Meteorological information determination method and device, electronic equipment and storage medium

Abstract

The invention provides a meteorological information determination method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, in particular to the fields of deep learning, large models and the like. According to the specific implementation scheme, feature extraction is carried out on meteorological raster data of a target region in a target time period to obtain meteorological feature vectors; inputting the to-be-processed meteorological data of the target region in the target time period into the large language model to obtain a text abstract including a meteorological information determination mode; performing information enhancement processing on the meteorological feature vector by using the text abstract to obtain an information enhancement result; and performing self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the to-be-processed meteorological data.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the fields of deep learning and large models, and specifically to a method, device, electronic device and storage medium for determining meteorological information. Background Art

[0002] Meteorology, including weather and climate information. With the frequent occurrence of extreme weather events and the impact of climate change around the world, the importance of weather and climate modeling has become increasingly prominent. Summary of the invention

[0003] The present disclosure provides a method, device, electronic device and storage medium for determining meteorological information.

[0004] According to one aspect of the present disclosure, a method for determining meteorological information is provided, comprising: extracting features from meteorological raster data of a target area within a target time period to obtain a meteorological feature vector; inputting the to-be-processed meteorological data of the target area within the target time period into a large language model to obtain a text summary including a method for determining meteorological information; performing information enhancement processing on the meteorological feature vector using the text summary to obtain an information enhancement result; and performing self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the to-be-processed meteorological data.

[0005] According to another aspect of the present disclosure, a meteorological information determination device is provided, including: a feature extraction module, used to extract features from meteorological raster data of a target area within a target time period to obtain a meteorological feature vector; a large language model module, used to input the to-be-processed meteorological data of the target area within the target time period into the large language model to obtain a text summary including a meteorological information determination method; an information enhancement module, used to perform information enhancement processing on the meteorological feature vector using the text summary to obtain an information enhancement result; and a self-attention processing module, used to perform self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the to-be-processed meteorological data.

[0006] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the meteorological information determination method of the present disclosure.

[0007] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the meteorological information determination method of the present disclosure.

[0008] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program is stored on at least one of a readable storage medium and an electronic device, and when the computer program is executed by a processor, the meteorological information determination method of the present disclosure is implemented.

[0009] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0011] Figure 1 An exemplary system architecture to which the method and device for determining meteorological information according to an embodiment of the present disclosure can be applied is schematically shown;

[0012] Figure 2 A flowchart of a method for determining meteorological information according to an embodiment of the present disclosure is schematically shown;

[0013] Figure 3A The framework diagram of the meteorological deep learning model based on LLM text enhancement according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 3B The overall architecture diagram of the meteorological deep learning model based on LLM text enhancement according to an embodiment of the present disclosure is schematically shown;

[0015] Figure 4 A block diagram schematically shows a weather information determination device according to an embodiment of the present disclosure; and

[0016] Figure 5 A block diagram of an electronic device that can be used to implement an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0017] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0018] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information involved comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0019] In the technical solution of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0020] Related technologies use atmospheric physics numerical models and data-driven deep learning models to model weather and climate.

[0021] Physics-based numerical models of the atmosphere are designed to simulate nonlinear dynamics and complex interactions between multiple variables. The most representative of these are general circulation models (GCMs), which are systems of differential equations related to the flow of energy and matter in the atmosphere, land, and oceans that can be integrated over time to obtain forecasts of atmospheric variables.

[0022] In the process of implementing the concept of the present disclosure, the inventors found that although the atmospheric physics numerical model is the most widely used in practice, this method is a computationally intensive model, which has poor ability to accurately represent physical processes and initial conditions at a fine granularity, and cannot be expanded in terms of data scale and lacks flexibility. These factors limit their use in many scenarios, especially in the rapid simulation of atmospheric variables on a very short time scale such as a few hours, or the accurate simulation of atmospheric variables on a long time scale such as a few hours.

[0023] Based on data-driven deep learning models, the key idea is to train deep neural networks to predict target atmospheric variables using reanalysis of decades of historical global datasets. Unlike GCMs, these networks are not explicitly based on physics but are trained for specific predictive modeling tasks.

[0024] In the process of implementing the concept of the present disclosure, the inventors found that the data-driven deep learning model was trained for specific predictive modeling tasks and could not meet the requirements of spatiotemporal task training for heterogeneous data sets, so it lacks the versatility of earth system science. In addition, homogeneous data set training makes it difficult to fully mine climate information and effectively utilize its hidden features. These reasons lead to the poor performance of data-driven deep learning models in the prediction of extreme weather events and long-term climate forecasts under limited spatiotemporal supervision and computing budget.

[0025] In summary, atmospheric physics-based numerical models only support computational expansion and cannot be flexibly applied to data sets of different sizes. Data-driven models are usually limited to specific tasks and lack versatility in different practical scenarios.

[0026] Therefore, it is critical to explore how to effectively use heterogeneous climate data and fully explore their characteristics.

