A method for generating high-density regions of lightning

Abstract

The present application relates to a kind of lightning high-density data generation method, belong to lightning nowcasting technical field.The method first establishes ground-based wide-area lightning location network;Lightning location point is regarded as a piece of area with probability of lightning, constructs the expression of probability distribution of lightning occurrence;Superimpose the lightning density corresponding to lightning location point in a time step, the data obtained after superimposition represents the number of lightning that can occur in a time step in the area regarded by the lightning location point in statistics.Lightning density in the area regarded by the lightning location point is obtained by mixing calculation using weight method, and the lightning density image and lightning high-density area in the time period of the monitoring range are obtained.The present application optimizes the problems of too large grid, rough resolution and serious numerical mutation in traditional lightning frequency statistical data, and finally solves the problem of lack of suitable data form, which has good application prospect in lightning nowcasting field.

Description

Technical Field

[0001] This invention belongs to the field of lightning nowcasting technology, specifically relating to a method for generating high-density lightning regions. Background Technology

[0002] Nowcasting is a weather forecasting method used to predict weather changes over the next few hours. It is based on the movement and evolution of weather systems, inputting past observational data into a forecasting model to obtain the forecast results. Extrapolation of high-density lightning areas is a crucial step in this process.

[0003] Deep neural networks are machine learning models that learn and extract complex features from input data through connections and weight adjustments between multiple layers of neurons. Although deep neural networks are widely used in nowcasting, under current technological conditions, there is still a lack of suitable data formats for extrapolating high-density lightning areas using them. Existing deep neural networks for nowcasting accept gridded data as input, which is incompatible with scattered lightning location data. Therefore, to utilize artificial intelligence techniques to apply discrete lightning location data to nowcasting, it is essential to grid the lightning data. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and apply discrete lightning location data to near-term warning and forecasting, and to propose a method for generating high-density lightning regions.

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

[0006] A method for generating high-density lightning regions includes the following steps:

[0007] Step (1): Within the monitoring range, establish a ground-based wide-area lightning location network. The ground-based wide-area lightning location network includes multiple base stations, each of which includes a very low frequency vertical polarized antenna and a fast electric field change detector for location.

[0008] Step (2): Based on the data collected by establishing a ground-based wide-area lightning location network, the lightning location point is regarded as an area where lightning is likely to occur, and a spatial distribution expression of lightning density is constructed.

[0009] Step (3): Superimpose the lightning density corresponding to the lightning location point in a time step. The superimposed data represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. Using the lightning occurrence threshold, judge based on the superimposed data to obtain the high-density lightning area within this time step.

[0010] Step (4): For a detection time period containing n time steps, the weighted method is used to mix the data obtained by superimposing the current time step with the data obtained by superimposing the previous n-1 time steps to obtain the final mixed lightning density of the region regarded as the lightning location point; based on the final mixed lightning density of the region regarded as each lightning location point, the lightning density image of the monitoring range in this time period is obtained, thereby obtaining the high-density lightning region from the lightning density image of the monitoring range.

[0011] Furthermore, preferably, the specific method of step (2) is as follows:

[0012] For any given lightning location, there is a probability distribution of lightning occurrence centered on that location; a two-dimensional Gaussian impulse function is used as the spatial distribution of lightning density.

[0013] Setting the peak value of the Gaussian distribution to 1 and the mean value to 0, we obtain the following spatial distribution expression for lightning density:

[0014] r = x 2 +y 2

[0015]

[0016] In this system, the origin of the coordinate system is the location where the lightning occurs, r represents the distance from any point to the origin, x and y represent the two-dimensional direct coordinates of that point, σ represents the Gaussian pulse width, and G(x,y) is the lightning density value.

[0017] Furthermore, preferably, r is less than 50km.

[0018] Furthermore, preferably, the specific method of step (3) is as follows:

[0019] The lightning density corresponding to the lightning location point within a time step is superimposed. The data obtained after superposition represents the number of times that lightning may occur statistically in the area regarded as the lightning location point within a time step.

[0020] Set a threshold for the number of lightning occurrences. If the data obtained after aggregation exceeds the threshold, the threshold represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. The area considered as the lightning location point is then the high-density lightning area within this time step.

