Time-space downscaling method, device and equipment for meteorological data under complex terrain and medium
Through the combination of multi-source heterogeneous data sources and deep learning network models, the problem of low weather forecasting accuracy under complex terrain is solved, and the meteorological data under complex terrain is achieved is achieved, and the accuracy and meticulousness of meteorological forecasting is improved. sex.
Patent Information
- Application Number
- CN202410965976.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to accurately capture the impact of terrain on the weather system under complex terrain, resulting in low weather forecast accuracy, and complex calculations and long time. It is difficult to fully consider the terrain's ups and downs, mountains, rivers and other factors on the airflow movement and Effects of humidity distribution.
Multi-source heterogeneous data source and deep learning network model are used to achieve fine prediction of meteorological data under complex terrain through data preprocessing, embedded vector transformation, feature extraction of multi-head attention network layer and spatiotemporal relationship modeling.
It improves the accuracy and accuracy of weather forecasts, can capture meteorological information under complex terrain more comprehensively, realizes fine prediction of meteorological data under complex terrain, and provides more detailed time and space information for meteorological forecasts.
Smart Images

Figure CN118940211A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment and medium for spatiotemporal downscaling of meteorological data under complex terrain. Background Art
[0002] The existing time-space downscaling methods for weather forecasting under complex terrain mainly include traditional numerical weather forecast models and meteorological simulation methods under the influence of terrain. Traditional numerical weather forecast models usually use a grid mode to divide the atmosphere into grid cells for simulation, but in complex terrain areas, this mode often cannot accurately capture the impact of terrain on the weather system, resulting in low forecast accuracy. On the other hand, meteorological simulation methods under the influence of terrain attempt to improve forecast accuracy by considering the impact of terrain on atmospheric movement, but these methods are often computationally complex and time-consuming. In addition, when processing meteorological data under complex terrain, the existing technology is difficult to fully consider the impact of factors such as terrain ups and downs, mountains, rivers, etc. on the movement of airflow and humidity distribution, so how to achieve the prediction of meteorological data under complex terrain has become a technical issue that cannot be underestimated. Summary of the invention
[0003] In view of this, the purpose of this application is to provide a method, device, equipment and medium for spatiotemporal downscaling of meteorological data under complex terrain. By utilizing multi-source heterogeneous data sources and deep learning network models, it is possible to more comprehensively capture meteorological information under complex terrain, thereby improving the precision and accuracy of weather forecasts. The deep learning network model can effectively learn the spatiotemporal relationship between different data sources, effectively improve the spatiotemporal resolution of meteorological data, achieve precise prediction of meteorological data under complex terrain, and provide more detailed information for meteorological forecasts.
[0004] The present application embodiment provides a temporal and spatial downscaling method for meteorological data under complex terrain, the temporal and spatial downscaling method comprising:
[0005] Acquire meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine processed meteorological data;
[0006] Inputting the processed meteorological data into multiple embedded vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, performing embedded vector conversion processing on the processed meteorological data, and outputting multiple meteorological data vectors under the complex terrain;
[0007] Input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then perform feature extraction and spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iterative training of a deep learning network model.
[0008] In a possible implementation manner, the preprocessing of the meteorological data under the complex terrain to determine the processed meteorological data includes:
[0009] Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
[0010] In a possible implementation manner, the processed meteorological data is input into the embedded vector network layer of the pre-trained meteorological data spatiotemporal downscaling model, the processed meteorological data is subjected to embedded vector conversion processing, and a plurality of meteorological data vectors under the complex terrain are output, including:
[0011] The processed meteorological data from different multi-source heterogeneous data sources are input into the corresponding embedding vector network layer, the meteorological data is mapped to a low-dimensional embedding space and the semantic information of the meteorological data is retained to generate the meteorological data vector; wherein the embedding vector network layers corresponding to different multi-source heterogeneous data sources are different.
