Lightning forecasting method and device based on multi-source data, electronic equipment and medium

Abstract

The invention provides a lightning forecasting method and device based on multi-source data, electronic equipment and a medium, and relates to the field of lightning prediction.The method comprises the steps that satellite data and radar data of a target area in the current time period are acquired; performing interpolation processing on the radar data based on a preset time resolution and a preset spatial resolution to obtain target radar data; the trained lightning forecasting model is adopted to process satellite data and target radar data, and the probability of lightning occurrence in the future period is obtained; the lightning forecasting model is a multi-source data fusion model which is established on the basis of Lightning-Segnet-LSTM (Long Short Term Memory). According to the method, lightning identification is carried out on FY4B AGRI channel 9, channels 11-15 and radar data through Lightning-Segnet, and extrapolation forecasting of lightning is carried out in combination with a memory network LSTM.

Description

Technical Field

[0001] The present application relates to the field of lightning prediction, and more particularly, to a lightning prediction method, apparatus, electronic device, and medium based on multi-source data. Background Art

[0002] Lightning prediction is an important research direction in the meteorological field, and it is of great significance for ensuring aviation safety and grid stability. There are mainly three existing lightning identification and prediction methods: statistical diagnosis, numerical prediction models, and AI-based modeling. However, these methods often rely on a single data source and cannot comprehensively capture the multi-faceted information required for lightning formation and development. Summary of the Invention

[0003] The purpose of the embodiments of the present application is to provide a lightning prediction method, apparatus, electronic device, and medium based on multi-source data, which solves the above problems existing in the prior art and can comprehensively capture the formation and development of lightning.

[0004] In a first aspect, a lightning prediction method based on multi-source data is provided. The method may include:

[0005] Obtain satellite data and radar data of a target area in the current period; the satellite data is data collected by a satellite according to a preset time resolution and a preset spatial resolution;

[0006] Based on the preset time resolution and the preset spatial resolution, perform interpolation processing on the radar data to obtain target radar data;

[0007] Use a trained lightning prediction model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in a future period; wherein, the lightning prediction model includes a three-dimensional encoding network, a three-dimensional decoding network, and a memory network.

[0008] In a possible implementation, the memory network includes an input layer, multiple LSTM layers, a Dropout layer, a dense connection layer, and a memory output layer.

[0009] In a possible implementation, using a trained lightning prediction model to process the satellite data and the target radar data includes:

[0010] Use the three-dimensional encoding network to extract features from the satellite data and the target radar data to obtain multiple feature maps;

[0011] Use the three-dimensional decoding network to process each of the multiple feature maps to obtain grid-based lightning data of the target area;

[0012] Using the memory network, the gridded lightning data is processed to obtain the probability of lightning occurrence in a future time period.

[0013] In a possible implementation, the three-dimensional encoding network includes: 13 first three-dimensional convolutional layers, 13 first batch normalization layers, 13 first activation function layers, and 5 downsampling pooling layers.

[0014] In a possible implementation, the convolutional kernel of the first three-dimensional convolutional layer is 3×3×3; the stride is 1.

[0015] In a possible implementation, the three-dimensional decoding network includes: 5 upsampling pooling layers, 13 second three-dimensional convolutional layers, 13 second batch normalization layers, 13 second activation function layers, and an output layer.

[0016] In a possible implementation, the convolutional kernel of the second three-dimensional convolutional layer is 3×3×3; the stride is 1.

[0017] In a second aspect, a lightning forecasting device based on multi-source data is provided. The device may include:

[0018] An acquisition unit, configured to acquire satellite data and radar data of a target area in the current time period; the satellite data is data collected by a satellite according to a preset time resolution and a preset spatial resolution;

[0019] An interpolation unit, configured to perform interpolation processing on the radar data based on the preset time resolution and the preset spatial resolution to obtain target radar data;

[0020] A processing unit, configured to use a trained lightning forecasting model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in a future time period; wherein, the lightning forecasting model includes a three-dimensional encoding network, a three-dimensional decoding network, and a memory network.

[0021] In a third aspect, an electronic device is provided. The electronic device includes a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus;

[0022] The memory is used to store a computer program;

[0023] The processor is configured to implement any of the method steps described in the first aspect when executing the program stored in the memory.

