Wind field quality control method based on single moment image structure
By identifying wind field anomalies through curve wave analysis and gradient thresholding, this method solves the problem that existing wind field quality control methods cannot adapt to the spatiotemporal continuity of wind fields, and achieves accurate quality control of wind fields at a single moment, especially in the effective identification and removal of abnormal data in strong rotating wind fields.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF INFORMATION SCI & TECH
- Filing Date
- 2025-11-06
- Publication Date
- 2026-04-10
AI Technical Summary
Existing wind field quality control methods are mostly based on statistical analysis of long-term data, which cannot be applied to the spatiotemporal continuity characteristics of wind fields. Furthermore, in strong rotating wind fields, they are prone to misjudging the true wind field in the core area of a typhoon as an anomaly, resulting in information loss.
The curve wave analysis method is used to decompose the wind field data into two parts: the main term and the residual term. Abnormal data are identified by gradient threshold and vorticity threshold. Combined with progressive quality control and interpolation processing, outliers are identified and removed.
It achieves accurate quality control of wind fields at a single moment, can identify and eliminate abnormal data, retain the typhoon's rotational structure, is suitable for strong rotating wind fields, and reduces misjudgments.
Smart Images
Figure CN121831968A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind speed prediction technology, and more specifically to a wind field quality control method based on a single-moment image structure. Background Technology
[0002] Research on quality control methods for automatic weather stations is constantly progressing, but existing methods are mostly based on statistical analysis of data over a certain time period to obtain the regularity characteristics of variables. For example, they utilize long-term data to statistically analyze extreme climate values and average 3-hour temperature variations, or employ complex statistical methods, such as EOF analysis, to extract continuity features. However, the spatiotemporal continuity of wind fields is far lower than that of variables like temperature and air pressure. This means that existing quality control methods based on strong spatiotemporal continuity are not suitable for wind field data. Furthermore, in strong rotating wind fields, single-component wind speeds may exhibit drastic spatial variations. Quality control based solely on wind speed may misjudge some of the true wind field in the typhoon core area as outliers, leading to the partial loss of typhoon structural information. Summary of the Invention
[0003] Purpose of the invention: The purpose of this invention is to provide a wind field quality control method based on a single-moment image structure, which utilizes the multidirectional characteristics of curve wave analysis to solve the problems existing in the background technology.
[0004] Technical solution: The wind field quality control method based on a single-moment image structure described in this invention includes the following steps: (1) Convert the original wind direction and wind speed data into U and V wind components, and use the nearest neighbor method combined with bilinear interpolation to interpolate the U and V wind components and wind direction station observation data into regular grid data; (2) The curve decomposition method is used to decompose the gridded wind field data into two parts: the main term containing the main structural information and the residual term which is mainly random. (3) For U and V wind speed and wind direction data, the main data after curve wave analysis is used to calculate the spatial gradient of the target grid point, and an anomaly detection is performed using a progressive gradient threshold sequence. That is, gradient thresholds of 2.5, 2.0, 1.5 and 1.0 are applied to U and V components in sequence, and gradient thresholds of 180°, 135°, 90° and 70° are applied to wind direction in sequence. Grid points with gradients greater than the current threshold sequence value are judged as anomalies. (4) Repeat steps (2)-(3) to perform multiple progressive quality control checks in order to identify discontinuous outliers in the main items; (5) For U and V wind speed and wind direction data, after completing the main term curve analysis, perform curve decomposition to obtain the residual term data, set the abnormal threshold based on the standard deviation of the residual term data, and mark the grid points that exceed the threshold as error values. (6) Based on the vorticity value of each station, the wind field area is divided into a weak rotating wind area and a strong rotating wind area. In the weak rotating wind area, if either the U or V component is marked as abnormal in the previous quality control, the wind field data at that point is determined to be incorrect. In the strong rotating wind area, the U or V component and the wind direction need to be marked as abnormal at the same time before the wind field data is determined to be incorrect.
[0005] Furthermore, in step (2), the curve decomposition method is as follows: the wind field data is decomposed into a series of curve coefficients by using curve basis functions in multiple scales and directions, and the wind field data is decomposed into a series of curve coefficients by using scale parameters and control direction parameters. Then, the coefficients are separated into principal coefficients and residual coefficients based on a predetermined modulus threshold, and the variable field represented by the principal and residual terms is reconstructed by inverse curve transformation.
[0006] Furthermore, the modulus threshold is 2.0, where the principal term coefficient corresponds to the large-scale spatial structure of the wind field, and the residual term coefficient corresponds to random disturbances.
[0007] Furthermore, in step (3), the progressive gradient threshold sequence is a decreasing sequence, used to gradually eliminate non-continuous outliers.
[0008] Furthermore, in step (5), the abnormal threshold is calculated based on the standard deviation of the remaining data, specifically three times the standard deviation.