[0027] The present disclosure provides a method, device, electronic device and storage medium for determining meteorological information, and relates to the field of artificial intelligence technology, and in particular to the fields of deep learning, large models, etc. The method for determining meteorological information includes: extracting features from meteorological raster data of a target area within a target time period to obtain a meteorological feature vector; inputting the to-be-processed meteorological data of the target area within the target time period into a large language model to obtain a text summary including a method for determining meteorological information; performing information enhancement processing on the meteorological feature vector using the text summary to obtain an information enhancement result; and performing self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the to-be-processed meteorological data.

[0028] Figure 1 An exemplary system architecture to which the method and apparatus for determining meteorological information according to an embodiment of the present disclosure can be applied is schematically shown.

[0029] It should be noted that Figure 1 The examples shown are only examples of system architectures to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, the exemplary system architecture to which the meteorological information determination method and apparatus can be applied may include a terminal device, but the terminal device may implement the meteorological information determination method and apparatus provided by the embodiments of the present disclosure without interacting with the server.

[0030] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0031] The user may use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only as examples).

[0032] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0033] The server 105 may be a server that provides various services, such as a background management server that provides support for the content browsed by the user using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (for example only). The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device. The server may also be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0034] It should be noted that the meteorological information determination method provided in the embodiment of the present disclosure can generally be executed by the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the content processing device provided in the embodiment of the present disclosure can also be set in the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0035] Alternatively, the meteorological information determination method provided in the embodiment of the present disclosure may also be generally performed by the server 105. Accordingly, the meteorological information determination device provided in the embodiment of the present disclosure may generally be arranged in the server 105. The meteorological information determination method provided in the embodiment of the present disclosure may also be performed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the meteorological information determination device provided in the embodiment of the present disclosure may also be arranged in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0036] For example, when determining meteorological information, the first terminal device 101, the second terminal device 102, and the third terminal device 103 can obtain meteorological grid data and meteorological data to be processed in the target area within the target time period, and then send the obtained meteorological grid data and meteorological data to be processed to the server 105, and the server 105 extracts features from the meteorological grid data to obtain a meteorological feature vector; inputs the meteorological data to be processed into a large language model to obtain a text summary including a method for determining meteorological information; uses the text summary to perform information enhancement processing on the meteorological feature vector to obtain an information enhancement result; and performs self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the meteorological data to be processed. Alternatively, a server or server cluster that can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105 analyzes the meteorological grid data and the meteorological data to be processed, and obtains a meteorological information determination result output for the meteorological data to be processed.

[0037] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0038] Figure 2 The flowchart of the method for determining meteorological information according to an embodiment of the present disclosure is schematically shown.

[0039] like Figure 2 As shown, the method includes operations S210 to S240.

[0040] In operation S210, feature extraction is performed on the meteorological grid data of the target area within the target time period to obtain a meteorological feature vector.

[0041] According to an embodiment of the present disclosure, the meteorological grid data may be grid data in NC (a grid common data format for meteorological data) format downloaded based on climate data sets such as CMIP6 (Coupled Model Intercomparison Project, the sixth international coupled model comparison program), ERA5 (the fifth generation atmospheric reanalysis data set), etc. The number of grid rows and columns of the grid data may be determined by the longitude and latitude information of the earth grid. Each grid in the meteorological grid data may include, for example, temperature, sea level pressure, humidity, wind speed, radiation, precipitation, snow depth, sunshine hours, etc. at each longitude and latitude position, and may not be limited thereto.

[0042] According to the embodiments of the present disclosure, any form of feature extraction module with feature extraction function can be used to extract features to obtain meteorological feature vectors. A pre-trained deep learning model can also be used to extract features to obtain meteorological feature vectors. This is not limited here. The deep learning model may, for example, include a pre-trained large unsupervised basic model or other neural network model, and may not be limited thereto. By using a pre-trained large unsupervised basic model for feature extraction, the supervision bottleneck problem can be alleviated and the generalization ability of the model can be improved.

[0043] In operation S220, the to-be-processed meteorological data of the target region within the target time period is input into the large language model to obtain a text summary including a method for determining the meteorological information.

[0044] According to the embodiments of the present disclosure, the meteorological data to be processed may represent basic meteorological data adapted to specific business needs, which is collected from climate data sets such as CMIP6 and EAR5 according to business needs. Since business needs are diverse, the data structure, data content, etc. of the basic meteorological data obtained for different business needs may be different. For example, the meteorological data to be processed may be meteorological raster data, or may be array-type basic meteorological data that only includes meteorological attributes required by business needs, or may include location information of the target area, which is not limited here.

[0045] According to an embodiment of the present disclosure, the large language model may be a large language model (LLM) generated by pre-training, which can output a text summary including a method for determining meteorological information based on the meteorological data to be processed as input. The text summary may only include text information representing a method for determining meteorological information, such as determining meteorological information for a certain period of time, a certain location, a certain meteorological attribute, etc. The text summary may also include text information marking a method for determining meteorological information and text information representing part or all of the meteorological information to be processed, and may not be limited thereto.

[0046] In operation S230, information enhancement processing is performed on the meteorological feature vector using the text summary to obtain an information enhancement result.

[0047] According to an embodiment of the present disclosure, the information enhancement result can represent the result of data enhancement of meteorological raster data using text summary. In this operation, the data enhancement method can be expressed as: fusing the text summary with the meteorological raster data, extracting features from the fusion result of the text summary and the meteorological raster data, and obtaining the information enhancement result. The data enhancement method can also be expressed as: extracting features from the text summary, fusing the feature extraction result of the text summary with the meteorological feature vector, and obtaining the information enhancement result.