[0024] In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements any of the method steps described in the first aspect.

[0025] The present application provides a lightning forecasting method based on multi-source data. The method includes: acquiring satellite data and radar data of a target area in the current period; the satellite data is data collected by a satellite according to a preset time resolution and a preset spatial resolution; based on the preset time resolution and the preset spatial resolution, interpolating the radar data to obtain target radar data; using a trained lightning forecasting model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in a future period; wherein, the lightning forecasting model includes a three-dimensional encoding network, a three-dimensional decoding network, and a memory network. The present application is based on Lightning-Segnet-LSTM to establish a lightning forecasting model with multi-source data fusion. Through Lightning-Segnet, lightning is identified for the FY4B AGRI channels 9, 11-15, and radar data, and combined with the memory network LSTM for extrapolation forecasting of lightning. Through multi-source fusion data, better feature extraction of the lightning occurrence mechanism is carried out, and the model is modified based on the semantic segmentation model Segnet to adapt to the input data, lightning is identified through feature extraction, and the identification result is input into the memory network for extrapolation forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0027] Figure 1 It is a system architecture diagram applied to a lightning forecasting method based on multi-source data provided by an embodiment of the present application;

[0028] Figure 2 It is a schematic flowchart of a lightning forecasting method based on multi-source data provided by an embodiment of the present application;

[0029] Figure 3 It is a schematic structural diagram of a lightning forecasting model provided by an embodiment of the present application;

[0030] Figure 4 It is a schematic structural diagram of a lightning forecasting device based on multi-source data provided by an embodiment of the present application;

[0031] Figure 5 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0033] The lightning forecasting method based on multi-source data provided by the embodiments of the present application can be applied in Figure 1 the system architecture shown in Figure 1 As shown, the system may include: a server and a terminal. The server may be a physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. The terminal may be a user equipment (UE) such as a mobile phone, a smart phone, a laptop computer, a digital broadcast receiver, a personal digital assistant (PDA), a tablet computer (PAD), a handheld device, a vehicle-mounted device, a wearable device, a computing device or other processing devices connected to a wireless modem, a mobile station (MS), a mobile terminal, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication methods, which are not limited in this application.

[0034] The terminal is used to obtain satellite data and radar data of the target area in the current period, and send the satellite data and radar data in the current period to the server;

[0035] The server is used to receive the satellite data and radar data in the current period to execute a lightning forecasting method based on multi-source data provided by the present application.

[0036] Lightning forecasting is an important research direction in the meteorological field, and it is of great significance for ensuring aviation safety and grid stability. There are mainly three existing lightning identification and forecasting methods: statistical diagnosis, numerical forecasting models, and AI-based modeling. However, these methods often rely on a single data source and cannot comprehensively capture the multi-faceted information required for lightning formation and development.

[0037] Single data source limitation: Using satellite data alone can monitor cloud body changes, but it is difficult to observe the inside of the convective system in depth; radar can provide high-resolution precipitation and storm information, but it cannot cover a large area.

[0038] Traditional fusion methods are backward: In the past, data fusion mostly used simple interpolation methods, ignoring the complex relationships between data from different sources, resulting in limited prediction accuracy.

[0039] This application provides a lightning prediction method based on multi-source data to solve the above problems existing in the prior art, which can comprehensively capture the formation and development of lightning and improve the lightning prediction accuracy.

[0040] The following describes the preferred embodiments of this application with reference to the accompanying drawings of the specification. It should be understood that the preferred embodiments described herein are only used to illustrate and explain this application, and are not used to limit this application. And without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0041] Figure 2 It is a schematic flowchart of a lightning prediction method based on multi-source data provided by an embodiment of this application. As Figure 2 shown, the method may include:

[0042] Step S210, obtain satellite data and radar data of the target area in the current period.

[0043] Among them, the satellite data is the data collected by the satellite corresponding to the target area according to the preset time resolution and preset space resolution.

