Deep learning sub-season day-by-day high temperature prediction method and related equipment

Abstract

The invention discloses a deep learning sub-season day-by-day high temperature prediction method and related equipment. The method comprises the following steps: acquiring a basic prediction set of high temperature prediction data in a preset period, wherein the basic prediction set is determined by carrying out arithmetic average on daily high temperature prediction in the preset period in a secondary seasonal mode; training a main model based on the basic prediction set and historical observation data to obtain a prediction correction model for performing high temperature prediction in a future preset period; and performing high temperature prediction in a future preset period through the prediction correction model according to the sub-seasonal mode prediction data and the latest observation data in the future preset period. The problem that due to the uncertainty of the mode and the incomplete accuracy of the initial condition, the prediction result often has magnitude errors or phase errors in the high-temperature process can be solved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of meteorological analysis, and more specifically, the present invention relates to a deep learning sub-seasonal daily high temperature prediction method and related equipment. Background Art

[0002] S2S (Subseasonal to Seasonal Prediction) refers to the prediction of weather and climate on a time scale of several weeks to several months. Such predictions are usually based on NWP (Numerical Weather Prediction Models), which simulate the state of the atmosphere by solving a set of physical equations. However, due to the uncertainty of the model and the inaccuracy of the initial conditions, the prediction results often have magnitude errors or phase errors of high temperature processes. When predicting high temperatures, the model may underestimate or overestimate the temperature, resulting in magnitude errors, causing the forecast results to be inconsistent with the actual situation. The difference between the high temperature occurrence time predicted by the model and the actual high temperature occurrence time results in a phase error of the high temperature process. This error will cause the predicted high temperature event to not accurately reflect the actual time of high temperature occurrence. Summary of the invention

[0003] A series of simplified concepts are introduced in the Summary of the Invention, which will be further described in detail in the Detailed Description of the Invention. The Summary of the Invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0004] In order to solve the problem that when predicting high temperature, the model may underestimate or overestimate the temperature, resulting in an order of magnitude error, which causes the forecast result to be inconsistent with the actual situation. The difference between the high temperature occurrence time predicted by the model and the actual high temperature occurrence time results in a phase error of the high temperature process. This error will cause the predicted high temperature event to not accurately reflect the actual high temperature occurrence time. In the first aspect, the present invention proposes a deep learning sub-seasonal daily high temperature prediction method, which includes:

[0005] Obtaining a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period;

[0006] Training the main model based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period;

[0007] The high temperature prediction for the future preset period is performed through the prediction correction model according to the sub-seasonal model forecast data and the latest observation data within the future preset period.

[0008] Specifically, before training the main model based on the basic prediction set and the historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period, it also includes:

[0009] The historical observation data are obtained by integrating the daily maximum temperature of CRA40 and the observed maximum temperature in the region.

[0010] Specifically, the training of the main model based on the basic prediction set and the historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period includes:

[0011] Taking the basic prediction set and historical observation data as the training data set, the 3D Swin-Transformer deep learning ViT model is trained by utilizing grid sample augmentation, regional model forecast and observation grid data forecast, and the magnitude error and the phase error of the high temperature process are corrected, finally obtaining a prediction correction model for high temperature forecast for future preset periods.

[0012] Specifically, it also includes:

[0013] The data that has not participated in the training are selected as independent sample data to perform model verification on the prediction and correction model.

[0014] Specifically, the selecting of data not involved in training as independent sample data to perform model verification on the prediction and correction model includes:

[0015] Select the data that did not participate in the training as independent sample data to predict the grid high temperature of the specified area;

[0016] The predicted results are compared with the actual observed values ​​to evaluate the prediction accuracy of the model.

[0017] Specifically, it also includes:

[0018] Through the GIS platform, geospatial analysis of observation data and model forecast data is carried out to identify high temperature hotspots and spatial distribution characteristics.

[0019] Specifically, it also includes:

[0020] The 3D Swin-Transformer deep learning model is used to combine geographic location and meteorological data to perform spatial weighted regression analysis on high temperature prediction.

[0021] In a second aspect, the present invention further proposes a deep learning sub-seasonal daily high temperature prediction device, comprising:

[0022] An acquisition unit, used to acquire a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period;

[0023] A training unit, used for training the main model based on the basic prediction set and observation data in the region to obtain a prediction correction model for high temperature forecasting in a future preset period;

[0024] The prediction unit is used to make high temperature predictions for a future preset period based on the sub-seasonal model forecast data for a future preset period and the latest regional observation data through the prediction correction model.

