Weather prediction method based on deep learning

Abstract

The invention discloses a weather prediction method and device based on deep learning, a medium and equipment, and the method comprises the steps: obtaining meteorological data and a weather prediction model based on deep learning, and enabling the weather prediction model based on deep learning to be determined according to n cascaded / parallel weather prediction sub-models based on semantic segmentation, the input end of each stage of weather prediction sub-model except the first stage of weather prediction sub-model is in communication connection with the output ends of the first n-1 stages of weather prediction sub-models; respectively inputting meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and correspondingly obtaining each piece of meteorological prediction data within a period of time after the current moment; and after each piece of weather prediction data is weighted, the final weather prediction data in the to-be-predicted space range is output, and through the application of the deep learning method, the calculation efficiency of the weather prediction model is greatly improved, so that the real-time prediction requirement is met, and a quick-response weather prediction result is provided for various application scenes.

Description

Technical Field

[0001] The present application relates to the technical field of meteorological forecasting, and in particular to a weather forecasting method, device, medium and equipment based on deep learning. Background Art

[0002] In the field of weather forecasting, there are currently two main methods: weather forecasting methods based on numerical weather forecasting and weather forecasting methods based on machine learning (especially deep learning).

[0003] The numerical weather forecast method is based on the principles of physics and has high scientificity and interpretability. By modeling the dynamic and thermodynamic processes of the atmosphere, the numerical weather forecast method can simulate the evolution of the weather system and obtain the forecast results by calculating complex physical equations. Because it is based on physical models, the numerical weather forecast method has good interpretability and scientificity. The forecast results of this method are more in line with the laws of atmospheric physics and have high application value for professional research and meteorological analysis. However, the numerical weather forecast method relies on a large amount of computing resources, and the reasoning speed is relatively slow, usually requiring hours or even days of computing time. At the same time, with the accumulation of historical data, although more input data can be provided, the prediction accuracy will not be significantly improved. At present, the numerical weather forecast method has been applied to the weather forecast systems of some countries and regions, such as the global forecast system of the United States and the ensemble forecast system of the European Medium-Term Weather Forecast Center. The existing weather forecast system performs poorly in terms of prediction effect, reasoning speed, and data utilization. Summary of the invention

[0004] The main purpose of this application is to provide a weather forecasting method, device, medium and equipment based on deep learning, aiming to combine the advantages of data-driven to provide a more efficient and accurate solution for weather forecasting.

[0005] To achieve the above objectives, the present application provides a weather forecasting method based on deep learning, comprising:

[0006] Acquire meteorological data within a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning, wherein the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of the weather prediction sub-model except the first level weather prediction sub-model is communicatively connected to the output end of each of the first n-1 levels of the weather prediction sub-model; input the meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and obtain corresponding meteorological prediction data within a period of time after the current moment; and output the final meteorological prediction data within the spatial range to be predicted after weighting each of the meteorological prediction data.

[0007] Optionally, the n cascaded semantic segmentation-based weather prediction sub-models include multiple 3D models and at least two of the multiple 3D models, the 2D model includes MedNeXt and PVT-CASCADE, and the 3D model includes MISSFormer, ConvNeXt, ConResNet and FourCastNet.

[0008] Optionally, the meteorological data includes multiple pressure layer variables and multiple ground variables, each of the pressure layer variables includes multiple pressure layers, and the ground variables are various surface meteorological indicators; the input shape of each of the 3D models includes: a first channel number, a first depth, a height and a width, the first channel number is equal to a plurality of sampling moments within a period of time before the current moment, and the first depth is the sum of the number of all pressure layers and the number of all surface meteorological indicators; the input shape of each of the 2D models includes: a second channel number, a height and a width, the second channel number is equal to the dimension after the first channel and the depth are combined; the shapes of each of the 3D models and each of the 2D models are: a third channel number, a second depth, a height and a width, the third channel number is a plurality of predicted moments within a period of time after the current moment, and the second depth is equal to the number of ground variables; wherein the height and the width respectively represent the grid points of the spatial range to be predicted.

