Meteorological data analysis method and system for hail monitoring and identification

Abstract

The invention discloses a meteorological data analysis method and system for hail monitoring and recognition, and relates to the technical field of civil aviation meteorological monitoring and warning, and the system comprises a data collection module, a feature extraction module, a model construction module and a model training module, echo information of meteorological radar and satellite remote sensing data. The beneficial effects of the application are that meteorological characteristics related to hail formation are collected and extracted, correlation between echo information of a meteorological radar and hail data in historical observation data is analyzed, a hail monitoring model is constructed, a prediction result of the model is combined with data fed back by meteorological early warning, and the hail early warning feedback is obtained according to feedback in practical application. According to the method, the model is continuously iterated and optimized, the monitoring accuracy and real-time performance are improved, the hail monitoring and recognition efficiency can be remarkably improved through implementation of the comprehensive technical route, and powerful technical support is provided for meteorological early warning, disaster prevention and emergency response.

Description

Technical Field

[0001] The present invention relates to the technical field of civil aviation meteorological monitoring and warning, in particular to a meteorological data analysis method and system for hail monitoring and identification. Background Art

[0002] Hail is a solid precipitation phenomenon that occurs under strong convection conditions. It is characterized by strong locality, short duration, significant influence of terrain, and high intensity. It is often accompanied by weather processes such as strong winds and heavy rains. Complex terrain causes frequent intersection of cold and warm air, and frequent hail disasters, which bring huge losses to agriculture, transportation and other aspects.

[0003] Due to the strong updraft, warm liquid particles are brought from the lower layer to the middle layer to form supercooled cloud rain. Most of the hailstorm warnings in the existing technology focus on the qualitative analysis and exploration of observation data at the macro level, but rarely involve the forecast indication of the hail embryo formation stage during the hailstorm process, and cannot accurately obtain the quantitative characteristic forecast indicators for the hail embryo stage during the development of hail clouds from the micro level. Summary of the invention

[0004] The purpose of this section is to provide a meteorological data analysis method and system for hail monitoring and identification, which can improve the efficiency of hail monitoring and identification.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a meteorological data analysis method for hail monitoring and identification, comprising the following steps: S100, collecting historical observation data of ground meteorological stations, echo information of meteorological radars and satellite remote sensing data; S200, extracting meteorological features related to hail formation from the collected data through image processing and signal processing technology, and constructing grid distribution data for training hail monitoring algorithms; S300, analyzing the correlation between the echo information of meteorological radars and hail data in historical observation data, and constructing a hail monitoring model; S400, combining the prediction results of the hail monitoring model with the feedback data of meteorological warnings, and continuously iterating and optimizing the hail monitoring model according to feedback in actual applications.

[0006] As a preferred embodiment of the meteorological data analysis method for hail monitoring and identification of the present invention, in S200, the meteorological characteristics related to hail formation include temperature distribution, humidity, wind speed, wind direction and radar reflectivity of the cloud layer.

[0007] As a preferred embodiment of the meteorological data analysis method for hail monitoring and identification of the present invention, in S200, the specific steps are as follows: S201, multiple grid points are set in an area, multiple observation points are set in each grid point, and a time period U is set. In the time period, whether a hail event occurs at each observation point is recorded to obtain discrete distribution data of whether a hail event occurs in the area; S202, a judgment value of whether hail occurs is set to L, L is in the numerical range [0,1], the larger the value of L, the larger the magnitude of hail, and when L=0, no hail event occurs; S203, any observation point in the grid where a hail event occurs is taken as the center point, the judgment value L is set to 1, and the corresponding attenuation value ΔL is calculated according to the distance r between the remaining observation points in the grid point and the center point, and the formula is:

[0008]

[0009] Among them, the value range of ΔL is between 0-1, and the unit of r is meter; S204, calculate and record the influence relationship between all observation points in each grid point according to the above formula, obtain multiple attenuation values ​​of each grid point from different observation points, take the arithmetic mean as the judgment value of the grid point, and obtain the grid distribution data with a value interval of [0,1].