[0044] The satellite data can be the meteorological observation information with large range, high spatio-temporal resolution provided by FY-4B AGRI channels 9, 11-15. This meteorological observation information can include key parameters such as cloud characteristics, atmospheric temperature, and water vapor distribution of the target area in the current period. Specifically, FY4B AGRI channels 9 and 11 are located in the water vapor channels. Channel 9 is usually used to detect high-level water vapor; Channel 11 is used to detect low-level water vapor. Channels 12-15 are located in the long-wave infrared channels and are usually used to detect clouds, surface temperature, etc.

[0045] The radar data is the precipitation and convective activity information with high spatio-temporal resolution provided by the radar.

[0046] Step S220, based on the preset time resolution and preset space resolution, perform interpolation processing on the radar data to obtain target radar data.

[0047] Specifically, based on the preset time resolution and preset spatial resolution of satellite data, the process of interpolating radar data can be as follows: If the preset time resolution of satellite data is 15 minutes and the preset spatial resolution is 4 km, and the time resolution of radar data is 6 minutes and the spatial resolution is 1 km, then the bilinear interpolation algorithm is used to perform longitude and latitude spatial matching on the radar data to obtain the initial radar data with the preset spatial resolution; thereafter, the bilinear interpolation algorithm or the spline interpolation algorithm can be used to process the initial radar data to obtain the target radar data with the preset time resolution. For example, to obtain the data value at a certain location at the 15th minute, it can be estimated by linearly interpolating the actual observed values at the 12th minute and the 18th minute.

[0048] In some embodiments, a generator network is trained to create new data points, and a discriminator network is used to evaluate the authenticity of these data points to generate more realistic interpolation results. Or by analyzing the frequency shift (i.e., Doppler shift) in the radar echo signal, the speed of the target relative to the radar can be determined. Using this principle, even the data collected in a short period of time can be used to infer the movement of the target through speed information, thereby indirectly changing the time resolution and spatial resolution of the radar data.

[0049] In another embodiment, the radar data is input into a spatio-temporal fusion model (SPSTFM model, remote sensing spatio-temporal fusion model), and the target radar data with the preset time resolution and preset spatial resolution can be obtained.

[0050] Step S230: Use the trained lightning forecast model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in the future period.

[0051] Before performing step S230, the method further includes training the pre-trained lightning forecast model;

[0052] The specific process of training can include: obtaining historical satellite data, historical radar data, and historical lightning data corresponding to multiple historical periods; among them, the historical lightning data can also be understood as historical lightning location data.

[0053] Among them, the historical radar data is the radar combined emissivity data provided by the Central Meteorological Observatory, with a data horizontal resolution of 1 km and a time resolution of 6 minutes.

[0054] The historical satellite data is the historical meteorological observation information corresponding to channels 9, 11 - 15 of FY4B AGRI, with a time resolution of 15 minutes and a spatial resolution of 4 km.

[0055] The time resolution of the historical lightning data is 1 minute.

[0056] Since the historical lightning data is scatter data, it is necessary to grid the scatter data and use the nearest neighbor interpolation algorithm to interpolate it onto the FY4B spatial longitude and latitude distribution. At the same time, the nearest neighbor interpolation algorithm is used to set the time resolution of the historical radar data to 15 minutes and the spatial resolution to 4 km.

[0057] After that, based on the historical satellite data, historical radar data, and historical lightning data corresponding to multiple historical periods, a training set is determined.

[0058] The historical satellite data and historical radar data of each historical period with unified time resolution and spatial resolution are determined as training samples;

[0059] The historical lightning data corresponding to each historical period is determined as a training label.

[0060] Using the training samples and training labels in the training set, the pre-trained lightning forecasting model is trained to obtain a trained lightning forecasting model.

[0061] Combined Figure 3 As shown, the specific structure of the lightning forecasting model may include: a three-dimensional encoding network, a three-dimensional decoding network, and a memory network.

[0062] Step S230 may specifically include:

[0063] Using the three-dimensional encoding network, feature extraction is performed on the satellite data and the target radar data to obtain multiple feature maps;

[0064] Using the three-dimensional decoding network, each of the multiple feature maps is processed to obtain the gridded lightning data of the target area;

[0065] Using the memory network, the gridded lightning data is processed to obtain the probability of lightning occurrence in the future period.