[0009] Optionally, the meteorological data are respectively input into each of the weather prediction sub-models, including: marking the meteorological data with sea and land types and / or soil types and / or altitudes to obtain first input data; obtaining the difference value of the meteorological data at sampling times within a period of time before the current moment, and concatenating the difference value of the meteorological data with the meteorological data at the current moment to obtain second input data; adding Gaussian noise to the second input data to obtain third input data; and inputting the first input data, the second input data or the third input data into each of the weather prediction sub-models.

[0010] Optionally, before the meteorological data is marked with sea and land types and / or soil types and / or altitude to obtain the first input data, the method further includes: obtaining the mean and standard deviation of the altitude data, and obtaining standardized altitude data according to the ratio of the mean and the standard deviation, and obtaining corresponding altitude mark data according to the meteorological data.

[0011] Optionally, the meteorological data are respectively input into the weather prediction sub-models of each of the weather prediction models based on deep learning, and the meteorological forecast data for a period of time after the current moment are output accordingly, including: determining a weather prediction model group according to the weather prediction sub-models in parallel, and determining the weather prediction model based on deep learning based on multiple weather prediction model groups; the meteorological data are respectively input into each of the weather prediction model groups in the weather prediction model based on deep learning, and the meteorological forecast data for each sub-time period in a period of time after the current moment are obtained accordingly, wherein each sub-time period is continuous in time.

[0012] Optionally, the method also includes: obtaining final meteorological forecast data and real label data, and determining a loss function based on the final meteorological forecast data and the real label data; according to the loss function, using the gradient descent method to train the deep learning-based weather forecast model to obtain a deep learning-based weather forecast model with updated parameters.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a weather prediction device based on deep learning, including: an acquisition module, used to obtain meteorological data within a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning, wherein the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of the weather prediction sub-model except the first level weather prediction sub-model is communicatively connected to the output end of the first n-1 levels of the weather prediction sub-model; a processing module, used to respectively input the meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and obtain the corresponding meteorological prediction data within a period of time after the current moment; an output module, used to weight each of the meteorological prediction data and output the final meteorological prediction data within the spatial range to be predicted.

[0014] To achieve the above objectives, the present application also provides a computer-readable storage medium, which includes instructions, which, when executed on a computer, enables the computer to execute the deep learning-based weather prediction method provided in the above embodiment.

[0015] An electronic device, comprising: at least one processor, a memory and an input-output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the deep learning-based weather prediction method provided in the aforementioned embodiment.

[0016] The embodiments of the present application propose a weather prediction method, device, medium and equipment based on deep learning, which obtains meteorological data within a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning. The weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of the weather prediction sub-model except the first level weather prediction sub-model is communicatively connected to the output end of the first n-1 levels of the weather prediction sub-model; the meteorological data is respectively input into each weather prediction sub-model of each weather prediction model based on deep learning, and corresponding meteorological prediction data within a period of time after the current moment is obtained; the final meteorological prediction data within the spatial range to be predicted is output after weighting each meteorological prediction data. The present application greatly improves the computational efficiency of the weather prediction model through the application of the deep learning method, so that it can complete the prediction operation within milliseconds to seconds, thereby meeting the real-time prediction needs and providing fast-response weather prediction results for various application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart diagram of an embodiment of a weather prediction method based on deep learning provided by the present application;

[0018] Figure 2 A schematic diagram of an embodiment of a weather prediction method based on deep learning provided by the present application;

[0019] Figure 3 Provided for an embodiment of a weather prediction method based on deep learning in this application;

[0020] Figure 4 A schematic diagram of functional modules provided for an embodiment of a weather prediction device based on deep learning in the present application.

