A method, system and computer device for forecasting the probability of approaching thunderstorm and gale

Abstract

The present invention relates to the field of computer prediction models, and proposes a method, system and computer equipment for predicting the near-probability of thunderstorms and gales. By designing a near-probability forecasting model for thunderstorms and gales, the system can predict the near-probability of thunderstorms and gales based on the weather radar. WILL The maximum value, duration and spatial distribution of the moving trajectory of the thunderstorm and gale within the near time are measured and processed in the figure, so that the near probability of the thunderstorm and gale is quickly and accurately calculated within the near time, so as to make a nowcast of the thunderstorm and gale. The present invention greatly improves the accuracy and timeliness of the forecast of the near probability of the thunderstorm and gale.

Description

Technical Field

[0001] The present invention relates to the field of computer prediction models, and in particular to a method, system and computer equipment for forecasting the impending probability of thunderstorms and gales. Background Art

[0002] Thunderstorms and gale are extreme weather caused by severe convective storms, which pose a great threat and harm to agricultural crops, factory buildings, residential buildings, vehicles and ships. What distinguishes them from conventional gale weather is that thunderstorms and gale are sudden, short-lived, strong and destructive, which usually results in thunderstorm and gale disasters being fleeting and difficult to prevent in advance, creating great difficulties for weather forecasts and warnings.

[0003] In the prior art, thunderstorms and gale are mostly predicted by meteorological forecasters through comprehensive analysis of radar images and meteorological data from meteorological stations and observation stations, and then subjective judgments are made manually and warning areas are drawn to achieve forecast and warning. However, due to the characteristics of thunderstorms and gale, this warning method based mainly on manual judgment gives meteorological forecasters too little time for analysis and judgment, and is too hasty, and is easily affected by the subjectivity of forecasters, which leads to the problems of low accuracy, long release time and poor timeliness in thunderstorm and gale forecasts. The existing method of identifying mesocyclones and downbursts or judging whether the shape of a thunderstorm has a bow-shaped feature through the reflectivity echo image and velocity image of the Doppler weather radar can only identify the structure that produces gale in a thunderstorm, but cannot determine whether this structure will produce gale weather that reaches the ground, nor can it quantitatively determine whether the gale can meet the issuance standard of the thunderstorm and gale warning signal. The existing near-term gale forecasting method has the problems of poor timeliness, complex calculation amount and large feature difference.

[0004] Therefore, how to design a probability forecasting method for thunderstorms and gale to improve the accuracy and timeliness of thunderstorm and gale forecasts has become an urgent problem to be solved. Summary of the invention

[0005] Based on this, the purpose of the present invention is to provide a method, system and computer equipment for predicting the probability of thunderstorm and gale approaching, by designing a thunderstorm and gale nowcasting model, in the meteorological radar WILL The maximum value, duration and spatial distribution of the moving trajectory of the thunderstorm and gale within the near time are measured and processed in the figure, so that the near probability of the thunderstorm and gale is quickly and accurately calculated within the near time, so as to make a nowcast of the thunderstorm and gale. The present invention greatly improves the accuracy and timeliness of the forecast of the near probability of the thunderstorm and gale.

[0006] The present invention provides a method for forecasting the impending probability of thunderstorm and gale, comprising:

[0007] Constructing a pre-training data set based on historical data of thunderstorms and gale in the current area, and pre-training a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model;

[0008] Real-time detection of the existence of thunderstorms and gale winds and matching the storm center to obtain the nearby data of thunderstorms and gale winds to be identified;

[0009] The thunderstorm and gale near-coming data to be identified is input into the pre-trained thunderstorm and gale near-coming forecast model to obtain the thunderstorm and gale near-coming forecast.

[0010] In summary, based on the above-mentioned thunderstorm and gale near-probability forecasting method, a thunderstorm and gale near-forecasting model is designed. WILL The maximum value, duration and spatial distribution of the moving trajectory of the thunderstorm and gale within the near time are measured and processed in the figure, so that the near probability of the thunderstorm and gale is quickly and accurately calculated within the near time, so as to make a nowcast of the thunderstorm and gale. The present invention greatly improves the accuracy and timeliness of the forecast of the near probability of the thunderstorm and gale. Specifically, a pre-training data set is constructed according to the historical data of thunderstorms and strong winds in the current area, and a thunderstorm and strong wind nowcasting model is pre-trained. The thunderstorm and strong wind nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model. The model is pre-trained by the historical data of thunderstorms and strong winds in the current area, and the accuracy of the thunderstorm and strong wind nowcasting model is further optimized and improved. The convolutional network sub-model is designed to extract basic features, which improves the processing speed and robustness of the overall model. The long short-term memory network sub-model is designed to enhance features. For various types of meteorological data with strong temporal sequence, the long-term dependency between data in time sequence is enhanced, thereby improving the overall accuracy of the thunderstorm and strong wind nowcasting model. The existence status of thunderstorms and strong winds is detected in real time and the storm center is matched to obtain the thunderstorm and strong wind near data to be identified. The thunderstorm and strong wind near data to be identified is input into the pre-trained thunderstorm and strong wind nowcasting model to obtain the thunderstorm and strong wind nowcasting forecast. The present invention greatly improves the accuracy and timeliness of the near probability forecast of thunderstorms and strong winds.

