The influence weights are determined by the confidence level of the observations, spatial distance, and temporal differences, and iterative correction is performed until convergence, thereby achieving the coordinated utilization of three types of heterogeneous data sources.

Benefits of technology

The fused wind field data retains the authenticity of the observation data and has the spatial continuity of the numerical model, improving the coverage and resolution of the low-altitude wind field data and enhancing the accuracy and reliability of wind field prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116187A_ABST
    Figure CN123116187A_ABST
Patent Text Reader

Abstract

This application relates to the field of meteorological data processing technology, specifically to a data fusion method for low-altitude multi-source wind fields. The method includes: acquiring surface wind observation data, profile wind observation data, and numerical model forecast data within the same spatial region and time window; performing data cleaning and quality control on the multi-source wind field data; interpolating the numerical model forecast data onto a three-dimensional target grid as a background field; obtaining the observation values ​​of each observation point based on the surface and profile observation data; determining the influence weight of each observation point on grid points within its spatial neighborhood; iteratively correcting the background field until convergence, recalculating the observation increment based on the current background field in each iteration, and updating the background field using the observation increment and influence weights; and generating a standardized wind field data file. This application can effectively fuse three types of heterogeneous data sources, overcoming the limitations of a single data source in terms of coverage, and improving the accuracy and reliability of low-altitude wind field data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of meteorological data processing technology, specifically to a data fusion method for low-altitude multi-source wind fields. Background Technology

[0002] With the rapid development of low-altitude economic sectors such as drone delivery and urban air traffic, low-altitude flight activities have placed unprecedented demands on meteorological support capabilities. As a key meteorological element affecting flight safety and operational efficiency, accurate forecasting of low-altitude wind fields is directly related to flight path planning, takeoff and landing safety assessment, and energy consumption optimization. Urban low-altitude areas are characterized by complex terrain and dense buildings, resulting in significant small-scale spatiotemporal variations in wind fields. Traditional large-scale wind field forecast products are insufficient to meet the refined requirements of this scenario.

[0003] Existing methods for acquiring low-level wind field information mainly include three types: First, wind field forecasting based on numerical weather prediction models, such as the Regional Rapid Update Cyclic Assimilation Forecasting System, which can provide three-dimensional gridded wind field forecast products with wide coverage and good spatial continuity; second, wind field monitoring based on ground meteorological observation stations, which acquires real-time wind direction and speed data near the ground through equipment such as urban ground micro-stations, offering high temporal resolution and accurate observations; and third, vertical wind field detection based on wind profilers or radiosondes, which can acquire the vertical variation characteristics of the wind field, providing data support for analyzing the vertical structure of the wind field. These technologies provide a data foundation for low-level wind field analysis from different dimensions.

[0004] However, each of the aforementioned existing technologies has limitations when used independently. While surface wind observation data offers high accuracy and timeliness, its spatial coverage is limited by the density of station deployments, making it difficult to reflect wind field conditions in observational gaps. Profile wind observation data can reveal the vertical structure of the wind field, but the limited number of devices and high deployment costs prevent the formation of high-density spatial coverage. Although numerical model forecast data has continuous three-dimensional spatial coverage capabilities, its horizontal resolution is typically at the kilometer level, and its vertical hierarchy is also relatively sparse, making it difficult to meet the stringent requirements for wind field spatial resolution in urban low-altitude flights. More importantly, the three data sources mentioned above differ significantly in terms of data format, spatiotemporal reference, and quality characteristics. Existing technologies lack a means to effectively fuse these three types of heterogeneous data, preventing the synergistic utilization of the wind field information contained in each data source. Furthermore, in existing assimilation methods, the weights of the influence of different sources of observation data on the background field are usually determined solely based on spatial distance, without fully considering the differences in confidence levels between different data sources and the impact of the deviation between the observation time and the target time on the weights. Regarding iterative correction strategies, existing methods typically apply the observation increment directly to all iterations after calculating it once, failing to dynamically adjust the observation increment according to the changes in the background field in each iteration. This results in insufficient coordination between the background field and the observation information, limiting the accuracy of the fused wind field data. Summary of the Invention

[0005] To address the technical challenges of existing low-altitude wind field information acquisition methods, which rely on three independent data sources—surface wind observations, profile wind observations, and numerical model forecasts—each making it difficult to simultaneously meet the coverage and spatial resolution requirements of low-altitude wind fields, and hindering direct collaborative utilization of different data types due to differences in format, benchmarks, and quality characteristics, this application provides a data fusion method for low-altitude multi-source wind fields. By simultaneously acquiring surface wind observation data, profile wind observation data, and numerical model forecast data, and constructing a background field using numerical model forecast data while using surface and profile observation data as observation fields for multi-source data assimilation and fusion, this method achieves collaborative utilization of the three heterogeneous data sources, overcomes the limitations of single data sources in terms of coverage, and ensures that the fused wind field data possesses both the authenticity of observational data and the spatial continuity of numerical models.

[0006] In a first aspect, this application provides a data fusion method for low-altitude multi-source wind fields, comprising the following steps: S1. Acquire multi-source wind field data for the same spatial region and within the same time window. The multi-source wind field data includes surface wind observation data, profile wind observation data, and numerical model forecast data. Among them, the numerical model forecast data is three-dimensional grid data. Each grid point is determined by its longitude, latitude, and vertical level coordinates, and contains a U component value and a V component value. S2. Perform data cleaning and quality control processing on the multi-source wind field data, including value range verification, outlier removal, and missing value check; S3. The cleaned and quality-controlled numerical model forecast data is interpolated onto the pre-constructed three-dimensional target grid as the background field. The cleaned and quality-controlled surface wind observation data and profile wind observation data are converted into U-component and V-component forms, respectively, to obtain the observation values ​​of each observation point. Among them, observation data from different data sources at the same spatial location are considered to belong to different observation points; S4. Determine the influence weight of each observation point on each grid point within its spatial neighborhood based on the preset observation confidence level, the spatial distance and time difference between the observation point location and each grid point on the three-dimensional target grid; Each data source type corresponds to an observation confidence level; S5. Iteratively correct the background field on the 3D target grid until convergence to obtain the fused 3D wind field grid data; In each iteration of correction, the observed value of each observation point is calculated relative to the observed value of the background field in the current round. Then, the background field is updated based on the observed value and the influence weight. The background field obtained after the last iteration of correction is the fused three-dimensional wind field grid data. S6. Generate a standardized format wind field data file based on the fused 3D wind field grid data.