[0021] If the threshold is not exceeded, the superimposed data represents the number of times a lightning strike could statistically occur within a time step in the area considered as the lightning location point.

[0022] Furthermore, preferably, the time step length is set to 60 minutes, with each 6 minutes considered as the same lightning density, and the threshold for the number of lightning occurrences is 25.

[0023] Furthermore, preferably, the specific method of step (4) is as follows:

[0024] For data at a certain time step, n-1 time steps are taken forward, for a total of n time steps, and then weighted and mixed. The calculation method is shown in the following formula:

[0025] (4.1) The attenuation function is set to e -x ;

[0026] (4.2) Following the time sequence, with the current time step as the first time step, normalize each time step and then multiply it by 10 to obtain the intermediate variable k. i ; i takes values ​​of 1, 2, 3, ..., n; that is, the intermediate variable of the current time step is k1, the intermediate variable of the previous time step is k2, the intermediate variable of the previous time step is k3, and so on;

[0027] (4.3) For the weighting coefficient w at the i-th time step i for:

[0028]

[0029] Where n represents the total number of time steps required for mixing, and w i k represents the weight of the i-th time period during mixing. i It is an intermediate variable, P i This represents the data obtained after superimposing the data at the i-th time step of the region considered as the lightning location point; P represents the final mixed lightning density of the region considered as the lightning location point.

[0030] Furthermore, preferably, the lightning density of the area ultimately mixed based on each lightning location point is assigned a corresponding color value. The higher the density, the darker the color, and this value is displayed on the map to obtain a lightning density image of the monitored area during that time period.

[0031] This invention also provides a method for forecasting lightning levels, which uses a deep neural network model for prediction. The lightning density image of the monitored area within the time period obtained by the above method is used as the input of the deep neural network model, and the regional lightning level regarded as the corresponding lightning location point is used as the output for training, thereby obtaining a deep neural network model for forecasting lightning levels.

[0032] In step (1) of this invention, multiple base stations are evenly deployed within the monitoring range, but this is not limited to them. They can also be deployed according to actual needs. This invention does not specifically limit the monitoring range. It can be the whole country, but it can also be a certain region, a certain province, etc. This invention does not limit this.

[0033] When superimposing in step (3) of this invention, the superimposed value can be displayed directly, or the superimposed value can be rounded and the corresponding integer value can be displayed.

[0034] This invention replaces the lightning location point with a two-dimensional Gaussian pulse function. The center position of the Gaussian pulse function (i.e., the location point) represents the place with the highest lightning density. The lightning density decreases as the distance from the center position increases.

[0035] This invention constructs a weighting function that decays over time, causing the intensity of the Gaussian pulse function corresponding to each lightning location point to gradually decrease over time. This invention considers the temporal nature of lightning data, meaning that lightning locations may change significantly over time. Therefore, this invention proposes a temporal hybrid processing method. Each discrete lightning location point is represented by a Gaussian pulse function, and a weighting function that decays over time is constructed, causing the intensity of the Gaussian pulse function corresponding to each lightning location point to gradually decrease over time. This achieves spatiotemporal continuity of lightning density distribution.

[0036] The total high-density distribution of lightning over a given period of time is obtained by superimposing the Gaussian pulse functions corresponding to all lightning location points within that period, taking into account the gradual decay over time.

[0037] In this invention, r can be set to less than 50 km, and σ is also set to a constant according to the actual situation. The larger σ is, the smaller the attenuation of lightning density distribution with increasing distance; conversely, the smaller σ is, the larger it is.

[0038] The purpose of this invention (4.2) is to map n to the range of 1 to 10.

[0039] Compared with the prior art, the beneficial effects of this invention are as follows:

[0040] This invention provides a method for generating high-density lightning data. In the extrapolation of high-density lightning regions, it overcomes the problem that lightning location data cannot be directly input into the prediction model. It also optimizes the problems of excessively large grids, coarse resolution, and severe numerical abrupt changes in traditional lightning frequency statistics data. Ultimately, it solves the problem of the lack of suitable data formats and has a promising application prospect in the field of lightning nowcasting. Attached Figure Description

[0041] Figure 1This is a flowchart illustrating the method for generating high-density lightning regions according to the present invention.

