A meteorological prediction method and system based on complex terrain wind field data

Abstract

The present invention discloses a meteorological prediction method and system based on complex terrain wind field data, which relates to the technical field of meteorological forecasting. The method includes obtaining wind measurement data, tracing the origin of the wind measurement data, obtaining the location information of the wind measurement station, obtaining regional geographical data, and performing data processing on the wind measurement data according to the regional geographical data and the location information of the wind measurement station to obtain wind measurement processed data. By removing abnormal data within the data time period of the wind measurement station, the reliability and accuracy of the data are ensured. By using the wind measurement station location index, wind measurement stations that are too concentrated are removed, ensuring a uniform distribution of the wind measurement station locations. By introducing the error of the wind measurement station into the loss function of the neural network model for the low-altitude wind field modeling, the distribution of the downscaled low-altitude wind field at the wind measurement station becomes more reasonable. By using the wind speed correction coefficient, the accuracy of meteorological prediction is ensured, and the deviation impact of complex terrain on the meteorological prediction results is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological forecasting, and specifically relates to a meteorological prediction method and system based on complex terrain wind field data. Background Art

[0002] As an important part of the natural environment, climate affects human life in all aspects, from personal life to agriculture, economy, and military. By measuring wind speed, temperature, and humidity to forecast the weather, it is convenient to arrange activities and work reasonably. Meteorological forecasting plays an indispensable role in modern society. By providing accurate weather information, it helps people make reasonable decisions, thereby improving the quality of life and ensuring the stable development of social economy.

[0003] At present, for meteorological prediction of complex terrain, there are still problems such as low output resolution of meteorological prediction, lack of detailed regional information, still relatively low forecasting accuracy for regions, still difficult to effectively characterize and describe the inherent complexity and uncertainty of weather systems at the present stage, still relatively large deviations in the output variables of dynamic downscaling models, still unable to meet the actual applications at small and medium scales, especially the impact assessment at the station scale, and the dynamic downscaling method takes a long time and consumes extremely high computing resources. Summary of the Invention

[0004] To solve the above technical problems, a meteorological prediction method and system based on complex terrain wind field data are provided. This technical solution solves the problems of low output resolution of meteorological prediction, lack of detailed regional information, still relatively low forecasting accuracy for regions, still difficult to effectively characterize and describe the inherent complexity and uncertainty of weather systems at the present stage, still relatively large deviations in the output variables of dynamic downscaling models, still unable to meet the actual applications at small and medium scales, especially the impact assessment at the station scale, and the dynamic downscaling method takes a long time and consumes extremely high computing resources as mentioned in the above background art.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A meteorological prediction method based on complex terrain wind field data, comprising:

[0007] Obtain wind measurement data, where the wind measurement data includes regional wind direction information and regional wind speed data;

[0008] Trace the source of the wind measurement data to obtain the location information of the wind measurement station;

[0009] Obtain regional geographical data, where the regional geographical data includes regional altitude information and regional terrain feature information;

[0010] According to the regional geographical data and the location information of the wind measurement stations, the wind measurement data is processed to obtain the processed wind measurement data;

[0011] Based on the WRF regional meteorological model, dynamic downscaling is performed on the processed wind measurement data to obtain the basic data of the downscaled wind field;

[0012] Analyze the basic data of the downscaled wind field and the processed wind measurement data to determine whether the basic data of the downscaled wind field meets the standards. If not, re-perform dynamic downscaling on the processed wind measurement data. If so, obtain the wind measurement characteristic data according to the processed wind measurement data;

[0013] Input the wind measurement characteristic data into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model;

[0014] Based on the low-altitude wind field downscaling model, perform downscaling on the processed wind measurement data to obtain the low-altitude wind field data;

[0015] According to the low-altitude wind field data and the regional geographical data, obtain the meteorological prediction information, where the meteorological prediction information includes wind speed information and wind direction information.

[0016] Preferably, the processing of the wind measurement data according to the regional geographical data and the location information of the wind measurement stations to obtain the processed wind measurement data specifically includes:

[0017] According to the regional geographical data and the location information of the wind measurement stations, obtain the geographical influence information of the wind measurement stations, where the geographical influence information of the wind measurement stations includes the geographical influence area information of the wind measurement stations and the geographical data within this area;

[0018] Obtain the historical data of the wind measurement stations, where the historical data of the wind measurement stations includes the wind speed data, wind direction information, and friction velocity measured by this wind measurement station;

[0019] According to the historical data of the wind measurement stations, obtain the surface roughness of the wind measurement stations;

[0020] Based on the geographical influence information of the wind measurement stations, obtain the standard surface roughness of the wind measurement stations;

[0021] According to the surface roughness of the wind measurement stations and the standard surface roughness of the wind measurement stations, obtain the data time period of the wind measurement stations;

[0022] Analyze the wind speed within each data time period of each wind measurement station. If the wind speed is abnormal, remove the wind measurement data for this data time period of this wind measurement station;

[0023] Take the wind measurement station with the smallest surface roughness of the wind measurement stations as the initial wind measurement station;

[0024] Based on the initial anemometry station, classify the anemometry stations according to the anemometry station location information to obtain the anemometry station classification information, where the anemometry station classification information includes the initial anemometry station, the basic anemometry station, and the auxiliary anemometry station;

[0025] Evaluate the auxiliary anemometry stations to obtain the anemometry station location index;

[0026] Filter the auxiliary anemometry stations according to the anemometry station location index;

[0027] Obtain the anemometry processing data according to the anemometry data of the filtered auxiliary anemometry stations, the initial anemometry station, and the basic anemometry station;

[0028] Among them, the calculation formula for the surface roughness of the anemometry station is:

[0029]

[0030] In the formula, Z0 is the surface roughness of the anemometry station, U(x) is the wind speed of the x-th anemometry station, U * is the friction velocity, h x is the vertical distance between the wind speed measurement position of the x-th anemometry station and the ground, and k is the von Kármán constant and k = 0.4.