[0066] The lightning forecasting model is the Lightning-Segnet-LSTM model: Since SegNet is a mature deep learning semantic segmentation model with an encoder-decoder structure. The lightning forecasting model is based on SegNet and modifies the two-dimensional max pooling layer to a three-dimensional pooling layer. This model can accurately extract features and perform feature fusion to better identify lightning;

[0067] Among them, A. The three-dimensional encoding network includes: 13 first three-dimensional convolutional layers, 13 first batch normalization layers, 13 first activation function layers, and 5 downsampling pooling layers.

[0068] The first 3D convolutional layer: It is used to perform 3D convolution operations on the input data (satellite data and target radar data) to extract features in the spatial and temporal dimensions. Each first 3D convolutional layer uses a 3*3*3 convolutional kernel with a stride of 1 to ensure that the size of the feature map remains unchanged.

[0069] The first batch normalization layer: It normalizes the feature map output by the first 3D convolutional layer to accelerate model training and improve stability; a first batch normalization layer is connected after each first 3D convolutional layer.

[0070] The first activation function layer: It performs a non-linear transformation on the feature map output by the first batch normalization layer to enhance the expressive ability of the model; a ReLU activation function is connected after each first batch normalization layer.

[0071] The downsampling pooling layer: It downsamples the feature map to reduce the size of the feature map (time, height, width) while retaining key information; a downsampling pooling layer is connected after each first functional partition (Stage); where the first functional partition is an area composed of at least one first 3D convolutional layer and a first batch normalization layer, and the first functional partition can be flexibly designed and adjusted according to needs, which is not limited in this application.

[0072] B. The 3D decoding network includes: 5 upsampling pooling layers, 13 second 3D convolutional layers, 13 second batch normalization layers, 13 second activation function layers, and an output layer.

[0073] The upsampling pooling layer: It is used to upsample the feature map output by the downsampling pooling layer to restore the size of the feature map (time, height, width). Each upsampling pooling layer is connected to a second functional partition, where the second functional partition is an area composed of at least one second 3D convolutional layer and a second batch normalization layer, and the second functional partition can be flexibly designed and adjusted according to needs, which is not limited in this application.

[0074] The second 3D convolutional layer: It is used to perform convolution operations on the upsampled feature map to gradually restore the detailed information of the feature map. Each convolutional layer uses a 3x3x3 convolutional kernel with a stride of 1 to ensure that the size of the feature map remains unchanged.

[0075] The second batch normalization layer: It is used to normalize the feature map output by the second 3D convolutional layer. A second batch normalization layer is connected after each second 3D convolutional layer.

[0076] The second activation function layer: It is used to perform a non-linear transformation on the feature map output by the second batch normalization layer to enhance the expressive ability of the model. A ReLU activation function is connected after each second batch normalization layer.

[0077] An output layer, which is used to map the feature map output by the second activation function layer to a target output space, generate grid data of lightning occurrence probability, that is, gridded lightning data.

[0078] C. Memory network: It is used to process multiple gridded lightning data and output the probability distribution of lightning occurrence at different longitude and latitude positions in the future time period.

[0079] Among them, the multiple gridded lightning data are time series data composed of the current time period and multiple time periods before the current time period.

[0080] The memory network includes: an input layer, multiple LSTM layers, a Dropout layer, a dense connection layer, and a memory output layer. Among them, the input layer of the memory network: is used to receive time series data; the multiple LSTM layers can include: a. The first-layer LSTM: At the beginning, one or more LSTM layers can be used. The number of units (i.e., the number of neurons) in each layer needs to be selected according to specific situations. More units can capture more complex patterns, but may also lead to overfitting. The initial setting can try 50 to 100 units. B. Stacked LSTM layers: To increase the depth of the memory network, multiple LSTM layers can be considered stacked. Usually, two to three LSTM layers are sufficient. It should be noted that except for the last layer, other LSTM layers should set return_sequences = True to ensure that the output sequence is passed to the next layer. To prevent overfitting, a Dropout layer can be added after the LSTM layer. Dropout is a regularization technique that reduces the model complexity by randomly "dropping out" a part of the neurons (i.e., setting their outputs to zero). The commonly used Dropout rate is 0.2 to 0.5. After the last LSTM layer, usually one or more fully connected layers (Dense) are connected to perform the final classification or regression task. For a regression problem (such as predicting the specific number of lightning strikes), the last Dense layer usually has only one unit. The memory output layer contains only one neuron and is used to output the probability distribution of lightning occurrence in the target area in the future time period.