[0011] Furthermore, the step of constructing a pre-training data set based on the historical data of thunderstorms and gales in the current area specifically includes:

[0012] Selecting historical data of thunderstorms and strong winds in the current area according to a preset thunderstorm and strong wind detection standard, wherein the preset thunderstorm and strong wind detection standard includes a preset wind speed threshold and a preset wind area range threshold;

[0013] Screening the inspection data in the thunderstorm and gale historical data to perform quality control on the remaining data in the thunderstorm and gale historical data according to the quality control code of the inspection data;

[0014] Obtain the weather radar of each thunderstorm and gale historical data instance in the thunderstorm and gale historical data WILL Figure 2 shows the weather radar WILL The graph is matched with the corresponding thunderstorm and gale historical data to complete the graph-by-graph and point-by-point preprocessing. The preprocessing formula is as follows:

[0015] ;

[0016] in, Represents an unprocessed data value, represents the data value after preprocessing, a Indicates weather radar WILL The figure number of the figure group, m represents the longitudinal grid point number, n represents the latitudinal grid point number, S A grid point set representing a single thunderstorm and gale historical data case after matching;

[0017] The preprocessed weather radar WILL The graph is constructed as a pre-training dataset for the thunderstorm and gale nowcasting model.

[0018] Furthermore, the thunderstorm and gale nowcasting model specifically includes:

[0019] The thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model;

[0020] The convolutional network sub-model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer. The convolutional network sub-model performs basic feature extraction of thunderstorm and gale data. The thunderstorm and gale data includes historical thunderstorm and gale data and nearby thunderstorm and gale data. The formula for extracting the basic features is as follows:

[0021] ;

[0022] in, y represents the feature output of the convolutional layer, l represents the number of convolutional layers, i represents the ordinal number of the feature output, j Represents the number of feature outputs, represents the activation function, k Represents the convolution kernel index value, K Represents the number of outputs of the convolutional layer, represents the convolution kernel weight, x represents the output vector of the previous layer, b represents the bias, p represents the output of the pooling layer, r represents the pooling window, R represents the pooling layer kernel size,T represents the step size of the pooling layer, represents the maximum value function;

[0023] The long short-term memory network sub-model includes an input gate, a forget gate and an output gate. The long short-term memory network sub-model performs feature enhancement. The formula for the feature enhancement is as follows:

[0024] ;

[0025] in, in represents the input gate, f represents the forget gate, o represents the output gate, Represents the input at the current moment, represents the activation function, c Represents a storage memory unit, t Indicates the time, W represents the weight term, U represents the bias term, h Indicates the output quantity, represents the hyperbolic tangent function, Represents the generated candidate value.

[0026] Furthermore, the step of pre-training the thunderstorm and gale nowcasting model further includes:

[0027] Conduct optimal testing on the pre-trained thunderstorm and gale nowcasting model;

[0028] The formula for the optimal test is as follows:

[0029] ;

[0030] in, CSI represents the critical success index, HSS Represents the rating index, TP It represents true yang. United Nations Indicates false vagina. FP Indicates false positive. TN Indicates true yin.

[0031] Furthermore, the step of detecting the existence of thunderstorms and gale in real time and matching the storm center to obtain the nearby data of thunderstorms and gale to be identified specifically includes:

[0032] Obtaining the gale observation data of all single observation stations in the current area, screening the detection data in the gale observation data, and performing quality control on the remaining data in the gale observation data according to the quality control code of the detection data;

[0033] Determine whether the high wind observation data after quality control is greater than or equal to a preset wind speed threshold, and if the high wind observation data is greater than or equal to the preset wind speed threshold, determine that there is high wind in the observation site area corresponding to the high wind observation data;

[0034] Get the weather radar for the current area WILL Figure 1. Weather radar WILL The graph is defined as follows:

[0035] ;

[0036] in, WILL represents the vertical cumulative liquid water content, Z represents the radar reflectivity numerator, H Indicates the maximum altitude detected by the radar;

[0037] According to the preset storm center matching rule, the weather radar in the observation station area with strong winds WILL The storm center is matched with the image. If the storm center is matched, it is determined that there is thunderstorm and strong wind in the observation site area, and the strong wind observation data corresponding to the observation site area and the meteorological radar are compared. WILL The figure is used as the nearby data of thunderstorm and gale to be identified.