[0007] It should be further noted that in step S1, the ground wind observation data is real-time wind field information at a height of 10 meters near the ground, which is obtained through urban ground micro-stations. Each record of the ground wind observation data includes wind speed, wind direction, longitude, latitude and corresponding timestamp. Profile wind observation data is acquired through wind profilers or radiosonde equipment. Each record of profile wind observation data includes wind speed, wind direction, altitude, longitude, latitude, and a corresponding timestamp.

[0008] It should be further noted that the numerical model forecast data comes from the RMAPS-ST d02 numerical model forecast system.

[0009] It should be further noted that in step S1, surface wind observation data is stored in CSV format, profile wind observation data is stored in NetCDF4 format, and numerical model forecast data is stored in GRIB2 format.

[0010] It should be further noted that in step S1, the time difference between the timestamps of the surface wind observation data, the timestamps of the profile wind observation data, and the effective time of the numerical model forecast data does not exceed 10 minutes. The forecast lead time for numerical model forecast data shall not exceed 6 hours. The effective time of numerical model forecast data is the time corresponding to the sum of the forecast start time and the forecast lead time of the numerical model forecast data.

[0011] It should be further explained that in step S2, the value range check is as follows: if the wind speed or wind direction value of a certain record in the multi-source wind field data exceeds the preset physical possibility range, then the record is removed. Outlier removal specifically involves identifying abnormal records where wind direction changes drastically in a short period of time under calm conditions through median filtering or adjacent time consistency checks, and then removing the identified abnormal records. The missing rate check specifically involves: calculating the missing percentage of key fields in each data source within a preset time window; if the missing percentage exceeds the preset missing rate threshold, the data source is deemed unusable and skipped; if the number of consecutive missing records in a data source exceeds the preset consecutive missing number threshold, the data source is automatically skipped and an alarm is recorded.

[0012] It should be further noted that, in the value range test, the preset range for wind speed values ​​of surface wind observation data is 0-40 m / s, the preset range for wind speed values ​​of profile wind observation data is 0-80 m / s, and the preset range for all wind direction values ​​is 0°-360°. In the missing rate check, the threshold for the missing rate of key fields is 5%, and the threshold for the number of consecutive missing occurrences is 3.

[0013] It should be further noted that the surface wind observation data and profile wind observation data after cleaning and quality control are organized in pandas.DataFrame format, and the numerical model forecast data after cleaning and quality control are organized in xarray.DataArray format. All intermediate data reside only in memory and are not written to disk.

[0014] It should be further noted that after step S2 is completed, data source availability assessment and degradation processing are performed, including: If the effective data ratios of both surface wind observation data and profile wind observation data are lower than the preset minimum effective data ratio, then skip the subsequent fusion steps and revert to using only numerical model forecast data as output. If numerical model forecast data is unavailable, the fusion task will be terminated. If surface wind observation data or profile wind observation data alone is unavailable, the remaining available data sources will be used to continue the subsequent fusion steps. The minimum percentage of valid data is 30%.

[0015] It should be further noted that in step S1, the multi-source wind field data is obtained based on the data source path configuration in the preset configuration file; In step S3, the 3D target mesh is constructed based on the fused mesh parameters in the configuration file; The configuration file is a pre-defined file independent of the method flow, stored in YAML format, and includes four parts: data source path configuration, fusion grid parameters, time window parameters, and quality control parameters. The data source path configuration includes the storage paths of surface wind observation data, profile wind observation data and numerical model forecast data, as well as the names of their respective wind speed, wind direction and time fields. The parameters for the merged grid include projection method, horizontal resolution, vertical resolution, maximum height, and latitude and longitude range; The time window parameter includes the maximum time deviation between the effective time of each data source and the numerical model forecast data, as well as the maximum allowable forecast lead time of the numerical model forecast data. The quality control parameters include the upper limit of wind speed for surface wind observation data, the upper limit of wind speed for profile wind observation data, the threshold for missing key fields, the threshold for consecutive missing data, and the minimum proportion of valid data.

[0016] It should be further noted that in step S3, the horizontal resolution of the three-dimensional target mesh is 100 m, and the vertical resolution is 50 m below a height of 1000 m.

[0017] It should be further noted that in step S3, the interpolation method used to interpolate the numerical model forecast data from the original grid to the pre-constructed three-dimensional target grid as the background field is bilinear interpolation.

[0018] It should be further noted that in step S3, the conversion formulas for converting the surface wind observation data and profile wind observation data from wind direction and wind speed forms to U-component and V-component forms, respectively, are as follows:

[0019]

[0020] in, The U component value at the current observation point. The V component value at the current observation point. This represents the wind speed value at the current observation point. This represents the wind direction at the current observation point.

[0021] It should be further noted that in step S4, the confidence level of the observed values ​​of the surface wind observation data is 0.8, and the confidence level of the observed values ​​of the profile wind observation data is 0.7.

[0022] It should be further noted that in step S4, the spatial proximity range of the observation value is: with the location of the observation point as the center, the horizontal distance does not exceed the horizontal influence radius. The vertical distance does not exceed the preset vertical influence radius. The three-dimensional spatial region.