[0025] In a third aspect, an electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the deep learning sub-seasonal daily high temperature prediction method as described in any one of the first aspects above when executing the computer program stored in the memory.

[0026] In a fourth aspect, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the deep learning sub-seasonal daily high temperature prediction method of any one of the above items in the first aspect is implemented.

[0027] In summary, the deep learning sub-seasonal daily high temperature prediction method proposed in this application obtains a basic prediction set of high temperature prediction data within a preset period, and the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of sub-seasonal patterns within the preset period; the main model is trained based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasts for future preset periods; and the high temperature forecast for future preset periods is predicted by the prediction correction model according to the sub-seasonal pattern forecast data and the latest observation data within the future preset period. Since the temperature forecasts for the next few weeks to months are based on numerical weather forecast models, they have certain randomness and errors. By arithmetically averaging the forecasts for each day, the random errors in the single-day forecasts can be reduced, and a more stable forecast can be obtained. By averaging day by day, the random errors in the forecasts are reduced, the data are smoothed, and a more reliable basic data set is provided for subsequent forecasts. The accuracy and stability of the model in high temperature forecasts are ensured. The overall scheme not only improves the magnitude and phase accuracy of high temperature forecasts, but also enhances the practicality of the forecast model, providing effective technical support for responding to climate change and extreme weather.

[0028] The deep learning sub-seasonal daily high temperature prediction method of the present invention, other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will also be understood by technical personnel in the field through research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present specification. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0030] Figure 1 A schematic diagram of a deep learning sub-seasonal daily high temperature prediction method provided in an embodiment of the present application;

[0031] Figure 2 A schematic diagram of another deep learning sub-seasonal daily high temperature prediction method provided in an embodiment of the present application;

[0032] Figure 3 A schematic diagram of the structure of a deep learning sub-seasonal daily high temperature prediction device provided in an embodiment of the present application;

[0033] Figure 4 A schematic diagram of the structure of an electronic device for deep learning sub-seasonal daily high temperature prediction provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0035] In order to solve the problem of high temperature prediction, the model may underestimate or overestimate the temperature, resulting in an order of magnitude error, which causes the forecast result to be inconsistent with the actual situation. The difference between the high temperature occurrence time predicted by the model and the actual high temperature occurrence time results in a phase error of the high temperature process. This error will cause the predicted high temperature event to not accurately reflect the actual high temperature occurrence time. Please refer to Figure 1 , is a schematic flow chart of a deep learning sub-seasonal daily high temperature prediction method provided in an embodiment of the present application, which may specifically include: steps S110 to S130.

[0036] S110, obtaining a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period.

[0037] For example, suppose there is a forecast model that generates a set of high temperature forecasts for each of the next 64 days. We average the forecasts for each of these 64 days to get a more stable set of forecasts.

[0038] Exemplarily, a multi-source fusion basic prediction training set is established to obtain high temperature prediction data within a preset prediction period of 1-60 days in the next. The basic prediction set is the prediction value and observation value of the sub-seasonal pattern within the preset period, wherein the prediction factor data sets include: ① the value determined by averaging the arithmetic median of multiple members of the daily high temperature forecast, ② the high temperature observation value of the previous 1-120 days in history, ③ the multi-year average of high temperature of the next 1-60 days in many years in history; prediction object data set: high temperature data set of the next 1-60 days.

[0039] S120, training the main model based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasting in a future preset period.

[0040] Exemplarily, a 3D Swin-Transformer deep learning model is used to establish a basic prediction set and historical observation data to train the main model, obtain daily grid high temperature values, and obtain a prediction correction model for high temperature forecasts for future preset periods.

[0041] S130, performing high temperature prediction for a future preset period according to the sub-seasonal model forecast data and the latest observation data within the future preset period through the prediction correction model.

[0042] In summary, the deep learning sub-seasonal daily high temperature prediction method provided by the embodiment of the present application obtains a basic prediction set of high temperature prediction data within a preset period, and the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of sub-seasonal patterns within the preset period; the main model is trained based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasts for future preset periods; and the high temperature forecast for future preset periods is performed through the prediction correction model according to the sub-seasonal pattern forecast data and the latest observation data within the future preset period. Since the temperature forecasts for the next few weeks to months are based on numerical weather forecast models, they have certain randomness and errors. By arithmetic averaging the forecasts for each day, the random errors in the single-day forecasts can be reduced, and a more stable forecast can be obtained. By averaging day by day, the random errors in the forecasts are reduced, the data are smoothed, and a more reliable basic data set is provided for subsequent forecasts. The accuracy and stability of the model in high temperature forecasts are ensured. The overall solution not only improves the magnitude and phase accuracy of high temperature forecasts, but also enhances the practicality of the forecast model, providing effective technical support for responding to climate change and extreme weather.