[0031] Preferably, the filtering of the auxiliary anemometry stations according to the anemometry station location index specifically includes:

[0032] Based on the initial anemometry station, connect the anemometry stations until the measurable area of the anemometry stations covers the entire anemometry demand area, and obtain the anemometry station distribution plan information;

[0033] According to the anemometry station distribution plan information, select the anemometry station distribution plan with the smallest number of connected anemometry stations to obtain the basic anemometry station distribution information;

[0034] Among them, the basic anemometry station distribution information includes the initial anemometry station information and the basic anemometry station information;

[0035] Classify the anemometry stations according to the basic anemometry station distribution information to obtain the anemometry station classification information;

[0036] Evaluate the auxiliary anemometry stations according to the anemometry station classification information to obtain the anemometry station location index;

[0037] Based on the actual measurement requirements of the regional wind field, obtain the anemometry station location index threshold;

[0038] According to the anemometry station location index and the anemometry station location index threshold, it is judged whether the auxiliary anemometry station meets the measurement requirements. If the anemometry station location index is lower than the anemometry station location index threshold, the auxiliary anemometry station does not meet the measurement requirements, and the anemometry data of the auxiliary anemometry station is removed. If the anemometry station location index is higher than the anemometry station location index threshold, the auxiliary anemometry station meets the measurement requirements;

[0039] Among them, the calculation formula of the anemometry station location index is:

[0040]

[0041] In the formula, Q(x) is the anemometry station location index, Z0(g) is the anemometry station surface roughness of the gth basic anemometry station overlapping with the measurable area of the xth auxiliary anemometry station, S g is the overlapping area of the measurable area of the xth auxiliary anemometry station and the measurable area of the gth basic anemometry station, and n is the total number of basic anemometry stations overlapping with the measurable area of the xth auxiliary anemometry station.

[0042] Preferably, analyzing the downscaled wind field basic data and the anemometry processed data to judge whether the downscaled wind field basic data meets the standard specifically includes:

[0043] Performing mathematical analysis on the downscaled wind field basic data and the anemometry processed data to obtain the means μ1, μ2 and standard deviations σ1, σ2 of the downscaled wind field basic data and the anemometry processed data;

[0044] Calculating the cumulative distribution functions of the downscaled wind field basic data and the anemometry processed data:

[0045]

[0046] In the formula, F1(y1) is the cumulative distribution function of the downscaled wind field basic data, y1 is the downscaled wind field basic data of the yth anemometry station, F2(y2) is the cumulative distribution function of the anemometry processed data, y2 is the anemometry processed data of the yth anemometry station, and erf is the error function;

[0047] Obtaining the maximum deviation value according to the cumulative distribution functions of the downscaled wind field basic data and the anemometry processed data:

[0048] D = max|F1(y1) - F2(y2)|;

[0049] In the formula, max means taking the maximum value;

[0050] Obtaining the significance level value based on the data distribution test requirements;

[0051] Obtaining the deviation critical value based on the K-S test table according to the anemometry processed data and the significance level value;

[0052] According to the maximum deviation value and the deviation critical value, determine whether the data distributions of the downscaled wind field basic data and the wind measurement processed data conform to data consistency. If the maximum deviation value exceeds the deviation critical value, the data distributions of the downscaled wind field basic data and the wind measurement processed data are inconsistent, and the dynamic downscaling of the wind measurement processed data is performed again. If the maximum deviation value does not exceed the deviation critical value, the data distributions of the downscaled wind field basic data and the wind measurement processed data are consistent;

[0053] According to the wind measurement processed data, obtain the wind measurement characteristic data, and the wind measurement characteristic data includes meridional wind information, zonal wind information, temperature, and humidity.

[0054] Preferably, inputting the wind measurement characteristic data into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model specifically includes:

[0055] Normalize the wind measurement characteristic data to obtain a wind measurement characteristic data set;

[0056] Divide the wind measurement characteristic data set into a training set and a test set, and perform iterative training on the neural network U-Net to obtain neural network training information, where the neural network training information includes the fluctuation data of the loss function of the neural network;

[0057] According to the neural network training information, adjust the number of convolution kernels, activation function, number of iterations, learning rate, weighting coefficient of the loss function, and regularization constraint;

[0058] Among them, the loss function is specifically:

[0059]

[0060] In the formula, Loss is the combined loss function, α is the weight constrained by the downscaled wind field basic data, β is the weight constrained by the wind measurement processed data, N gird is the amount of downscaled wind field basic data, N s is the amount of wind measurement processed data, is the neural network output value of the i-th downscaled wind field basic data, y i is the actual value of the i-th downscaled wind field basic data, YM j is the neural network output value of the j-th wind measurement processed data, S j is the actual value of the i-th wind measurement processed data.

[0061] Preferably, obtaining the meteorological prediction information according to the low-altitude wind field data and the regional geographical data specifically includes:

[0062] Obtain historical meteorological data and historical wind measurement data, and the historical meteorological data is low-altitude meteorological information, including low-altitude wind speed information and wind direction information;

[0063] Process the historical wind measurement data to obtain processed historical wind measurement data;

[0064] According to the low-altitude wind field downscaling model, perform meteorological prediction on the processed historical wind measurement data to obtain historical low-altitude wind field data;

[0065] Based on the historical low-altitude wind field data and through visualization processing, obtain the historical low-altitude wind field map;

[0066] Based on the historical low-altitude wind field map, with the due north direction as the positive direction, obtain the historical low-altitude wind direction angle information;

[0067] Based on the historical low-altitude wind field data and historical meteorological data, obtain the historical wind speed difference data;

[0068] Based on the historical wind speed difference data and historical low-altitude wind direction angle information, and using a linear regression equation, obtain the wind speed correction coefficient;

[0069] Based on the wind speed correction coefficient and the low-altitude wind field data, obtain meteorological prediction information;

[0070] Among them, the meteorological prediction wind speed is:

[0071]

[0072] In the formula, V corr is the meteorological prediction wind speed, V is the low-altitude wind field wind speed output by the neural network, is the wind speed correction coefficient, is the wind speed non-linear correction coefficient, and θ is the wind direction angle.

[0073] Furthermore, a meteorological prediction system based on complex terrain wind field data is proposed to implement the above prediction method, including:

[0074] The main control module, which is used to judge whether the auxiliary wind measurement station meets the measurement requirements according to the wind measurement station position index and the wind measurement station position index threshold, judge whether the data distribution of the downscaled wind field basic data and the processed wind measurement data meets data consistency according to the maximum deviation value and the deviation critical value, obtain the wind measurement characteristic data according to the processed wind measurement data, perform iterative training on the neural network U-Net according to the wind measurement characteristic data to obtain the low-altitude wind field downscaling model, perform downscaling on the processed wind measurement data based on the low-altitude wind field downscaling model to obtain the low-altitude wind field data, and obtain meteorological prediction information according to the low-altitude wind field data and the regional geographical data;

[0075] An information acquisition module, which is used to acquire wind measurement data, regional wind direction information, regional wind speed data, regional geographical data, regional altitude information and regional terrain feature information, trace the wind measurement data, obtain the location information of the wind measurement station, obtain the historical data of the wind measurement station, historical meteorological data and historical wind measurement data, process the historical wind measurement data, obtain the processed historical wind measurement data, and transmit it to the calculation module;

[0076] A calculation module, which is used to process the wind measurement data according to the regional geographical data and the location information of the wind measurement station to obtain the processed wind measurement data, obtain the surface roughness of the wind measurement station according to the historical data of the wind measurement station, evaluate the auxiliary wind measurement station according to the classification information of the wind measurement station to obtain the location index of the wind measurement station, conduct mathematical analysis on the downscaled wind field basic data and the processed wind measurement data to obtain the mean and standard deviation of the downscaled wind field basic data and the processed wind measurement data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the processed wind measurement data;

[0077] A display module, which interacts with the main control module and is used to output and display meteorological prediction information.