[0021] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0022] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0023] Due to the existing technology, weather forecasting methods based on deep learning have gradually become an important means. This method uses models such as deep neural networks to learn and recognize patterns of historical weather data to quickly generate forecast results. Weather forecasting methods based on deep learning have excellent forecasting performance and great potential. This method requires relatively few computing resources and has a fast inference speed, which can reach a computing speed of milliseconds or even seconds. In addition, deep learning methods can effectively utilize historical input data to improve the accuracy of predictions. However, the current deep learning methods have certain shortcomings in interpretability, and the model output results are often difficult to fully understand, which to a certain extent limits the wide application of this method in the field of meteorology. At present, some weather forecasting systems based on deep learning have been put into use, such as the Pangu model, the Fuxi model, and the GraphCast system. These systems make full use of the advantages of deep learning in data processing and pattern recognition, and provide efficient solutions for short-term and medium-term weather forecasts. Although the numerical weather forecast method has high scientificity and interpretability, it requires huge computing resources, has a slow prediction speed, and the use of historical data is not significant.

[0024] The present application provides a solution that greatly improves the computational efficiency of the weather forecast model through the application of deep learning methods, so that it can complete the prediction operation within milliseconds to seconds, thereby meeting the real-time prediction needs and providing fast-response weather forecast results for various application scenarios. The present application uses a deep learning model to learn and extract patterns from a large amount of historical meteorological data, thereby improving the prediction accuracy, enabling the model to capture the complex changes in the weather system and provide technical support for more accurate weather forecasts. The present application makes full use of the deep learning model's ability to learn and mine historical data other than the current moment, and can discover the complex patterns and potential relationships of the weather system in large-scale data, so that the model can show higher prediction accuracy under different meteorological conditions. Compared with traditional numerical weather forecasting methods, the deep learning method proposed in this application can achieve efficient training and prediction under conditions of relatively few computing resources, is suitable for application in resource-constrained environments, and can reduce operating costs.

[0025] The technical solution proposed in this application is described in detail below with reference to the accompanying drawings.

[0026] Reference Figure 1 , the weather prediction method based on deep learning provided in the first embodiment of the present application, the weather prediction method based on deep learning may include:

[0027] S10, obtaining meteorological data for a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning, wherein the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of weather prediction sub-model except the first level of weather prediction sub-model is communicatively connected to the output end of each of the first n-1 levels of weather prediction sub-models;

[0028] refer to Figure 2 Specifically, this application adopts a phased prediction method, which means using the input and prediction results of the current stage as the input of the next stage, which can significantly improve the prediction accuracy. Model integration improves the accuracy and robustness of the overall prediction by combining the prediction results of multiple weather prediction sub-models based on semantic segmentation, thereby reducing the error and bias of a single model. Figure 3 The present invention adopts a weighted summation integration method, that is, the weight of each weather prediction sub-model based on semantic segmentation is set according to its prediction effect, and finally the prediction results of all models are multiplied and summed by their weights to obtain the fusion result. Among them, the sum of the weights of different weather prediction sub-models based on semantic segmentation is 1.

[0029] In the process of selecting sub-models, we mainly base our selection on six 3D and 2D models in the field of semantic segmentation: MISSFormer, ConvNeXt, ConResNet, MedNeXt, FourCastNet, and PVT-CASCADE. Among them, the 2D models include MedNeXt and PVT-CASCADE, and the 3D models include MISSFormer, ConvNeXt, ConResNet, and FourCastNet.

[0030] During the data definition process, the meteorological data includes multiple pressure layer variables and multiple ground variables, each pressure layer variable includes multiple pressure layers, and the ground variables are various surface meteorological indicators; the input shape of each 3D model includes: the first number of channels, the first depth, the height and the width, the first number of channels is equal to the multiple sampling moments within a period of time before the current moment, and the first depth is the sum of the number of all pressure layers and the number of all surface meteorological indicators; the input shape of each 2D model includes: the second number of channels, the height and the width, the second number of channels is equal to the dimension after the first channel and the depth are combined; the shapes of each 3D model and each 2D model are: the third number of channels, the second depth, the height and the width, the third number of channels is the multiple predicted moments within a period of time after the current moment, and the second depth is equal to the number of ground variables; wherein the height and the width respectively represent the grid points of the spatial range to be predicted.