[0038] Furthermore, the step of inputting the to-be-identified thunderstorm and gale near-falling data into the pre-trained thunderstorm and gale near-falling forecast model to obtain the thunderstorm and gale near-falling forecast specifically includes:

[0039] The thunderstorm and gale near-coming data to be identified is input into the pre-trained thunderstorm and gale near-coming forecast model, and the thunderstorm and gale near-coming forecast model calculates and outputs thunderstorm and gale forecast data, and the thunderstorm and gale forecast data includes the thunderstorm and gale maximum spatial distribution data A , spatial distribution data of thunderstorm and gale duration B , spatial distribution data of the moving trajectory of the storm center C The formula for calculating and outputting the thunderstorm and gale forecast data is as follows:

[0040] ;

[0041] in, WIND represents the intermediate discriminant, Indicates continuous output of graph groups WILL quantity, TRACK represents the grid points covered by the storm's moving track, a Indicates weather radar WILL The figure number of the figure group, m represents the longitudinal grid point number, n Indicates the latitude grid point number;

[0042] The probability of approaching thunderstorm and gale is calculated based on the thunderstorm and gale forecast data to obtain a thunderstorm and gale nowcast.

[0043] Furthermore, the step of calculating the probability of approaching thunderstorm and gale according to the thunderstorm and gale forecast data to obtain a thunderstorm and gale nowcast specifically includes:

[0044] The probability of approaching thunderstorm and gale forecast data is calculated according to the preset probability division rule, and the preset probability division rule is as follows:

[0045] ;

[0046] in, Pwind Indicates the probability of approaching thunderstorms and gales. Represents the straight-line distance function from the grid point to the trajectory;

[0047] Then, a thunderstorm and gale nowcast is obtained based on the preset wind force level standard and the preset thunderstorm and gale warning signal standard.

[0048] The present invention provides a thunderstorm and gale approach probability forecasting system, comprising:

[0049] A pre-training module, used to construct a pre-training data set based on the historical data of thunderstorms and gale in the current area, and to pre-train a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model;

[0050] The data processing module is used to detect the existence of thunderstorms and gale in real time and match the storm center to obtain the nearby data of thunderstorms and gale to be identified;

[0051] The forecast acquisition module is used to input the thunderstorm and gale near-coming data to be identified into the pre-trained thunderstorm and gale near-coming forecast model to obtain the thunderstorm and gale near-coming forecast.

[0052] The present invention also provides a storage medium, wherein the storage medium stores one or more programs, and when the programs are executed by a processor, the above-mentioned method for forecasting the impending probability of thunderstorms and gales is implemented.

[0053] The present invention also provides a computer device, the computer device comprising a memory and a processor, wherein:

[0054] The memory is used to store computer programs;

[0055] When the processor is used to execute the computer program stored in the memory, the above-mentioned method for forecasting the impending probability of thunderstorms and gales is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1A flow chart of a method for forecasting the impending probability of thunderstorm and gale proposed in the first embodiment of the present invention;

[0057] Figure 2 A flow chart of a method for forecasting the impending probability of thunderstorm and gale proposed in a second embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the structure of a thunderstorm and gale approach probability forecasting system proposed in the third embodiment of the present invention.

[0059] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0060] In order to facilitate understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are provided in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0061] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0063] See also Figure 1 , which is a flow chart of a method for forecasting the approaching probability of a thunderstorm and gale proposed in a first embodiment of the present invention, wherein the method for forecasting the approaching probability of a thunderstorm and gale comprises steps S01 to S03, wherein:

[0064] Step S01: constructing a pre-training data set based on historical data of thunderstorms and gale in the current area, and pre-training a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model;

[0065] It should be noted that the thunderstorm and gale historical data in this embodiment selects 50 thunderstorm and gale cases, and the thunderstorm and gale historical data in the current area is selected according to the preset thunderstorm and gale detection standard, and the preset thunderstorm and gale detection standard includes a preset wind speed threshold and a preset wind area range threshold;

[0066] Screening the inspection data in the thunderstorm and gale historical data to perform quality control on the remaining data in the thunderstorm and gale historical data according to the quality control code of the inspection data;

[0067] Obtain the weather radar of each thunderstorm and gale historical data instance in the thunderstorm and gale historical data WILL Figure 2 shows the weather radar WILL The graph is matched with the corresponding thunderstorm and gale historical data to complete the graph-by-graph and point-by-point preprocessing. The preprocessing formula is as follows:

[0068] ;

[0069] in, Represents an unprocessed data value, represents the data value after preprocessing, a Indicates weather radar WILL The figure number of the figure group, m represents the longitudinal grid point number, n represents the latitudinal grid point number, S A grid point set representing a single thunderstorm and gale historical data case after matching;

[0070] The preprocessed weather radar WILL The graph is constructed as a pre-training dataset for the thunderstorm and gale nowcasting model.

[0071] It should be noted that the thunderstorm and gale nowcasting model in this embodiment specifically includes:

[0072] The thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model. The two sub-models are designed in series. The number of convolutional layers and pooling layers of the convolutional network are set to 5, and the number of layers of the long short-term memory network is 20.