[0078] Optionally, the main control module specifically includes:

[0079] A control unit, which is used to obtain wind measurement characteristic data according to the processed wind measurement data, perform iterative training on the neural network U-Net according to the wind measurement characteristic data to obtain a low-altitude wind field downscaling model, downscale the processed wind measurement data based on the low-altitude wind field downscaling model to obtain low-altitude wind field data, and obtain meteorological prediction information according to the low-altitude wind field data and the regional geographical data;

[0080] An information receiving unit, which interacts with the information acquisition module and the calculation module and is used to receive data and transmit it to the judgment unit;

[0081] A judgment unit, which is used to judge whether the auxiliary wind measurement station meets the measurement requirements according to the location index of the wind measurement station and the location index threshold of the wind measurement station, and judge whether the data distribution of the downscaled wind field basic data and the processed wind measurement data conforms to data consistency according to the maximum deviation value and the deviation critical value.

[0082] Optionally, the information acquisition module specifically includes:

[0083] A first acquisition unit, which is used to acquire wind measurement data, regional wind direction information, regional wind speed data, regional geographical data, regional altitude information and regional terrain feature information, trace the wind measurement data, and obtain the location information of the wind measurement station;

[0084] A second acquisition unit, which is used to acquire historical data of the anemometer station, historical meteorological data, and historical anemometry data, process the historical anemometry data to obtain historical anemometry processed data, and transmit it to the calculation module.

[0085] Optionally, the calculation module specifically includes:

[0086] A data processing unit, which is used to process the anemometry data according to the regional geographical data and the location information of the anemometer station to obtain anemometry processed data, obtain the surface roughness of the anemometer station according to the historical data of the anemometer station, and evaluate the auxiliary anemometer station according to the anemometer station classification information to obtain the anemometer station location index;

[0087] A data inspection unit, which is used to perform mathematical analysis on the downscaled wind field basic data and the anemometry processed data to obtain the mean and standard deviation of the downscaled wind field basic data and the anemometry processed data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the anemometry processed data.

[0088] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0089] The present invention proposes a meteorological prediction method and system based on complex terrain wind field data. By removing abnormal data within the data time period of the anemometer station, the reliability and accuracy of the data are ensured. By using the anemometer station location index, anemometer stations that are too concentrated are removed, ensuring a uniform distribution of anemometer station locations. By introducing the error of the anemometer station into the loss function of the neural network model for the low-altitude wind field modeling, the distribution of the downscaled low-altitude wind field at the anemometer station is made more reasonable. By using the wind speed correction coefficient, the accuracy of meteorological prediction is ensured, and the deviation influence of complex terrain on meteorological prediction results is reduced. Description of the Drawings

[0090] Figure 1 It is a flowchart of a meteorological prediction method based on complex terrain wind field data proposed by the present invention;

[0091] Figure 2 It is a flowchart for obtaining anemometry processed data in the present invention;

[0092] Figure 3 It is a flowchart for obtaining the anemometer station location index in the present invention;

[0093] Figure 4 It is a flowchart for obtaining meteorological prediction information in the present invention;

[0094] Figure 5 It is a comparison diagram of the low-altitude wind field in the present invention;

[0095] Figure 6Block diagram of a meteorological prediction system based on complex terrain wind field data proposed by the present invention. Detailed implementation manners

[0096] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and other obvious variations can be conceived by those skilled in the art.

[0097] Referring to Figure 1 - Figure 4 As shown, a meteorological prediction method based on complex terrain wind field data in an embodiment of the present invention includes:

[0098] Obtain wind measurement data, where the wind measurement data includes regional wind direction information and regional wind speed data;

[0099] Trace the source of the wind measurement data to obtain the location information of the wind measurement station;

[0100] Obtain regional geographical data, where the regional geographical data includes regional altitude information and regional terrain feature information;

[0101] According to the regional geographical data and the location information of the wind measurement station, process the wind measurement data to obtain processed wind measurement data;

[0102] Specifically, according to the regional geographical data and the location information of the wind measurement station, processing the wind measurement data to obtain processed wind measurement data specifically includes:

[0103] According to the regional geographical data and the location information of the wind measurement station, obtain the geographical influence information of the wind measurement station, where the geographical influence information of the wind measurement station includes the geographical influence area information of the wind measurement station and the geographical data within this area;

[0104] Obtain the historical data of the wind measurement station, where the historical data of the wind measurement station includes the wind speed data, wind direction information and friction velocity measured by this wind measurement station;

[0105] According to the historical data of the wind measurement station, obtain the surface roughness of the wind measurement station;

[0106] Based on the geographical influence information of the wind measurement station, obtain the standard surface roughness of the wind measurement station;

[0107] According to the surface roughness of the wind measurement station and the standard surface roughness of the wind measurement station, obtain the data time period of the wind measurement station;

[0108] Analyze the wind speed within each data time period of each wind measurement station. If the wind speed is abnormal, remove the wind measurement data of this data time period of this wind measurement station;

[0109] Take the wind measurement station with the smallest surface roughness of the wind measurement station as the initial wind measurement station;

[0110] Based on the initial wind measurement station, classify the wind measurement stations according to the position information of the wind measurement stations to obtain wind measurement station classification information, where the wind measurement station classification information includes the initial wind measurement station, the basic wind measurement station, and the auxiliary wind measurement station;

[0111] Evaluate the auxiliary wind measurement stations to obtain the wind measurement station position index;

[0112] Screen the auxiliary wind measurement stations according to the wind measurement station position index;

[0113] Obtain wind measurement processing data according to the wind measurement data of the screened auxiliary wind measurement stations, the initial wind measurement station, and the basic wind measurement station;

[0114] Among them, the calculation formula for the surface roughness of the wind measurement station is:

[0115]

[0116] In the formula, Z0 is the surface roughness of the wind measurement station, U(x) is the wind speed of the x-th wind measurement station, and U * is the friction velocity, h x is the vertical distance between the wind speed measurement position of the x-th wind measurement station and the ground, and k is the von Kármán constant and k = 0.4.