[0048] It should be noted that the above-mentioned information enhancement method is only an exemplary embodiment, but is not limited thereto, and may also include other information enhancement methods known in the art, as long as an information enhancement result can be obtained.

[0049] In operation S240, self-attention processing is performed on the information enhancement result to obtain a meteorological information determination result output for the meteorological data to be processed.

[0050] According to an embodiment of the present disclosure, a transformer (a neural network model based on a self-attention mechanism) can be used to perform an encoding operation based on a self-attention mechanism on the information enhancement result to obtain an encoding vector. Then, a decoding operation based on a self-attention mechanism can be performed on the encoding vector to obtain a meteorological information determination result.

[0051] It should be noted that using transformer for self-attention processing is only an exemplary embodiment, but is not limited to this. Other models based on the self-attention mechanism known in the art can also be used as long as they can perform self-attention processing.

[0052] According to an embodiment of the present disclosure, the meteorological information determination result may represent the result obtained after processing the meteorological data to be processed based on the meteorological raster data and the meteorological information determination method in the text summary.

[0053] Through the above-mentioned embodiments of the present disclosure, the large language model is used to process the meteorological data to be processed, which helps to better extract the text modal information of the meteorological data to be processed. The obtained text summary can contain the core points and key information in the meteorological data to be processed. By using the text summary as an enhanced modality and performing information enhancement processing on the meteorological feature vector, it is possible to focus on these core points and key information on the basis of the meteorological feature vector, accurately capture the intention and theme in the text summary, better mine the potential features in the meteorological raster data and the data to be processed, and obtain the meteorological information determination result that conforms to the intention and theme in the text summary output for the meteorological data to be processed. In addition, since the text summary is obtained by processing the meteorological data to be processed using a large language model, when implementing the relevant operations of this method, the data modality of the meteorological data to be processed can be not limited, and the full mining of various types of heterogeneous meteorological data can be effectively realized, so that the method can be flexibly applied to data sets of different sizes and realize universality in multimodal data scenarios.

[0054] In conjunction with specific embodiments, Figure 2 The method shown is further explained.

[0055] According to an embodiment of the present disclosure, based on the advantage of using LLM to deeply mine text information in the above method, a meteorological deep learning model based on LLM text enhancement can also be constructed. By pre-training the meteorological deep learning model based on LLM text enhancement, the meteorological information determination method can be accurately performed, including performing the above operations S210~S240.

[0056] Figure 3A The framework diagram of the meteorological deep learning model based on LLM text enhancement according to an embodiment of the present disclosure is schematically shown.

[0057] like Figure 3A As shown, the meteorological deep learning model 300 based on LLM text enhancement can be a model obtained by modeling using ViT (Vision Transformers, a deep learning model based on Transformer architecture), including a deep learning module 310, a text generation enhancement module 320, and a transformer 330.

[0058] According to an embodiment of the present disclosure, the deep learning module 310 can receive the meteorological raster data 311_0 collected based on EAR5 and CMIP6 as input, perform feature extraction, and output a meteorological feature vector 315. The text generation enhancement module 320 can receive the meteorological data to be processed 321 as input, and output a text summary or its vector representation 323. The transformer 330 can receive the meteorological feature vector 315 and the text summary or its vector representation 323, perform self-attention processing, and obtain a meteorological information determination result 331.

[0059] To achieve the above purpose, we can use the self-supervision objectives of climate datasets such as CMIP6 and ERA5 to Figure 3A The meteorological deep learning model architecture shown in the figure is pre-trained, including: the text summary of the meteorological data to be processed generated by LLM is used as the enhanced modality, and is input into the ViT architecture together with the meteorological feature vector for pre-training. It can also be fine-tuned based on various downstream tasks related to weather and climate, so as to make full use of data to improve the prediction effect.

[0060] Figure 3B The overall architecture diagram of the meteorological deep learning model based on LLM text enhancement according to an embodiment of the present disclosure is schematically shown.

[0061] like Figure 3B As shown, the deep learning module 310 may include a variable labeling module 311 and a variable aggregation module 312, but is not limited thereto. The deep learning module 310 is designed as a generalizable deep learning model that can input heterogeneous data sets of different variables and process heterogeneous data of different dimensions.

[0062] According to an embodiment of the present disclosure, after obtaining meteorological raster data, the above operation S210 can be performed in combination with the deep learning module 310 in the meteorological deep learning model 300 based on LLM text enhancement obtained by pre-training. The above operation S210 may include: marking the meteorological raster data into multiple sequence fragments, and the spatial resolution represented by the multiple sequence fragments is the same. Linearly transforming the sequence fragments to obtain a sequence vector representation. Aggregating the sequence vector representation to obtain a sequence aggregation vector. Fusion of multiple sequence aggregation vectors corresponding to the multiple sequence fragments to obtain a raster aggregation vector. Determine the meteorological feature vector based on the raster aggregation vector.