[0012] In a possible implementation, the step of inputting the plurality of meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fusing the plurality of meteorological data vectors, then performing feature extraction and spatiotemporal relationship modeling on the fused meteorological data vectors, and outputting the spatiotemporal downscaled meteorological data includes:
[0013] Inputting the plurality of meteorological data vectors into the multi-head attention network layer, performing vector concatenation on the plurality of meteorological data vectors, and determining the fused meteorological data vector;
[0014] Performing feature extraction on the fused meteorological data vector to determine a plurality of feature vectors;
[0015] Based on the multi-head attention mechanism, the spatiotemporal relationship modeling process is performed between the multiple feature vectors, and the spatiotemporal downscaled meteorological data under the complex terrain is output.
[0016] In a possible implementation manner, the spatiotemporal downscaling model of the meteorological data is determined by the following steps:
[0017] Acquire sample meteorological data under multiple complex terrains and label information of each sample meteorological data;
[0018] Input each of the sample meteorological data into the deep learning network model, perform spatiotemporal downscaling on the sample meteorological data, and determine the sample meteorological data after spatiotemporal downscaling;
[0019] Determining a loss value of the deep learning network model based on each of the sample meteorological data after the spatiotemporal downscaling and label information of each of the sample meteorological data;
[0020] The deep learning network model is iteratively trained based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data.
[0021] In a possible implementation manner, if the loss value is less than or equal to a preset loss threshold, the iterative training of the deep learning network model is stopped, and the deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data;
[0022] If the loss value is greater than the preset loss threshold, the network parameters of the deep learning network model are changed and the deep learning network model is iteratively trained until the loss value is less than the preset loss threshold, and the training is stopped. The deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data.
[0023] The embodiment of the present application also provides a temporal and spatial downscaling device for meteorological data under complex terrain, the temporal and spatial downscaling device comprising:
[0024] A data processing module, used to obtain meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine the processed meteorological data;
[0025] An embedding vector generation module is used to input the processed meteorological data into multiple embedding vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, perform embedding vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain;
[0026] The prediction module is used to input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then extract features and perform spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model.
[0027] In a possible implementation manner, when the data processing module is used to perform data preprocessing on the meteorological data under the complex terrain and determine the processed meteorological data, the data processing module is specifically used to:
[0028] Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
[0029] An embodiment of the present application also provides an electronic device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the spatiotemporal downscaling method for meteorological data under complex terrain as described above are performed.
[0030] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the spatiotemporal downscaling method for meteorological data under complex terrain as described above are executed.
[0031] The embodiments of the present application provide a method, device, equipment and medium for spatiotemporal downscaling of meteorological data under complex terrain. The spatiotemporal downscaling method includes: acquiring meteorological data under complex terrain based on multi-source heterogeneous data sources, performing data preprocessing on the meteorological data under complex terrain, and determining processed meteorological data; inputting the processed meteorological data into multiple embedded vector network layers of a pre-trained spatiotemporal downscaling model for meteorological data, performing embedded vector conversion processing on the processed meteorological data, and outputting multiple meteorological data vectors under the complex terrain; inputting multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fusing the multiple meteorological data vectors, then performing feature extraction and spatiotemporal relationship modeling processing on the fused meteorological data vectors, and outputting the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model. By utilizing multi-source heterogeneous data sources and deep learning network models, it is possible to more comprehensively capture meteorological information in complex terrain, thereby improving the precision and accuracy of weather forecasts. In addition, deep learning network models can effectively learn the spatiotemporal relationship between different data sources, effectively improve the spatiotemporal resolution of meteorological data, and achieve precise prediction of meteorological data in complex terrain, providing more detailed information for weather forecasts.
[0032] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0034] Figure 1 A flow chart of a method for spatiotemporal downscaling of meteorological data under complex terrain provided in an embodiment of the present application;
[0035] Figure 2 One of the structural schematic diagrams of a spatiotemporal downscaling device for meteorological data under complex terrain provided in an embodiment of the present application;
[0036] Figure 3 A second structural schematic diagram of a spatiotemporal downscaling device for meteorological data under complex terrain provided in an embodiment of the present application;
[0037] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0038] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme 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. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art, under the guidance of the content of the present application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.