[0009] Furthermore, in step (6), the vorticity threshold is 10. -5 / s is used to distinguish between weak and strong rotating wind zones.
[0010] The wind field quality control system based on a single-moment image structure as described in this invention includes: Interpolation module: Used to convert raw wind direction and wind speed data into U and V wind components, and uses a combination of nearest neighbor method and bilinear interpolation to interpolate U, V wind components and wind direction station observation data into regular grid data; Curve decomposition module: Used to decompose gridded wind field data into two parts: a main term containing the main structural information and a residual term mainly consisting of random variations, using the curve decomposition method. The calculation module is used to calculate the spatial gradient of the target grid points for U and V wind speed and direction data using the principal data after curve wave analysis, and to perform anomaly detection using a progressive gradient threshold sequence: gradient thresholds of 2.5, 2.0, 1.5 and 1.0 are applied to the U and V components respectively, and gradient thresholds of 180°, 135°, 90° and 70° are applied to the wind direction respectively. Grid points with gradients greater than the current threshold sequence value are judged as anomalies. Quality control module: Used to perform multiple progressive quality controls to identify discontinuous outliers in the main item; The residual module is used to perform curve decomposition to obtain residual data after completing the main curve analysis for U and V wind speed and direction data. It sets an anomaly threshold based on the standard deviation of the residual data and marks grid points that exceed the threshold as error values. Judgment module: Based on the vorticity value of each station, the wind field area is divided into weak rotating wind area and strong rotating wind area; in the weak rotating wind area, if either the U or V component is marked as abnormal in the previous quality control, the wind field data at that point is judged to be incorrect; in the strong rotating wind area, both the U or V component and the wind direction need to be marked as abnormal before the wind field data is judged to be incorrect.
[0011] An electronic device according to the present invention includes a memory and a processor. The memory stores a computer program, and the processor executes the program to implement the steps of the method.
[0012] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method.
[0013] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention introduces the curve wave analysis method to decompose the wind field at a single moment into a main term containing spatial continuous change characteristics at different scales and a residual term mainly characterized by random change. Based on the performance characteristics of extreme and general erroneous data in different components, the erroneous data is accurately identified. Finally, a quality control method that can be applied to wind fields at a single moment and is not easily affected by high-intensity weather systems such as typhoons is established. Attached Figure Description
[0014] Figure 1 This is a flowchart of the present invention;
[0015] Figure 2 This is a map showing the distribution of site data in the research area of this invention;
[0016] Figure 3 This is a schematic diagram of multiple main item quality control of the U-wind component of the present invention; wherein Figure 3 (a) in the diagram is a schematic diagram of the first main item quality control of the U-wind component; Figure 3 (b) in the diagram is a schematic diagram of the second main item quality control of the U-wind component; Figure 3 (c) in the diagram is the third main item quality control diagram of the U-wind component; Figure 3 (d) in the figure is a schematic diagram of the fourth main item quality control of the U-wind component;
[0017] Figure 4 This is a schematic diagram of the multiple main quality control of wind direction in this invention; wherein Figure 4 (a) in the diagram is a schematic diagram of the first main quality control of wind direction; Figure 4 (b) in the diagram is a schematic diagram of the second main quality control of wind direction; Figure 4(c) in the diagram is a schematic diagram of the third main quality control item for wind direction; Figure 4 (d) in the diagram is the fourth main quality control diagram for wind direction;
[0018] Figure 5 These are schematic diagrams before and after the quality control of the U-wind component residual term of the present invention; wherein Figure 5 (a) in the diagram is a schematic diagram of the U-wind component residual term before quality control; Figure 5 (b) in the diagram is a schematic diagram after quality control of the U-wind component residual term;
[0019] Figure 6 This is a schematic diagram of the wind field after comprehensive quality control according to the present invention. Detailed Implementation
[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0021] like Figure 1 As shown, this embodiment of the invention provides a wind field quality control method based on a single-moment image structure, comprising the following steps:
[0022] Step 1: Preprocessing of observation data: Convert the raw wind direction and wind speed data into U and V wind components, and interpolate the U and V wind components and wind direction station observation data into regular grid data;
[0023] Step 2: Using the curve decomposition method, the gridded data is decomposed into two parts: the main term containing the main structural information and the residual term which is mainly characterized by random variation.
[0024] Step 3: Quality Control of Main Items: For U and V wind speed and direction data, anomaly detection is performed on the spatial gradient of the target grid point values using the main item data after curve wave analysis. Grid point values with gradients greater than a threshold are judged as anomalies.
[0025] Step 4: Repeat steps 2 and 3, and through four progressive quality control processes, ensure that discontinuous outliers in the main item are effectively identified.