[0081] In some embodiments, the lightning forecasting model may only include a three-dimensional encoding network and a three-dimensional decoding network, and the memory network exists independently.

[0082] In another embodiment, a multi-modal fusion model is adopted to fuse satellite data and radar data to obtain a fusion result. The multi-modal fusion model is used to fuse satellite data and target radar data to obtain the regional data of the fused target area. A convolutional neural network is used for multi-modal fusion, automatically adjusting the weight size according to recent data, and effectively adjusting the weight according to the quality of satellite data and radar data, so as to obtain a fusion result with high accuracy. Alternatively, a pixel-based method is used to directly merge the target radar data and satellite data at the pixel level to obtain regional data. Then, the trained lightning prediction model is used to process the fused regional data and output the probability distribution of lightning occurrence in the target area in the future period. It can be understood that the training samples used in the training stage of the currently used lightning prediction model are also historical regional data.

[0083] This application provides a lightning prediction method based on multi-source data. The method includes: obtaining satellite data and radar data of a target area in the current period; the satellite data is data collected by a satellite according to a preset time resolution and a preset space resolution; based on the preset time resolution and the preset space resolution, interpolating the radar data to obtain target radar data; using the trained lightning prediction model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in the future period; wherein, the lightning prediction model includes a three-dimensional encoding network, a three-dimensional decoding network and a memory network. This application is based on Lightning-Segnet-LSTM to establish a lightning prediction model for multi-source data fusion. Through Lightning-Segnet, lightning is identified from the FY4B AGRI channels 9, 11-15 and radar data, and the memory network LSTM is combined for extrapolation prediction of lightning. That is, better feature extraction of the lightning occurrence mechanism is carried out through multi-source fusion data, the model is modified based on the semantic segmentation model Segnet to adapt to the input data, lightning is identified through feature extraction, and the identification result is input into the memory network for extrapolation prediction.

[0084] Corresponding to the above method, an embodiment of this application further provides a lightning prediction device based on multi-source data, as Figure 4 shown. The device includes:

[0085] An acquisition unit 410, configured to acquire satellite data and radar data of a target area in the current period; the satellite data is data collected by a satellite according to a preset time resolution and a preset space resolution;

[0086] An interpolation unit 420, configured to interpolate the radar data based on the preset time resolution and the preset space resolution to obtain target radar data;

[0087] A processing unit 430 is configured to process the satellite data and the target radar data by using a trained lightning forecasting model to obtain the probability of lightning occurrence in a future period; wherein, the lightning forecasting model includes a three-dimensional encoding network, a three-dimensional decoding network, and a memory network.

[0088] The functions of the functional units of a lightning forecasting device based on multi-source data provided in the above embodiments of the present application can be implemented by the above method steps. Therefore, the specific working processes and beneficial effects of each unit in a lightning forecasting device based on multi-source data provided in the embodiments of the present application will not be repeated here.

[0089] The embodiments of the present application also provide an electronic device, as Figure 5 shown, including a processor 510, a communication interface 520, a memory 530, and a communication bus 540. Among them, the processor 510, the communication interface 520, and the memory 530 complete communication with each other through the communication bus 540.

[0090] The memory 530 is used to store a computer program;

[0091] When the processor 510 is configured to execute the program stored on the memory 530, the following steps are implemented:

[0092] Obtain satellite data and radar data of a target area in the current period; the satellite data is data collected by a satellite according to a preset time resolution and a preset space resolution;

[0093] Based on the preset time resolution and the preset space resolution, perform interpolation processing on the radar data to obtain target radar data;

[0094] Use a trained lightning forecasting model to process the satellite data and the target radar data to obtain the probability of lightning occurrence in a future period; wherein, the lightning forecasting model includes a three-dimensional encoding network, a three-dimensional decoding network, and a memory network.

[0095] The above-mentioned communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of simplicity, only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0096] The communication interface is used for communication between the above electronic device and other devices.