[0042] Figure 2 This is a schematic diagram of the weighting function used in this invention;

[0043] Figure 3 Four-hour lightning density images generated for an application example of the present invention; wherein, (a) is from 7:00 to 8:00 on November 28, 2022; (b) is from 8:00 to 9:00 on November 28, 2022; (c) is from 9:00 to 10:00 on November 28, 2022; and (d) is from 10:00 to 11:00 on November 28, 2022.

[0044] Figure 4 The lightning density image generated on November 27, 2022, is an application example of this invention. Detailed Implementation

[0045] The present invention will now be described in further detail with reference to the embodiments.

[0046] Those skilled in the art will understand that the following embodiments are for illustrative purposes only and should not be construed as limiting the scope of the invention. Where specific techniques or conditions are not specified in the embodiments, they are performed in accordance with the techniques or conditions described in the literature in the field or according to the product instructions. Materials or equipment whose manufacturers are not specified are all conventional products that can be obtained by purchase.

[0047] like Figure 1 As shown, a method for generating high-density lightning regions includes the following steps:

[0048] Step (1): Within the monitoring range, establish a ground-based wide-area lightning location network. The ground-based wide-area lightning location network includes multiple base stations, each of which includes a very low frequency vertical polarized antenna and a fast electric field change detector for location.

[0049] Step (2): Based on the data collected by establishing a ground-based wide-area lightning location network, the lightning location point is regarded as an area where lightning is likely to occur, and a spatial distribution expression of lightning density is constructed.

[0050] Step (3): Superimpose the lightning density corresponding to the lightning location point in a time step. The superimposed data represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. Using the lightning occurrence threshold, judge based on the superimposed data to obtain the high-density lightning area within this time step.

[0051] Step (4): For a detection time period containing n time steps, the weighted method is used to mix the data obtained by superimposing the current time step with the data obtained by superimposing the previous n-1 time steps to obtain the final mixed lightning density of the region regarded as the lightning location point; based on the final mixed lightning density of the region regarded as each lightning location point, the lightning density image of the monitoring range in this time period is obtained, thereby obtaining the high-density lightning region from the lightning density image of the monitoring range.

[0052] The specific method for step (2) is as follows:

[0053] For any given lightning location, there is a probability distribution of lightning occurrence centered on that location; a two-dimensional Gaussian impulse function is used as the spatial distribution of lightning density.

[0054] Setting the peak value of the Gaussian distribution to 1 and the mean value to 0, we obtain the following spatial distribution expression for lightning density:

[0055] r = x 2 +y 2

[0056]

[0057] In this system, the origin of the coordinate system is the location where the lightning occurs, r represents the distance from any point to the origin, x and y represent the two-dimensional direct coordinates of that point, σ represents the Gaussian pulse width, and G(x,y) is the lightning density value.

[0058] Preferably, r is less than 50km.

[0059] The specific method for step (3) is as follows:

[0060] The lightning density corresponding to the lightning location point within a time step is superimposed. The data obtained after superposition represents the number of times that lightning may occur statistically in the area regarded as the lightning location point within a time step.

[0061] Set a threshold for the number of lightning occurrences. If the data obtained after aggregation exceeds the threshold, the threshold represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. The area considered as the lightning location point is then the high-density lightning area within this time step.

[0062] If the threshold is not exceeded, the superimposed data represents the number of times a lightning strike could statistically occur within a time step in the area considered as the lightning location point.

[0063] The time step length is set to 60 minutes, with each 6-minute interval considered as the same lightning density, and the threshold for the number of lightning occurrences is 25.

[0064] The specific method for step (4) is as follows:

[0065] For data at a certain time step, n-1 time steps are taken forward, for a total of n time steps, and then weighted and mixed. The calculation method is shown in the following formula:

[0066] (4.1) The attenuation function is set to e -x ;

[0067] (4.2) Following the time sequence, with the current time step as the first time step, normalize each time step and then multiply it by 10 to obtain the intermediate variable k. i ; i takes values ​​of 1, 2, 3, ..., n; that is, the intermediate variable of the current time step is k1, the intermediate variable of the previous time step is k2, the intermediate variable of the previous time step is k3, and so on;

[0068] (4.3) For the weighting coefficient w at the i-th time step i for:

[0069]

[0070] Where n represents the total number of time steps required for mixing, and w i k represents the weight of the i-th time period during mixing. i It is an intermediate variable, P i This represents the data obtained after superimposing the data at the i-th time step of the region considered as the lightning location point; P represents the final mixed lightning density of the region considered as the lightning location point.