[0010] As a preferred embodiment of the meteorological data analysis method for hail monitoring and identification of the present invention, in S300, the specific steps are as follows: S301, respectively establish a data group for each grid point, each data group includes the hail magnitude x monitored by the meteorological radar and the hail magnitude y observed in real time by the algorithm, and the difference between the two is set to d, according to the formula:

[0011]

[0012] Calculate the correlation coefficient ρ between x and y, where d iis the difference in hail magnitude in the ith data group, and n is the observation sample size; S302, set the threshold m. When the observation sample size is not less than m, compare the value of ρ with 0.05 to test the correlation between x and y; S303, set the large radar and the small radar, take the timestamp of the large radar as the reference point, and perform data optical flow extrapolation for several minutes on the small radar for time alignment. Set the distance weight based on the distance between each observation point and different radars, and calculate the average value based on the distance weight to ensure the continuity of data distribution, and unify the radar data of different bands into a single input matrix; S304, apply convolutional neural network (CNN) and long short-term memory network (LSTM) to build a multi-layer neural network model, use a large amount of historical hail event data for supervised learning, and expand the training set through data enhancement technology to prevent overfitting. Adopt the attention mechanism to build a module that integrates global and local feature information, and finally carry out hail monitoring and recognition based on the adaptively improved semantic segmentation model.

[0013] As a preferred embodiment of the meteorological data analysis method for hail monitoring and identification described in the present invention, in S400, the feedback data of the meteorological warning refers to the meteorological information released by the meteorological department for reminding and alerting the public and relevant departments.

[0014] As a preferred solution of the meteorological data analysis system for hail monitoring and identification described in the present invention, it includes: a data acquisition module, a feature extraction module, a model construction module and a model training module.

[0015] As a preferred solution of the meteorological data analysis system for hail monitoring and identification described in the present invention, the data acquisition module is used to collect historical observation data of ground meteorological stations, echo information of meteorological radars and satellite remote sensing data.

[0016] As a preferred solution of the meteorological data analysis system for hail monitoring and identification described in the present invention, the feature extraction module is used to extract meteorological features related to hail formation in the collected data.

[0017] As a preferred solution of the meteorological data analysis system for hail monitoring and identification described in the present invention, the model building module is used to analyze the correlation between the echo information of the meteorological radar and the hail data in the historical observation data, and to build a hail monitoring model.

[0018] As a preferred solution of the meteorological data analysis system for hail monitoring and identification described in the present invention, the model training module trains the hail monitoring algorithm according to the grid distribution data to optimize the hail monitoring model.

[0019] Beneficial effects of the present invention:

[0020] 1. The present invention collects and extracts meteorological features related to hail formation, analyzes the correlation between the echo information of the meteorological radar and the hail data in the historical observation data, constructs a hail monitoring model, combines the prediction results of the model with the data of meteorological warning feedback, and continuously iterates and optimizes the model according to the feedback in practical applications to improve the accuracy and real-time performance of monitoring. Through the implementation of this comprehensive technical route, the monitoring and identification efficiency of hail can be significantly improved, providing strong technical support for meteorological warning, disaster prevention and emergency response.

[0021] 2. The present invention integrates high-precision observation data from ground weather stations, echo information from weather radars, and satellite remote sensing data such as wind and cloud. Using these diverse data sources, the meteorological changes in the basin can be effectively captured, providing a comprehensive meteorological background for subsequent analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creativity and labor. Among them:

[0023] Figure 1 is a flow chart of the analysis method of the present invention;

[0024] Figure 2 Flow chart of the analysis system of the present invention. DETAILED DESCRIPTION

[0025] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below with reference to the accompanying drawings.

[0026] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from this description. The "embodiment" referred to here refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention.

[0027] Embodiment 1

[0028] Reference Figure 1This embodiment provides a meteorological data analysis method for hail monitoring and identification, which specifically includes the following steps: S100, collecting historical observation data of ground meteorological stations, echo information of meteorological radars, and satellite remote sensing data; S200, extracting meteorological features related to hail formation from the collected data through image processing and signal processing technology, and constructing grid distribution data for training hail monitoring algorithms; S300, analyzing the correlation between the echo information of meteorological radars and hail data in historical observation data, and constructing a hail monitoring model; S400, combining the prediction results of the hail monitoring model with the feedback data of meteorological warnings, and continuously iterating and optimizing the hail monitoring model based on feedback in actual applications.

[0029] In S200 , meteorological characteristics related to hail formation include temperature distribution, humidity, wind speed, wind direction, and radar reflectivity of the cloud layer.

[0030] In S200, the specific steps are as follows: S201, multiple grid points are set in an area, multiple observation points are set in each grid point, and a time period U is set. In the time period, whether a hail event occurs at each observation point is recorded to obtain discrete distribution data of whether a hail event occurs in the area; S202, a judgment value of whether hail occurs is set to L, and L is in the numerical range [0,1]. The larger the value of L, the larger the magnitude of hail. When L=0, no hail event occurs; S203, any observation point in the grid where a hail event occurs is taken as the center point, the judgment value L is set to 1, and the corresponding attenuation value ΔL is calculated according to the distance r between the remaining observation points in the grid and the center point. The formula is:

[0031]

[0032] Among them, the value range of ΔL is between 0-1, and the unit of r is meter; S204, calculate and record the influence relationship between all observation points in each grid point according to the above formula, obtain multiple attenuation values ​​of each grid point from different observation points, take the arithmetic mean as the judgment value of the grid point, and obtain the grid distribution data with a value interval of [0,1].