[0117] Specifically, screening the auxiliary wind measurement stations according to the wind measurement station position index specifically includes:

[0118] Based on the initial wind measurement station, connect the wind measurement stations until the measurable area of the wind measurement stations covers the entire wind measurement requirement area, and obtain the wind measurement station distribution plan information;

[0119] According to the wind measurement station distribution plan information, select the wind measurement station distribution plan with the smallest number of connected wind measurement stations to obtain the basic wind measurement station distribution information;

[0120] Among them, the basic wind measurement station distribution information includes the initial wind measurement station information and the basic wind measurement station information;

[0121] Classify the wind measurement stations according to the basic wind measurement station distribution information to obtain the wind measurement station classification information;

[0122] Evaluate the auxiliary wind measurement stations according to the wind measurement station classification information to obtain the wind measurement station position index;

[0123] Based on the actual measurement requirements of the regional wind field, obtain the wind measurement station position index threshold;

[0124] According to the anemometer station location index and the anemometer station location index threshold, it is judged whether the auxiliary anemometer station meets the measurement requirements. If the anemometer station location index is lower than the anemometer station location index threshold, the auxiliary anemometer station does not meet the measurement requirements, and the anemometer data of the auxiliary anemometer station is removed. If the anemometer station location index is higher than the anemometer station location index threshold, the auxiliary anemometer station meets the measurement requirements;

[0125] Among them, the calculation formula of the anemometer station location index is:

[0126]

[0127] In the formula, Q(x) is the anemometer station location index, Z0(g) is the anemometer station surface roughness of the g-th basic anemometer station overlapping with the measurable area of the x-th auxiliary anemometer station, S g is the overlapping area of the measurable area of the x-th auxiliary anemometer station and the measurable area of the g-th basic anemometer station, and n is the total number of basic anemometer stations overlapping with the measurable area of the x-th auxiliary anemometer station.

[0128] In this solution, through the historical data of the anemometer station, the anemometer station surface roughness is obtained. Through the anemometer station surface roughness and the standard surface roughness of the anemometer station, the anemometer station data time period is obtained. By analyzing the wind speed within each anemometer station data time period of each anemometer station, the data that significantly deviates (more than twice) from the 99% quantile in the anemometer data is identified and removed. By evaluating the auxiliary anemometer station, the anemometer station location index is obtained. According to the anemometer station location index and the anemometer station location index threshold, it is judged whether the auxiliary anemometer station meets the measurement requirements, and the overly concentrated stations are removed to make the distribution of the anemometer stations more uniform.

[0129] It should be noted that the standard surface roughness of the anemometer station varies with the change of surface characteristics, which is 0.065 for flat grassland, 1.5 for urban areas, and 2.5 for forest areas. The basic time period of the anemometer station data is 2 hours. Through the anemometer station surface roughness and the standard surface roughness of the anemometer station, the length of the basic time period is adjusted, that is:

[0130]

[0131] In the formula, T is the length of the anemometer station data time period, Z is the standard surface roughness of the anemometer station, and T1 is the basic time period of the anemometer station data.

[0132] Based on the WRF regional meteorological model, dynamic downscaling is performed on the anemometry processed data to obtain the basic data of the downscaled wind field;

[0133] Analyze the downscaled wind field basic data and the wind measurement processed data, and determine whether the downscaled wind field basic data meets the standards. If not, perform dynamic downscaling on the wind measurement processed data again. If so, obtain the wind measurement characteristic data based on the wind measurement processed data;

[0134] Specifically, analyze the downscaled wind field basic data and the wind measurement processed data, and determine whether the downscaled wind field basic data meets the standards, which specifically includes:

[0135] Perform mathematical analysis on the downscaled wind field basic data and the wind measurement processed data to obtain the means μ1, μ2 and standard deviations σ1, σ2 of the downscaled wind field basic data and the wind measurement processed data;

[0136] Calculate the cumulative distribution functions of the downscaled wind field basic data and the wind measurement processed data:

[0137]

[0138] In the formula, F1(y1) is the cumulative distribution function of the downscaled wind field basic data, y1 is the downscaled wind field basic data of the y-th wind measurement station, F2(y2) is the cumulative distribution function of the wind measurement processed data, y2 is the wind measurement processed data of the y-th wind measurement station, and erf is the error function;

[0139] Obtain the maximum deviation value according to the cumulative distribution functions of the downscaled wind field basic data and the wind measurement processed data:

[0140] D = max|F1(y1) - F2(y2)|;

[0141] In the formula, max represents taking the maximum value;

[0142] Obtain the significance level value based on the data distribution test requirements;

[0143] Obtain the deviation critical value based on the wind measurement processed data and the significance level value according to the K-S test table;

[0144] Judge whether the data distributions of the downscaled wind field basic data and the wind measurement processed data are in data consistency according to the maximum deviation value and the deviation critical value. If the maximum deviation value exceeds the deviation critical value, the data distributions of the downscaled wind field basic data and the wind measurement processed data are inconsistent, and perform dynamic downscaling on the wind measurement processed data again. If the maximum deviation value does not exceed the deviation critical value, the data distributions of the downscaled wind field basic data and the wind measurement processed data are consistent;

[0145] Obtain the wind measurement characteristic data based on the wind measurement processed data, and the wind measurement characteristic data includes zonal wind information, meridional wind information, temperature and humidity.

[0146] In this solution, by calculating the cumulative distribution functions of the downscaled wind field basic data and the wind measurement processed data, and based on the cumulative distribution functions of the downscaled wind field basic data and the wind measurement processed data, the maximum deviation value is obtained. The data distribution consistency of the downscaled wind field basic data and the wind measurement processed data is tested through the maximum deviation value, ensuring the stability and reliability of the data.

[0147] It should be noted that the variables for which data collection and processing are completed in the wind measurement characteristic data include the meridional wind, zonal wind, temperature, humidity, the altitude of the underlying surface, and the boundary layer height. They are all three-dimensional variable fields within the range of 10m to 200m above the ground. These are divided into 20 layers according to the geometric height. Each variable in each layer is taken as a key feature, and the results input by the dynamic downscaling WRF model are resampled onto the geometric coordinates by means of interpolation.