[0063] During the application process, the deep learning module 310 can adopt the shape of Input, Refers to the number of input variables, which can be weather conditions or meteorological factors. and Refers to the spatial resolution of the input data, which depends on how densely the Earth grid is gridded. This general representation allows flexibility in handling a wide variety of downstream tasks in Earth system science.

[0064] Reference Figure 3B As shown, the variable labeling module 311 can firstly convert the shape into The input token is a sequence ,get sequence segments 311_1, and each segment can be linearly embedded into a The output dimension is The sequence vector representation 311_2 is used as the output of the variable labeling module 311. Among them, is the patch size, which represents the image block.

[0065] In the process of implementing the concept of the present disclosure, the inventors found that while the variable labeling module alleviates the training of irregular data sets, it also brings problems such as the computational complexity increasing with the input variables and the sequence easily containing labels of different variables with different physical bases.

[0066] According to an embodiment of the present disclosure, the output of the variable labeling module 311 may be further processed in combination with the variable aggregation module 312. The variable aggregation module 312 may perform aggregation processing in a cross-attention manner, but is not limited thereto.

[0067] Reference Figure 3B As shown, the variable aggregation module 312 can A cross-attention operation is performed once for each spatial position in . In the process of performing the cross-attention operation, Q (query) can be a learnable vector, K (Key) and V (Value) can be the vectors of the position. variables The variable aggregation module 312 can output a single vector for each spatial position, thereby The sequence length is reduced to , and the sequence length is The variable aggregation module 312 may then fuse multiple sequence aggregation vectors 312_1 to obtain a grid aggregation vector 312_2 as the output of the variable aggregation module 312.

[0068] According to an embodiment of the present disclosure, after obtaining the raster aggregation vector, when the leading time embedding module 313 and the leading position embedding module 314 are not added to the deep learning module 310, the raster aggregation vector 312_2 can be determined as the meteorological feature vector 315 obtained by extracting the meteorological raster data.

[0069] Through the above-mentioned embodiments of the present disclosure, by marking the meteorological raster data into multiple sequence segments with the same spatial resolution, flexible processing of irregular data sets can be achieved. By aggregating the sequence vector representations of multiple training segments, an embedding vector of equal size can be output for each spatial position, thereby reducing the computational cost.

[0070] Reference Figure 3B As shown, at least one of the leading time embedding module 313 and the leading position embedding module 314 may also be added to the deep learning module 310. The leading time embedding module 313 and the leading position embedding module 314 may be a single-layer MLP.

[0071] According to an embodiment of the present disclosure, by adding a pre-time embedding module 313 in the deep learning module 310, a pre-trained model generally applicable to various time prediction tasks can be constructed. The time prediction task can represent the prediction of meteorological data for a certain time period. By adding a pre-position embedding module 314 in the deep learning module 310, a pre-trained model generally applicable to various regional prediction tasks can be constructed. The regional prediction task can represent the prediction of meteorological data for a certain region.

[0072] Taking the pre-training model for implementing the time prediction task as an example, during the pre-training period, the latitude weighted mean square error can be used to construct the loss function shown in formula (1), and the lead time information provided by the lead time embedding module 313 can be randomized from 6 hours to 168 hours (one week), and is not limited to this, to support the implementation of the time prediction task.

[0073] Formula (1)

[0074] In formula (1), is the information of the target time period, is the lead time information, which can be generated randomly. is the latitude weighting factor, for The forecast results for the time period, for The actual result of the time period is determined as shown in formula (2).

[0075] Formula (2)

[0076] In formula (2), For the grid The latitude corresponding to the row, Represents an index variable that is used to traverse all latitude information in the grid when calculating the latitude weighting factor. The latitude weighting factor can account for regional non-uniformity when drawing the earth grid. For example, the grid cells close to the equator have larger areas than the grid cells close to the poles, so more weight should be assigned.

[0077] The pre-training process of the pre-training model for implementing the regional prediction task is similar to the pre-training process of the pre-training model for implementing the time prediction task. It is only necessary to replace the leading time information in the loss function with the leading position information, which will not be repeated here.

[0078] After the pre-training process of the above framework, a deep learning model that learns to perceive a large amount of meteorological data can be obtained. When it is necessary to perform predictions for specific tasks for different downstream meteorological tasks, it is only necessary to select the corresponding loss function for fine-tuning. The prediction head can select models such as MLP, but is not limited to this.

[0079] It should be noted that if Figure 3A and Figure 3B The pre-training framework shown is only an exemplary embodiment, but the present disclosure is not limited to this pre-training method. The architecture and pre-training strategies of other large meteorological models can be applied to the meteorological information determination method of the present disclosure.

[0080] According to an embodiment of the present disclosure, determining a meteorological feature vector based on a sequence aggregation vector may further include: fusing at least one of the leading time embedding information and the leading position embedding information with the grid aggregation vector to obtain the meteorological feature vector.

[0081] Reference Figure 3B As shown, when at least one of the leading time embedding module 313 and the leading position embedding information 314 has been added to the deep learning module 310, at least one of the leading time embedding information provided by the leading time embedding module 313 and the leading position embedding information provided by the leading position embedding module 314 can be added to the grid aggregation vector 312_2 to obtain the meteorological feature vector 315. In order to perform the time prediction task and the regional prediction task in combination with at least one of the leading time embedding information and the leading position embedding information in the future.