[0031] Exemplarily, the data set used in this application is mainly derived from the ERA5 reanalysis data set, covering meteorological historical data and meteorological variables required for future predictions, specifically including 5 pressure layer variables (each variable has 13 pressure layers), 5 ground variables, a total of 70 variables. The 5 pressure layer variables are geopotential height, temperature, zonal wind, meridional wind, and relative humidity, and the 5 ground variables are common surface meteorological indicators, namely 2-meter temperature, 10-meter zonal wind, 10-meter meridional wind, mean sea level pressure, and 6-hour cumulative precipitation. The submodel predicts the values ​​of 5 ground variables at the next 20 moments by inputting historical data (70 variables) at 2 moments, and the time interval between 2 adjacent moments is 6 hours. The spatial range covered by the data is 10 to 50 degrees north latitude and 100 to 140 degrees east longitude, and the spatial resolution is 0.25 degrees, so a grid point of 161×161 is obtained. The shape of the input data is [2,70,161,161] (2 historical moments, 70 input variables, 161×161 grid points), and the shape of the prediction results and true labels is [20,5,161,161] (20 future moments, 5 ground variables, 161×161 grid points).

[0032] For 3D models, the input shape is [2,70,161,161], which represents the number of channels, depth, height, and width of the 3D model input, respectively. The output channel is set to 20 (representing the next 20 moments), and finally the depth is mapped to 5 (representing 5 ground variables) through a fully connected layer, resulting in an output shape of [20,5,161,161]. For 2D models, the first two dimensions of the input data are first merged to obtain data of shape [140,161,161]. At this time, 140, 161, and 161 are used as the number of channels, height, and width of the 2D model input, respectively. The output channel is set to 100, and then expanded to 20×5, resulting in an output shape of [20,5,161,161].

[0033] The present application also includes an optimization process for input data. In the optimization process for input data, the weather prediction method based on deep learning proposed in the present application may also include: marking the meteorological data for sea and land type and / or soil type and / or altitude to obtain the first input data; obtaining the difference value of the meteorological data at the sampling time within a period of time before the current moment, splicing the difference value of the meteorological data with the meteorological data at the current moment to obtain the second input data; adding Gaussian noise to the second input data to obtain the third input data; inputting the first input data, the second input data or the third input data into each weather prediction sub-model. Specifically, before the meteorological data is marked for sea and land type and / or soil type and / or altitude to obtain the first input data, the data optimization process also includes: obtaining the subtracted mean and standard deviation of the altitude data, and obtaining the standardized altitude data according to the ratio of the subtracted mean and standard deviation, and obtaining the corresponding altitude mark data according to the meteorological data. Because the meteorological changes in different regions are affected by factors such as the sea and land type, soil type, and altitude of the region. Therefore, the present application will splice additional supplementary data in the original input, specifically including sea and land type marks, soil type marks, and altitude. The range of altitude data varies greatly and needs to be standardized by subtracting the mean and dividing by the standard deviation before use. In addition, in order to make the model more explicitly model the changing trend of historical data, the difference value of the data at two historical moments will be spliced ​​with the original input. Data enhancement can increase data diversity and improve the generalization performance of the model. Here, learnable random Gaussian noise is added to the input data. The specific implementation is to use Gaussian noise as a learnable weight for the model and optimize it along with the training process.

[0034] In an embodiment of the present application, the weather prediction method based on deep learning proposed in the present application may also include:

[0035] Obtain the final weather forecast data and the real label data, and determine the loss function based on the final weather forecast data and the real label data;

[0036] According to the loss function, the gradient descent method is used to train the weather forecast model based on deep learning to obtain the weather forecast model based on deep learning with updated parameters.