[0071] like Figure 2 The diagram shows a temporal weighting function, where the horizontal axis represents the dimension of time and the vertical axis represents the decay coefficient. Multiplying this temporal weighting function by a Gaussian impulse function reveals the continuity of the spatiotemporal distribution of lightning density. Furthermore, multiplying this temporal weighting function by a Gaussian impulse function shows that the density cloud corresponding to each lightning strike gradually decays over time.

[0072] This invention allows for the setting of a threshold based on lightning density images of the monitored area within a given time period. Areas exceeding this threshold are considered high-density lightning regions. This invention does not impose specific limitations on the threshold; it can be set according to actual needs.

[0073] Based on the lightning density of the area ultimately mixed by each lightning location point, corresponding color values ​​are generated. The higher the density, the darker the color, and these values ​​are displayed on the map to obtain a lightning density image of the monitored area during that time period.

[0074] A method for forecasting lightning intensity employs a deep neural network model. The lightning density image of the monitored area within the specified time period, obtained by the aforementioned high-density lightning region generation method, is used as the input to the deep neural network model. The corresponding lightning location point is used as the regional lightning intensity as the output for training, thereby obtaining a deep neural network model for forecasting lightning intensity.

[0075] Application Examples

[0076] 1. Data Source

[0077] The lightning location data used in this experiment comes from the Very Low Frequency Wide Area Lightning Location Network (VLF-LLN), which has been deployed and gradually improved in China since September 2021 by Nanjing University of Information Science and Technology. This experiment collected a total of 54,394 data points from November 27th and 28th, 2022. These data cover key parameters for each lightning event, such as the time of occurrence and geographical coordinates. Figure 2 The figure shows the spatial distribution of lightning density.

[0078] 2. A method for generating high-density lightning regions, comprising the following steps:

[0079] Step (1): Within the monitoring range, establish a ground-based wide-area lightning location network. The ground-based wide-area lightning location network includes multiple base stations, each of which includes a very low frequency vertical polarized antenna and a fast electric field change detector for location.

[0080] Step (2): Based on the data collected by establishing a ground-based wide-area lightning location network, the lightning location point is regarded as an area where lightning is likely to occur, and a spatial distribution expression of lightning density is constructed.

[0081] Step (3): Superimpose the lightning density corresponding to the lightning location point in a time step. The superimposed data represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. Using the lightning occurrence threshold, judge based on the superimposed data to obtain the high-density lightning area within this time step.

[0082] Step (4): For a detection time period containing n time steps, the weighted method is used to mix the data obtained by superimposing the current time step with the data obtained by superimposing the previous n-1 time steps to obtain the final mixed lightning density of the region regarded as the lightning location point; based on the final mixed lightning density of the region regarded as each lightning location point, the lightning density image of the monitoring range in this time period is obtained, thereby obtaining the high-density lightning region from the lightning density image of the monitoring range.

[0083] Step (1) specifically includes:

[0084] A nationwide ground-based wide-area lightning location network was constructed, with each base station including a very low frequency vertically polarized antenna (bandwidth within 30kHz) directly used for location. All base stations continuously acquire data at a sampling rate of 1MHz, with a data recording length of 1ms and a pre-trigger time of 300μs. Raw waveform data exceeding a minimum threshold is saved (the minimum threshold is generally set according to actual conditions, mainly for noise reduction). Currently, the ground-based wide-area lightning location network exhibits good location quality, providing good location results for various processes such as single thunderstorms, multi-single thunderstorms, and squall lines at different times.

[0085] Step (2) specifically includes:

[0086] For any given lightning location, there exists a corresponding Gaussian distribution centered at that location. Since the Gaussian distribution is the most common and fundamental probability distribution in nature, its one-dimensional form as the probability distribution for lightning occurrence can be described by the following formula:

[0087]

[0088] Where σ represents the variance, which can also represent the concentration and peak value of the distribution, and μ represents the mean. For ease of calculation, when converting the location points to a Gaussian distribution, we set the peak value of the Gaussian distribution to 1 (i.e., representing "1" lightning bolt) and the mean to 0, and use its two-dimensional form as follows:

[0089] r = x 2 +y 2

[0090]

[0091] In this system, the origin of the coordinate system is the location where the lightning occurs, r represents the distance from any point to the origin, x and y represent the two-dimensional direct coordinates of that point, σ represents the Gaussian pulse width, and G(x,y) is the lightning density value.