[0082] Through the above-mentioned embodiments of the present disclosure, by providing leading time embedded information and leading position embedded information, the task of determining meteorological information in the time dimension and the geographical dimension can be achieved, which is conducive to expanding the application scenarios of the meteorological information determination method.

[0083] According to an embodiment of the present disclosure, after aggregating multimodal meteorological raster data in different time and space based on the deep learning module 310, it is possible to consider pre-training the generated LLM on a large text corpus to better mine the potential information of the meteorological data to be processed and improve the effect of the pre-trained model.

[0084] According to an embodiment of the present disclosure, the LLM may be provided in the text generation enhancement module 320 .

[0085] like Figure 3B As shown, a LLM 322 can be defined in the text generation enhancement module 320, and the LLM 332 can capture the meteorological data to be processed. Contextual understanding and text summarization .in, means LLM 322, is to instruct LLM to Contextual prompts.

[0086] According to an embodiment of the present disclosure, after obtaining the meteorological data to be processed, the above operation S220 can be performed in combination with the text generation enhancement module 320 in the meteorological deep learning model 300 based on LLM text enhancement obtained by pre-training. The above operation S220 may include: generating target text information representing the derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target region. Generate a text summary according to the context information of the basic meteorological data and the target text information.

[0087] According to the embodiments of the present disclosure, the derived meteorological data can represent data that cannot be directly collected from climate data sets such as CMIP6 and EAR5 and need to be calculated based on basic meteorological data. For example, the derived meteorological data can represent the overall meteorological information of a region that needs to be comprehensively determined based on the basic meteorological information of each location in the region, and can also represent the meteorological change information of the region within a preset time period that needs to be predicted based on the basic meteorological information of the region within the target time period, and can also represent the meteorological change information in a region adjacent to the region that needs to be predicted based on the basic meteorological information of the region within the target region, and it is not limited to this.

[0088] According to an embodiment of the present disclosure, at least one of the basic meteorological data, the information of the target time period, and the location information of the target region can be input into a pre-trained large language model, and the output represents the target text information used to determine the derived meteorological data. The text summary can include only the target text information, or it can include the context information of the basic meteorological data and the target text information, which is not limited here.

[0089] To achieve the above purpose, for example, basic meteorological data can be used as input of LLM 322, and information including text information of "derived meteorological data of certain attributes" can be used as output of the large language model to train the large language model. When put into use, the unprocessed meteorological data 321 collected based on climate data sets such as CMIP6 and EAR5 can be input into LLM 322, and a text summary 323_1 including the above target text information can be output.

[0090] Based on this text summary, a prediction task can be generated.

[0091] According to an embodiment of the present disclosure, since meteorological data has typical time series, basic meteorological data may include meteorological time series data. The above-mentioned generating target text information representing the target for determining derived meteorological data according to at least one of the basic meteorological data, information of the target time period, and location information of the target region may include: generating first text information according to information of the target time period, the first text information representing meteorological changes within a preset time period after the target time period. Determine the target text information according to the first text information.

[0092] To achieve this goal, for example, historical meteorological time series data can be used as the input of the large language model, and information including the text information "predicting meteorological changes in the future time period" can be used as the output of the large language model to train the large language model. When put into use, the meteorological time series data in the first time period can be input into the large language model, and the text summary of the first text information including "predicting meteorological changes in the second time period" can be output.

[0093] Based on this embodiment, the above-mentioned time prediction task can be constructed.

[0094] According to an embodiment of the present disclosure, considering that there is a business demand for predicting meteorological information in an adjacent region, the above-mentioned generating target text information representing the target for determining derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target region may include: generating second text information according to the location information of the target region, the second text information representing the meteorological changes in an extended region adjacent to the target region. Determine the target text information according to the second text information.

[0095] To achieve this goal, for example, the basic meteorological data of the target area can be used as the input of the large language model, and the information including the text information "the meteorological changes in the extended area adjacent to the target area" can be used as the output of the large language model to train the large language model. When put into use, the location information and basic meteorological data of the target area can be input into the large language model, and the text summary including the second text information "predicting the meteorological changes in a certain extended area" can be output.

[0096] Based on this embodiment, the above-mentioned regional prediction task can be constructed.

[0097] Through the above-mentioned embodiments of the present disclosure, the large language model can make full use of its rich semantic and contextual understanding capabilities, discover features that are ignored by traditional time series data processing models, provide richer information for the overall meteorological information determination method, and improve the accuracy of downstream prediction tasks.

[0098] According to an embodiment of the present disclosure, after obtaining the text summary and the meteorological feature vector, the above operation S230 can be performed in combination with the text generation enhancement module 320 and the transformer 330 in the meteorological deep learning model 300 based on LLM text enhancement obtained by pre-training. The above operation S230 may include: performing a linear transformation on the text summary to obtain a summary vector representation, and the vector dimension of the summary vector representation is the same as the vector dimension of the meteorological feature vector. The meteorological feature vector and the summary vector representation are fused to obtain a fused feature vector. The fused feature vector is determined as the information enhancement result.