[0148] Since the target variable is the low-altitude (i.e., the height above the ground is 10 - 200m) wind field, when selecting key features, more meteorological variables related to the near-surface wind field in the output of the WRF numerical model should be selected according to physical principles, while meteorological variables with less impact on the accuracy of predicting the low-altitude wind field are abandoned. Selecting more meteorological variables related to the near-surface wind field according to physical principles includes the meridional wind, zonal wind, temperature, humidity, the altitude of the underlying surface, and the boundary layer height. Because both the input and output are in the format of two-dimensional pictures, for these three-dimensional variable fields, they need to be divided into 20 layers in the geometric height. Each variable in each layer is taken as a key feature, so there are a total of 20×7 = 140 variables, which is equivalent to a total of 140 input data channels for the neural network.

[0149] Since the vertical coordinates of the output variables of the WRF model are not in the geometric height, it is necessary to resample them to the geometric coordinates by means of interpolation in the vertical direction. In this process, factors such as terrain undulation, surface type, and roughness that may affect the wind field need to be considered. In addition, different interpolation methods are adopted at different points, including bilinear interpolation, cubic spline interpolation, inverse distance weighting method, Kriging interpolation, etc.

[0150] Input the wind measurement characteristic data into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model.

[0151] Specifically, inputting the wind measurement characteristic data into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model specifically includes:

[0152] Normalize the wind measurement characteristic data to obtain the wind measurement characteristic data set.

[0153] The wind measurement feature dataset is divided into a training set and a test set, and iterative training is performed on the neural network U-Net to obtain neural network training information, where the neural network training information includes the fluctuation data of the loss function of the neural network.

[0154] According to the neural network training information, adjust the number of convolutional kernels, activation function, number of iterations, learning rate, weighting coefficient of the loss function, and regularization constraint.

[0155] Among them, the loss function is specifically:

[0156]

[0157] In the formula, Loss is the combined loss function, α is the weight of the downscaled wind field basic data constraint, β is the weight of the wind measurement processed data constraint, N gird is the amount of downscaled wind field basic data, N s is the amount of wind measurement processed data, is the neural network output value of the i-th downscaled wind field basic data, y i is the actual value of the i-th downscaled wind field basic data, YM j is the neural network output value of the j-th wind measurement processed data, S j is the actual value of the i-th wind measurement processed data.

[0158] In this solution, the U-Net neural network is composed of an Encoder-Decoder structure with skip connections added in the middle. The Encoder is a downsampling process composed of convolutional layers and downsampling layers. The unit in the downsampling process is a convolutional block, and each convolutional block is composed of two convolutional layers and a pooling layer. The convolutional layer is used to extract information such as texture, content, and color in the data, and the pooling layer is used to reduce the resolution, thereby capturing more abstract features and reducing the dimension of the data, thus reducing the amount of computation and the number of parameters. The Decoder includes an upsampling layer and convolutional layers, which are used to reconstruct the spatial information of the image, that is, to map the feature information in the network back to the spatial position of the original image. The convolutional block in the upsampling process is similar to the downsampling, and is composed of two convolutional blocks and a transposed convolution process. Each downsampling will double the coarseness of the data resolution, and each upsampling will double the resolution. Using 4 layers of downsampling combined with 6 layers of upsampling can increase the data resolution by 4 times. The combined loss function of the neural network is the weighted mean square error sum of two parts, one part is the output of the high-resolution dynamic downscaling model, and the other part is the observed data of the wind measurement station.

[0159] It can be understood that in the iterative training of a neural network, during the normalization operation, data with different units and magnitudes need to be unified to sizes close to each other (close to between 0 and 1 or -1 and 1). Variable normalization greatly affects the prediction accuracy and generalization ability of the neural network model. This step is closely related to the extreme value elimination step. It is necessary to prevent overly large extreme values or variable upper bounds from participating in the multi-variable normalization operation. Alternative normalization methods include Min-max normalization, mean normalization, Z-score normalization, and non-linear normalization. These different normalization methods have their own unique adaptability for achieving the optimal downscaling accuracy of the low-altitude wind field at different points. After the data is normalized, the data set is divided into a training set, a validation set, and a test set according to a ratio of 6:2:2. The existence of the validation set is to ensure that the model is not affected by the set itself when predicting the test set. The loss function constructed for U-Net, where Loss is the combined loss function, is jointly constrained by the spatial error between the neural network output and the dynamically downscaled model output and the error between the neural network output and the station observations. The first term on the right side of the loss function equation is the L2 norm between the neural network output and the high-resolution output obtained by WRF dynamic downscaling, which is the sum of the squares of the errors between each grid point. The second term on the right side of the loss function equation is the L2 norm between the result of interpolating the neural network output to the station and the station observations;

[0160] Since the output of the neural network is a two-dimensional grid point distribution, while the wind measurement stations are single-point distributions, interpolating the output of the neural network to the wind measurement stations is equivalent to multiplying the network output by a two-dimensional mask matrix M. The size of the mask matrix is a×b, where a is the width of the data and b is the number of wind measurement stations. The output of the model is reduced to b×1, and the values of the mask matrix are obtained by interpolation.

[0161] α is the weight for the constraint of the downscaled wind field basic data, and β is the weight for the constraint of the wind measurement processed data, which are determined by human convention. This loss function not only considers the error with the high-resolution model output but also the constraint of the wind measurement station observations on the downscaling model, which can improve the accuracy of the model.

[0162] After establishing the loss function, iterative training begins to converge the loss function of the neural network. According to the changing trends of the loss functions of the training set and the validation set, the model is tuned and optimized. If the loss function on the training set decreases rapidly, but the loss function on the test set increases instead after decreasing to a certain level, it indicates that the model has overfitting. Parameter optimization can be carried out by reducing the number of convolutional kernels, reducing the number of iterations, adjusting the learning rate, increasing regularization constraints, or adding a normalization layer. If the loss functions on both the training set and the validation set are large, it indicates that the model is underfitting. At this time, optimization can be carried out by increasing the complexity of the model. Analyze the changes in the loss function, perform parameter optimization to converge the loss function of the network, improve the convergence speed and accuracy of the model, and enhance the generalization ability of the model.

[0163] As Figure 5 shown, Figure (a) is the wind field map with a coarse resolution output by the WRF model, with a resolution of 1000m, and Figure (b) is the wind field map after downscaling using the neural network U-Net, with a resolution of 250m.

[0164] Based on the low-altitude wind field downscaling model, downscale the wind measurement processed data to obtain low-altitude wind field data;

[0165] According to the low-altitude wind field data and regional geographical data, obtain meteorological prediction information, where the meteorological prediction information includes wind speed information and wind direction information.