[0039] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0040] In order to enable those skilled in the art to use the contents of this application, the following implementation method is given in combination with the specific application scenario of "temporal and spatial downscaling of meteorological data under complex terrain". For those skilled in the art, the general principles defined here can be applied to other embodiments and application scenarios without departing from the spirit and scope of this application.
[0041] First, the application scenarios to which the present application is applicable are introduced. The present application can be applied in the field of data processing technology.
[0042] According to research, the existing time-space downscaling methods for weather forecasting under complex terrain mainly include traditional numerical weather forecast models and meteorological simulation methods under the influence of terrain. Traditional numerical weather forecast models usually use a grid mode to divide the atmosphere into grid cells for simulation, but in complex terrain areas, this mode often cannot accurately capture the impact of terrain on the weather system, resulting in low forecast accuracy. On the other hand, meteorological simulation methods under the influence of terrain attempt to improve forecast accuracy by considering the impact of terrain on atmospheric movement, but these methods are often computationally complex and time-consuming. In addition, when processing meteorological data under complex terrain, the existing technology is difficult to fully consider the impact of factors such as terrain ups and downs, mountains, rivers, etc. on the movement of airflow and humidity distribution, so how to achieve the prediction of meteorological data under complex terrain has become a technical issue that cannot be underestimated.
[0043] Based on this, an embodiment of the present application provides a spatiotemporal downscaling method for meteorological data under complex terrain. By utilizing multi-source heterogeneous data sources and deep learning network models, it is possible to more comprehensively capture meteorological information under complex terrain, thereby improving the precision and accuracy of weather forecasts. In addition, the deep learning network model can effectively learn the spatiotemporal relationship between different data sources, effectively improve the spatiotemporal resolution of meteorological data, achieve precise prediction of meteorological data under complex terrain, and provide more detailed information for meteorological forecasts.
[0044] See also Figure 1 , Figure 1 A flow chart of a method for spatiotemporal downscaling of meteorological data in complex terrain provided in an embodiment of the present application. Figure 1 As shown in , the spatiotemporal downscaling method provided in the embodiment of the present application includes:
[0045] S101: Acquire meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine processed meteorological data.
[0046] In this step, meteorological data under complex terrain are acquired from multi-source heterogeneous data sources, data preprocessing is performed on the meteorological data under complex terrain, and processed meteorological data are determined.
[0047] In a possible implementation manner, the preprocessing of the meteorological data under the complex terrain to determine the processed meteorological data includes:
[0048] Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
[0049] Here, data cleaning, data normalization and feature extraction are performed on a plurality of meteorological data to determine the processed meteorological data.
[0050] Among them, multiple meteorological data are preprocessed, including data cleaning, normalization, feature extraction, etc. Each heterogeneous data source may have different characteristics and distributions, and they need to be converted into a unified representation for subsequent processing.
[0051] S102: Input the processed meteorological data into multiple embedded vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, perform embedded vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain.
[0052] In this step, the processed meteorological data is input into multiple embedded vector network layers of the meteorological data spatiotemporal downscaling model, the processed meteorological data is subjected to embedded vector conversion processing, and multiple meteorological data vectors under complex terrain are output.
[0053] Here, multiple embedding vector network layers are added to the input of the meteorological data spatiotemporal downscaling model to convert the meteorological data into a low-dimensional dense vector representation. For each data source, an independent embedding vector network layer can be designed to map the meteorological data into a low-dimensional embedding space. This can retain the semantic information of the data and reduce the dimension of the data, which is conducive to improving the model learning and generalization capabilities.