[0099] According to an embodiment of the present disclosure, after obtaining the text summary, the text summary needs to be linearly transformed to obtain a summary representation vector. The linear transformation process can be implemented by setting a linear transformation module in the text generation enhancement module 320, or by setting a linear transformation module in the transformer 330, which is not limited here.

[0100] Reference Figure 3BAs shown, in order to use the text summary for the ViT framework, a linear transformation module can be constructed in the text generation enhancement module 320 by using a combination of CLS Token and Linear Classifier, and a linear transformation process is performed on the text summary output by LLM 322 to obtain a summary vector representation 323_2, and the vector dimension of the summary vector representation 323_2 is kept the same as the vector dimension of the meteorological feature vector 315. After that, the meteorological feature vector 315 and the summary vector representation 323_2 can be input into transformer 330 for fusion to obtain a fused feature vector. The fused feature vector can be used as an information enhancement result, and after the self-attention-based encoding and self-attention-based decoding process in transformer 330, the meteorological information determination result output for the meteorological data to be processed can be obtained.

[0101] According to an embodiment of the present disclosure, corresponding to the above-mentioned time prediction task, in the process of performing the above-mentioned fusion of the meteorological feature vector and the summary vector representation, it can also include: obtaining time information representing the preset time period to be determined in the text summary from the summary vector representation. Fusion of the time information with the leading time embedding information in the meteorological feature vector.

[0102] For example, given the shape At a specific time Short-term weather in , the text summary output by the large language model can include "determine The weather after ”, which can be obtained from the summary vector representation of the text summary , and in the fusion process The fused feature vector can include the meteorological feature vector enhanced by the summary vector representation. Combined with the subsequent transformer self-attention processing, we can get The weather after .

[0103] According to an embodiment of the present disclosure, corresponding to the above-mentioned regional prediction task, in the process of performing the above-mentioned fusion of the meteorological feature vector and the summary vector representation, it can also include: obtaining the location information of the extended region to be determined in the representation text summary from the summary vector representation. The location information of the extended region is fused with the pre-position embedded information in the meteorological feature vector.

[0104] For example, given the shape In a specific area Weather in , the text summary output by the large language model can include "determine Weather in the area ”, which can be obtained from the summary vector representation of the text summary , and in the fusion process The fused feature vector can include the meteorological feature vector enhanced by the summary vector representation. Combined with the subsequent transformer self-attention processing, we can get Weather in the area .

[0105] It should be noted that if Figure 3A and Figure 3B Each module in the model architecture shown can be implemented using any neural network unit with corresponding functions. The meteorological information determination method disclosed in the present invention is not limited to using Figure 3A and Figure 3B The model architecture shown can also be implemented by training deep learning models of other structures, which is not limited here.

[0106] Through the above-mentioned embodiments of the present disclosure, it is possible to effectively achieve the expansion of meteorological data scale and meteorological data modality, and utilize a unique generated text generation enhancement module to optimize data feature extraction, so as to achieve more accurate prediction of downstream meteorological tasks. Specifically, the biggest innovation of this solution is the text generation enhancement module based on a large language model. The text summary is generated based on the meteorological data to be processed, and usually contains the core points and key information in the text, which are often crucial for the prediction task. By focusing on these key information, the model can more accurately capture the intention and theme of the text, better explore the potential features of the meteorological data to be processed, enhance the training and reasoning effects of the large meteorological model, and improve the performance of downstream prediction tasks. Combined with labeling processing and cross-attention operations, it can process various types of irregular data and effectively improve the efficiency and versatility of calculations.

[0107] Figure 4 A block diagram of a weather information determination device according to an embodiment of the present disclosure is schematically shown.

[0108] like Figure 4 As shown, the meteorological information determination device 400 includes a feature extraction module 410, a large language model module 420, an information enhancement module 430 and a self-attention processing module 440.

[0109] The feature extraction module 410 is used to extract features from the meteorological grid data of the target area within the target time period to obtain a meteorological feature vector.

[0110] The large language model module 420 is used to input the to-be-processed meteorological data of the target region within the target time period into the large language model to obtain a text summary including a method for determining the meteorological information.

[0111] The information enhancement module 430 is used to perform information enhancement processing on the meteorological feature vector using the text summary to obtain an information enhancement result.

[0112] The self-attention processing module 440 is used to perform self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the meteorological data to be processed.

[0113] According to an embodiment of the present disclosure, the meteorological data to be processed includes basic meteorological data and location information of a target region. The large language model module includes a target text generation unit and a text summary generation unit.

[0114] The target text generation unit is used to generate target text information representing the derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target area.

[0115] The text summary generating unit is used to generate a text summary according to the context information of the basic meteorological data and the target text information.

[0116] According to an embodiment of the present disclosure, the basic meteorological data includes meteorological time series data. The target text generation unit includes a first text generation subunit and a first target text determination subunit.

[0117] The first text generation subunit is used to generate first text information according to the information of the target time period, wherein the first text information represents the meteorological changes within a preset time period after the target time period.

[0118] The first target text determination subunit is used to determine target text information according to the first text information.

[0119] According to an embodiment of the present disclosure, the target text generating unit includes a second text generating subunit and a second target text determining subunit.

