A wind speed prediction method based on the relationship between spatial distance and altitude

Abstract

This invention belongs to the field of meteorological forecasting technology, and specifically relates to a wind speed prediction method based on the relationship between spatial distance and altitude. Traditional regional wind speed prediction only provides wind speed and latitude / longitude features. This invention, through wind speed feature processing, adds two new wind speed components—meridian wind speed and zonal wind speed—based on the collected wind direction data and the spatial geometric relationship between wind speed and direction. While interpolating the instantaneous wind speed, the two wind speed components are also interpolated. The wind speed is then calculated again using the mathematical relationship between the interpolated wind speed components. Since the accuracy of these two wind speeds varies under different conditions, assigning different weights to them according to different situations can significantly improve the overall accuracy.

Description

Technical Field

[0001] This invention belongs to the field of meteorological forecasting technology, and in particular relates to a wind speed forecasting method based on the relationship between spatial distance and altitude. Background Technology

[0002] In the power sector, wind speed prediction at key points along transmission lines is crucial for safe operation and provides a reference for line inspections. Wind speed prediction for transmission lines doesn't rely on historical data to forecast future wind speeds; instead, it predicts the wind speed at the predicted point at a given moment based on the surrounding wind conditions. Given the wide area covered by transmission lines and the limited number of meteorological monitoring stations, reliable wind speed prediction is essential for power line inspection departments to conduct timely and effective inspections, reducing costs. Currently, there are two main methods: universal Kriging interpolation and CressMan interpolation. Both algorithms are spatial interpolation algorithms that utilize a spatial grid of coordinate positions. They perform interpolation calculations based on spatial distance correlations to obtain gridded data, which is then compared with the coordinates of the predicted point to arrive at the interpolation result.

[0003] The core of the Kriging wind speed interpolation algorithm is a method for predicting regional wind speed using the Kriging algorithm. The process involves obtaining wind speed data from several points within a given area, inputting the latitude and longitude coordinates and instantaneous wind speed of each point. Based on the known latitude and longitude coordinates, the maximum and minimum values ​​of longitude and latitude are found. A 1° x 1° spatial grid is then formed based on these maximum and minimum values. According to spatial correlation, a covariance function is used to model the stochastic process / random field. The wind speed values ​​at the known coordinate locations are extrapolated to surrounding coordinate points. A linear, unbiased optimal estimate is then performed on the sample points, ultimately yielding the wind speed grid data for the entire region.

[0004] The core of the CressMan spatial interpolation algorithm is the interpolation operation performed using the Cressman algorithm. Specifically, it involves taking wind speed data from several points within a region, inputting the latitude and longitude coordinates of these points, and the instantaneous wind speed. Based on the range of longitude and latitude, a 1° x 1° spatial grid is formed. An optimization interpolation method is used with a successive correction approach: the initial field is modified and corrected using the actual data and the initial field to obtain a new field. The difference between the new field and the actual value is then calculated to correct the previous field, continuing until the new field closely approximates the actual value. Ultimately, this yields the wind speed grid data for the entire region.

[0005] Kriging wind speed interpolation considers spatial correlation and is considered one of the more accurate spatial interpolation algorithms. However, due to the complexity and variability of wind speed data, which is easily affected by terrain and altitude, Kriging interpolation fails to account for these factors, resulting in inaccurate wind speed predictions that only provide approximate values ​​and regional wind speed references. Cressman spatial interpolation is the most widely used spatial interpolation algorithm for meteorological data. Given the discrete nature and unique diversity and complexity of meteorological data distribution, this algorithm can provide relatively good fitting interpolation. However, because it uses progressive optimization and revision, it does not adequately consider spatial correlation. When the range value is not given appropriately, it often leads to significant deviations in the prediction results. Therefore, this algorithm can only provide regional wind speed references.

[0006] Therefore, the two existing methods can only make a rough estimate of wind speed over a large area, and cannot achieve an accurate estimate of wind speed over a small area. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention proposes a wind speed prediction method based on the relationship between spatial distance and altitude.