[0037] Specifically, during the training phase, the model receives input data, obtains prediction results through a certain prediction method, calculates the loss between the prediction results and the true label, calculates the gradient of the loss to the model parameters, and updates the model parameters through the gradient descent method to make the model's prediction results closer to the true label. In order to capture the complex laws contained in meteorological data, six models with different architectures are trained during the training phase, and each model models a different pattern relationship. During the inference phase, the prediction results of the six models are combined to obtain the final prediction result, that is, the accuracy of weather forecasts is further increased through model integration.

[0038] S20, respectively inputting meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and correspondingly obtaining each meteorological prediction data within a period of time after the current moment;

[0039] In an optional embodiment, meteorological data is respectively input into the weather prediction sub-model of each weather prediction model based on deep learning, and the corresponding output of each meteorological prediction data within a period of time after the current moment includes: determining a weather prediction model group according to the parallel weather prediction sub-models, and determining a weather prediction model based on deep learning based on multiple weather prediction model groups; meteorological data is respectively input into each weather prediction model group in the weather prediction model based on deep learning, and the corresponding meteorological prediction data for each sub-time period in a period of time after the current moment is obtained, wherein each sub-time period is continuous in time.

[0040] In another optional embodiment, the weather forecasting method may also include: the classical autoregressive method, which is simple and can flexibly cope with different prediction durations, but the inference time is relatively long and there is an error accumulation problem; directly predicting all future moments, this method does not have the disadvantage of error accumulation, but it is difficult for a single model to accurately predict the next 20 moments by relying only on 2 historical moments.

[0041] In another optional embodiment, the weather forecasting method may also include: variable grouping + time grouping + phased forecasting method. When the forecasting time is longer, the correlation between the future and the input data is greatly weakened, and it will rely more on the data at adjacent moments, and the direct forecasting method cannot solve this problem.

[0042] Reference Figure 4 The second embodiment of the present application provides a weather prediction device based on deep learning. The weather prediction device 100 based on deep learning may include an acquisition module 1001, a processing module 1002 and an output module 1003, wherein the acquisition module 1001 is used to acquire meteorological data and a weather prediction model based on deep learning within a period of time before the current moment within the spatial range to be predicted, and the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of weather prediction sub-model except the first level weather prediction sub-model is communicatively connected to the output end of each of the first n-1 levels of weather prediction sub-models; the processing module 1002 is used to input meteorological data respectively into each weather prediction sub-model of each weather prediction model based on deep learning, and obtain each meteorological prediction data within a period of time after the current moment; the output module 1003 is used to weight each meteorological prediction data and output the final meteorological prediction data within the spatial range to be predicted.

[0043] Based on the above embodiments, the present application also provides a computer-readable storage medium, which includes instructions, which, when executed on a computer, enables the computer to execute the deep learning-based weather prediction method provided in the above embodiments.

[0044] Based on the above embodiments, an electronic device includes: at least one processor, a memory and an input-output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the deep learning-based weather prediction method provided in the above embodiments.

[0045] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A weather prediction method based on deep learning, characterized in that: include: Acquire meteorological data for a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning, wherein the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of the weather prediction sub-model except the first level of the weather prediction sub-model is communicatively connected to the output end of each of the first n-1 levels of the weather prediction sub-model; Input the meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and obtain meteorological prediction data for a period of time after the current moment; After weighting each of the weather forecast data, the final weather forecast data within the spatial range to be predicted is output.

2. The weather prediction method based on deep learning according to claim 1, characterized in that: The n cascaded semantic segmentation-based weather prediction sub-models include multiple 3D models and at least two of the multiple 3D models, the 2D models include MedNeXt and PVT-CASCADE, and the 3D models include MISSFormer, ConvNeXt, ConResNet and FourCastNet.