[0120] The second text generation subunit is used to generate second text information according to the location information of the target area, and the second text information represents the weather changes in the extended area adjacent to the target area.

[0121] The second target text determination subunit is used to determine the target text information according to the second text information.

[0122] According to an embodiment of the present disclosure, the feature extraction module includes a labeling unit, a first linear transformation unit, an aggregation unit, a first fusion unit and a meteorological feature vector determination unit.

[0123] The marking unit is used to mark the meteorological grid data into a plurality of sequence segments, and the spatial resolution represented by the plurality of sequence segments is the same.

[0124] The first linear transformation unit is used to perform linear transformation on the sequence fragments to obtain a sequence vector representation.

[0125] The aggregation unit is used to aggregate the sequence vector representation to obtain a sequence aggregation vector.

[0126] The first fusion unit is used to fuse multiple sequence aggregation vectors corresponding to multiple sequence segments to obtain a grid aggregation vector.

[0127] The meteorological feature vector determining unit is used to determine the meteorological feature vector according to the grid aggregation vector.

[0128] According to an embodiment of the present disclosure, the meteorological feature vector determination unit includes a fusion subunit.

[0129] The fusion subunit is used to fuse at least one of the leading time embedding information and the leading position embedding information with the grid aggregation vector to obtain a meteorological feature vector.

[0130] According to an embodiment of the present disclosure, the information enhancement module includes a second linear transformation unit, a second fusion unit and an information enhancement result determination unit.

[0131] The second linear transformation unit is used to perform linear transformation on the text summary to obtain a summary vector representation, and the vector dimension of the summary vector representation is the same as the vector dimension of the meteorological feature vector.

[0132] The second fusion unit is used to fuse the meteorological feature vector and the summary vector representation to obtain a fused feature vector.

[0133] The information enhancement result determination unit is used to determine the fused feature vector as the information enhancement result.

[0134] According to an embodiment of the present disclosure, the meteorological feature vector includes leading time embedding information. The second fusion unit includes a time information acquisition subunit and a time information embedding subunit.

[0135] The time information acquisition subunit is used to acquire time information representing a preset time period to be determined in the text summary from the summary vector representation.

[0136] The time information embedding subunit is used to fuse the time information with the preceding time embedding information.

[0137] According to an embodiment of the present disclosure, the meteorological feature vector includes pre-position embedding information. The second fusion unit includes a position information acquisition subunit and a position information embedding subunit.

[0138] The location information acquisition subunit is used to obtain the location information of the extended region to be determined in the text summary from the summary vector representation.

[0139] The location information embedding subunit is used to merge the location information of the extended area with the pre-position embedding information.

[0140] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0141] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the meteorological information determination method of the present disclosure.

[0142] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the meteorological information determination method of the present disclosure.

[0143] According to an embodiment of the present disclosure, a computer program product includes a computer program, wherein the computer program is stored on at least one of a readable storage medium and an electronic device, and when the computer program is executed by a processor, the meteorological information determination method of the present disclosure is implemented.

[0144] Figure 5 A block diagram of an electronic device that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0145] like Figure 5 As shown, the device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0146] Multiple components in the device 500 are connected to an input / output (I / O) interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0147] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the meteorological information determination method. For example, in some embodiments, the meteorological information determination method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the meteorological information determination method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the meteorological information determination method in any other appropriate manner (e.g., by means of firmware).

[0148] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0149] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0150] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0151] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0152] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0153] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0154] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0155] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining meteorological information, comprising: Extract features from the meteorological grid data of the target area within the target time period to obtain a meteorological feature vector; Inputting the to-be-processed meteorological data of the target region within the target time period into a large language model to obtain a text summary including a method for determining the meteorological information; Using the text summary to perform information enhancement processing on the meteorological feature vector to obtain an information enhancement result; as well as The information enhancement result is subjected to self-attention processing to obtain a meteorological information determination result output for the meteorological data to be processed.

2. The method according to claim 1, wherein: The meteorological data to be processed includes basic meteorological data and location information of the target area; the step of inputting the meteorological data to be processed of the target area within the target time period into a large language model to obtain a text summary including a method for determining meteorological information includes: generating target text information representing target data for determining derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target region; and The text summary is generated according to the context information of the basic meteorological data and the target text information.

3. The method according to claim 2, wherein: The basic meteorological data includes meteorological time series data; the generating of target text information representing the target for determining the derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target region includes: generating first text information according to the information of the target time period, wherein the first text information represents the meteorological changes within a preset time period after the target time period; and The target text information is determined according to the first text information.

4. The method according to claim 2, wherein: The generating, based on at least one of the basic meteorological data, the information of the target time period, and the location information of the target region, the target text information representing the target for determining the derived meteorological data comprises: generating second text information according to the location information of the target area, wherein the second text information represents the weather changes in an extended area adjacent to the target area; and The target text information is determined according to the second text information.