[0026] Step 5, Quality control of remaining items: For U and V wind speed and direction data, after completing the main item curve analysis, perform the fifth curve decomposition on the data and take the remaining items for quality control: calculate three times the standard deviation, and grid points exceeding this threshold are marked as error values.
[0027] Step 6, Comprehensive Quality Control: Using the vorticity of each station as the judgment standard, when the vorticity is less than 10... -5 A wind speed of / s is considered a weak or non-rotating wind region. Any component of U or V within this region is marked as abnormal during previous quality control, indicating an error in the wind field data at that point. For regions with rotating winds (vorticity ≥ 10...), ... -5 For wind field data to be considered erroneous, the U or V component and the wind direction must be marked as abnormal simultaneously.
[0028] Firstly, since the algorithm requires input data to be regular grid data, this invention selects an appropriate grid spacing based on the station density of the analysis area and then uses spatial interpolation to convert discrete station observation data into regular grid data. This interpolation method ensures that each grid point corresponds to only one station, thus facilitating accurate location of erroneous stations during quality control. During interpolation, the nearest neighbor method is first used to allocate all station data to grid points, and then bilinear interpolation is used to fill in the remaining missing grid points. It should be noted that the interpolated data is only used for curvelet decomposition and does not participate in subsequent quality control; therefore, the interpolation process does not affect the quality control results.
[0029] Secondly, the key to this invention lies in ensuring that the principal terms after curvelet decomposition fully retain the large-scale spatial structure while effectively reflecting the characteristics of random disturbances in the residual terms. Therefore, the selection of the decomposition threshold is crucial. Based on relevant statistical analysis and taking into account the differences in wind speed magnitude, this scheme uses 2.0 as the decomposition threshold for the curvelet coefficient.
[0030] Furthermore, the errors in the main data are primarily characterized by spatial discontinuities, and spatial gradient is a crucial indicator for measuring these discontinuities. To reduce the mutual influence between extreme values, a progressive quality control approach is adopted for the main data: multiple outlier removals are performed by setting a decreasing gradient threshold sequence. Specifically, the U and V components undergo four quality control iterations using gradient thresholds of 2.5, 2.0, 1.5, and 1.0, respectively, while wind direction is controlled at 180°, 135°, 90°, and 70°.
[0031] This invention utilizes curvelet decomposition to decompose instantaneous wind fields into components of different scales. By setting a threshold, the wind field is divided into a large-scale structural main term with strong continuity and a residual term dominated by random variation. Based on the large-scale continuity characteristic, abnormal data in the main term can be effectively identified; for the residual term, outliers are detected from the perspective of probability distribution based on the 3σ principle of random variables.
[0032] The specific experiment is as follows:
[0033] This embodiment uses automatic weather station wind field observation data from the nearshore area of eastern China (117°E~122°E, 28°N~33°N) in September 2024 as an example. Figure 2 This section introduces the specific implementation of this method. The region has a high density of stations and is easily affected by strong weather such as sea and land breezes and typhoons, making it more representative.
[0034] S1: Convert the original wind direction and wind speed data into U and V wind components, and interpolate the U and V wind components and wind direction station observation data onto a regular grid with a grid spacing of 0.04°.
[0035] S2: Using the curve decomposition method, the gridded U, V wind and wind direction data are decomposed into two parts: a main term containing the main structural information and a residual term mainly consisting of random variations.
[0036] S3: Quality control of the main items of U, V wind and wind direction data.
[0037] S4: Repeat steps S2 and S3. Perform four quality control iterations on the U and V components using gradient thresholds of 2.5, 2.0, 1.5, and 1.0 respectively, while for wind direction, use thresholds of 180°, 135°, 90°, and 70° respectively. The U-wind component at 12:00 on September 16, 2024 is given here. Figure 3 ) and wind direction ( Figure 4 Taking the decomposition results of the main item as an example, it can be seen that as quality control progresses, the number of singularities in the main item reconstruction results gradually decreases, exhibiting a smoother spatial structure.
[0038] S5: Quality control of residual items: For U and V wind speed and direction data, after completing the main item curve analysis, the data is decomposed into a fifth curve and the residual items are used for quality control: calculate three times the standard deviation, and grid points exceeding this threshold are marked as error values. Figure 5 The spatial distribution of the U-wind field before and after quality control is given. The figure shows that the erroneous data identified are randomly distributed.
[0039] S6: Comprehensive Quality Control: Using the vorticity of each station as the judgment standard, when the vorticity is less than 10... -5 A wind speed of / s is considered a weak or non-rotating wind region. Any component of U or V within this region is marked as abnormal during previous quality control, indicating an error in the wind field data at that point. For regions with rotating winds (vorticity ≥ 10...), ... -5 For wind field data to be considered erroneous, the U or V component and the wind direction must be marked as abnormal simultaneously. Figure 6 The quality control results of 10-meter wind station data within the range of 119~122°E and 30~33°N, after comprehensive quality control, are presented. For example... Figure 6 As shown, this invention effectively identifies disordered and abnormal data such as wind speed or wind direction mismatch with the surrounding wind field while completely preserving the typhoon's rotating structure.