[0166] Specifically, obtaining meteorological prediction information according to the low-altitude wind field data and regional geographical data specifically includes:

[0167] Obtain historical meteorological data and historical wind measurement data, where the historical meteorological data is low-altitude meteorological information, including low-altitude wind speed information and wind direction information;

[0168] Perform data processing on the historical wind measurement data to obtain historical wind measurement processed data;

[0169] According to the low-altitude wind field downscaling model, perform meteorological prediction on the historical wind measurement processed data to obtain historical low-altitude wind field data;

[0170] According to the historical low-altitude wind field data, obtain the historical low-altitude wind field map based on visualization processing;

[0171] According to the historical low-altitude wind field map, with the due north direction as the positive direction, obtain historical low-altitude wind direction angle information;

[0172] According to the historical low-altitude wind field data and historical meteorological data, obtain historical wind speed difference data;

[0173] According to the historical wind speed difference data and historical low-altitude wind direction angle information, obtain the wind speed correction coefficient based on the linear regression equation;

[0174] Obtain meteorological prediction information according to the wind speed correction coefficient and low-altitude wind field data;

[0175] Among them, the meteorological predicted wind speed is:

[0176]

[0177] In the formula, V corr is the meteorological predicted wind speed, V is the low-altitude wind field wind speed output by the neural network, is the wind speed correction coefficient, is the wind speed non-linear correction coefficient, and θ is the wind direction angle.

[0178] In this solution, through the low-altitude wind field downscaling model, meteorological prediction is carried out on the historical wind measurement processed data to obtain historical low-altitude wind field data. According to the historical low-altitude wind field data and historical meteorological data, historical wind speed difference data is obtained. According to the historical wind speed difference data and historical low-altitude wind direction angle information, based on the linear regression equation, the wind speed correction coefficient is obtained. According to the wind speed correction coefficient and low-altitude wind field data, meteorological prediction information is obtained, ensuring the accuracy of meteorological prediction and reducing the deviation impact of complex terrain on meteorological prediction results.

[0179] Referring to Figure 6 As shown, further, in combination with the above-mentioned meteorological prediction method based on complex terrain wind field data, a meteorological prediction system based on complex terrain wind field data is proposed, including:

[0180] The main control module is used to judge whether the auxiliary wind measurement station meets the measurement requirements according to the wind measurement station position index and the wind measurement station position index threshold, judge whether the data distribution of the downscaled wind field basic data and the wind measurement processed data meets the data consistency according to the maximum deviation value and the deviation critical value, obtain the wind measurement characteristic data according to the wind measurement processed data, perform iterative training on the neural network U-Net according to the wind measurement characteristic data to obtain the low-altitude wind field downscaling model, perform downscaling on the wind measurement processed data based on the low-altitude wind field downscaling model to obtain the low-altitude wind field data, and obtain meteorological prediction information according to the low-altitude wind field data and regional geographical data;

[0181] The information acquisition module is used to acquire wind measurement data, regional wind direction information, regional wind speed data, regional geographical data, regional altitude information and regional terrain feature information, trace the source of the wind measurement data to obtain the wind measurement station position information, obtain the wind measurement station historical data, historical meteorological data and historical wind measurement data, perform data processing on the historical wind measurement data to obtain the historical wind measurement processed data, and transmit it to the calculation module;

[0182] A calculation module, which is used to process the wind measurement data according to the regional geographical data and the location information of the wind measurement stations to obtain the processed wind measurement data, obtain the surface roughness of the wind measurement stations according to the historical data of the wind measurement stations, evaluate the auxiliary wind measurement stations according to the classification information of the wind measurement stations to obtain the location index of the wind measurement stations, perform mathematical analysis on the downscaled wind field basic data and the processed wind measurement data to obtain the mean and standard deviation of the downscaled wind field basic data and the processed wind measurement data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the processed wind measurement data;

[0183] A display module, which interacts with the main control module and is used to output and display meteorological prediction information.

[0184] The main control module specifically includes:

[0185] A control unit, which is used to obtain the wind measurement characteristic data according to the processed wind measurement data, perform iterative training on the neural network U-Net according to the wind measurement characteristic data to obtain a low-altitude wind field downscaling model, downscale the processed wind measurement data based on the low-altitude wind field downscaling model to obtain low-altitude wind field data, and obtain meteorological prediction information according to the low-altitude wind field data and the regional geographical data;

[0186] An information receiving unit, which interacts with the information acquisition module and the calculation module and is used to receive data and transmit it to the judgment unit;

[0187] A judgment unit, which is used to judge whether the auxiliary wind measurement station meets the measurement requirements according to the location index of the wind measurement station and the location index threshold of the wind measurement station, and judge whether the data distribution of the downscaled wind field basic data and the processed wind measurement data conforms to data consistency according to the maximum deviation value and the deviation critical value.

[0188] The information acquisition module specifically includes:

[0189] A first acquisition unit, which is used to acquire wind measurement data, regional wind direction information, regional wind speed data, regional geographical data, regional altitude information and regional terrain feature information, trace the source of the wind measurement data to obtain the location information of the wind measurement stations;

[0190] A second acquisition unit, which is used to acquire the historical data of the wind measurement stations, historical meteorological data and historical wind measurement data, process the historical wind measurement data to obtain historical processed wind measurement data, and transmit it to the calculation module.

[0191] The calculation module specifically includes:

[0192] A data processing unit, which is used to process the wind measurement data according to the regional geographical data and the wind measurement station location information to obtain the processed wind measurement data, obtain the surface roughness of the wind measurement station according to the historical data of the wind measurement station, and evaluate the auxiliary wind measurement stations according to the wind measurement station classification information to obtain the wind measurement station location index;

[0193] A data inspection unit, which is used to perform mathematical analysis on the downscaled wind field basic data and the processed wind measurement data to obtain the mean and standard deviation of the downscaled wind field basic data and the processed wind measurement data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the processed wind measurement data.

[0194] In summary, the advantages of the present invention are as follows: the surface roughness of the wind measurement station is obtained through the historical data of the wind measurement station, the data time period of the wind measurement station is obtained through the surface roughness of the wind measurement station and the standard surface roughness of the wind measurement station, and the abnormal data within the data time period of the wind measurement station is removed, ensuring the reliability and accuracy of the data. The wind measurement station location index is obtained by evaluating the auxiliary wind measurement stations, and the overly concentrated wind measurement stations are removed through the wind measurement station location index, ensuring the uniform distribution of the wind measurement station locations. By introducing the error of the wind measurement station into the loss function of the neural network model of the low-altitude wind field, the distribution of the downscaled low-altitude wind field at the wind measurement stations is made more reasonable. The wind speed correction coefficient is obtained based on the linear regression equation through the historical wind speed difference data and the historical low-altitude wind direction angle information, and the meteorological prediction information is obtained through the wind speed correction coefficient and the low-altitude wind field data, ensuring the accuracy of the meteorological prediction and reducing the deviation impact of complex terrain on the meteorological prediction results.