5. The method according to any one of claims 1 to 4, wherein: The feature extraction of the meteorological grid data of the target area within the target time period to obtain the meteorological feature vector includes: Marking the meteorological grid data into a plurality of sequence segments, wherein the plurality of sequence segments represent the same spatial resolution; Performing a linear transformation on the sequence fragments to obtain a sequence vector representation; Aggregating the sequence vector representation to obtain a sequence aggregation vector; fusing a plurality of sequence aggregation vectors corresponding to the plurality of sequence segments to obtain a grid aggregation vector; and The meteorological feature vector is determined according to the grid aggregation vector.

6. The method according to claim 5, wherein: The determining the meteorological feature vector according to the grid aggregation vector comprises: At least one of the leading time embedding information and the leading position embedding information is fused with the grid aggregation vector to obtain the meteorological feature vector.

7. The method according to any one of claims 1 to 6, wherein: The information enhancement processing of the meteorological feature vector using the text summary to obtain the information enhancement result includes: Performing a linear transformation on the text summary to obtain a summary vector representation, wherein the vector dimension of the summary vector representation is the same as the vector dimension of the meteorological feature vector; fusing the meteorological feature vector and the summary vector representation to obtain a fused feature vector; and The fused feature vector is determined as the information enhancement result.

8. The method according to claim 7, wherein: The meteorological feature vector includes leading time embedded information; and the fusing of the meteorological feature vector and the summary vector representation includes: Acquiring time information representing a preset time period to be determined in the text summary from the summary vector representation; and The time information is merged with the preceding time embedding information.

9. The method according to claim 7, wherein: The meteorological feature vector includes pre-position embedded information; and the fusing of the meteorological feature vector and the summary vector representation includes: Acquire, from the summary vector representation, location information representing the extended region to be determined in the text summary; and The location information of the extended area is merged with the pre-position embedding information.

10. A device for determining meteorological information, comprising: A feature extraction module is used to extract features from the meteorological grid data of the target area within the target time period to obtain a meteorological feature vector; A large language model module, used for inputting the to-be-processed meteorological data of the target region within the target time period into the large language model to obtain a text summary including a method for determining the meteorological information; An information enhancement module, used to perform information enhancement processing on the meteorological feature vector using the text summary to obtain an information enhancement result; as well as The self-attention processing module is used to perform self-attention processing on the information enhancement result to obtain a meteorological information determination result output for the meteorological data to be processed.

11. The device according to claim 10, wherein: The meteorological data to be processed includes basic meteorological data and location information of the target area; The large language model module includes: A target text generation unit, configured to generate target text information representing the target data used to determine the derived meteorological data according to at least one of the basic meteorological data, the information of the target time period, and the location information of the target region; as well as A text summary generating unit is used to generate the text summary according to the context information of the basic meteorological data and the target text information.

12. The device according to claim 11, wherein The basic meteorological data includes meteorological time series data; The target text generation unit comprises: A first text generation subunit, configured to generate first text information according to the information of the target time period, wherein the first text information represents the meteorological changes within a preset time period after the target time period; as well as The first target text determination subunit is used to determine the target text information according to the first text information.

13. The device according to claim 11, wherein: The target text generation unit comprises: A second text generation subunit is configured to generate second text information according to the location information of the target area, wherein the second text information represents the weather changes in an extended area adjacent to the target area; and The second target text determination subunit is used to determine the target text information according to the second text information.

14. The device according to any one of claims 10 to 13, wherein: The feature extraction module comprises: A marking unit, used for marking the meteorological grid data into a plurality of sequence segments, wherein the spatial resolution represented by the plurality of sequence segments is the same; A first linear transformation unit, used for performing a linear transformation on the sequence fragment to obtain a sequence vector representation; an aggregation unit, used for aggregating the sequence vector representation to obtain a sequence aggregation vector; A first fusion unit is used to fuse multiple sequence aggregation vectors corresponding to the multiple sequence segments to obtain a grid aggregation vector; and A meteorological feature vector determining unit is used to determine the meteorological feature vector according to the grid aggregation vector.

15. The device according to claim 14, wherein: The meteorological feature vector determination unit comprises: The fusion subunit is used to fuse at least one of the leading time embedding information and the leading position embedding information with the grid aggregation vector to obtain the meteorological feature vector.

16. The device according to any one of claims 10 to 15, wherein: The information enhancement module includes: A second linear transformation unit is used to perform a linear transformation on the text summary to obtain a summary vector representation, wherein the vector dimension of the summary vector representation is the same as the vector dimension of the meteorological feature vector; A second fusion unit is used to fuse the meteorological feature vector and the summary vector representation to obtain a fused feature vector; and The information enhancement result determining unit is used to determine the fused feature vector as the information enhancement result.

17. The device according to claim 16, wherein: The meteorological feature vector includes leading time embedding information; the second fusion unit includes: a time information acquisition subunit, configured to acquire time information representing a preset time period to be determined in the text summary from the summary vector representation; and The time information embedding subunit is used to merge the time information with the preceding time embedding information.

18. The device according to claim 16, wherein: The meteorological feature vector includes pre-position embedding information; the second fusion unit includes: a location information acquisition subunit, configured to acquire location information representing an extended region to be determined in the text summary from the summary vector representation; and The location information embedding subunit is used to merge the location information of the extended area with the pre-position embedding information.

19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

21. A computer program product, comprising a computer program, wherein the computer program is stored on at least one of a readable storage medium and an electronic device, and when the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images