[0023] It should be further explained that the first The observation point has its spatial neighborhood range of the first... The influence weight of each grid point is calculated using the following formula:

[0024] in, For the index of the observation point; For the first The index of grid points within the spatial neighborhood of each observation point; For the first Each observation point to the grid point Influence weight; For the first The confidence weight corresponding to the data source type of each observation point; For the first Observation points and grid points The spatial distance weight between them is calculated using the following formula:

[0025] For the first The time difference weight of each observation point is calculated using the following formula:

[0026] For the first Observation points and grid points The horizontal distance between them; For the first Observation points and grid points The vertical distance between them; The horizontal radius of influence; The radius of vertical influence; For the first The time difference between the timestamp corresponding to each observation point and the effective time of the numerical model forecast data; This is the preset time decay scale.

[0027] It should be further explained that in step S4, It is 5 km long. It is 200 m. It takes 10 minutes.

[0028] It should be further explained that in step S5, the background field on the 3D target mesh is iteratively corrected until convergence, specifically as follows: Iterate through each observation point in sequence. For each observation point, determine whether the observation point is within the range of the 3D target grid. If it is, calculate the observation increment of the observation point with respect to the current background field. Then, update the background field values ​​of each grid point in the spatial neighborhood of the observation point based on the observation increment and the influence weight. If not, skip the observation point. After all observation points have been processed, one round of iteration is completed. This process is repeated multiple times until convergence.

[0029] It should be further noted that in step S5, the observation increment includes the U-component observation increment and the V-component observation increment. In one round of iterative correction, the specific calculation of the observation increment of each observation point relative to the background field of the current round is as follows: The background field of the current round is interpolated to the location of each observation point to obtain the background field value at each observation point. The background field value includes the background field U component value and the background field V component value. Calculate the difference between the U-component observation value and the background field U-component value, and the difference between the V-component observation value and the background field V-component value at each observation point. These differences are used as the U-component observation increment and V-component observation increment for the current round, respectively. The U-component observation increment and V-component observation increment are combined as the observation increment. The formula for calculating the observation increment of an observation point in the current round is:

[0030]

[0031] in, For the first The U-component observation increment of each observation point in the current round. For the first U-component observations at each observation point For the first The background field U component values ​​of each observation point in the current round For the first The V component observation increment of each observation point in the current round For the first V component observation values ​​at each observation point For the first The background field V component values ​​of each observation point in the current round.

[0032] It should be further explained that in step S5, the first The observation point has its spatial neighborhood range of the first... A single update of the background field value for each grid point is calculated using the following formula:

[0033] in, For the first The background field value of each grid point before the current observation point is updated; For the first Background field values ​​updated for each grid point by the current observation point; For the first The observation point for the first The influence weight of each grid point; For the first The observation increment of each observation point in the current round, including the observation increment of the U component. and V component observation increment .

[0034] It should be further noted that after each iteration, the updated background field is checked to see if it meets the physical consistency constraint. If it does not, it is corrected before proceeding with the next steps. The physical consistency constraints include: Horizontal divergence constraint: For any grid point on the 3D target mesh, if its horizontal divergence... If the absolute value exceeds the preset horizontal divergence threshold, it is determined that the horizontal divergence constraint is not satisfied. Vertical velocity coordination constraint, for any grid point on the 3D target mesh, if the vertical velocity... Coordinated vertical velocity calculated from horizontal divergence If the difference exceeds the preset vertical velocity deviation threshold, it is determined that the vertical velocity coordination constraint is not met. The formula for calculating the horizontal divergence is as follows:

[0035] Coordinate vertical speed The calculation formula is:

[0036] in, The horizontal divergence of the current grid point; The background field value of the U component at the current grid point in the current round; The background field value of the V component at the current grid point in the current round; The longitude coordinates of the current grid point; The latitude and lateral coordinates of the current grid point; The vertical velocity of the current grid point; The vertical coordinates of the current grid point; The reference height represents the lowest vertical level of the 3D target mesh. The vertical column of the current grid point at the reference height The vertical velocity at that point, i.e., the vertical velocity value of the corresponding grid point at the lowest vertical level of that column, is 0.

[0037] It should be further noted that the preset horizontal divergence threshold is... ; The preset vertical velocity deviation threshold is .

[0038] It should be further noted that if the physical consistency constraint is not met, it will be corrected in the following way: For grid points that do not satisfy the horizontal divergence constraint, reduce the influence weight of each observation point in the spatial neighborhood of that grid point; For grid points that do not satisfy the vertical velocity coordination constraint, the vertical velocity... Apply relaxation correction to vertical velocity Adjust to current vertical speed Coordinate vertical speed The weighted average value, coordinating vertical speed The weight is set to 0.3, and the current vertical velocity is... The weight is set to 0.7; If the physical consistency constraint is still not met after the above adjustments, the single update step size is reduced, and the update values ​​of all observation points to neighboring grid points in the current round are multiplied by the relaxation coefficient α. The initial value of α is 1.0, and it is multiplied by 0.5 each time the step size is reduced, until the physical consistency constraint is met or α is less than 0.1.

[0039] It should be further noted that in step S5, the convergence criterion is that any one of the following conditions is met: The maximum change in background field value of all grid points on the 3D target mesh after this iteration. Less than the preset maximum change threshold ; The root mean square of the observation increments for all observation points after this iteration. Less than the preset root mean square threshold ; Among them, the maximum change in background field value The calculation formula is:

[0040] Root mean square of observed increments The calculation formula is:

[0041] in, This is the current round number; For the first The grid point at the th grid point Background field value after round of iteration; For the first The grid point at the th grid point Background field value after round of iteration; This represents the total number of grid points on the 3D target mesh. For the first The observation point at the ... The observation increment in the round iteration; This represents the total number of observation points.

[0042] It should be further noted that the maximum change threshold ε is set to... ; The root mean square threshold δ is taken as: .

[0043] It should be further noted that in step S5, the number of iterations for correction is 3.