[0092] The lightning location point is considered as an area where lightning is likely to occur. The probability of lightning occurring in the area is closer to 100% the closer it is to the center. According to the experiment, r is set to be less than 25km.

[0093] Step (3) specifically includes:

[0094] The Gaussian distributions corresponding to all lightning location points within a time step are superimposed onto a zero-filled gridded dataset. The time step length is set to six minutes. Each grid point has a maximum value of 25, representing a statistically probable 25 lightning strikes within that area in six minutes. Values ​​exceeding 25 are clipped to 25. If a grid point's value reaches 25, the area it occupies is considered a high-density lightning region; otherwise, the value still represents lightning density information. For moderate-intensity thunderstorms, the transition from a high-density region (value 25) to a region with no possibility of lightning (value 0) is relatively smooth; however, for severe thunderstorms, this transition is steeper because a large number of Gaussian distributions are superimposed on the same area, followed by clipping to a maximum value of 25. Since the data input to the deep neural network first needs to be normalized, and the location and number of lightning strikes are random, it is impossible to determine the upper limit of the number of lightning strikes in a given area. Therefore, the clipped data can not only represent high-density areas, but also prevent the relatively weak strong thunderstorms from being affected by the values ​​of the strongest thunderstorms after normalization. At the same time, the steepness of the edges can reflect the original intensity of a certain high-density area, and no information is lost.

[0095] Step (4) specifically includes:

[0096] For data within a specific time period, it is time-series mixed with data from previous time periods according to certain weights, with earlier time periods receiving smaller weights. The weights are taken from the negative half-axis of the natural exponential function because the area between the negative half-axis and the x-axis of the natural exponential function is 1. Taking values ​​from such a function ensures that the sum of the weights is 1, meaning that the numerical range of the mixed data remains 100% of the original range. Figure 2 As shown, the horizontal axis represents time information, and the vertical axis represents intensity decay.

[0097] The specific algorithm for time series mixing is as follows: For data in a certain time period, n-1 time steps should be obtained in advance, for a total of n time steps, and then weighted mixing is performed. The weight calculation method is shown in the following formula.

[0098] (4.1) The attenuation function is set to e -x ;

[0099] (42) Following the time sequence, with the current time step as the first time step, normalize each time step and then multiply it by 10 to obtain the intermediate variable k. i ; i takes values ​​of 1, 2, 3, ..., n; that is, the intermediate variable of the current time step is k1, the intermediate variable of the previous time step is k2, the intermediate variable of the previous time step is k3, and so on;

[0100] (4.3) For the weighting coefficient w at the i-th time stepi for:

[0101]

[0102] Where n represents the total number of time steps required for mixing, and w i k represents the weight of the i-th time period during mixing. i It is an intermediate variable, P i This represents the data obtained after superimposing the data at the i-th time step of the region considered as the lightning location point; P represents the final mixed lightning density of the region considered as the lightning location point.

[0103] k i It is an intermediate variable, its purpose being to map 1 to n to 1 to 10, while e -10 =4×10 -5 ≈0, which makes integration easier in actual calculations.

[0104] 3. Test Results

[0105] Using the high-density lightning zone generation method, this experiment successfully identified four severe weather events, specifically the periods from 7:00 to 8:00, 8:00 to 9:00, 9:00 to 10:00, and 10:00 to 11:00 on November 28, 2022, as follows: Figure 3 As shown in the image, the spatiotemporal distribution characteristics of lightning activity over these four hours can be clearly observed through the intuitive display of the image. The shades of color represent the density and intensity changes of lightning events in different time periods, effectively revealing the dynamic process of thunderstorm system evolution.

[0106] At the same time, based on the above method, we obtain as follows: Figure 4 The image shows the lightning density and high-density distribution on November 27, 2022. The high-density distribution is represented by red. Each small dot represents a lightning location; the darker the color, the higher the lightning density in that area; the lighter the color, the lower the lightning density. In the upper left corner of the image, the lightning locations are more dispersed, resulting in a lower lightning density and a lighter color. In the center of the image, the lightning locations are more concentrated, indicating a higher lightning density and a darker red color.