[0008] The technical solution of this invention is as follows:

[0009] A wind speed prediction method based on the relationship between spatial distance and altitude, characterized by the following steps:

[0010] S1. Obtain wind speed data features through sampling points, including instantaneous wind speed, instantaneous wind direction, longitude, latitude, and collection time;

[0011] S2. Expand the features of the acquired wind speed data by calculating the meridional wind component and the zonal wind component based on the instantaneous wind speed and instantaneous wind direction, and adding them as new features to the wind speed data features.

[0012] S3. Based on the location of the prediction point, the sampling data obtained from the sampling points around the prediction point are used as training data through the set sampling interval;

[0013] S4. Obtain the altitude of the prediction point and all sampling points, and correct the data obtained from each sampling point based on the altitude difference between the prediction point and each sampling point. Specifically, set a high-altitude threshold and a low-altitude threshold according to the altitude of the prediction point. If the altitude of the sampling point is less than the low-altitude threshold, increase the wind speed of the sampling point according to the first bias value; if the altitude of the sampling point is greater than the high-altitude threshold, decrease the wind speed of the sampling point according to the second bias value; if the altitude of the sampling point is between the high-altitude threshold and the low-altitude threshold, adjust the wind speed of the sampling point according to the altitude difference between the prediction point and the sampling point, and set a weight based on the distance relationship.

[0014] S5. Calculate the wind speed at the prediction point using the calibrated sampling data:

[0015] a) Process the sampling data into two independent sample data. The first sample data features: wind speed value, longitude value, latitude value; the second sample data features: meridional wind speed value, zonal wind speed value, longitude value, latitude value;

[0016] b) Both sample data are processed through the universal Kriging interpolation algorithm, and two predicted wind speed values (v1, v2) will be obtained at the prediction point. Threshold judgment is performed on these two wind speed values, specifically;

[0017] Calculate the difference between these two wind speed values, set a threshold, and compare the absolute value of the difference with the threshold. When the difference is greater than the threshold, the larger one of the two wind speed values will be removed, and the smaller one will be retained; when the difference is lower than the threshold, both wind speed values will be retained, and then their mean value will be calculated; finally, it will be used as the first predicted value (V1).

[0018] c) Similarly, both sample data are calculated through the Cressman spatial interpolation algorithm, and two predicted wind speed values (v3, v4) will also be obtained at the prediction point. Threshold judgment is also performed on the difference between these two wind speed values, and its final result will be used as the second predicted value (V2).

[0019] S6. Perform fusion calculation on the obtained first predicted value (V1) and second predicted value (V2) to obtain the final predicted value, specifically:

[0020] Take the radius parameter (r) of the Cressman algorithm as the screening condition and the parameter for weight calculation. The weight values are: ω1 = e r , ω2 = e -r , and then set the radius threshold r1, and perform weighted average calculation under different conditions. The specific calculation formula is as follows;

[0021] When r < r1:

[0022]

[0023] When r >= r1:

[0024]

[0025] The beneficial effect of the present invention is: The present invention can significantly improve the accuracy of wind speed prediction. Brief Description of the Drawings

[0026] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0027] Figure 2 This is a flowchart illustrating an embodiment. Detailed Implementation

[0028] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] like Figure 1 The diagram illustrates the overall process of this invention. Traditional regional wind speed prediction only provides wind speed and latitude / longitude features. This invention, through wind speed feature processing, adds two new wind speed components—meridian wind speed and zonal wind speed—based on the collected wind direction data and the spatial geometric relationship between wind speed and direction. While interpolating the instantaneous wind speed, the two wind speed components are also interpolated. The wind speed is then calculated again using the mathematical relationship between the interpolated wind speed components. Since the accuracy of these two wind speeds varies under different conditions, assigning different weights to them based on different situations can significantly improve the overall accuracy.