[0054] In a possible implementation manner, the processed meteorological data is input into the embedded vector network layer of the pre-trained meteorological data spatiotemporal downscaling model, the processed meteorological data is subjected to embedded vector conversion processing, and a plurality of meteorological data vectors under the complex terrain are output, including:
[0055] The processed meteorological data from different multi-source heterogeneous data sources are input into the corresponding embedding vector network layer, the meteorological data is mapped to a low-dimensional embedding space and the semantic information of the meteorological data is retained to generate the meteorological data vector; wherein the embedding vector network layers corresponding to different multi-source heterogeneous data sources are different.
[0056] Here, the processed meteorological data from different multi-source heterogeneous data sources are input into the corresponding embedding vector network layer, the meteorological data is mapped into a low-dimensional embedding space and the semantic information of the meteorological data is retained to generate a meteorological data vector.
[0057] In this application, due to the use of a variety of complex terrain information and multi-source data sources, it can better meet the needs of weather forecasting under complex terrain conditions. The spatiotemporal downscaling model of meteorological data can fully consider the impact of terrain on meteorological phenomena, improve the understanding and prediction capabilities of meteorological changes under complex terrain, and provide reliable support for weather forecasting under different terrain conditions.
[0058] S103: Input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then extract features and perform spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model.
[0059] In this step, multiple meteorological data vectors are input into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, and the multiple meteorological data vectors are first fused, and then the fused meteorological data vectors are subjected to feature extraction and spatiotemporal relationship modeling, and the spatiotemporal downscaled meteorological data are output.
[0060] Among them, the meteorological data after spatiotemporal downscaling is the downscaling of meteorological data in time and space, and the meteorological data after spatiotemporal downscaling is more detailed. For example, the meteorological data of a complex area is reported once every three hours, while the meteorological data after spatiotemporal downscaling can be reported once every hour for a specific location in the complex area.
[0061] In a possible implementation, the step of inputting the plurality of meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fusing the plurality of meteorological data vectors, then performing feature extraction and spatiotemporal relationship modeling on the fused meteorological data vectors, and outputting the spatiotemporal downscaled meteorological data includes:
[0062] A: Input the multiple meteorological data vectors into the multi-head attention network layer, perform vector splicing on the multiple meteorological data vectors, and determine the fused meteorological data vector.
[0063] Here, multiple meteorological data vectors are input into the multi-head attention network layer, and the multiple meteorological data vectors are vector-concatenated to determine the fused meteorological data vector.
[0064] Among them, in addition to the vector splicing method, multiple meteorological data vectors can also be fused through other methods such as vector addition and vector weighted average, which are not specifically limited in this part.
[0065] B: Extract features from the fused meteorological data vector to determine multiple feature vectors.
[0066] Here, feature extraction is performed on the fused meteorological data vector to determine a plurality of feature vectors.
[0067] C: Based on the multi-head attention mechanism, the spatiotemporal relationship between the multiple feature vectors is modeled and processed, and the spatiotemporal downscaled meteorological data under the complex terrain is output.
[0068] Here, the spatiotemporal relationship between multiple feature vectors is modeled and processed according to the multi-head attention mechanism, and the spatiotemporal downscaled meteorological data under complex terrain is output.
[0069] The spatiotemporal relationship modeling is to model multiple feature vectors in the time dimension and the space dimension, so that the spatiotemporal information between the multiple feature vectors can be understood.
[0070] Here, the multi-head attention network layer is used for feature extraction and spatiotemporal relationship modeling, which can effectively learn the complex relationship between meteorological data and realize the prediction ability of spatiotemporal downscaling. The multi-head attention mechanism can capture the dependency relationship and spatiotemporal information between meteorological data, and realize efficient feature extraction and modeling. The multi-head attention network layer has strong representation and generalization capabilities, can handle the complex relationship of multi-source heterogeneous data, and improve the prediction precision and accuracy of the model.
[0071] In a possible implementation manner, the spatiotemporal downscaling model of the meteorological data is determined by the following steps:
[0072] (1): Obtain sample meteorological data under multiple complex terrains and label information of each sample meteorological data.