[0044] It should be further noted that step S6, which involves generating a standardized format wind field data file based on the fused 3D wind field grid data, specifically includes: Horizontal interpolation: The fused 3D wind field grid data is interpolated from the current computation grid to the target projection grid through bilinear interpolation. The target projection grid is a 2D coordinate grid. Vertical interpolation: The fused 3D wind field grid data is interpolated from the current vertical level to the preset target vertical level through spline interpolation. The target vertical level is a set of multiple discrete height levels set at preset intervals. The U and V components in the three-dimensional wind field grid data after horizontal and vertical interpolation are converted into wind direction and wind speed, resulting in three-dimensional wind field grid data containing wind direction and wind speed values. Organize the three-dimensional wind field grid data containing wind direction and wind speed values ​​into a standardized wind field data file and output it.

[0045] It should be further noted that the target projection grid adopts the WGS-84 geographic coordinate system or the UTM projection coordinate system; The vertical resolution of the target vertical layer below 1000m is 50m.

[0046] It should be further explained that the conversion formulas for the U and V components to wind direction and wind speed are as follows:

[0047]

[0048] in, This represents the wind speed value. This represents the wind direction value, with due north as 0°, increasing clockwise. For the U component value; For the V component value.

[0049] It should be further noted that the standardized wind field data file is in NetCDF4 format and contains data variables, coordinate dimensions, and metadata variables; The data variables include wind speed, wind direction, U component, and V component; Coordinate dimensions include time, altitude, latitude, and longitude; Metadata variables include the crs variable used to record projection parameters; The filenames of the standardized wind field data files are set according to the preset naming rules, and the filenames include the timestamp of the valid time of the numerical model forecast data.

[0050] Secondly, this application provides a data fusion system for low-altitude multi-source wind fields to implement the above-mentioned data fusion method, including: The data acquisition module is used to acquire multi-source wind field data for the same spatial area and within the same time window. The multi-source wind field data includes surface wind observation data, profile wind observation data, and numerical model forecast data. The data preprocessing module is used to perform data cleaning and quality control on multi-source wind field data, including value range verification, outlier removal and missing rate checking. The background field and observation generation module is used to interpolate the cleaned and quality-controlled numerical model forecast data onto a pre-constructed three-dimensional target grid as the background field, and to convert the cleaned and quality-controlled surface wind observation data and profile wind observation data into U-component and V-component forms, respectively, to obtain the observation values ​​of each observation point. The influence weight calculation module is used to determine the influence weight of each observation point on each grid point in its spatial neighborhood based on the preset observation confidence level, the spatial distance and time difference between the observation point location and each grid point on the three-dimensional target grid. The iterative correction module is used to iteratively correct the background field on the 3D target mesh until convergence, and obtain the fused 3D wind field grid data. The standard file generation module is used to generate standardized wind field data files based on the fused 3D wind field grid data.

[0051] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described data fusion method.

[0052] Fourthly, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described data fusion method.

[0053] As can be seen from the above technical solutions, this application has the following advantages: 1. This application overcomes the inherent limitations of single data sources in terms of coverage by simultaneously acquiring three types of data: surface wind observation data, profile wind observation data, and numerical model forecast data. It constructs a three-dimensional background field using numerical model forecast data and assimilates and fuses multi-source data using surface and profile observation data as the observation field. Surface wind observations provide high-precision real-time information about the near-surface layer, profile wind observations supplement the vertical variation characteristics, and numerical model forecasts provide a spatially continuous three-dimensional background field. These three sources complement each other, and the fused wind field data retains the authenticity of the observational data while filling in observational gaps with the help of numerical models, achieving a more comprehensive characterization of the low-altitude three-dimensional wind field.

[0054] 2. This application determines the influence weight of each observation point on each grid point in its spatial neighborhood based on the preset observation confidence level, the spatial distance between the observation point location and each grid point on the three-dimensional target grid, and the time difference between the timestamp of the observation point and the effective time of the numerical model forecast data. This overcomes the shortcomings of existing methods where the weight depends only on spatial distance and improves the quality of the fused wind field data.

[0055] 3. In each iteration of this application, the observation increments of each observation point are recalculated based on the background field of the current iteration. The background field is then updated using the recalculated observation increments and influence weights. After multiple iterations until convergence, the fused three-dimensional wind field grid data is obtained, overcoming the shortcomings of insufficient coordination between the background field and observation information in existing methods. Since the background field changes after each iteration, the deviation of the observation values ​​relative to the updated background field also changes. This invention, by recalculating the observation increments in each iteration, ensures that each correction is based on the most accurate background field state. The background field gradually approaches the observation field during the iteration process, achieving sufficient coordination between the observation information and the background field, effectively reducing the error in wind field prediction, and improving the accuracy and reliability of the fusion results. Attached Figure Description

[0056] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart of a data fusion method for low-altitude multi-source wind fields in one embodiment of this application.

[0058] Figure 2 This is a schematic block diagram of a low-altitude multi-source wind field data fusion system in one embodiment of this application.

[0059] Figure 3 This is a schematic diagram of the hardware structure of an electronic device in one embodiment of this application. Detailed Implementation

[0060] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] The data fusion method involved in this application will be described in detail below. Specific details such as particular system structures and technologies are presented for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0062] In the data fusion methods involved in this application, the term "comprising" indicates the presence of the described feature, whole, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or collections thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0063] To facilitate a clear description of the technical solutions of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.

[0064] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0065] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0066] The data fusion method provided in this application embodiment is executed by a computer device, and correspondingly, the data fusion system for low-altitude multi-source wind fields runs in the computer device.

[0067] Figure 1 This is a flowchart of a data fusion method for low-altitude multi-source wind fields according to an embodiment of this application. Figure 1 The executing entity can be a data fusion system. Depending on different needs, the order of the steps in this flowchart can be changed, and some can be omitted.

[0068] like Figure 1 As shown, the data fusion method for low-altitude multi-source wind fields includes: Step S1: Obtain multi-source wind field data for the same spatial region and within the same time window. The multi-source wind field data includes surface wind observation data, profile wind observation data, and numerical model forecast data. The numerical model forecast data consists of three-dimensional grid data. Each grid point is determined by its longitude, latitude, and vertical level coordinates, and contains a U component value and a V component value.