3. The weather prediction method based on deep learning as claimed in claim 2, characterized in that: The meteorological data includes a plurality of pressure layer variables and a plurality of ground variables, each of the pressure layer variables includes a plurality of pressure layers, and the ground variables include various surface meteorological indicators; The input shapes of each of the 3D models include: The first channel number, the first depth, the height and the width, the first channel number is equal to a plurality of sampling moments in a period of time before the current moment, and the first depth is the sum of the number of all the air pressure layers and the number of all the surface meteorological indicators; The input shapes of each of the 2D models include: a second number of channels, a height, and a width, wherein the second number of channels is equal to the dimensions of the first channel combined with the first depth; The shapes of the 3D models and the 2D models are: a third channel number, a second depth, a height and a width, wherein the third channel number is a plurality of predicted moments within a period of time after the current moment, and the second depth is equal to the number of ground variables; The height and the width respectively represent the grid points of the spatial range to be predicted.

4. The weather forecasting method based on deep learning according to claim 1, characterized in that: Inputting the meteorological data into each of the weather forecast sub-models respectively includes: Marking the meteorological data by sea and land type and / or soil type and / or altitude to obtain first input data; Obtaining a difference value of meteorological data at sampling moments within a period of time before the current moment, and concatenating the difference value of meteorological data with the meteorological data at the current moment to obtain second input data; Adding Gaussian noise to the second input data to obtain third input data; The first input data, the second input data or the third input data is input into each of the weather prediction sub-models.

5. The weather prediction method based on deep learning according to claim 1, characterized in that: Before the meteorological data is subjected to sea and land type marking and / or soil type marking and / or altitude marking to obtain the first input data, the method further comprises: The mean value and standard deviation of the altitude data are obtained, and the standardized altitude data is obtained according to the ratio of the mean value to the standard deviation, and the corresponding altitude mark data is obtained according to the meteorological data.

6. The weather prediction method based on deep learning according to claim 1, characterized in that: The meteorological data are respectively input into the weather prediction sub-models of each weather prediction model based on deep learning, and the meteorological prediction data for a period of time after the current moment are correspondingly output, including: Determine a weather prediction model group according to the parallel weather prediction sub-models, and determine the deep learning-based weather prediction model based on multiple weather prediction model groups; The meteorological data is respectively input into each of the weather prediction model groups in the deep learning-based weather prediction model, and the meteorological forecast data for each sub-time period in a period of time after the current moment is correspondingly obtained, wherein each sub-time period is continuous in time.

7. The weather prediction method based on deep learning according to claim 1, characterized in that: The method further comprises: Acquire final weather forecast data and real label data, and determine a loss function based on the final weather forecast data and the real label data; According to the loss function, the weather prediction model based on deep learning is trained by using the gradient descent method to obtain the weather prediction model based on deep learning with updated parameters.

8. A weather prediction device based on deep learning, characterized in that: include: An acquisition module, used to acquire meteorological data within a period of time before the current moment within the spatial range to be predicted and a weather prediction model based on deep learning, wherein the weather prediction model based on deep learning is determined according to n cascaded / parallel semantic segmentation-based weather prediction sub-models, wherein the input end of each level of the weather prediction sub-model except the first level of the weather prediction sub-model is communicatively connected to the output end of each of the first n-1 levels of the weather prediction sub-model; A processing module, used to input the meteorological data into each weather prediction sub-model of each weather prediction model based on deep learning, and obtain meteorological prediction data within a period of time after the current moment; The output module is used to output the final meteorological forecast data within the spatial range to be predicted after weighting each of the meteorological forecast data.

9. A computer-readable storage medium, characterized in that: It includes instructions, which, when executed on a computer, enable the computer to execute the weather forecasting method based on deep learning as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the weather forecasting method based on deep learning according to any one of claims 1 to 7.

Images