[0073] Here, sample meteorological data under multiple complex terrains and label information of each sample meteorological data are obtained.
[0074] The label information of the sample meteorological data is the real sample meteorological data after spatiotemporal downscaling.
[0075] (2): Each of the sample meteorological data is input into the deep learning network model, and the sample meteorological data is subjected to spatiotemporal downscaling to determine the sample meteorological data after spatiotemporal downscaling.
[0076] Here, each sample meteorological data is input into the deep learning network model, the sample meteorological data is temporally and spatially downscaled, and the sample meteorological data after temporal and spatial downscaling is determined.
[0077] The spatiotemporal downscaling process is the process of first performing fusion processing, then performing feature extraction and spatiotemporal relationship modeling processing as described in the above step S102, which will not be described in detail in this part.
[0078] (3): Based on each of the sample meteorological data after spatiotemporal downscaling and the label information of each of the sample meteorological data, the loss value of the deep learning network model is determined.
[0079] Here, the loss value of the deep learning network model is determined based on the sample meteorological data after each spatiotemporal downscaling and the label information of each sample meteorological data.
[0080] Among them, any loss function in the prior art can determine the above loss value, which will not be described in detail in this part.
[0081] (4): Iteratively train the deep learning network model based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data.
[0082] Here, the deep learning network model is iteratively trained according to the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data.
[0083] In a possible implementation manner, the iterative training of the deep learning network model based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data includes:
[0084] a: If the loss value is less than or equal to a preset loss threshold, the iterative training of the deep learning network model is stopped, and the deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data.
[0085] Here, if the loss value is less than or equal to the preset loss threshold, the iterative training of the deep learning network model is stopped, and the deep learning network model is determined as a spatiotemporal downscaling model of meteorological data.
[0086] b: If the loss value is greater than the preset loss threshold, the network parameters of the deep learning network model are changed and the deep learning network model is continued to be iteratively trained until the loss value is less than the preset loss threshold, and the training is stopped, and the deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data.
[0087] Here, if the loss value is greater than the preset loss threshold, the network parameters of the deep learning network model are changed and the deep learning network model is iteratively trained until the loss value is less than the preset loss threshold, and the training is stopped. The deep learning network model is determined as a spatiotemporal downscaling model for meteorological data.
[0088] In a specific embodiment, step 1: preprocess the meteorological data under complex terrain, including data cleaning, normalization, feature extraction, etc. Each data source may have different characteristics and distributions, and they need to be converted into a unified representation for subsequent processing. Step 2: Input the processed meteorological data into the corresponding embedding vector network layer to convert the original data into a low-dimensional dense vector representation. For each data source, an independent Embedding layer can be designed to map the original data into a low-dimensional embedding space. This can retain the semantic information of the data and reduce the dimension of the data, which is conducive to the improvement of model learning and generalization capabilities. Step 3: Fusion of meteorological data vectors from various data sources can be achieved by simple splicing, addition, weighted averaging, etc. This can integrate information from different data sources and provide richer features for the multi-head attention network layer. Step 4: The multi-head attention network layer extracts features and models the spatiotemporal relationship of the fused meteorological data vector to achieve the prediction capability of spatiotemporal downscaling, and outputs the spatiotemporal downscaled meteorological data.
[0089] In this scheme, by utilizing multi-source heterogeneous data sources and the multi-head attention network layer in the deep learning network model, it is possible to capture meteorological information under complex terrain more comprehensively, thereby improving the precision and accuracy of weather forecasts. The model can effectively learn the spatiotemporal relationship between different data sources, extract key features and perform efficient downscaling processing, so that the forecast results are more in line with the actual situation. Through the spatiotemporal relationship modeling capability of the multi-head attention network layer, the spatiotemporal resolution of weather forecasts can be effectively improved. The model can capture local meteorological phenomena under complex terrain conditions, realize fine prediction of local meteorological changes, and provide more detailed spatiotemporal information for meteorological forecasts. Due to the use of terrain information and multi-source data sources, it can better adapt to the needs of weather forecasts under complex terrain conditions. The model can fully consider the impact of terrain on meteorological phenomena, improve the understanding and prediction capabilities of meteorological changes under complex terrain, and provide reliable support for meteorological forecasts under different terrain conditions.