Claims
1. A wind field quality control method based on a single-moment image structure, characterized in that, Includes the following steps: (1) Convert the original wind direction and wind speed data into U and V wind components, and use the nearest neighbor method combined with bilinear interpolation to interpolate the U and V wind components and wind direction station observation data into regular grid data; (2) The curve decomposition method is used to decompose the gridded wind field data into two parts: the main term containing the main structural information and the residual term which is mainly random. (3) For U and V wind speed and wind direction data, the spatial gradient of the target grid points is calculated using the main data after curve analysis, and anomaly detection is performed using a progressive gradient threshold sequence: gradient thresholds of 2.5, 2.0, 1.5 and 1.0 are applied to U and V components respectively, and gradient thresholds of 180°, 135°, 90° and 70° are applied to wind direction respectively. Grid points with gradients greater than the current threshold sequence value are judged as anomalies. (4) Repeat steps (2)-(3) to perform multiple progressive quality control checks in order to identify discontinuous outliers in the main items; (5) For U and V wind speed and wind direction data, after completing the main term curve analysis, perform curve decomposition to obtain the residual term data, set the abnormal threshold based on the standard deviation of the residual term data, and mark the grid points that exceed the threshold as error values. (6) Based on the vorticity value of each station, the wind field area is divided into a weak rotating wind area and a strong rotating wind area. In the weak rotating wind area, if either the U or V component is marked as abnormal in the previous quality control, the wind field data is judged to be incorrect. In the strong rotating wind area, the U or V component and the wind direction need to be marked as abnormal at the same time before the wind field data is judged to be incorrect.
2. The wind field quality control method based on a single-moment image structure according to claim 1, characterized in that, In step (2), the curve wave decomposition method is as follows: the wind field data is decomposed into a series of curve wave coefficients by using curve wave basis functions in multiple scales and directions. Then, based on the predetermined modulus threshold, the coefficients are separated into principal coefficients and residual coefficients, and the variable field represented by the principal and residual terms is reconstructed by inverse curve wave transformation.
3. The wind field quality control method based on a single-moment image structure according to claim 2, characterized in that, The modulus threshold is 2.0, where the principal coefficient corresponds to the large-scale spatial structure of the wind field, and the residual coefficient corresponds to random disturbances.
4. The wind field quality control method based on a single-moment image structure according to claim 1, characterized in that, In step (3), the progressive gradient threshold sequence is a decreasing sequence used to gradually eliminate non-continuous outliers.
5. The wind field quality control method based on a single-moment image structure according to claim 1, characterized in that, In step (5), the abnormal threshold is calculated based on the standard deviation of the remaining data, specifically three times the standard deviation.
6. The wind field quality control method based on a single-moment image structure according to claim 1, characterized in that, In step (6), the vorticity threshold is 10. -5 / s is used to distinguish between weak and strong rotating wind zones.
7. A wind field quality control system based on a single-moment image structure, characterized in that, include: Interpolation module: Used to convert raw wind direction and wind speed data into U and V wind components, and uses a combination of nearest neighbor method and bilinear interpolation to interpolate U, V wind components and wind direction station observation data into regular grid data; Curve decomposition module: Used to decompose gridded wind field data into two parts: a main term containing the main structural information and a residual term mainly consisting of random variations, using the curve decomposition method. The calculation module is used to calculate the spatial gradient of the target grid points for U and V wind speed and direction data using the principal data after curve wave analysis, and to perform anomaly detection using a progressive gradient threshold sequence: gradient thresholds of 2.5, 2.0, 1.5 and 1.0 are applied to the U and V components respectively, and gradient thresholds of 180°, 135°, 90° and 70° are applied to the wind direction respectively. Grid points with gradients greater than the current threshold sequence value are judged as anomalies. Quality control module: Used to perform multiple progressive quality controls to identify discontinuous outliers in the main item; The residual module is used to perform curve decomposition to obtain residual data after completing the main curve analysis for U and V wind speed and direction data. It sets an anomaly threshold based on the standard deviation of the residual data and marks grid points that exceed the threshold as error values. Judgment module: Based on the vorticity value of each station, the wind field area is divided into weak rotating wind area and strong rotating wind area; in the weak rotating wind area, if either the U or V component is marked as abnormal in the previous quality control, the wind field data is judged to be incorrect; in the strong rotating wind area, the U or V component and the wind direction need to be marked as abnormal at the same time before the wind field data is judged to be incorrect.
8. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the program to implement the steps of the method according to claims 1-6.
9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the method described in claims 1-6.