[0073] The convolutional network sub-model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer. The convolutional layer is used to perform convolution calculations on the data input by the input layer to locally perceive the data to achieve feature extraction. The pooling layer is used to reduce the size of the prediction model, improve the calculation speed and retain the original important features to improve the robustness of the data features. The fully connected layer is used to integrate the extracted feature values ​​into vector outputs to achieve specific tasks. The convolutional network sub-model performs basic feature extraction of thunderstorm and gale data. The thunderstorm and gale data includes historical thunderstorm and gale data and nearby thunderstorm and gale data. The formula for basic feature extraction is as follows:

[0074] ;

[0075] in, y represents the feature output of the convolutional layer, l represents the number of convolutional layers, i represents the ordinal number of the feature output, j Represents the number of feature outputs, represents the activation function, k Represents the convolution kernel index value, K Represents the number of outputs of the convolutional layer, represents the convolution kernel weight, x represents the output vector of the previous layer, b represents the bias, p represents the output of the pooling layer, r represents the pooling window, R represents the pooling layer kernel size, T represents the step size of the pooling layer, represents the maximum value function;

[0076] The LSTM sub-model includes an input gate, a forget gate, and an output gate, and introduces a set of storage units to record time. t All the information stored in the , input gate, forget gate and output gate are three logical functions, whose element values ​​are in the range of [0, 1]. The forget gate is used to control the amount of information in each storage unit, and the output gate is used to control the output of each storage unit. The long short-term memory network sub-model performs feature enhancement, and the formula for the feature enhancement is as follows:

[0077] ;

[0078] in, in represents the input gate, f represents the forget gate, o represents the output gate, Represents the input at the current moment, represents the activation function, c Represents a storage memory unit, t Indicates the time, W represents the weight term, U represents the bias term, h Indicates the output quantity, represents the hyperbolic tangent function, Represents the generated candidate value.

[0079] Step S02: Real-time detection of the existence of thunderstorms and gale winds and matching the storm center to obtain the nearby data of thunderstorms and gale winds to be identified;

[0080] It should be noted that in this embodiment, the steps of detecting the existence of thunderstorms and gale in real time and matching the storm center to obtain the nearby data of thunderstorms and gale to be identified specifically include:

[0081] Obtaining the gale observation data of all single observation stations in the current area, screening the detection data in the gale observation data, and performing quality control on the remaining data in the gale observation data according to the quality control code of the detection data;

[0082] Determine whether the high wind observation data after quality control is greater than or equal to a preset wind speed threshold, and if the high wind observation data is greater than or equal to the preset wind speed threshold, determine that there is high wind in the observation site area corresponding to the high wind observation data;

[0083] Get the weather radar for the current area WILL Figure 1. Weather radar WILL The graph is defined as follows:

[0084] ;

[0085] in, WILL represents the vertical cumulative liquid water content, Z represents the radar reflectivity numerator, H Indicates the maximum altitude detected by the radar;

[0086] According to the preset storm center matching rule, the weather radar in the observation station area with strong winds WILL The storm center is matched with the image. If the storm center is matched, it is determined that there is thunderstorm and strong wind in the observation site area, and the strong wind observation data corresponding to the observation site area and the meteorological radar are compared. WILL The figure is used as the nearby data of thunderstorm and gale to be identified.

[0087] Step S03: inputting the thunderstorm and gale near-falling data to be identified into the pre-trained thunderstorm and gale near-falling forecast model to obtain the thunderstorm and gale near-falling forecast;

[0088] It should be noted that in this embodiment, the step of inputting the to-be-identified thunderstorm and gale near-falling data into the pre-trained thunderstorm and gale near-falling forecast model to obtain the thunderstorm and gale near-falling forecast specifically includes:

[0089] The thunderstorm and gale near-coming data to be identified is input into the pre-trained thunderstorm and gale near-coming forecast model, and the thunderstorm and gale near-coming forecast model calculates and outputs thunderstorm and gale forecast data, and the thunderstorm and gale forecast data includes the thunderstorm and gale maximum spatial distribution data A , spatial distribution data of thunderstorm and gale duration B , spatial distribution data of the moving trajectory of the storm center CThe formula for calculating and outputting the thunderstorm and gale forecast data is as follows:

[0090] ;

[0091] in, WIND represents the intermediate discriminant, Indicates continuous output of graph groups WILL quantity, TRACK represents the grid points covered by the storm's moving track, a Indicates weather radar WILL The figure number of the figure group, m represents the longitudinal grid point number, n Indicates the latitude grid point number;

[0092] The thunderstorm and gale nowcasting model in this embodiment can continuously and gradually process the VIL forecast graphs of different time steps, and define WILL The threshold is 35kg / m 2 , the grid points that are greater than or equal to the threshold are recorded as 1, and the grid points that are less than the threshold are recorded as 0. All grid points in the current area are processed one by one. As long as a group of images has a situation that is greater than or equal to the threshold, the intermediate discriminant WIND If the judgment is 1, then the grid point value is 1. After all grid points are processed, the spatial distribution of the maximum value of thunderstorm and gale is obtained. A , process all grid points in the current area one by one, and accumulate the points in this set of images that have appeared and satisfied a threshold value or more, that is, the intermediate discriminant WIND The accumulated value is used as the point value. After all grid points are processed, the spatial distribution of the duration of thunderstorms and gales is obtained. B Each step in the forecast process is WILL Find the maximum value and connect the maximum values ​​in space in sequence to get the trajectory. The grid points covered by the trajectory are recorded as TRACK ,in accordance with TRACK Determine the spatial distribution of the movement trajectory of the storm center within the nowcast time C , calculating the probability of approaching thunderstorm and gale according to the thunderstorm and gale forecast data to obtain a thunderstorm and gale nowcast, specifically including:

[0093] The probability of approaching thunderstorm and gale forecast data is calculated according to the preset probability division rule, and the preset probability division rule is as follows:

[0094] ;

[0095] in, Pwind Indicates the probability of approaching thunderstorms and gales. Represents the straight-line distance function from the grid point to the trajectory;

[0096] Then, a thunderstorm and gale nowcast is obtained based on the preset wind force level standard and the preset thunderstorm and gale warning signal standard.

[0097] In summary, based on the above-mentioned thunderstorm and gale near-probability forecasting method, a thunderstorm and gale near-forecasting model is designed. WILL The maximum value, duration and spatial distribution of the moving trajectory of the thunderstorm and gale within the near time are measured and processed in the figure, so that the near probability of the thunderstorm and gale is quickly and accurately calculated within the near time, so as to make a nowcast of the thunderstorm and gale. The present invention greatly improves the accuracy and timeliness of the forecast of the near probability of the thunderstorm and gale. Specifically, a pre-training data set is constructed according to the historical data of thunderstorms and strong winds in the current area, and a thunderstorm and strong wind nowcasting model is pre-trained. The thunderstorm and strong wind nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model. The model is pre-trained by the historical data of thunderstorms and strong winds in the current area, and the accuracy of the thunderstorm and strong wind nowcasting model is further optimized and improved. The convolutional network sub-model is designed to extract basic features, which improves the processing speed and robustness of the overall model. The long short-term memory network sub-model is designed to enhance features. For various types of meteorological data with strong temporal sequence, the long-term dependency between data in time sequence is enhanced, thereby improving the overall accuracy of the thunderstorm and strong wind nowcasting model. The existence status of thunderstorms and strong winds is detected in real time and the storm center is matched to obtain the thunderstorm and strong wind near data to be identified. The thunderstorm and strong wind near data to be identified is input into the pre-trained thunderstorm and strong wind nowcasting model to obtain the thunderstorm and strong wind nowcasting forecast. The present invention greatly improves the accuracy and timeliness of the near probability forecast of thunderstorms and strong winds.

[0098] See also Figure 2 , which is a flow chart of a method for forecasting the approaching probability of thunderstorm and gale proposed in a second embodiment of the present invention, the method for forecasting the approaching probability of thunderstorm and gale comprises steps S11 to S15, wherein:

[0099] Step S11: Select the thunderstorm and gale historical data in the current area according to the preset thunderstorm and gale detection standard, screen the test data in the thunderstorm and gale historical data, perform quality control on the remaining data in the thunderstorm and gale historical data according to the quality control code of the test data, and obtain the meteorological radar of each thunderstorm and gale historical data case in the thunderstorm and gale historical data. WILL Figure 2. Weather radar WILL The image is matched with the corresponding historical data of thunderstorm and gale to complete the preprocessing of each image and point by point. WILL The graph is constructed as a pre-training dataset for the thunderstorm and gale nowcasting model;

[0100] It should be noted that the preset thunderstorm and gale detection standard in this embodiment presets the wind speed threshold and the wind zone range threshold, the preset wind speed threshold is 17.2 m / s, the preset wind zone range threshold is detected by five or more observation stations, and the quality control code in this embodiment is as shown in Table 1:

[0101] Table 1

[0102]

[0103] The test data in this embodiment is the 1h maximum wind speed observation data of the automatic weather station. WILL The thunderstorm and gale historical data are matched with the preset storm center matching rules. The storm grid point corresponding to the current storm center is the target storm with thunderstorm and gale. WILL Value greater than 5kg / m 2 The value of is used to distinguish the boundary of other storms to determine the spatial scope of the target storm.

[0104] Step S12: pre-training the thunderstorm and gale nowcasting model, and performing optimal testing on the pre-trained thunderstorm and gale nowcasting model;

[0105] It should be noted that in this embodiment, the pre-trained thunderstorm and gale nowcasting model is optimally tested;

[0106] The formula for the optimal test is as follows:

[0107] ;

[0108] in, CSI represents the critical success index, HSS Represents the rating index, TP It represents true yang. United Nations Indicates false vagina. FP Indicates false positive. TN Indicates true yin.