[0030] The present invention also incorporates the altitude of the actual location point, introducing the influence of altitude on wind speed changes. The correction for wind speed based on altitude is combined with the spatial distance between wind speed points, and is divided into three cases;

[0031] The first method: If the target point's altitude is at the highest point within the selected range, then the wind speed at surrounding points will be weighted, correspondingly increasing the wind speed values ​​at all points within the range. The weighting is calculated using the formula:

[0032] ω=e αθ

[0033] Where θ represents the points within the range sorted in descending order based on altitude, and the wind speed is calculated as a function of altitude difference; α is a manually added correction value.

[0034] In the second scenario, if the altitude of the target point is the lowest point within the selected range, the wind speed values ​​of surrounding points will be weighted, and the wind speed values ​​of all points within the range will be reduced. The weighting calculation is the same as in the first method.

[0035] In the third case, if the altitude of the target point is within the range of the altitude values ​​of the sample points selected, a distance correction value will be added to the weight calculation. That is, in addition to considering the influence of altitude, the distance relationship between the sample point and the target point will also be considered, and a distance threshold will be set. Sample points with a distance less than the threshold will have a greater weight.

[0036] Finally, the solution of this invention obtains the final predicted value by fusing the interpolation results of the generalized Kriging interpolation algorithm and the CressMan spatial interpolation algorithm.

[0037] Since the algorithm interpolates wind speed into two values, it first compares these two values ​​by setting a difference threshold. If the difference exceeds this threshold, the maximum value is discarded, and the smaller wind speed value is used as the predicted value. If the difference is below the threshold, the average of the two wind speeds is calculated as the predicted value.

[0038] The next stage of processing involves setting filtering conditions and weighting parameters for the Cressman algorithm's range radius. First, a radius threshold is set, derived from the spatial distance distribution of actual sample points. Different weights are assigned to the two wind speed values ​​under different threshold conditions. Finally, a weighted average is used to obtain the final wind speed value.

[0039] This invention innovatively integrates the generalized Kriging interpolation algorithm and the Cressman spatial interpolation algorithm, and innovatively incorporates the influence of wind speed point altitude into the algorithm. This not only reduces the interference of random wind speed fluctuations on prediction, but also considers the spatial distance correlation of wind speed locations in conjunction with altitude. This significantly improves the accuracy of wind speed interpolation, providing more comprehensive wind speed data for transmission line inspection.

[0040] Example

[0041] like Figure 2 As shown, this example specifically includes:

[0042] Step 1: Collect wind speed data features, including instantaneous wind speed, instantaneous wind direction, longitude, latitude, and collection time.

[0043] Step 2: Calculate the meridional and zonal wind components from the instantaneous wind speed and instantaneous wind direction features in the wind speed data using the formulas (μ=-ωsinα, v=-ωcosα). Add these two components as new wind speed features to the overall wind speed data.

[0044] Step 3: Based on the location of the predicted point, select training sample points near the predicted target point at intervals of K, using latitude and longitude. A threshold is set to determine whether the number of data points within the range reaches the threshold.

[0045] The data input features are: longitude, latitude, instantaneous wind direction, meridional wind component, zonal wind component, and instantaneous wind speed.

[0046] Step 4: Read the altitude of the sample points and test points based on their latitude and longitude coordinates, and record the altitude value of each coordinate point.

[0047] Step 5: According to the first theorem of geography: everything is related, and things that are close together are more closely related. Therefore, taking each test sample point as the center, training sample points near the target point are selected at intervals of K, using latitude and longitude. A threshold is set to determine whether the number of data points within the range reaches the threshold.

[0048] Step 6: Determine whether the number of sample points within the selected range reaches the set threshold. If it is greater than the threshold, only the sample points within the selected range will be used for interpolation calculation. If it is less than the threshold, the entire dataset at the same time will be used as sample points for interpolation calculation.