[0069] In some specific embodiments, the ground wind observation data is real-time wind field information at a height of 10 meters near the ground, which is obtained through urban ground micro-stations. Each record of the ground wind observation data includes wind speed, wind direction, longitude, latitude and corresponding timestamp. Profile wind observation data is acquired through wind profilers or radiosonde equipment. Each record of profile wind observation data includes wind speed, wind direction, altitude, longitude, latitude, and a corresponding timestamp.

[0070] In some specific embodiments, the numerical model forecast data are derived from the RMAPS-ST d02 numerical model forecast system.

[0071] In some specific embodiments, surface wind observation data is stored in CSV format, profile wind observation data is stored in NetCDF4 format, and numerical model forecast data is stored in GRIB2 format.

[0072] In some specific embodiments, the time difference between the timestamp of the surface wind observation data, the timestamp of the profile wind observation data, and the effective time of the numerical model forecast data does not exceed 10 minutes. The forecast lead time for numerical model forecast data shall not exceed 6 hours. The effective time of numerical model forecast data is the time corresponding to the sum of the forecast start time and the forecast lead time of the numerical model forecast data.

[0073] Step S2 involves performing data cleaning and quality control on the multi-source wind field data, including value range verification, outlier removal, and missing value check.

[0074] In some specific embodiments, in step S2, the value range check is specifically as follows: if the wind speed or wind direction value of a certain record in the multi-source wind field data exceeds the preset physical possibility range, then the record is removed. Outlier removal specifically involves identifying abnormal records where wind direction changes drastically in a short period of time under calm conditions through median filtering or adjacent time consistency checks, and then removing the identified abnormal records. The missing rate check specifically involves: calculating the missing percentage of key fields in each data source within a preset time window; if the missing percentage exceeds the preset missing rate threshold, the data source is deemed unusable and skipped; if the number of consecutive missing records in a data source exceeds the preset consecutive missing number threshold, the data source is automatically skipped and an alarm is recorded.

[0075] In some specific embodiments, during the value range test, the preset range for wind speed values ​​of surface wind observation data is 0-40 m / s, the preset range for wind speed values ​​of profile wind observation data is 0-80 m / s, and the preset range for all wind direction values ​​is 0°-360°. In the missing rate check, the threshold for the missing rate of key fields is 5%, and the threshold for the number of consecutive missing occurrences is 3.

[0076] In some specific embodiments, the cleaned and quality-controlled surface wind observation data and profile wind observation data are organized in the form of pandas.DataFrame, and the cleaned and quality-controlled numerical model forecast data are organized in the form of xarray.DataArray. All intermediate data reside only in memory and are not written to disk.

[0077] In some specific embodiments, after step S2 is completed, data source availability is assessed and downgraded, including: If the effective data ratios of both surface wind observation data and profile wind observation data are lower than the preset minimum effective data ratio, then skip the subsequent fusion steps and revert to using only numerical model forecast data as output. If numerical model forecast data is unavailable, the fusion task will be terminated. If surface wind observation data or profile wind observation data alone is unavailable, the remaining available data sources will be used to continue the subsequent fusion steps. The minimum percentage of valid data is 30%.

[0078] Step S3: The numerical model forecast data after cleaning and quality control is interpolated onto the pre-constructed three-dimensional target grid as the background field. The surface wind observation data and profile wind observation data after cleaning and quality control are converted into U-component and V-component forms, respectively, to obtain the observation values ​​of each observation point. In this context, observation data from different data sources at the same spatial location are considered to belong to different observation points.

[0079] In some specific embodiments, in step S1, the multi-source wind field data is obtained based on the data source path configuration in the preset configuration file; In step S3, the 3D target mesh is constructed based on the fused mesh parameters in the configuration file; The configuration file is a pre-defined file independent of the method flow, stored in YAML format, and includes four parts: data source path configuration, fusion grid parameters, time window parameters, and quality control parameters. The data source path configuration includes the storage paths of surface wind observation data, profile wind observation data and numerical model forecast data, as well as the names of their respective wind speed, wind direction and time fields. The parameters for the merged grid include projection method, horizontal resolution, vertical resolution, maximum height, and latitude and longitude range; The time window parameter includes the maximum time deviation between the effective time of each data source and the numerical model forecast data, as well as the maximum allowable forecast lead time of the numerical model forecast data. The quality control parameters include the upper limit of wind speed for surface wind observation data, the upper limit of wind speed for profile wind observation data, the threshold for missing key fields, the threshold for consecutive missing data, and the minimum proportion of valid data.

[0080] In some specific embodiments, in step S3, the horizontal resolution of the three-dimensional target mesh is 100 m, and the vertical resolution is 50 m below a height of 1000 m.

[0081] In some specific embodiments, in step S3, the interpolation method for interpolating numerical model forecast data from the original grid to the pre-constructed three-dimensional target grid as the background field is bilinear interpolation.

[0082] In some specific embodiments, in step S3, the conversion formulas for converting the surface wind observation data and profile wind observation data from wind direction and wind speed forms to U-component and V-component forms, respectively, are as follows:

[0083]

[0084] in, The U component value at the current observation point. The V component value at the current observation point. This represents the wind speed value at the current observation point. This represents the wind direction at the current observation point.

[0085] Step S4: Determine the influence weight of each observation point on each grid point in its spatial neighborhood based on the preset observation confidence level, the spatial distance and time difference between the observation point location and each grid point on the three-dimensional target grid; Each data source type corresponds to an observation confidence level.

[0086] In some specific embodiments, the confidence level of the observations for surface wind observation data is 0.8, and the confidence level of the observations for profile wind observation data is 0.7.

[0087] In some specific embodiments, the spatial proximity range of the observation is defined as follows: with the location of the observation point as the center, the horizontal distance does not exceed the horizontal radius of influence. The vertical distance does not exceed the preset vertical influence radius. The three-dimensional spatial region.