[0043] According to some embodiments, before training the main model based on the basic prediction set and the historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period, the method further includes:

[0044] The historical observation data are obtained by integrating the daily maximum temperature of CRA40 and the observed maximum temperature in the region.

[0045] For example, CRA40 (China Reanalysis) is a high-resolution reanalysis dataset that contains meteorological data such as global daily temperatures. Reanalysis data are generated based on observational data and numerical weather forecast models. The observed maximum temperature in the region can be the daily maximum temperature data actually observed in the Chinese region, which is more accurate and specific. Integrating CRA40 and observational data in the Chinese region can provide more comprehensive and accurate high temperature data and enhance the input data quality of the model. For example, the daily maximum temperature data of CRA40 is integrated with the maximum temperature data actually observed at Chinese meteorological stations, such as through interpolation or data fusion methods, so that the dataset has both global coverage and regional accuracy.

[0046] In some examples, the training of the main model based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period includes:

[0047] The 3D Swin-TransformerViT model is trained using the basic prediction set and historical observation data as training data sets to obtain a prediction correction model for high temperature forecasting for a future preset period.

[0048] Exemplarily, the 3D Swin-Transformer ViT (Vision Transformer) model is a deep learning model based on the self-attention mechanism, which is good at processing images and spatiotemporal data. The long-distance dependencies of the input data can be captured through the self-attention mechanism to improve the prediction accuracy. Then, the powerful feature extraction and spatiotemporal relationship capture capabilities of the 3DSwin-Transformer ViT model are used to enhance the accuracy of high temperature prediction. For example, using the basic prediction set and historical observation data as training data sets, the 3DSwin-Transformer ViT model is trained using grid sample augmentation, regional pattern forecasts, and observation grid data forecasts, and the magnitude error and the error of the high temperature process phase are corrected, and finally a prediction correction model for high temperature forecasting for future preset periods is obtained.

[0049] In some examples, this also includes:

[0050] The data that has not participated in the training are selected as independent sample data to perform model verification on the prediction and correction model.

[0051] In some examples, the selecting data that has not participated in the training as independent sample data to perform model verification on the prediction and correction model includes:

[0052] Select the data that did not participate in the training as independent sample data to predict the grid high temperature of the specified area;

[0053] The predicted results are compared with the actual observed values ​​to evaluate the prediction accuracy of the model.

[0054] For example, independent sample tests are used to verify the actual prediction ability of the model and ensure the robustness and accuracy of the model on different data sets. For example, an area that has not participated in the training is selected, and the historical observation data of the area is input into the model to predict the high temperature in the next 64 days, and then compared with the actual observations to evaluate the model performance.

[0055] It is understandable that the training samples are enriched, data noise is reduced, and the robustness of the model is improved through daily arithmetic averaging and data integration. Combining observational data and model forecast data, the advanced 3D Swin-Transformer ViT model is used for feature extraction and high temperature prediction, which improves the accuracy and stability of the model. Through supervised learning and independent sample verification, the robustness and practical application effect of the model on different data sets are ensured, providing a reliable technical means for high temperature prediction.

[0056] In some examples, this also includes:

[0057] Through the GIS platform, geospatial analysis of observation data and model forecast data is carried out to identify high temperature hotspots and spatial distribution characteristics.

[0058] For example, geospatial analysis of observation data and model forecast data can be performed through GIS platforms (such as ArcGIS and QGIS) to identify high temperature hotspots and spatial distribution characteristics. Combining geographic information for data preprocessing and feature extraction can improve the spatial resolution capability of the model.

[0059] In some examples, this also includes:

[0060] The 3D Swin-Transformer ViT model was used to combine geographic location and meteorological data to perform spatial weighted regression analysis for high temperature prediction.

[0061] For example, the 3D Swin-Transformer ViT model is used to combine geographic location and meteorological data to perform spatial weighted regression analysis on high temperature prediction. For example, in the prediction of urban heat island effect, urban geographic information and meteorological data are combined to improve the spatial accuracy and reliability of the prediction.

[0062] In some examples, this also includes:

[0063] Acquire image data uploaded by a user on a network platform with a geographical location marker located near the adjacent reference site and the site to be processed;

[0064] Selecting image data with outdoor clothing information as target supplementary training data, wherein the image data is associated with shooting time;

[0065] The prediction and correction model is optimized based on the target supplementary training data.