[0195] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A meteorological forecasting method based on complex terrain wind field data, characterized in that: include: Acquire wind measurement data, wherein the wind measurement data includes regional wind direction information and regional wind speed data; Trace the wind measurement data to obtain the location information of the wind measurement station; Acquiring regional geographic data, wherein the regional geographic data includes regional altitude information and regional terrain feature information; According to the regional geographic data and the location information of the wind measuring station, the wind measuring data is processed to obtain the wind measuring processed data; Based on the WRF regional meteorological model, the wind measurement data is dynamically downscaled to obtain the downscaled wind field basic data; Analyze the downscaled wind field basic data and wind measurement processing data to determine whether the downscaled wind field basic data meets the standards. If not, dynamically downscale the wind measurement processing data again. If yes, obtain wind measurement characteristic data based on the wind measurement processing data. The wind measurement characteristic data is input into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model; Based on the low-altitude wind field downscaling model, the wind measurement data is downscaled to obtain low-altitude wind field data; Obtaining weather forecast information based on low-altitude wind field data and regional geographic data, wherein the weather forecast information includes wind speed information and wind direction information; The wind measurement data is processed according to the regional geographic data and the wind measurement station location information to obtain the wind measurement processed data, specifically including: According to the regional geographic data and the location information of the wind measuring station, the geographic impact information of the wind measuring station is obtained, wherein the geographic impact information of the wind measuring station includes the geographic impact area information of the wind measuring station and the geographic data in the area; Acquire historical data of a wind measuring station, wherein the historical data of the wind measuring station includes wind speed data, wind direction information, and friction speed measured by the wind measuring station; According to the historical data of the wind measuring station, the surface roughness of the wind measuring station is obtained; Based on the geographical impact information of the wind measuring station, the standard surface roughness of the wind measuring station is obtained; According to the surface roughness of the wind measuring station and the standard surface roughness of the wind measuring station, the data time period of the wind measuring station is obtained; Analyze the wind speed in each wind measuring station data time period of each wind measuring station. If the wind speed is abnormal, remove the wind measurement data in the wind measuring station data time period. The wind measuring station with the smallest surface roughness is selected as the initial wind measuring station; Taking the initial wind measuring station as a reference and classifying the wind measuring stations according to the location information of the wind measuring stations, the classification information of the wind measuring stations is obtained, wherein the classification information of the wind measuring stations includes the initial wind measuring station, the basic wind measuring station and the auxiliary wind measuring station; Evaluate auxiliary wind measuring stations and obtain wind measuring station location index; According to the wind station location index, the auxiliary wind stations are screened; Obtain wind measurement processing data based on the wind measurement data of the screened auxiliary wind measurement stations, the initial wind measurement stations and the basic wind measurement stations; The calculation formula of the surface roughness of the wind measuring station is: ， In the formula, is the surface roughness of the wind station, is the wind speed at the xth wind measuring station, is the friction speed, is the vertical distance between the wind speed measurement position of the xth wind measuring station and the ground, k is the Kelvin constant and ; The auxiliary wind measuring stations are screened according to the wind measuring station location index, specifically including: Taking the initial wind measuring station as the benchmark, connect the wind measuring stations until the measurable area of ​​the wind measuring station covers the entire wind measurement demand area, and obtain the wind measuring station distribution plan information; According to the wind measuring station distribution plan information, a wind measuring station distribution plan with the smallest number of connected wind measuring stations is selected to obtain basic distribution information of wind measuring stations; The basic distribution information of wind measuring stations includes initial wind measuring station information and basic wind measuring station information; According to the basic distribution information of wind measuring stations, the wind measuring stations are classified and the classification information of wind measuring stations is obtained; According to the classification information of wind measuring stations, the auxiliary wind measuring stations are evaluated to obtain the wind measuring station location index; Based on the actual measurement requirements of the regional wind field, obtain the index threshold of the wind measuring station location; According to the wind measuring station location index and the wind measuring station location index threshold, determine whether the auxiliary wind measuring station meets the measurement requirements. If the wind measuring station location index is lower than the wind measuring station location index threshold, the auxiliary wind measuring station does not meet the measurement requirements, and the wind measurement data of the auxiliary wind measuring station is removed. If the wind measuring station location index is higher than the wind measuring station location index threshold, the auxiliary wind measuring station meets the measurement requirements. The calculation formula of the wind station location index is: In the formula, is the wind station location index, is the surface roughness of the g-th basic wind station that overlaps with the measurable area of ​​the x-th auxiliary wind station, is the overlapping area of ​​the measurable area of ​​the xth auxiliary wind measuring station and the measurable area of ​​the gth basic wind measuring station, and n is the total number of basic wind measuring stations overlapping with the measurable area of ​​the xth auxiliary wind measuring station; The wind measurement characteristic data is input into the neural network U-Net for iterative training to obtain the low-altitude wind field downscaling model, specifically including: Normalizing the wind measurement characteristic data to obtain a wind measurement characteristic data set; The wind measurement characteristic data set is divided into a training set and a test set, and the neural network U-Net is iteratively trained to obtain neural network training information, wherein the neural network training information includes loss function fluctuation data of the neural network; According to the neural network training information, adjust the number of convolution kernels, activation function, number of iterations, learning rate, weight coefficient of loss function and regularization constraints; Among them, the loss function is specifically: In the formula, Loss is the joint loss function, is the weight of downscaling wind field basic data constraint, is the weight of the wind measurement processing data constraint, To downscale the basic data of wind field, The amount of data processed for wind measurement, is the neural network output value of the i-th downscaled wind field basic data, is the actual value of the basic data of the i-th downscaled wind field, is the neural network output value of the jth wind measurement processing data, is the actual value of the i-th wind measurement processing data; The obtaining of weather forecast information based on low-altitude wind field data and regional geographic data specifically includes: Acquire historical meteorological data and historical wind measurement data, wherein the historical meteorological data is low-altitude meteorological information, including low-altitude wind speed information and wind direction information; Process the historical wind measurement data to obtain the historical wind measurement processing data; According to the low-altitude wind field downscaling model, meteorological forecasts are made on the historical wind measurement data to obtain historical low-altitude wind field data; According to the historical low-altitude wind field data, based on visualization processing, obtain the historical low-altitude wind field map; According to the historical low-altitude wind field map, taking the north direction as the positive direction, obtain the historical low-altitude wind direction angle information; Obtain historical wind speed difference data based on historical low-altitude wind field data and historical meteorological data; According to the historical wind speed difference data and the historical low-altitude wind direction angle information, the wind speed correction coefficient is obtained based on the linear regression equation; Obtain weather forecast information based on wind speed correction coefficient and low-altitude wind field data; Among them, the meteorological forecast wind speed is: In the formula, For weather forecasts, is the low-altitude wind speed output by the neural network, is the wind speed correction factor, is the wind speed nonlinear correction coefficient, is the wind direction angle.