[0088] In some specific embodiments, the first The observation point has its spatial neighborhood range of the first... The influence weight of each grid point is calculated using the following formula:

[0089] in, For the index of the observation point; For the first The index of grid points within the spatial neighborhood of each observation point; For the first Each observation point to the grid point Influence weight; For the first The confidence weight corresponding to the data source type of each observation point; For the first Observation points and grid points The spatial distance weight between them is calculated using the following formula:

[0090] For the first The time difference weight of each observation point is calculated using the following formula:

[0091] For the first Observation points and grid points The horizontal distance between them; For the first Observation points and grid points The vertical distance between them; The horizontal radius of influence; The radius of vertical influence; For the first The time difference between the timestamp corresponding to each observation point and the effective time of the numerical model forecast data; This is the preset time decay scale.

[0092] In some specific embodiments, It is 5 km long. It is 200 m. It takes 10 minutes.

[0093] Step S5: Iteratively correct the background field on the 3D target grid until convergence to obtain the fused 3D wind field grid data; In each iteration of the correction process, the observed value of each observation point is calculated relative to the observed value of the background field in the current iteration. Then, the background field is updated based on the observed value and the influence weight. The background field obtained after the last iteration of the correction process is the fused three-dimensional wind field grid data.

[0094] In some specific embodiments, the background field on the 3D target mesh is iteratively corrected until convergence, specifically as follows: Iterate through each observation point in sequence. For each observation point, determine whether the observation point is within the range of the 3D target grid. If it is, calculate the observation increment of the observation point with respect to the current background field. Then, update the background field values ​​of each grid point in the spatial neighborhood of the observation point based on the observation increment and the influence weight. If not, skip the observation point. After all observation points have been processed, one round of iteration is completed. This process is repeated multiple times until convergence.

[0095] In some specific embodiments, the observation increment includes the U-component observation increment and the V-component observation increment. In one round of iterative correction, the specific calculation of the observation increment of each observation point relative to the background field of the current round is as follows: The background field of the current round is interpolated to the location of each observation point to obtain the background field value at each observation point. The background field value includes the background field U component value and the background field V component value. Calculate the difference between the U-component observation value and the background field U-component value, and the difference between the V-component observation value and the background field V-component value at each observation point. These differences are used as the U-component observation increment and V-component observation increment for the current round, respectively. The U-component observation increment and V-component observation increment are combined as the observation increment. The formula for calculating the observation increment of an observation point in the current round is:

[0096]

[0097] in, For the first The U-component observation increment of each observation point in the current round. For the first U-component observations at each observation point For the first The background field U component values ​​of each observation point in the current round For the first The V component observation increment of each observation point in the current round For the first V component observation values ​​at each observation point For the first The background field V component values ​​of each observation point in the current round.

[0098] In some specific embodiments, the first The observation point has its spatial neighborhood range of the first... A single update of the background field value for each grid point is calculated using the following formula:

[0099] in, For the first The background field value of each grid point before the current observation point is updated; For the first Background field values ​​updated for each grid point by the current observation point; For the first The observation point for the first The influence weight of each grid point; For the first The observation increment of each observation point in the current round, including the observation increment of the U component. and V component observation increment .

[0100] In some specific embodiments, after each iteration, it is checked whether the updated background field satisfies the physical consistency constraint. If it does not, it is corrected, and then the subsequent steps are executed. The physical consistency constraints include: Horizontal divergence constraint: For any grid point on the 3D target mesh, if its horizontal divergence... If the absolute value exceeds the preset horizontal divergence threshold, it is determined that the horizontal divergence constraint is not satisfied. Vertical velocity coordination constraint, for any grid point on the 3D target mesh, if the vertical velocity... Coordinated vertical velocity calculated from horizontal divergence If the difference exceeds the preset vertical velocity deviation threshold, it is determined that the vertical velocity coordination constraint is not met. The formula for calculating the horizontal divergence is as follows:

[0101] Coordinate vertical speed The calculation formula is:

[0102] in, The horizontal divergence of the current grid point; The background field value of the U component at the current grid point in the current round; The background field value of the V component at the current grid point in the current round; The longitude coordinates of the current grid point; The latitude and lateral coordinates of the current grid point; The vertical velocity of the current grid point; The vertical coordinates of the current grid point; The reference height represents the lowest vertical level of the 3D target mesh. The vertical column of the current grid point at the reference height The vertical velocity at that point, i.e., the vertical velocity value of the corresponding grid point at the lowest vertical level of that column, is 0.

[0103] In some specific embodiments, the preset horizontal divergence threshold is: ; The preset vertical velocity deviation threshold is .

[0104] In some specific embodiments, if the physical consistency constraint is not met, it is corrected in the following way: For grid points that do not satisfy the horizontal divergence constraint, reduce the influence weight of each observation point in the spatial neighborhood of that grid point; For grid points that do not satisfy the vertical velocity coordination constraint, the vertical velocity... Apply relaxation correction to vertical velocity Adjust to current vertical speed Coordinate vertical speed The weighted average value, coordinating vertical speed The weight is set to 0.3, and the current vertical velocity is... The weight is set to 0.7; If the physical consistency constraint is still not met after the above adjustments, the single update step size is reduced, and the update values ​​of all observation points to neighboring grid points in the current round are multiplied by the relaxation coefficient α. The initial value of α is 1.0, and it is multiplied by 0.5 each time the step size is reduced, until the physical consistency constraint is met or α is less than 0.1.

[0105] In some specific embodiments, the convergence criterion is that any one of the following is satisfied: The maximum change in background field value of all grid points on the 3D target mesh after this iteration. Less than the preset maximum change threshold ; The root mean square of the observation increments for all observation points after this iteration. Less than the preset root mean square threshold ; Among them, the maximum change in background field value The calculation formula is:

[0106] Root mean square of observed increments The calculation formula is:

[0107] in, This is the current round number; For the first The grid point at the th grid point Background field value after round of iteration; For the first The grid point at the th grid point Background field value after round of iteration; This represents the total number of grid points on the 3D target mesh. For the first The observation point at the ... The observation increment in the round iteration; This represents the total number of observation points.