Claims

1. A method for generating high-density lightning regions, characterized in that, Includes the following steps: Step (1): Within the monitoring range, establish a ground-based wide-area lightning location network. The ground-based wide-area lightning location network includes multiple base stations, each of which includes a very low frequency vertical polarized antenna and a fast electric field change detector for location. Step (2): Based on the data collected by establishing a ground-based wide-area lightning location network, the lightning location point is regarded as an area where lightning is likely to occur, and a spatial distribution expression of lightning density is constructed. Step (3): Superimpose the lightning density corresponding to the lightning location point in a time step. The superimposed data represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. Using the lightning occurrence threshold, judge based on the superimposed data to obtain the high-density lightning area within this time step. Step (4): For a time period to be detected containing n time steps, the weighting method is used to mix the data obtained by superimposing the current time step with the data obtained by superimposing the previous n-1 time steps to obtain the final mixed lightning density of the area regarded as the lightning location point; based on the final mixed lightning density of the area regarded as each lightning location point, the lightning density image of the area to be monitored in this time period is obtained, thereby obtaining the high-density lightning area from the lightning density image of the area to be monitored; The specific method for step (2) is as follows: For any given lightning location, there is a probability distribution of lightning occurrence centered on that location; a two-dimensional Gaussian impulse function is used as the spatial distribution of lightning density. Setting the peak value of the Gaussian distribution to 1 and the mean value to 0, we obtain the following spatial distribution expression for lightning density: In this system, the origin of the coordinate system is the location where the lightning occurs, r represents the distance from any point to the origin, x and y represent the two-dimensional direct coordinates of that point, and σ represents the Gaussian pulse width. This represents the lightning density value. The specific method for step (3) is as follows: The lightning density corresponding to the lightning location point within a time step is superimposed. The data obtained after superposition represents the number of times that lightning may occur statistically in the area regarded as the lightning location point within a time step. Set a threshold for the number of lightning occurrences. If the data obtained after aggregation exceeds the threshold, the threshold represents the number of times that lightning may occur statistically within a time step in the area considered as the lightning location point. The area considered as the lightning location point is then the high-density lightning area within this time step. If the threshold is not exceeded, the superimposed data represents the number of times a lightning strike could statistically occur within a time step in the area considered as the lightning location point. The specific method for step (4) is as follows: For data at a certain time step, n-1 time steps are taken forward, for a total of n time steps, and then weighted and mixed. The calculation method is shown in the following formula: (4.1) The attenuation function is set to e -x ; (4.2) Following the time sequence, with the current time step as the first time step, normalize each time step and then multiply it by 10 as an intermediate variable. ; i takes values ​​of 1, 2, 3, ..., n; that is, the intermediate variable of the current time step is The intermediate variable one time step ahead is The intermediate variable one time step ahead is And so on; (4.3) Weighting coefficients for the i-th time step for: Where n represents the total number of time steps required for mixing. This represents the weight of the i-th time period during the mixing process. It is an intermediate variable. This represents the data obtained after superimposing the data at the i-th time step of the region considered as the lightning location point; P represents the final mixed lightning density of the region considered as the lightning location point.

2. The method for generating high-density lightning regions according to claim 1, characterized in that, r is less than 50km.

3. The method for generating high-density lightning regions according to claim 1, characterized in that, The time step length is set to 60 minutes, with each 6-minute interval considered as the same lightning density, and the threshold for the number of lightning occurrences is 25.

4. The method for generating high-density lightning regions according to claim 1, characterized in that, Based on the lightning density of the area ultimately mixed by each lightning location point, corresponding color values ​​are generated. The higher the density, the darker the color, and these values ​​are displayed on the map to obtain a lightning density image of the monitored area during that time period.

5. A method for forecasting lightning levels, employing a deep neural network model for prediction, characterized in that, The lightning density image of the monitoring range within the time period obtained by the lightning high-density region generation method described in claim 1 is used as the input of the deep neural network model, and the corresponding lightning location point is regarded as the regional lightning level as the output for training, thereby obtaining a deep neural network model for predicting lightning level.

Images