[0090] The embodiment of the present application provides a spatiotemporal downscaling method for meteorological data under complex terrain, and the spatiotemporal downscaling method includes: acquiring meteorological data under complex terrain based on multi-source heterogeneous data sources, performing data preprocessing on the meteorological data under complex terrain, and determining processed meteorological data; inputting the processed meteorological data into multiple embedded vector network layers of a pre-trained spatiotemporal downscaling model for meteorological data, performing embedded vector conversion processing on the processed meteorological data, and outputting multiple meteorological data vectors under the complex terrain; inputting multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fusing the multiple meteorological data vectors, then performing feature extraction and spatiotemporal relationship modeling processing on the fused meteorological data vectors, and outputting the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model. By utilizing multi-source heterogeneous data sources and deep learning network models, it is possible to more comprehensively capture meteorological information in complex terrain, thereby improving the precision and accuracy of weather forecasts. In addition, deep learning network models can effectively learn the spatiotemporal relationship between different data sources, effectively improve the spatiotemporal resolution of meteorological data, and achieve precise prediction of meteorological data in complex terrain, providing more detailed information for weather forecasts.
[0091] See also Figure 2 , Figure 3 , Figure 2 One of the structural schematic diagrams of a spatiotemporal downscaling device for meteorological data under complex terrain provided in an embodiment of the present application; Figure 3 This is a second structural diagram of a spatiotemporal downscaling device for meteorological data in complex terrain provided by an embodiment of the present application. Figure 2 As shown in , the spatiotemporal downscaling device 200 for meteorological data under complex terrain includes:
[0092] The data processing module 210 is used to obtain meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine the processed meteorological data;
[0093] The embedding vector generation module 220 is used to input the processed meteorological data into multiple embedding vector network layers of the pre-trained meteorological data spatiotemporal downscaling model, perform embedding vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain;
[0094] The prediction module 230 is used to input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then extract features and perform spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model.
[0095] Furthermore, when the data processing module 210 is used to perform data preprocessing on the meteorological data under the complex terrain and determine the processed meteorological data, the data processing module 210 is specifically used to:
[0096] Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
[0097] Further, when the prediction module 230 is used to input the processed meteorological data into the embedded vector network layer of the pre-trained meteorological data spatiotemporal downscaling model, perform embedded vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain, the prediction module 230 is specifically used to:
[0098] The processed meteorological data from different multi-source heterogeneous data sources are input into the corresponding embedding vector network layer, the meteorological data is mapped to a low-dimensional embedding space and the semantic information of the meteorological data is retained to generate the meteorological data vector; wherein the embedding vector network layers corresponding to different multi-source heterogeneous data sources are different.
[0099] Further, when the prediction module 230 is used to input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, firstly fuse the multiple meteorological data vectors, then perform feature extraction and spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data, the prediction module 230 is specifically used to:
[0100] Inputting the plurality of meteorological data vectors into the multi-head attention network layer, performing vector concatenation on the plurality of meteorological data vectors, and determining the fused meteorological data vector;
[0101] Performing feature extraction on the fused meteorological data vector to determine a plurality of feature vectors;
[0102] Based on the multi-head attention mechanism, the spatiotemporal relationship modeling process is performed between the multiple feature vectors, and the spatiotemporal downscaled meteorological data under the complex terrain is output.