[0108] In some specific embodiments, the maximum change threshold ε is set to a value of ; The root mean square threshold δ is taken as: .

[0109] In some specific embodiments, the number of iterations for correction is 3.

[0110] Step S6: Generate a standardized format wind field data file based on the fused 3D wind field grid data.

[0111] In some specific embodiments, generating standardized format wind field data files based on the fused 3D wind field grid data specifically includes: Horizontal interpolation: The fused 3D wind field grid data is interpolated from the current computation grid to the target projection grid through bilinear interpolation. The target projection grid is a 2D coordinate grid. Vertical interpolation: The fused 3D wind field grid data is interpolated from the current vertical level to the preset target vertical level through spline interpolation. The target vertical level is a set of multiple discrete height levels set at preset intervals. The U and V components in the three-dimensional wind field grid data after horizontal and vertical interpolation are converted into wind direction and wind speed, resulting in three-dimensional wind field grid data containing wind direction and wind speed values. Organize the three-dimensional wind field grid data containing wind direction and wind speed values ​​into a standardized wind field data file and output it.

[0112] In some specific embodiments, the target projection grid adopts the WGS-84 geographic coordinate system or the UTM projection coordinate system; The vertical resolution of the target vertical layer below 1000m is 50m.

[0113] In some specific embodiments, the conversion formulas for converting the U and V components into wind direction and wind speed are as follows:

[0114]

[0115] in, This represents the wind speed value. This represents the wind direction value, with due north as 0°, increasing clockwise. For the U component value; For the V component value.

[0116] In some specific embodiments, the standardized wind field data file is in NetCDF4 format and includes data variables, coordinate dimensions, and metadata variables; The data variables include wind speed, wind direction, U component, and V component; Coordinate dimensions include time, altitude, latitude, and longitude; Metadata variables include the crs variable used to record projection parameters; The filenames of the standardized wind field data files are set according to the preset naming rules, and the filenames include the timestamp of the valid time of the numerical model forecast data.

[0117] The following are embodiments of the low-altitude multi-source wind field data fusion system provided in this application. This low-altitude multi-source wind field data fusion system and the data fusion methods of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the data fusion system, please refer to the embodiments of the low-altitude multi-source wind field data fusion method described above.

[0118] like Figure 2 As shown, the data fusion system for low-altitude multi-source wind fields includes: Secondly, this application provides a data fusion system for low-altitude multi-source wind fields to implement the above-mentioned data fusion method, including: The data acquisition module is used to acquire multi-source wind field data for the same spatial area and within the same time window. The multi-source wind field data includes surface wind observation data, profile wind observation data, and numerical model forecast data. The data preprocessing module is used to perform data cleaning and quality control on multi-source wind field data, including value range verification, outlier removal and missing rate checking. The background field and observation generation module is used to interpolate the cleaned and quality-controlled numerical model forecast data onto a pre-constructed three-dimensional target grid as the background field, and to convert the cleaned and quality-controlled surface wind observation data and profile wind observation data into U-component and V-component forms, respectively, to obtain the observation values ​​of each observation point. The influence weight calculation module is used to determine the influence weight of each observation point on each grid point in its spatial neighborhood based on the preset observation confidence level, the spatial distance and time difference between the observation point location and each grid point on the three-dimensional target grid. The iterative correction module is used to iteratively correct the background field on the 3D target mesh until convergence, and obtain the fused 3D wind field grid data. The standard file generation module is used to generate standardized wind field data files based on the fused 3D wind field grid data.

[0119] The data fusion system in this embodiment is used to implement a data fusion method for low-altitude multi-source wind fields.

[0120] This application also provides an electronic device for implementing the various embodiments of this application. Figure 3 To illustrate the hardware structure of an electronic device according to various embodiments of this application, as shown in the following diagram... Figure 3 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor.

[0121] Those skilled in the art will understand that the electronic device structure involved in the embodiments of this application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0122] In embodiments of this application, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0123] In this application embodiment, the processor can be implemented using at least one of an Application-Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a processor, a controller, a microcontroller, a microprocessor, or an electronic unit designed to perform the functions described herein. In some cases, such implementations can be implemented within a controller. For software implementations, implementations such as processes or functions can be implemented with separate software modules that allow the performance of at least one function or operation. The software code can be implemented by a software application (or program) written in any suitable programming language, and the software code can be stored in memory and executed by the controller.

[0124] In addition, the electronic device includes some functional modules not shown, which will not be described in detail here.

[0125] Those skilled in the art will understand that the various aspects of the electronic device provided in this application can be implemented as a system, method, or program product. Therefore, the various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "system."

[0126] This application also provides a storage medium storing a program product capable of implementing a data fusion method for low-altitude multi-source wind fields. In some possible embodiments, various aspects of this application can also be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this application.

[0127] The storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0128] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data fusion method for low-altitude multi-source wind fields, characterized in that, include: S1. Acquire multi-source wind field data for the same spatial region and within the same time window. The multi-source wind field data includes surface wind observation data, profile wind observation data, and numerical model forecast data. Among them, the numerical model forecast data is three-dimensional grid data. Each grid point is determined by its longitude, latitude, and vertical level coordinates, and contains a U component value and a V component value. S2. Perform data cleaning and quality control processing on the multi-source wind field data, including value range verification, outlier removal, and missing value check; S3. The cleaned and quality-controlled numerical model forecast data is interpolated onto the pre-constructed three-dimensional target grid as the background field. The cleaned and quality-controlled surface wind observation data and profile wind observation data are converted into U-component and V-component forms, respectively, to obtain the observation values ​​of each observation point. S4. Determine the influence weight of each observation point on each grid point within its spatial neighborhood based on the preset observation confidence level, the spatial distance and time difference between the observation point location and each grid point on the three-dimensional target grid; Each data source type corresponds to an observation confidence level; S5. Iteratively correct the background field on the 3D target grid until convergence to obtain the fused 3D wind field grid data; In each iteration of correction, the observed value of each observation point is calculated relative to the observed value of the background field in the current round. Then, the background field is updated based on the observed value and the influence weight. The background field obtained after the last iteration of correction is the fused three-dimensional wind field grid data. S6. Generate a standardized format wind field data file based on the fused 3D wind field grid data.