[0033] In S300, the specific steps are as follows: S301, establish a data group for each grid point, each data group includes the hail magnitude x monitored by the meteorological radar and the hail magnitude y observed in real time by the algorithm, and the difference between the two is set to d, according to the formula:

[0034]

[0035] Calculate the correlation coefficient ρ between x and y, where d i is the difference in hail magnitude in the ith data group, and n is the observation sample size.

[0036] The above formula is a simplified formula. The correlation measurement method in this algorithm is the Spearman correlation coefficient, which is a non-parametric indicator used to measure the dependence of two variables. The monotonic equation is used to evaluate the correlation of two statistical variables, that is, to check to what extent the two variables keep pace in their changing trends. If there are no repeated values ​​in the data and when the two variables are completely monotonically correlated, the Spearman correlation coefficient is +1 or -1. For a sample with a sample size of n, n original data are converted into hailstone magnitude data, and the calculation formula of the correlation coefficient ρ is:

[0037]

[0038] Among them, x i and i are the hail magnitude monitored by the meteorological radar and the hail magnitude observed in real time in the i-th data group, and Respectively represent the average magnitude.

[0039] S302. Set the threshold m. When the observation sample capacity is not less than m, compare the value of ρ with 0.05 to check the correlation between x and y. S303. Set the large radar and the small radar. Take the timestamp of the large radar as the reference point. The small radar performs data optical flow extrapolation for several minutes for time alignment. Set the distance weight according to the distance between each observation point and different radars. Calculate the average value according to the distance weight to ensure the continuity of data distribution and unify radar data of different bands into a single input matrix. S304. Apply convolutional neural network (CNN) and long short-term memory network (LSTM) to build a multi-layer neural network model. Use a large amount of historical hail event data for supervised learning. Use data enhancement technology to expand the training set to prevent overfitting. Use the attention mechanism to build a module that integrates global and local feature information. Finally, hail monitoring and identification is carried out based on the adaptively improved semantic segmentation model.

[0040] This algorithm combines correlation analysis with feature selection and uses an embedded feature selection method to perform data correlation analysis. This method combines feature selection with the model training process and achieves feature selection and model optimization at the same time by penalizing features during the model training process.

[0041] The input data grid is established by multi-band radar data fusion, and the encoder-decoder structure is based on the Efficient-ResNet efficient residual network. The attention enhancement gate and attention enhancement module are specially added for the project task to extract the features of radar data.

[0042] The main components of Efficient-ResNet include (the convolutional layer and residual connection are mainly used to extract radar data features): Convolutional Layer: used to learn image features. Residual Connection: used to connect the input and output layers to avoid gradient disappearance. Pooling Layer: used to reduce the size of the image. Fully Connected Layer: used to classify the image. Scaling Factor: used to adjust the width and depth of the network.

[0043] The input data of each physical quantity of radar data are of the same dimension. Therefore, taking one physical quantity as an example, the radar echo data within a certain period of time is input as a three-dimensional matrix. The matrix is ​​assumed to be 224*224*3 (indicating that the radar data is a 224*224 grid, and a total of 3 data are selected within the time period. This value will be adjusted based on the actual situation). After passing through a 7x7 convolution kernel and a convolution with a step size of 2, a convolution layer with a size of 112x112 and 64 channels is obtained.

[0044] In S400, the feedback data of the meteorological warning refers to the meteorological information released by the meteorological department for reminding and alerting the public and relevant departments.

[0045] Embodiment 2

[0046] Reference Figure 2 This embodiment provides a meteorological data analysis system for hail monitoring and identification, which specifically includes a data acquisition module, a feature extraction module, a model construction module and a model training module.

[0047] The data acquisition module is used to collect historical observation data from ground meteorological stations, echo information from meteorological radars, and satellite remote sensing data.

[0048] The feature extraction module is used to extract meteorological features related to hail formation from the collected data.

[0049] The model building module is used to analyze the correlation between the echo information of the meteorological radar and the hail data in the historical observation data, and to build a hail monitoring model.

[0050] The model training module trains the hail monitoring algorithm based on the grid distribution data and optimizes the hail monitoring model.

[0051] Importantly, although only a few embodiments are described in detail in this disclosure, it should be readily understood by those who refer to this disclosure that many modifications are possible without substantially departing from the subject matter described in this application, such as the size, structure, shape and proportion of the various elements, as well as temperature, pressure, mounting arrangements, use of materials, color, directional changes, etc.; for example, an element shown as integrally formed may be composed of multiple parts or elements, and the position of the element may be inverted or otherwise changed; therefore, all such modifications should be included within the scope of the present invention, and other substitutions, modifications, changes and omissions may be made in the design, operating conditions and arrangement of the exemplary embodiments without departing from the scope of the present invention.