[0109] Step S13: Obtain the strong wind observation data of all single observation stations in the current area, screen the detection data in the strong wind observation data, perform quality control on the remaining data in the strong wind observation data according to the quality control code of the detection data, and determine whether the strong wind observation data after quality control is greater than or equal to the preset wind speed threshold. If the strong wind observation data is greater than or equal to the preset wind speed threshold, it is determined that there is strong wind in the observation station area corresponding to the strong wind observation data, and obtain the meteorological radar of the current area. WILL Figure 1. Weather radars in observation station areas with strong winds according to the preset storm center matching rules. WILLThe storm center is matched with the map. If the storm center is matched, it is determined that there is thunderstorm and strong wind in the observation station area, and the strong wind observation data corresponding to the observation station area and the meteorological radar are combined. WILL Figure 2 is used as the approaching data of thunderstorm and gale to be identified;

[0110] It should be noted that the preset storm center matching rule in this embodiment is specifically that for any strong wind observation data, according to its longitude and latitude, the corresponding meteorological radar WILL In the image, a search is conducted within a 10km radius around the meteorological station to determine whether there is a particle with a mass greater than 40kg / m 2 If the value is greater than 40kg / m 2 If the value is greater than , the result is that the storm center is matched, otherwise it is not matched.

[0111] Step S14: inputting the to-be-identified thunderstorm and gale near-falling data into the pre-trained thunderstorm and gale near-falling forecast model, wherein the thunderstorm and gale near-falling forecast model comprises a convolutional network sub-model and a long short-term memory network sub-model, wherein the convolutional network sub-model extracts basic features of the thunderstorm and gale data, and the long short-term memory network sub-model performs feature enhancement, calculates and outputs the thunderstorm and gale forecast data;

[0112] It should be noted that the thunderstorm and gale nowcasting model in this embodiment outputs three types of thunderstorm and gale forecast data, including the spatial distribution of the maximum value of thunderstorm and gale A 3. Spatial distribution of duration of thunderstorm and gale B , Spatial distribution of the moving trajectory of the storm center C The spatial distribution of the maximum value of thunderstorm and gale is used to judge whether thunderstorm and gale occur within a spatial range. It is generally believed that thunderstorm and gale are closely related to the main body of the storm. WILL The value is large, so it can be considered WILL The range greater than or equal to 35 is the area where thunderstorms and gale may occur. The duration of thunderstorms and gale and the spatial distribution of the moving trajectory of the storm center are the main basis for probability judgment. In a storm, the greater the wind speed, the faster the storm moves. WILL The duration greater than or equal to 35 is negatively correlated with the probability of strong winds. The location of strong winds is also closely related to the storm center. The closer to the center, the more likely strong winds are to occur. Therefore, it is also negatively correlated with distance.

[0113] Step S15: Calculate the probability of approaching thunderstorm and gale forecast data according to the preset probability division rule, and then obtain the thunderstorm and gale nowcast according to the preset wind force level standard and the preset thunderstorm and gale warning signal standard;

[0114] It should be noted that the preset wind force level standards in this embodiment are as shown in Table 2 below:

[0115] Table 2

[0116]

[0117] The preset thunderstorm and gale warning signal standards in this embodiment include:

[0118] Yellow warning signal: The area may be affected by thunderstorm and gale within 6 hours, with average wind speed of 8 or above, or gusts of 9 or above accompanied by strong lightning, or the area has been affected by thunderstorm and gale, with average wind speed of 8-9, or gusts of 9-10 accompanied by strong lightning, and may continue;

[0119] Orange warning signal: The area may be affected by thunderstorm and gale within 2 hours, with average wind speed of 10 or above, or gusts of 11 or above, accompanied by strong lightning; or the area has been affected by thunderstorm and gale, with average wind speed of 10-11, or gusts of 11-12, accompanied by strong lightning, and may continue;

[0120] Red warning signal: The area may be affected by thunderstorms and strong winds within 2 hours, with the average wind speed reaching above level 12 accompanied by strong lightning; or the area has been affected by thunderstorms and strong winds, with the average wind speed being above level 12 accompanied by strong lightning, and may continue.

[0121] In summary, based on the above-mentioned thunderstorm and gale near-probability forecasting method, a thunderstorm and gale near-forecasting model is designed. WILL The maximum value, duration and spatial distribution of the moving trajectory of the thunderstorm and gale within the near time are measured and processed in the figure, so that the near probability of the thunderstorm and gale is quickly and accurately calculated within the near time, so as to make a nowcast of the thunderstorm and gale. The present invention greatly improves the accuracy and timeliness of the forecast of the near probability of the thunderstorm and gale. Specifically, a pre-training data set is constructed according to the historical data of thunderstorms and strong winds in the current area, and a thunderstorm and strong wind nowcasting model is pre-trained. The thunderstorm and strong wind nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model. The model is pre-trained by the historical data of thunderstorms and strong winds in the current area, and the accuracy of the thunderstorm and strong wind nowcasting model is further optimized and improved. The convolutional network sub-model is designed to extract basic features, which improves the processing speed and robustness of the overall model. The long short-term memory network sub-model is designed to enhance features. For various types of meteorological data with strong temporal sequence, the long-term dependency between data in time sequence is enhanced, thereby improving the overall accuracy of the thunderstorm and strong wind nowcasting model. The existence status of thunderstorms and strong winds is detected in real time and the storm center is matched to obtain the thunderstorm and strong wind near data to be identified. The thunderstorm and strong wind near data to be identified is input into the pre-trained thunderstorm and strong wind nowcasting model to obtain the thunderstorm and strong wind nowcasting forecast. The present invention greatly improves the accuracy and timeliness of the near probability forecast of thunderstorms and strong winds.