[0103] Further, such as Figure 3 As shown, the spatiotemporal downscaling device for meteorological data under complex terrain further includes a model training module 240. The model training module 240 determines the spatiotemporal downscaling model of the meteorological data through the following steps:
[0104] Acquire sample meteorological data under multiple complex terrains and label information of each sample meteorological data;
[0105] Input each of the sample meteorological data into the deep learning network model, perform spatiotemporal downscaling on the sample meteorological data, and determine the sample meteorological data after spatiotemporal downscaling;
[0106] Determining a loss value of the deep learning network model based on each of the sample meteorological data after the spatiotemporal downscaling and label information of each of the sample meteorological data;
[0107] The deep learning network model is iteratively trained based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data.
[0108] Further, when the model training module 240 is used to iteratively train the deep learning network model based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data, the model training module 240 is specifically used to:
[0109] If the loss value is less than or equal to a preset loss threshold, stopping iterative training of the deep learning network model, and determining the deep learning network model as the spatiotemporal downscaling model of the meteorological data;
[0110] If the loss value is greater than the preset loss threshold, the network parameters of the deep learning network model are changed and the deep learning network model is iteratively trained until the loss value is less than the preset loss threshold, and the training is stopped. The deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data.
[0111] The embodiment of the present application provides a spatiotemporal downscaling device for meteorological data under complex terrain, and the spatiotemporal downscaling device includes: a data processing module, which is used to obtain meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under complex terrain, and determine the processed meteorological data; an embedding vector generation module, which is used to input the processed meteorological data into multiple embedding vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, perform embedding vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain; a prediction module, which is used to input multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first perform fusion processing on the multiple meteorological data vectors, then perform feature extraction and spatiotemporal relationship modeling processing on the fused meteorological data vectors, and output the meteorological data after spatiotemporal downscaling; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model. By utilizing multi-source heterogeneous data sources and deep learning network models, it is possible to more comprehensively capture meteorological information in complex terrain, thereby improving the precision and accuracy of weather forecasts. In addition, deep learning network models can effectively learn the spatiotemporal relationship between different data sources, effectively improve the spatiotemporal resolution of meteorological data, and achieve precise prediction of meteorological data in complex terrain, providing more detailed information for weather forecasts.
[0112] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown in , the electronic device 400 includes a processor 410 , a memory 420 and a bus 430 .
[0113] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the above-mentioned Figure 1 The steps of the spatiotemporal downscaling method for meteorological data under complex terrain in the method embodiment shown in the figure, and the specific implementation method can be found in the method embodiment, which will not be repeated here.
[0114] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The steps of the spatiotemporal downscaling method for meteorological data under complex terrain in the method embodiment shown in the figure, and the specific implementation method can be found in the method embodiment, which will not be repeated here.
[0115] 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.
[0116] 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. The device embodiments described above are merely schematic. For example, the division of the 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 through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.
[0117] 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.
[0118] 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.
[0119] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application can essentially be embodied in the form of a software product, or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including several instructions for 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 method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0120] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. A spatiotemporal downscaling method for meteorological data in complex terrain, characterized in that: The spatiotemporal downscaling method comprises: Acquire meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine processed meteorological data; Inputting the processed meteorological data into multiple embedded vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, performing embedded vector conversion processing on the processed meteorological data, and outputting multiple meteorological data vectors under the complex terrain; Input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then perform feature extraction and spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iterative training of a deep learning network model.
2. The spatiotemporal downscaling method according to claim 1, characterized in that: The step of preprocessing the meteorological data under the complex terrain to determine the processed meteorological data includes: Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
3. The spatiotemporal downscaling method according to claim 1, characterized in that: The processed meteorological data is input into a plurality of embedded vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, the processed meteorological data is subjected to embedded vector conversion processing, and a plurality of meteorological data vectors under the complex terrain are output, including: The processed meteorological data from different multi-source heterogeneous data sources are input into the corresponding embedding vector network layer, the meteorological data is mapped to a low-dimensional embedding space and the semantic information of the meteorological data is retained to generate the meteorological data vector; wherein the embedding vector network layers corresponding to different multi-source heterogeneous data sources are different.