[0049] Step 7: After selecting values ​​in Step 6, the altitude values ​​recorded in Step 4 are also input, and different weights are assigned to the sample points. There are three cases: If the altitude of the test point is the highest altitude within the selected range, a bias value will be artificially added to the wind speed value of the selected sample point to increase the wind speed at the test point; similarly, if the altitude of the test point is the lowest altitude within the selected range, a bias value will be artificially added to decrease the wind speed at the test point; if the test point is within the altitude range of the sample points within the selected range, the height difference between the altitude of the sample point and the altitude of the test point will be calculated, and combined with the distance relationship, different weights will be assigned to the wind speed values ​​of the sample points within the selected range.

[0050] Step 8: Universal Kriging modeling. The original wind speed data has a longitude range of (100.3614, 103.7928) and a latitude range of (26.2247, 29.2497), with relatively small overall intervals. When establishing grid data points, an n x n grid is formed with an interval of n. Furthermore, the actual spatial distance between adjacent grid points in the n-interval grid is approximately within 1 km, which will result in a strong correlation between the wind speeds of adjacent points.

[0051] Step 9: Substitute the output data from Step 7 into the grid established in Step 8, and use the universal kriging algorithm to perform interpolation calculations on the grid data. The spline parameter (nlag) in the universal kriging function is set to a range value, automatically selecting an appropriate parameter value. Since the grid data is established using latitude and longitude, the coordinate system used in the universal kriging algorithm is a spherical coordinate system.

[0052] Step 10: After obtaining the grid point data for the entire region through interpolation, the grid data format is as follows: a grid is constructed using longitude and latitude values ​​as the horizontal and vertical coordinates, where the data for each grid point is the interpolated data. Interpolation will generate three grids: wind speed grid data, meridional wind speed grid data, and zonal wind speed grid data.

[0053] Step 11: Based on the latitude and longitude data of the test point, find the latitude and longitude of the test point or the latitude and longitude of the grid point closest to the test point in the grid data, and take the corresponding interpolation value from the three grids in turn. The test point will have three values, namely: instantaneous wind speed, meridional wind speed, and zonal wind speed.

[0054] Step 12: Apply the formula to the meridional and zonal wind speeds calculated by the pan-kriging algorithm. The wind speed value is calculated, and the instantaneous wind speed value and the calculated wind speed value are saved during the process for subsequent calculations.

[0055] Step 13: CressMan spatial interpolation algorithm modeling. The input data for this model is still the data from Step 7. The difference is that when building the grid point data, the grid interval is m, that is, the final grid is m x m. This interval is beneficial for the selection of radius values ​​in the algorithm.

[0056] Step 14: Substitute the output data from Step 7 into the grid coordinates of Step 13, and use the Cressman algorithm to interpolate the instantaneous wind speed, meridional wind speed, and zonal wind speed. During this process, the radius parameter of the algorithm is set to a range of values, the optimal parameter is automatically selected, and the optimal radius parameter value is recorded for weight reference in subsequent wind speed calculations.

[0057] Step 15: After interpolation, three grids are still generated. The grid format remains longitude on the horizontal axis and latitude on the vertical axis. The value of each grid point is the wind speed interpolation. The three grids are: wind speed grid data, meridional wind speed grid data, and zonal wind speed grid data.

[0058] Step 16: Perform the operation in Step 11, and retrieve the wind speed values ​​at the same test points after interpolation using the Cressman algorithm. In addition to obtaining the instantaneous wind speed values, the same formula is still used. Wind speed is calculated using radial and zonal wind speeds. Both values ​​are then recorded.