[0122] See also Figure 3, which is a schematic diagram of the structure of a thunderstorm and gale approach probability forecasting system proposed in the third embodiment of the present invention, and the system comprises:

[0123] A pre-training module 10 is used to construct a pre-training data set based on the historical data of thunderstorms and gale in the current area, and to pre-train a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model;

[0124] The data processing module 20 is used to detect the existence of thunderstorms and gale in real time and match the storm center to obtain the nearby data of thunderstorms and gale to be identified;

[0125] The forecast acquisition module 30 is used to input the thunderstorm and gale near-falling data to be identified into the pre-trained thunderstorm and gale near-falling forecast model to obtain the thunderstorm and gale near-falling forecast.

[0126] The present invention also provides a computer storage medium on which one or more programs are stored. When the programs are executed by a processor, the above-mentioned method for forecasting the impending probability of thunderstorms and gales is implemented.

[0127] The present invention also proposes a computer device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the above-mentioned method for predicting the impending probability of thunderstorms and gales.

[0128] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, "computer-readable medium" may be any device that can contain storage, communication, propagation or transmission of a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0129] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0130] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0131] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0132] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for predicting the probability of approaching thunderstorm and gale, characterized in that: include: Constructing a pre-training data set based on historical data of thunderstorms and gale in the current area, and pre-training a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model; The step of constructing a pre-training data set based on the historical data of thunderstorms and gales in the current area specifically includes: Selecting historical data of thunderstorms and strong winds in the current area according to a preset thunderstorm and strong wind detection standard, wherein the preset thunderstorm and strong wind detection standard includes a preset wind speed threshold and a preset wind area range threshold; Screening the inspection data in the thunderstorm and gale historical data to perform quality control on the remaining data in the thunderstorm and gale historical data according to the quality control code of the inspection data; Obtain the weather radar of each thunderstorm and gale historical data instance in the thunderstorm and gale historical data VIL Figure 2 shows the weather radar VIL The graph is matched with the corresponding thunderstorm and gale historical data to complete the graph-by-graph and point-by-point preprocessing. The preprocessing formula is as follows: ； in, Represents an unprocessed data value, represents the preprocessed data value, a Indicates weather radar VIL The figure number of the figure group, m represents the longitudinal grid point number, n represents the latitudinal grid point number, S A grid point set representing a single thunderstorm and gale historical data case after matching; The preprocessed weather radar VIL The graph is constructed as a pre-training dataset for the thunderstorm and gale nowcasting model; Real-time detection of the existence of thunderstorms and gale winds and matching the storm center to obtain the nearby data of thunderstorms and gale winds to be identified; The thunderstorm and gale near-coming data to be identified is input into the pre-trained thunderstorm and gale near-coming forecast model to obtain the thunderstorm and gale near-coming forecast.

2. The method for predicting the impending probability of thunderstorm and gale according to claim 1, characterized in that: The thunderstorm and gale nowcasting model specifically includes: The thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model; The convolutional network sub-model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer. The convolutional network sub-model performs basic feature extraction of thunderstorm and gale data. The thunderstorm and gale data includes historical thunderstorm and gale data and nearby thunderstorm and gale data. The formula for extracting the basic features is as follows: ； in, y represents the feature output of the convolutional layer, l represents the number of convolutional layers, i represents the ordinal number of the feature output, j Represents the number of feature outputs, represents the activation function, k Represents the convolution kernel index value, K Represents the number of outputs of the convolutional layer, represents the convolution kernel weight, x represents the output vector of the previous layer, b Indicates bias, p represents the output of the pooling layer, r represents the pooling window, R represents the pooling layer kernel size, T represents the step size of the pooling layer, represents the maximum value function; The long short-term memory network sub-model includes an input gate, a forget gate and an output gate. The long short-term memory network sub-model performs feature enhancement. The formula for the feature enhancement is as follows: ； in, in represents the input gate, f represents the forget gate, o represents the output gate, Represents the input at the current moment, represents the activation function, c Represents a storage memory unit, t Indicates the time, W represents the weight term, U represents the bias term, h Indicates the output quantity, represents the hyperbolic tangent function, Represents the generated candidate value.

3. The method for predicting the impending probability of thunderstorm and gale according to claim 1, characterized in that: The step of pre-training the thunderstorm and gale nowcasting model further includes: Conduct optimal testing on the pre-trained thunderstorm and gale nowcasting model; The formula for the optimal test is as follows: ； in, CSI represents the critical success index, HSS Represents the rating index, TP It represents true yang. FN Indicates false vagina. FP Indicates false positive. TN Indicates true yin.