[0097] The memory may include a Random Access Memory (RAM), or may also include a Non-Volatile Memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0098] The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0099] Since the implementation manners and beneficial effects of the components of the electronic device in the above embodiments for solving problems can be seen from Figure 2 the steps in the shown embodiments, therefore, the specific working process and beneficial effects of the electronic device provided in the embodiments of the present application will not be elaborated herein.

[0100] In another embodiment provided by the present application, there is also provided a computer-readable storage medium, in which instructions are stored, and when it runs on a computer, it causes the computer to execute any one of the above-described lightning forecasting methods based on multi-source data.

[0101] In another embodiment provided by the present application, there is also provided a computer program product containing instructions, and when it runs on a computer, it causes the computer to execute any one of the above-described lightning forecasting methods based on multi-source data.

[0102] Those skilled in the art should understand that the embodiments in the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the embodiments in the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments in the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0103] Embodiments of the present application are described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0104] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0106] Unless otherwise defined, the technical terms or scientific terms used in the present application should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs. The terms "first", "second", and similar words used in the present application do not denote any order, quantity, or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connected", "coupled", or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Words such as "up", "down", "left", and "right" are only used to indicate relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0107] Although the preferred embodiments in the embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the embodiments of the present application are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0108] Obviously, those skilled in the art can make various changes and modifications to the embodiments in the embodiments of the present application without departing from the spirit and scope of the embodiments in the embodiments of the present application. Thus, if these modifications and variations of the embodiments in the embodiments of the present application fall within the scope of the embodiments of the present application and their equivalent technologies, the embodiments of the present application are also intended to include these changes and modifications.

Claims

1. A lightning forecasting method based on multi-source data, characterized in that: The method comprises: Acquire satellite data and radar data of the target area in the current period; the satellite data is data collected by the satellite according to a preset time resolution and a preset spatial resolution; Based on the preset time resolution and the preset spatial resolution, interpolate the radar data to obtain target radar data; The satellite data and the target radar data are processed using a trained lightning prediction model to obtain the probability of lightning occurring in a future period; wherein the lightning prediction model includes a three-dimensional encoding network, a three-dimensional decoding network and a memory network.

2. The method according to claim 1, characterized in that The memory network includes an input layer, multiple LSTM layers, a Dropout layer, a dense connection layer and a memory output layer.

3. The method according to claim 1, characterized in that The satellite data and the target radar data are processed using the trained lightning forecast model, including: Using the three-dimensional coding network, extracting features from the satellite data and the target radar data to obtain a plurality of feature maps; Using the three-dimensional decoding network, a plurality of feature maps are processed respectively to obtain grid lightning data of the target area; The memory network is used to process the gridded lightning data to obtain the probability of lightning occurrence in the future period.

4. The method according to claim 1, characterized in that The three-dimensional coding network includes: 13 first three-dimensional convolutional layers, 13 first batch normalization layers, 13 first activation function layers and 5 downsampling pooling layers.

5. The method according to claim 4, characterized in that The convolution kernel of the first three-dimensional convolutional layer is 3×3×3, and the stride is 1.

6. The method according to claim 1, characterized in that The three-dimensional decoding network includes: 5 upsampling pooling layers, 13 second three-dimensional convolutional layers, 13 second batch normalization layers, 13 second activation function layers and an output layer.

7. The method according to claim 6, characterized in that The convolution kernel of the second three-dimensional convolution layer is 3×3×3, and the stride is 1.

8. A lightning forecasting device based on multi-source data, characterized in that: The device comprises: An acquisition unit, used to acquire satellite data and radar data of the target area in the current period; the satellite data is data collected by the satellite according to a preset time resolution and a preset spatial resolution; an interpolation unit, configured to perform interpolation processing on the radar data based on the preset time resolution and the preset spatial resolution to obtain target radar data; The processing unit is used to process the satellite data and the target radar data using a trained lightning prediction model to obtain the probability of lightning occurring in a future period; wherein the lightning prediction model includes a three-dimensional encoding network, a three-dimensional decoding network and a memory network.

9. An electronic device, characterized in that: The electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 7 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of claims 1 to 7 are implemented.

Images