2. A meteorological forecasting method based on complex terrain wind field data according to claim 1, characterized in that: The downscaled wind field basic data and wind measurement processing data are analyzed to determine whether the downscaled wind field basic data meets the standards, specifically including: Perform mathematical analysis on the downscaled wind field basic data and wind measurement processing data to obtain the mean of the downscaled wind field basic data and wind measurement processing data and standard deviation ; The cumulative distribution function of the downscaled wind field basic data and wind measurement processing data is calculated: In the formula, is the cumulative distribution function of the downscaled wind field basic data, is the downscaled wind field basic data of the yth wind measuring station, is the cumulative distribution function of wind measurement processing data, is the wind measurement processing data of the yth wind measurement station, is the error function; According to the cumulative distribution function of the downscaled wind field basic data and wind measurement processing data, the maximum deviation value is obtained: In the formula, max means taking the maximum value; Obtain significance level value based on data distribution test requirements; According to the wind measurement processing data and the significance level value, based on the KS test table, the deviation critical value is obtained; According to the maximum deviation value and the deviation critical value, it is judged whether the data distribution of the downscaled wind field basic data and the wind measurement processing data meets the data consistency. If the maximum deviation value exceeds the deviation critical value, the data distribution of the downscaled wind field basic data and the wind measurement processing data is inconsistent, and the wind measurement processing data is dynamically downscaled again. If the maximum deviation value does not exceed the deviation critical value, the data distribution of the downscaled wind field basic data and the wind measurement processing data is consistent; According to the wind measurement processing data, wind measurement characteristic data is obtained, and the wind measurement characteristic data includes meridional wind information, latitudinal wind information, temperature and humidity.

3. A meteorological forecasting system based on complex terrain wind field data, used to implement the forecasting method according to any one of claims 1-2, characterized in that: include: A main control module, wherein the main control module is used to determine whether the auxiliary wind measuring station meets the measurement requirements according to the wind measuring station location index and the wind measuring station location index threshold, determine whether the data distribution of the downscaled wind field basic data and the wind measurement processing data meets the data consistency according to the maximum deviation value and the deviation critical value, obtain wind measurement feature data according to the wind measurement processing data, iteratively train the neural network U-Net according to the wind measurement feature data, obtain a low-altitude wind field downscaling model, downscale the wind measurement processing data based on the low-altitude wind field downscaling model, obtain low-altitude wind field data, and obtain meteorological forecast information according to the low-altitude wind field data and regional geographic data; An information acquisition module, wherein the information acquisition module is used to obtain wind measurement data, regional wind direction information, regional wind speed data, regional geographic data, regional altitude information and regional terrain feature information, trace the wind measurement data, obtain the location information of the wind measurement station, obtain the historical data of the wind measurement station, historical meteorological data and historical wind measurement data, process the historical wind measurement data, obtain the historical wind measurement processing data, and transmit it to the calculation module; A calculation module, wherein the calculation module is used to process the wind measurement data according to the regional geographic data and the wind measurement station location information, obtain the wind measurement processing data, obtain the surface roughness of the wind measurement station according to the wind measurement station historical data, evaluate the auxiliary wind measurement station according to the wind measurement station classification information, obtain the wind measurement station location index, perform mathematical analysis on the downscaled wind field basic data and the wind measurement processing data, obtain the mean and standard deviation of the downscaled wind field basic data and the wind measurement processing data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the wind measurement processing data; The display module interacts with the main control module and is used to output and display weather forecast information.

4. A meteorological forecasting system based on complex terrain wind field data according to claim 3, characterized in that: The main control module specifically includes: A control unit, the control unit is used to obtain wind measurement characteristic data according to the wind measurement processing data, iteratively train the neural network U-Net according to the wind measurement characteristic data, obtain a low-altitude wind field downscaling model, downscale the wind measurement processing data based on the low-altitude wind field downscaling model, obtain low-altitude wind field data, and obtain meteorological forecast information according to the low-altitude wind field data and regional geographic data; An information receiving unit, which interacts with the information acquisition module and the calculation module to receive data and transmit it to the judgment unit; A judgment unit is used to judge whether the auxiliary wind measuring station meets the measurement requirements according to the wind measuring station location index and the wind measuring station location index threshold, and to judge whether the data distribution of the downscaled wind field basic data and the wind measurement processing data meets the data consistency according to the maximum deviation value and the deviation critical value.

5. A meteorological forecasting system based on complex terrain wind field data according to claim 3, characterized in that: The information acquisition module specifically includes: A first acquisition unit, the first acquisition unit is used to acquire wind measurement data, regional wind direction information, regional wind speed data, regional geographic data, regional altitude information and regional terrain feature information, trace the wind measurement data, and acquire the location information of the wind measurement station; The second acquisition unit is used to acquire historical data of the wind measurement station, historical meteorological data and historical wind measurement data, process the historical wind measurement data, acquire historical wind measurement processing data, and transmit the data to the calculation module.

6. A meteorological forecasting system based on complex terrain wind field data according to claim 3, characterized in that: The computing module specifically includes: A data processing unit, the data processing unit is used to process the wind measurement data according to the regional geographic data and the wind measurement station location information, obtain the wind measurement processing data, obtain the surface roughness of the wind measurement station according to the wind measurement station historical data, evaluate the auxiliary wind measurement station according to the wind measurement station classification information, and obtain the wind measurement station location index; A data verification unit is used to perform mathematical analysis on the downscaled wind field basic data and the wind measurement processing data, obtain the mean and standard deviation of the downscaled wind field basic data and the wind measurement processing data, and obtain the maximum deviation value according to the cumulative distribution function of the downscaled wind field basic data and the wind measurement processing data.

Images