Pollutant Assimilation Method, Device and Equipment Integrating 3D Variational and Deep Learning
By integrating three-dimensional variation and deep learning methods, the HATNet assimilation model is constructed, which solves the challenges of the computational complexity and execution time of existing assimilation methods, and achieves more efficient and accurate forecasts of pollutant concentrations.
Patent Information
- Application Number
- CN202510270497.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-07
- Publication Date
- 2025-06-24
- Estimated Expiration
- 2045-03-07
AI Technical Summary
The existing three-dimensional variational assimilation and EnKF assimilation methods have challenges in computational complexity and execution time, resulting in reduced accuracy of pollutant concentration forecasts and air quality predictions.
Using a pollutant assimilation method that integrates three-dimensional variation and deep learning, the HATNet assimilation model is constructed. This model combines the Hadamah attention module and the Transformer module to use the three-dimensional variation assimilation analysis field as a constraint for deep learning model training to generate a more accurate pollutant concentration analysis field.
It improves the accuracy and calculation efficiency of pollutant concentration forecasts, reduces the demand for computing resources, and can significantly improve the speed of assimilation calculation based on traditional methods.
Smart Images
Figure CN119808596B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of the integration of air pollutant data assimilation and artificial intelligence, and particularly to a pollutant assimilation method, device and equipment that integrate three-dimensional variational and deep learning. Background Art
[0002] Atmospheric chemical models are crucial for predicting and simulating air quality and can estimate the three-dimensional distribution of gas and aerosol concentrations. However, due to the uncertainties in pollutant emissions, initial conditions (ICs), boundary conditions (BCs), and chemical process parameterizations, accurate atmospheric chemical predictions remain a challenge. Among these factors, ICs provide key atmospheric and chemical variables and play an important role in determining the accuracy of atmospheric chemical forecasts. Data assimilation (DA) methods, such as variational assimilation and ensemble Kalman filter (EnKF) methods, are widely regarded as effective techniques for reducing IC uncertainty and improving the accuracy of atmospheric chemical forecasts.
[0003] Among them, the three-dimensional variational assimilation (3Dvar) method is one of the most widely applied variational assimilation methods. It provides a detailed and accurate initial state and improves the air pollution prediction ability. The EnKF method achieves assimilation by generating a set of multi-model predictions and providing uncertainty quantification, and it is suitable for solving non-linear systems. However, both the 3Dvar and EnKF assimilation methods require a large amount of computing resources and have a slow execution time. Moreover, with the increase in the dimensionality and sample size of state variables in recent years, the computational complexity of these methods is also increasing, resulting in longer calculation times for pollutant concentration forecasts and air quality predictions and reduced accuracy. Summary of the Invention
[0004] Based on this, it is necessary to provide a pollutant assimilation method, device and equipment that integrate three-dimensional variational and deep learning for the above technical problems.
[0005] A pollutant assimilation method that integrates three-dimensional variational and deep learning, the method comprising:
[0006] Step 1, collecting pollutant observation data and background fields in the study area;
[0007] Step 2, inputting the pollutant observation data and background fields into a three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants hourly;
[0008] Step 3: Construct a HATNet (Hadamard Attention Transformer Network) assimilation model with a Unet structure integrating the Hadamard attention module and the Transformer module, and input the pollutant observation data, background field, and three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained. Among them, the Hadamard attention module extracts the feature information of the model input data by highlighting key features through weighted summation, and the Transformer module further integrates and transforms the features by capturing long-range dependencies.
[0009] Step 4: Input the pollutant observation data and background field to be predicted into the trained HATNet assimilation model, predict and generate an hourly pollutant concentration analysis field, and apply the pollutant concentration analysis field to numerical weather prediction.
[0010] In one embodiment, collect the pollutant observation data and background field in the study area, including:
[0011] Collect the pollutant observation data at ground stations in the study area and perform quality control. Among them, the quality control includes: performing extreme value tests and consistency tests on the pollutant observation data, filling in missing values, removing outliers, and performing grid processing on the pollutant observation data.
[0012] Collect the hourly background field of three-dimensional pollutant concentrations provided by the operation results of the WRF-Chem (Weather Research and Forecasting - Chemistry) numerical model.
[0013] In one embodiment, input the pollutant observation data and background field into the three-dimensional variational assimilation system to generate an hourly three-dimensional variational assimilation analysis field of pollutants, including:
[0014] Construct a three-dimensional variational assimilation system based on the three-dimensional variational assimilation theory, and input the pollutant observation data and background field into the three-dimensional variational assimilation system. Among them, the pollutant observation data is used as a constraint, and the background field is used as a prior field. Construct a three-dimensional variational assimilation loss function and solve the objective functional to generate an hourly three-dimensional variational assimilation analysis field of pollutants.
[0015] In one embodiment, the HATNet assimilation model consists of a symmetric encoder and decoder, and the encoder and decoder are connected through a Transformer module.
[0016] Among them, the encoder contains a convolutional block and multiple Hadamard attention modules. The convolutional block is used to extract the initial pollutant concentration features in the input pollutant observation data and background field through convolutional operations, and the Hadamard attention modules are used to perform weighted processing on the initial pollutant concentration features output by the convolutional block at different levels.