[0059] Step 17: Calculate a weighted average of the four wind speed values ​​obtained from the two algorithms to arrive at the final wind speed prediction. The weights for the weighted average are derived from the radius in the Cressman algorithm. Furthermore, this part will select different calculation methods based on the difference between the two wind speed values ​​obtained from the pan-kriging algorithm. This part is crucial for obtaining a more accurate wind speed value, and is explained in detail below:

[0060] First, since both of the above algorithms are spatial interpolation algorithms, they rely on the wind speed values ​​of sample points around the test point and include latitude and longitude wind speed components. Therefore, there may be extreme cases in this part of the calculation. We will first determine the difference between the directly predicted wind speed value and the predicted value of the wind speed component, take a reasonable wind speed value, discard extreme values, and then calculate the wind speed value by averaging.

[0061] Secondly, the radius parameter of the Cressman algorithm is chosen to calculate the weights. This is because the Cressman algorithm itself performs optimal unbiased estimation for grid points within a specified range. Since the distribution and number of wind speed points near different test points vary, the algorithm uses the test point's coordinates as the center point and selects wind speed points within a specified radius for optimal unbiased estimation. Given the spatial correlation of wind speeds, the size of the radius directly affects the calculation of wind speed interpolation; therefore, the radius value is included in the wind speed weight calculation.

[0062] Finally, since both algorithms predict the same test point and use the same input sample data, this radius value has the same reference value for the wind speed value of the pan-kriging algorithm. Therefore, a threshold is set for the radius value here, and different weights are assigned to the wind speed values ​​of the two algorithms according to the relationship between the radius value and the threshold.

Claims

1. A wind speed prediction method based on the relationship between spatial distance and altitude, characterized in that, It includes the following steps: S1. Obtain the wind speed data characteristics through sampling points, including instantaneous wind speed, instantaneous wind direction, longitude, latitude, and collection time; S2. Expand the obtained wind speed data characteristics. Specifically, calculate the meridional wind component and zonal wind component based on the instantaneous wind speed and instantaneous wind direction, and supplement them as new features to the wind speed data characteristics; S3. Based on the position of the prediction point, use the set sampling interval to obtain the sampling data of the sampling points around the prediction point as training data; S4. Obtain the altitude of the prediction point and all sampling points, and based on the altitude difference between the prediction point and each sampling point, correct the sampling data obtained from each sampling point. Specifically: Set a high altitude threshold and a low altitude threshold according to the altitude of the prediction point. If the altitude of the sampling point is less than the low altitude threshold, increase the wind speed of the sampling point according to the first offset value; If the altitude of the sampling point is greater than the high altitude threshold, reduce the wind speed of the sampling point according to the second offset value; If the altitude of the sampling point is between the high altitude threshold and the low altitude threshold, set a weight according to the height difference between the prediction point and the sampling point and the distance relationship to adjust the wind speed of the sampling point; S5. Calculate the wind speed of the prediction point using the corrected sampling data: a) Process the sampling data into two independent sample data. The characteristics of the first sample data are: wind speed value, longitude value, latitude value; The characteristics of the second sample data are: meridional wind speed value, zonal wind speed value, longitude value, latitude value; b) Pass both sample data through the universal kriging interpolation algorithm to obtain two predicted wind speed values (v1, v2), and perform a threshold judgment on these two wind speed values. Specifically: Calculate the difference between these two wind speed values, set a threshold, and compare the absolute value of the difference with the threshold. When the difference is greater than the threshold, remove the larger one from the two wind speed values and retain the smaller one as the first predicted value; When the difference is lower than the threshold, retain both wind speed values and then calculate their average value as the first predicted value V1; c) Calculate both sample data through the Cressman spatial interpolation algorithm to obtain two predicted wind speed values (v3, v4), and use the same threshold judgment method as in step b to obtain the second predicted value V2; S6. Perform a fusion calculation on the obtained first predicted value V1 and second predicted value V2 to obtain the final predicted value. Specifically: When r < r1: Using the radius parameter r of the CressMan algorithm as the filtering condition and weight calculation parameter, the weight value is: ω1 = e r ,ω2=e -r Then, a radius threshold r1 is set, and a weighted average is calculated under different conditions. The specific calculation formula is as follows: When r >= r1: Where speed is the obtained final predicted value. ​

Images