4. The method for predicting the impending probability of thunderstorm and gale according to claim 1, characterized in that: The step of detecting the existence state of thunderstorm and gale in real time and matching the storm center to obtain the nearby data of thunderstorm and gale to be identified specifically includes: Obtaining the gale observation data of all single observation stations in the current area, screening the detection data in the gale observation data, and performing quality control on the remaining data in the gale observation data according to the quality control code of the detection data; Determine whether the high wind observation data after quality control is greater than or equal to a preset wind speed threshold, and if the high wind observation data is greater than or equal to the preset wind speed threshold, determine that there is high wind in the observation site area corresponding to the high wind observation data; Get the weather radar for the current area VIL Figure 1. Weather radar VIL The graph is defined as follows: ； in, VIL represents the vertical cumulative liquid water content, Z represents the radar reflectivity numerator, H Indicates the maximum altitude detected by the radar; According to the preset storm center matching rule, the weather radar in the observation station area with strong winds VIL The storm center is matched with the image. If the storm center is matched, it is determined that there is thunderstorm and strong wind in the observation site area, and the strong wind observation data corresponding to the observation site area and the meteorological radar are compared. VIL The figure is used as the nearby data of thunderstorm and gale to be identified.

5. The method for predicting the impending probability of thunderstorm and gale according to claim 1, characterized in that: The step of inputting the to-be-identified thunderstorm and gale near-falling data into the pre-trained thunderstorm and gale near-falling forecast model to obtain the thunderstorm and gale near-falling forecast specifically includes: The thunderstorm and gale near-coming data to be identified is input into the pre-trained thunderstorm and gale near-coming forecast model, and the thunderstorm and gale near-coming forecast model calculates and outputs thunderstorm and gale forecast data, and the thunderstorm and gale forecast data includes the thunderstorm and gale maximum spatial distribution data A , spatial distribution data of thunderstorm and gale duration B , spatial distribution data of the moving track of the storm center C The formula for calculating and outputting the thunderstorm and gale forecast data is as follows: ； in, WIND represents the intermediate discriminant, Indicates continuous output of graph groups VIL quantity, TRACK represents the grid points covered by the storm's moving track, a Indicates weather radar VIL The figure number of the figure group, m represents the longitudinal grid point number, n Indicates the latitude grid point number; The probability of approaching thunderstorm and gale is calculated based on the thunderstorm and gale forecast data to obtain a thunderstorm and gale nowcast.

6. The method for predicting the impending probability of thunderstorm and gale according to claim 5, characterized in that: The step of calculating the probability of approaching thunderstorm and gale according to the thunderstorm and gale forecast data to obtain a thunderstorm and gale nowcast specifically includes: The probability of approaching thunderstorm and gale forecast data is calculated according to the preset probability division rule, and the preset probability division rule is as follows: ； in, Pwind Indicates the probability of approaching thunderstorms and gales. Represents the straight-line distance function from the grid point to the trajectory; Then, a thunderstorm and gale nowcast is obtained based on the preset wind force level standard and the preset thunderstorm and gale warning signal standard.

7. A thunderstorm and gale approach probability forecasting system, characterized in that: include: A pre-training module, used to construct a pre-training data set based on the historical data of thunderstorms and gale in the current area, and to pre-train a thunderstorm and gale nowcasting model, wherein the thunderstorm and gale nowcasting model includes a convolutional network sub-model and a long short-term memory network sub-model; The step of constructing a pre-training data set based on the historical data of thunderstorms and gales in the current area specifically includes: Selecting historical data of thunderstorms and strong winds in the current area according to a preset thunderstorm and strong wind detection standard, wherein the preset thunderstorm and strong wind detection standard includes a preset wind speed threshold and a preset wind area range threshold; Screening the inspection data in the thunderstorm and gale historical data to perform quality control on the remaining data in the thunderstorm and gale historical data according to the quality control code of the inspection data; Obtain the weather radar of each thunderstorm and gale historical data instance in the thunderstorm and gale historical data VIL Figure 2 shows the weather radar VIL The graph is matched with the corresponding thunderstorm and gale historical data to complete the graph-by-graph and point-by-point preprocessing. The preprocessing formula is as follows: ； in, Represents an unprocessed data value, represents the preprocessed data value, a Indicates weather radar VIL The figure number of the figure group, m represents the longitudinal grid point number, n represents the latitudinal grid point number, S A grid point set representing a single thunderstorm and gale historical data case after matching; The preprocessed weather radar VIL The graph is constructed as a pre-training dataset for the thunderstorm and gale nowcasting model; The data processing module is used to detect the existence of thunderstorms and gale in real time and match the storm center to obtain the nearby data of thunderstorms and gale to be identified; The forecast acquisition module is used to input the thunderstorm and gale near-coming data to be identified into the pre-trained thunderstorm and gale near-coming forecast model to obtain the thunderstorm and gale near-coming forecast.

8. A storage medium, characterized in that: The storage medium stores one or more programs, which, when executed by the processor, implement the method for probabilistic forecasting of thunderstorms and gales as described in any one of claims 1 to 6.

9. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein: The memory is used to store computer programs; When the processor is used to execute the computer program stored in the memory, the method for probabilistic forecasting of thunderstorms and gales described in any one of claims 1 to 6 is implemented.

Images