Claims

1. A meteorological data analysis method for hail monitoring and identification, characterized in that: The following steps are involved: S100, collect historical observation data from ground weather stations, echo information from weather radars, and satellite remote sensing data; S200, extracting meteorological features related to hail formation from the collected data through image processing and signal processing technology, and constructing grid distribution data for training hail monitoring algorithms; S300, analyzing the correlation between the echo information of the meteorological radar and the hail data in the historical observation data, and constructing a hail monitoring model; S400, combining the prediction results of the hail monitoring model with the feedback data of the meteorological warning, and continuously iterating and optimizing the hail monitoring model according to the feedback in the actual application.

2. The meteorological data analysis method for hail monitoring and identification according to claim 1, characterized in that: In S200 , meteorological characteristics related to hail formation include temperature distribution, humidity, wind speed, wind direction, and radar reflectivity of the cloud layer.

3. The meteorological data analysis method for hail monitoring and identification according to claim 2, characterized in that: In S200, the specific steps are as follows: S201, setting a plurality of grid points in an area, setting a plurality of observation points in each grid point, setting a time period U, recording whether a hail event occurs at each observation point in the time period, and obtaining discrete distribution data of whether a hail event occurs in the area; S202, setting a judgment value of whether hail occurs to L, where L is in the numerical range [0,1]. The larger the value of L is, the larger the magnitude of the hail is. When L=0, no hail event occurs. S203, taking any observation point where a hail event occurs in the grid as the center point, setting the determination value L to 1, and calculating the corresponding attenuation value ΔL according to the distance r between the remaining observation points in the grid and the center point, the formula is: Among them, the value range of ΔL is between 0 and 1, and the unit of r is meter; S204. Calculate and record the influence relationship between all observation points in each grid point according to the above formula, obtain multiple attenuation values ​​of each grid point from different observation points, take the arithmetic mean as the judgment value of the grid point, and obtain grid distribution data with a value interval of [0,1].

4. The meteorological data analysis method for hail monitoring and identification according to claim 3, characterized in that: In S300, the specific steps are as follows: S301, respectively establish a data group for each grid point, each data group includes the hail magnitude x monitored by the weather radar and the hail magnitude y observed in real time by the algorithm, and the difference between the two is set to d, according to the formula: Calculate the correlation coefficient ρ between x and y, where d i is the difference in hail magnitude in the ith data set, and n is the observation sample size; S302, set a threshold m, and when the observed sample size is not less than m, compare the value of ρ with 0.05 to test the correlation between x and y; S303, setting a large radar and a small radar, taking the timestamp of the large radar as the reference point, and performing data optical flow extrapolation for several minutes on the small radar for time alignment, setting distance weights based on the distances between each observation point and different radars, and calculating the average value based on the distance weights to ensure the continuity of data distribution, and unifying radar data of different bands into a single input matrix; S304. A multi-layer neural network model was constructed using convolutional neural network (CNN) and long short-term memory network (LSTM). A large amount of historical hail event data was used for supervised learning. The training set was expanded through data enhancement technology to prevent overfitting. An attention mechanism was used to construct a module that integrated global and local feature information. Finally, hail monitoring and identification was performed based on an adaptively improved semantic segmentation model.

5. The meteorological data analysis method for hail monitoring and identification according to claim 1, characterized in that: In S400, the feedback data of the meteorological warning refers to the meteorological information released by the meteorological department for reminding and alerting the public and relevant departments.

6. A meteorological data analysis system for hail monitoring and identification, using the meteorological data analysis method for hail monitoring and identification according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, feature extraction module, model building module and model training module.

7. The meteorological data analysis system for hail monitoring and identification according to claim 6, characterized in that: The data acquisition module is used to collect historical observation data from ground weather stations, echo information from weather radars, and satellite remote sensing data.

8. The meteorological data analysis system for hail monitoring and identification according to claim 6, characterized in that: The feature extraction module is used to extract meteorological features related to hail formation from the collected data.

9. The meteorological data analysis system for hail monitoring and identification according to claim 6, characterized in that: The model building module is used to analyze the correlation between the echo information of the weather radar and the hail data in the historical observation data, and to build a hail monitoring model.

10. The meteorological data analysis system for hail monitoring and identification according to claim 6, characterized in that: The model training module trains the hail monitoring algorithm according to the grid distribution data and optimizes the hail monitoring model.

Images