4. The spatiotemporal downscaling method according to claim 1, characterized in that: The step of inputting the plurality of meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, firstly fusing the plurality of meteorological data vectors, then extracting features and performing spatiotemporal relationship modeling on the fused meteorological data vectors, and outputting the spatiotemporal downscaled meteorological data includes: Inputting the plurality of meteorological data vectors into the multi-head attention network layer, performing vector concatenation on the plurality of meteorological data vectors, and determining the fused meteorological data vector; Performing feature extraction on the fused meteorological data vector to determine a plurality of feature vectors; Based on the multi-head attention mechanism, the spatiotemporal relationship modeling process is performed on the multiple feature vectors, and the spatiotemporal downscaled meteorological data under the complex terrain is output.
5. The spatiotemporal downscaling method according to claim 1, characterized in that: The spatiotemporal downscaling model of the meteorological data is determined by the following steps: Acquire sample meteorological data under multiple complex terrains and label information of each sample meteorological data; Input each of the sample meteorological data into the deep learning network model, perform spatiotemporal downscaling on the sample meteorological data, and determine the sample meteorological data after spatiotemporal downscaling; Determining a loss value of the deep learning network model based on each of the sample meteorological data after the spatiotemporal downscaling and label information of each of the sample meteorological data; The deep learning network model is iteratively trained based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data.
6. The spatiotemporal downscaling method according to claim 5, characterized in that: The iterative training of the deep learning network model based on the loss value until the iterative training is completed to determine the spatiotemporal downscaling model of the meteorological data includes: If the loss value is less than or equal to a preset loss threshold, stopping iterative training of the deep learning network model, and determining the deep learning network model as the spatiotemporal downscaling model of the meteorological data; If the loss value is greater than the preset loss threshold, the network parameters of the deep learning network model are changed and the deep learning network model is iteratively trained until the loss value is less than the preset loss threshold, and the training is stopped. The deep learning network model is determined as the spatiotemporal downscaling model of the meteorological data.
7. A temporal and spatial downscaling device for meteorological data in complex terrain, characterized in that: The space-time downscaling device comprises: A data processing module, used to obtain meteorological data under complex terrain based on multi-source heterogeneous data sources, perform data preprocessing on the meteorological data under the complex terrain, and determine the processed meteorological data; An embedding vector generation module is used to input the processed meteorological data into multiple embedding vector network layers of a pre-trained meteorological data spatiotemporal downscaling model, perform embedding vector conversion processing on the processed meteorological data, and output multiple meteorological data vectors under the complex terrain; The prediction module is used to input the multiple meteorological data vectors into the multi-head attention network layer of the meteorological data spatiotemporal downscaling model, first fuse the multiple meteorological data vectors, then extract features and perform spatiotemporal relationship modeling on the fused meteorological data vectors, and output the spatiotemporal downscaled meteorological data; wherein the meteorological data spatiotemporal downscaling model is obtained by iteratively training a deep learning network model.
8. The spatiotemporal downscaling device according to claim 7, characterized in that: When the data processing module is used to perform data preprocessing on the meteorological data under the complex terrain and determine the processed meteorological data, the data processing module is specifically used to: Data cleaning, data normalization and feature extraction are performed on the plurality of meteorological data to determine the processed meteorological data.
9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the spatiotemporal downscaling method for meteorological data under complex terrain as described in any one of claims 1 to 6.
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 steps of the spatiotemporal downscaling method for meteorological data under complex terrain as claimed in any one of claims 1 to 6 are executed.
Citation Information
Patent Citations
Transformer-based power load prediction method
CN113592185A
Multi-source heterogeneous data fusion-based gale prediction algorithm along high-speed rail
CN117150430A
Short temporary rainfall prediction method based on multi-scale attention and convolution fusion
CN118298222A
Meteorological big data fusion method based on deep learning
US20230351164A1
Cited By
Multi-channel meteorological risk early warning information adaptive targeted publishing system and method based on low-altitude flight
CN120564389A