[0066] For example, since there is a huge amount of data information in social media currently, the image data uploaded by platform users covers a very wide range. Image data with geographic location tags and located near the neighboring reference sites and the sites to be processed can be used. If the image data has outdoor clothing information, since outdoor clothing information and temperature have a very high correlation, the outdoor clothing information can be used to further optimize the prediction correction model parameters to provide high-precision temperature predictions.

[0067] See also Figure 3 , an embodiment of the deep learning sub-seasonal daily high temperature prediction device in the embodiment of the present application may include:

[0068] An acquisition unit 21 is used to acquire a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period;

[0069] A training unit 22, used for training the main model based on the basic prediction set and the observation data in the region to obtain a prediction correction model for high temperature forecasting in a future preset period;

[0070] The prediction unit 23 is used to make high temperature predictions for a future preset period based on the sub-seasonal model forecast data for a future preset period and the latest regional observation data through the prediction correction model.

[0071] In summary, the deep learning sub-seasonal daily high temperature prediction device provided by the embodiment of the present application obtains a basic prediction set of high temperature prediction data within a preset period, and the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of sub-seasonal patterns within the preset period; the main model is trained based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasts for future preset periods; and the high temperature forecast for future preset periods is performed through the prediction correction model according to the sub-seasonal pattern forecast data and the latest observation data within the future preset period. Since the temperature forecasts for the next few weeks to months are based on numerical weather forecast models, they have certain randomness and errors. By arithmetic averaging the forecasts for each day, the random errors in the single-day forecasts can be reduced, and a more stable forecast can be obtained. By averaging day by day, the random errors in the forecasts are reduced, the data are smoothed, and a more reliable basic data set is provided for subsequent forecasts. The accuracy and stability of the model in high temperature forecasts are ensured. The overall solution not only improves the magnitude and phase accuracy of high temperature forecasts, but also enhances the practicality of the forecast model, providing effective technical support for responding to climate change and extreme weather.

[0072] like Figure 4As shown, the embodiment of the present application further provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 320 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any one of the above-mentioned deep learning methods for sub-seasonal daily high temperature prediction are implemented:

[0073] Obtaining geographical environment data, time data and temperature of a reference station adjacent to the station to be processed, wherein the time data is determined based on a time range of actual observation data of the station to be processed;

[0074] Based on the data set constructed based on the geographic environment data, time data and temperature of the adjacent reference station and the actual observation data of the station to be processed, the XGBoost algorithm model based on the gradient boosting tree is trained to obtain a temperature data interpolation and extension model;

[0075] According to the air temperature of the adjacent reference station, the data interpolation and extension model is used to estimate some missing data or data of the unobserved period of the station to be processed to complete the data interpolation and extension.

[0076] Specifically, by calculating the site similarity index, neighboring reference sites of the site to be processed are selected from the meteorological stations with long-sequence observation data.

[0077] Specifically, the geographical environment data includes at least one of longitude, latitude, altitude, slope and aspect, and the slope and aspect are determined by interpolation based on high-precision DEM data using a neighbor interpolation algorithm.

[0078] Specifically, it also includes:

[0079] Obtain at least one of relative humidity, sunshine and precipitation of a reference station adjacent to the station to be processed,

[0080] The actual observation data of the site to be processed is used as the label, and the data of the neighboring reference sites is used as the input feature quantity to construct the training data set and the test data set.

[0081] Specifically, it also includes:

[0082] The model parameters are optimized by cross-validation method to optimize the temperature data interpolation and extension model.

[0083] Specifically, it also includes:

[0084] Acquire image data uploaded by a user on a network platform with a geographical location marker located near the adjacent reference site and the site to be processed;

[0085] Selecting image data with outdoor clothing information as target supplementary training data, wherein the image data is associated with shooting time;

[0086] The temperature data interpolation and extension model is optimized based on the target supplementary training data.

[0087] Since the electronic device introduced in this embodiment is a device used to implement a deep learning sub-seasonal daily high temperature prediction device in the embodiment of the present application, based on the method introduced in the embodiment of the present application, the technical personnel in this field can understand the specific implementation mode of the electronic device of this embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiment of the present application is not introduced in detail here. As long as the equipment used by the technical personnel in this field to implement the method in the embodiment of the present application is within the scope of protection of this application.