2. The data fusion method as described in claim 1, characterized in that, In step S1, the ground wind observation data is real-time wind field information at a height of 10 meters near the ground, which is obtained through urban ground micro-stations. Each record of the ground wind observation data includes wind speed, wind direction, longitude, latitude and corresponding timestamp. Profile wind observation data is acquired through wind profilers or radiosonde equipment. Each record of profile wind observation data includes wind speed, wind direction, altitude, longitude, latitude, and a corresponding timestamp.

3. The data fusion method as described in claim 1, characterized in that, In step S2, the value range check is specifically as follows: if the wind speed or wind direction value of a certain record in the multi-source wind field data exceeds the preset physical possibility range, then the record is removed. Outlier removal specifically involves identifying abnormal records where wind direction changes drastically in a short period of time under calm conditions through median filtering or adjacent time consistency checks, and then removing the identified abnormal records. The missing rate check specifically involves: calculating the missing percentage of key fields in each data source within a preset time window; if the missing percentage exceeds the preset missing rate threshold, the data source is deemed unusable and skipped. If the number of consecutive missing records in a data source exceeds the preset threshold for consecutive missing records, the data source will be automatically skipped and an alarm will be recorded.

4. The data fusion method as described in claim 1, characterized in that, In step S4, the spatial proximity range of the observation is defined as follows: with the location of the observation point as the center, the horizontal distance does not exceed the horizontal influence radius. The vertical distance does not exceed the preset vertical influence radius. The three-dimensional spatial region.

5. The data fusion method as described in claim 1, characterized in that, No. The observation point has its spatial neighborhood range of the first... The influence weight of each grid point is calculated using the following formula: in, For the index of the observation point; For the first The index of grid points within the spatial neighborhood of each observation point; For the first Each observation point to the grid point Influence weight; For the first The confidence weight corresponding to the data source type of each observation point; For the first Observation points and grid points The spatial distance weight between them is calculated using the following formula: For the first The time difference weight of each observation point is calculated using the following formula: For the first Observation points and grid points The horizontal distance between them; For the first Observation points and grid points The vertical distance between them; The horizontal radius of influence; The radius of vertical influence; For the first The time difference between the timestamp corresponding to each observation point and the effective time of the numerical model forecast data; This is the preset time decay scale.

6. The data fusion method as described in claim 1, characterized in that, In step S5, the background field on the 3D target mesh is iteratively corrected until convergence, specifically as follows: Iterate through each observation point in sequence. For each observation point, determine whether the observation point is within the range of the 3D target grid. If it is, calculate the observation increment of the observation point with respect to the current background field. Then, update the background field values ​​of each grid point in the spatial neighborhood of the observation point based on the observation increment and the influence weight. If not, skip the observation point. After all observation points have been processed, one round of iteration is completed. This process is repeated multiple times until convergence.

7. The data fusion method as described in claim 6, characterized in that, In step S5, the observation increment includes the U-component observation increment and the V-component observation increment. In one round of iterative correction, the specific calculation of the observation increment of each observation point relative to the background field of the current round is as follows: The background field of the current round is interpolated to the location of each observation point to obtain the background field value at each observation point. The background field value includes the background field U component value and the background field V component value. Calculate the difference between the U-component observation value and the background field U-component value, and the difference between the V-component observation value and the background field V-component value at each observation point. These differences are used as the U-component observation increment and V-component observation increment for the current round, respectively. The U-component observation increment and V-component observation increment are combined as the observation increment. The formula for calculating the observation increment of an observation point in the current round is: in, For the first The U-component observation increment of each observation point in the current round. For the first U-component observations at each observation point For the first The background field U component values ​​of each observation point in the current round For the first The V component observation increment of each observation point in the current round For the first V component observation values ​​at each observation point For the first The background field V component values ​​of each observation point in the current round.

8. The data fusion method as described in claim 6, characterized in that, In step S5, the first The observation point has its spatial neighborhood range of the first... A single update of the background field value for each grid point is calculated using the following formula: in, For the first The background field value of each grid point before the current observation point is updated; For the first Background field values ​​updated for each grid point by the current observation point; For the first The observation point for the first The influence weight of each grid point; For the first The observation increment of each observation point in the current round, including the observation increment of the U component. and V component observation increment .

9. The data fusion method as described in claim 1, characterized in that, In step S5, the convergence criterion is that any one of the following is satisfied: The maximum change in background field value of all grid points on the 3D target mesh after this iteration. Less than the preset maximum change threshold ; The root mean square of the observation increments for all observation points after this iteration. Less than the preset root mean square threshold ; Among them, the maximum change in background field value The calculation formula is: Root mean square of observed increments The calculation formula is: in, This is the current round number; For the first The grid point at the th grid point Background field value after round of iteration; For the first The grid point at the th grid point Background field value after round of iteration; This represents the total number of grid points on the 3D target mesh. For the first The observation point at the ... The observation increment in the round iteration; This represents the total number of observation points.

10. The data fusion method as described in claim 1, characterized in that, In step S6, a standardized format wind field data file is generated based on the fused 3D wind field grid data, specifically including: Horizontal interpolation: The fused 3D wind field grid data is interpolated from the current computation grid to the target projection grid through bilinear interpolation. The target projection grid is a 2D coordinate grid. Vertical interpolation: The fused 3D wind field grid data is interpolated from the current vertical level to the preset target vertical level through spline interpolation. The target vertical level is a set of multiple discrete height levels set at preset intervals. The U and V components in the three-dimensional wind field grid data after horizontal and vertical interpolation are converted into wind direction and wind speed, resulting in three-dimensional wind field grid data containing wind direction and wind speed values. Organize the three-dimensional wind field grid data containing wind direction and wind speed values ​​into a standardized wind field data file and output it.