[0088] In the specific implementation process, when the computer program 311 is executed by the processor, it can achieve Figure 1 Any implementation method in the corresponding embodiment:

[0089] Obtaining geographical environment data, time data and temperature of a reference station adjacent to the station to be processed, wherein the time data is determined based on a time range of actual observation data of the station to be processed;

[0090] Based on the data set constructed based on the geographic environment data, time data and temperature of the adjacent reference station and the actual observation data of the station to be processed, the XGBoost algorithm model based on the gradient boosting tree is trained to obtain a temperature data interpolation and extension model;

[0091] According to the air temperature of the adjacent reference station, the data interpolation and extension model is used to estimate some missing data or data of the unobserved period of the station to be processed to complete the data interpolation and extension.

[0092] Specifically, by calculating the site similarity index, neighboring reference sites of the site to be processed are selected from the meteorological stations with long-sequence observation data.

[0093] Specifically, the geographical environment data includes at least one of longitude, latitude, altitude, slope and aspect, and the slope and aspect are determined by interpolation based on high-precision DEM data using a neighbor interpolation algorithm.

[0094] Specifically, it also includes:

[0095] Obtain at least one of relative humidity, sunshine and precipitation of a reference station adjacent to the station to be processed,

[0096] The actual observation data of the site to be processed is used as the label, and the data of the neighboring reference sites is used as the input feature quantity to construct the training data set and the test data set.

[0097] Specifically, it also includes:

[0098] The model parameters are optimized by cross-validation method to optimize the temperature data interpolation and extension model.

[0099] Specifically, it also includes:

[0100] Acquire image data uploaded by a user on a network platform with a geographical location marker located near the adjacent reference site and the site to be processed;

[0101] Selecting image data with outdoor clothing information as target supplementary training data, wherein the image data is associated with shooting time;

[0102] The temperature data interpolation and extension model is optimized based on the target supplementary training data.

[0103] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0104] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0105] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0106] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0108] The present application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device is caused to execute the following Figure 1 The process of deep learning sub-seasonal daily high temperature prediction in the corresponding embodiment.

[0109] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or a data center that includes one or more available media integration. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)), etc.

[0110] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0111] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0112] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0113] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0114] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0115] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A deep learning sub-seasonal daily high temperature prediction method, characterized in that: include: Obtaining a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period; Training the main model based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasting for a future preset period; The high temperature prediction for the future preset period is performed through the prediction correction model according to the sub-seasonal model forecast data and the latest observation data within the future preset period.

2. The method according to claim 1, characterized in that Before training the main model based on the basic prediction set and the historical observation data to obtain a prediction correction model for high temperature forecasting in a future preset period, the method further includes: The historical observation data are obtained by integrating the daily maximum temperature of CRA40 and the observed maximum temperature in the region.

3. The method according to claim 1, characterized in that The main model is trained based on the basic prediction set and historical observation data to obtain a prediction correction model for high temperature forecasting in a future preset period, including: The 3D ViT model is trained using the basic prediction set and historical observation data as training data sets to obtain a prediction correction model for high temperature forecasting for a future preset period.

4. The method according to claim 1, characterized in that Also includes: The data that has not participated in the training are selected as independent sample data to perform model verification on the prediction and correction model.

5. The method according to claim 4, characterized in that The selecting of data not involved in training as independent sample data to perform model verification on the prediction and correction model includes: Select the data that did not participate in the training as independent sample data to predict the grid high temperature of the specified area; The predicted results are compared with the actual observed values ​​to evaluate the prediction accuracy of the model.

6. The method according to any one of claims 1 to 5, characterized in that Also includes: Through the GIS platform, geospatial analysis of observation data and model forecast data is carried out to identify high temperature hotspots and spatial distribution characteristics.

7. The method according to any one of claims 1 to 5, characterized in that Also includes: The 3D Swin-Transformer deep learning model is used to combine geographic location and meteorological data to perform spatial weighted regression analysis on high temperature prediction.

8. A deep learning sub-seasonal daily high temperature prediction device, characterized in that: include: An acquisition unit, used to acquire a basic prediction set of high temperature prediction data within a preset period, wherein the basic prediction set is determined by arithmetic averaging of daily high temperature forecasts of a sub-seasonal model within the preset period; A training unit, used for training the main model based on the basic prediction set and observation data in the region to obtain a prediction correction model for high temperature forecasting in a future preset period; The prediction unit is used to make high temperature predictions for a future preset period based on the sub-seasonal model forecast data for a future preset period and the latest regional observation data through the prediction correction model.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the deep learning sub-seasonal daily high temperature prediction method as described in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the deep learning sub-seasonal daily high temperature prediction method according to any one of claims 1 to 7 is implemented.

Images