[0017] The Transformer module is used to further integrate and transform the encoded features finally output by the encoder, and output the processed features to the decoder;
[0018] The decoder also includes a convolutional block and multiple Hadamard attention modules, which are used to gradually decode and restore the features output by the Transformer module to generate a pollutant concentration analysis field; among them, the number of Hadamard attention modules in the encoder and the decoder is the same and they correspond one by one, and between the encoder and the decoder, the outputs of each layer of the encoder and the masks generated by each layer of the decoder are fused through the DSB (Double-Scale Bridge) module to achieve multi-scale feature extraction.
[0019] In one embodiment, pollutant observation data, background field, and three-dimensional variational assimilation analysis field are input into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained, including:
[0020] After interpolating the pollutant observation data into a grid form that is consistent with the research area and resolution of the background field, it is input into the HATNet assimilation model together with the background field and the three-dimensional variational assimilation analysis field, and the three-dimensional variational assimilation analysis field is used as a label, and the HATNet assimilation model is trained in a supervised learning manner until a trained HATNet assimilation model is obtained.
[0021] In one embodiment, the loss function during the training of the HATNet assimilation model is defined as
[0022] ;
[0023] Among them, represents the three-dimensional variational assimilation analysis field, represents the prediction result of the HATNet assimilation model, and MSE represents the mean square error.
[0024] A pollutant assimilation device that fuses three-dimensional variational and deep learning, the device includes:
[0025] A data acquisition module, which is used to acquire pollutant observation data and background field in the research area;
[0026] A variational assimilation module, which is used to input the pollutant observation data and background field into a three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants hourly;
[0027] A deep learning module for constructing a HATNet assimilation model with a Unet structure integrating a Hadamard attention module and a Transformer module, and inputting pollutant observation data, a background field, and a three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained. Among them, the Hadamard attention module extracts the feature information of the model input data by weighting and highlighting key features, and the Transformer module further integrates and transforms the features by capturing long-range dependencies.
[0028] A prediction output module for inputting the pollutant observation data and background field to be predicted into the trained HATNet assimilation model, predicting and generating an hourly pollutant concentration analysis field, and applying the pollutant concentration analysis field to numerical weather prediction.
[0029] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0030] Step 1, collecting pollutant observation data and a background field in the research area;
[0031] Step 2, inputting the pollutant observation data and the background field into a three-dimensional variational assimilation system to generate an hourly three-dimensional variational assimilation analysis field of pollutants;
[0032] Step 3, constructing a HATNet assimilation model with a Unet structure integrating a Hadamard attention module and a Transformer module, and inputting the pollutant observation data, the background field, and the three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained. Among them, the Hadamard attention module extracts the feature information of the model input data by weighting and highlighting key features, and the Transformer module further integrates and transforms the features by capturing long-range dependencies.
[0033] Step 4, inputting the pollutant observation data and the background field to be predicted into the trained HATNet assimilation model, predicting and generating an hourly pollutant concentration analysis field, and applying the pollutant concentration analysis field to numerical weather prediction.
[0034] The above-mentioned pollutant assimilation method, device, and equipment integrating three-dimensional variational and deep learning have the following beneficial effects compared with the prior art:
[0035] 1. By using the three-dimensional variational assimilation analysis field output by the three-dimensional variational assimilation system as a constraint for deep learning model training, the integration of three-dimensional variational assimilation and deep learning is achieved. Thus, the ability of three-dimensional variational assimilation to fuse and calibrate pollutant observation data and background fields is effectively utilized, providing a more accurate and physically consistent data basis for the deep learning model and improving the accuracy and reliability of the model output results.
[0036] 2. Construct a HATNet assimilation model with a Unet structure integrating Hadamard attention module and Transformer module. Among them, based on the Hadamard attention module and Transformer module, the learning ability of the spatial distribution characteristics of pollutant concentration can be enhanced, and based on the symmetric encoder and decoder in the Unet structure, spatial information can be captured more effectively, thereby improving the accuracy of pollutant concentration prediction.
[0037] 3. The trained HATNet assimilation model is an end-to-end assimilation model that can directly assimilate the pollutant concentration analysis field based on the input pollutant observation data and background field data, avoiding the design of large matrices such as background error and covariance and complex solution calculations, and there is no iterative solution process in three-dimensional variational assimilation. It has low demand for computing resources and can greatly improve the speed of three-dimensional data assimilation. Description of the Drawings
[0038] Figure 1 It is a schematic flow chart of a pollutant assimilation method integrating three-dimensional variational and deep learning in an embodiment;
[0039] Figure 2 It is a schematic logical architecture diagram of a pollutant assimilation method integrating three-dimensional variational and deep learning in an embodiment;
[0040] Figure 3 It is a schematic structural diagram of a HATNet assimilation model in an embodiment;
[0041] Figure 4 It is a spatial distribution diagram of the concentration of a certain pollutant at 18:00 on November 19, 2023, taking SO2 as an example; among them, Figure 4 (a) is the concentration distribution of the SO2 analysis field after three-dimensional variational assimilation, Figure 4 (b) is the concentration distribution of the SO2 analysis field after assimilation by the HATNet assimilation model;
[0042] Figure 5 It is a schematic diagram of the vertical direction increment distribution of SO2 concentration at 18:00 on November 19, 2023 in an embodiment. Among them, Figure 5 (a) is the vertical direction increment distribution of SO2 concentration after three-dimensional variational assimilation, Figure 5(b) Vertical increment distribution of SO2 concentration after assimilation by the HATNet assimilation model;
[0043] Figure 6 For the SO2 data in the study area at 18:00 on November 19, 2023 in an embodiment, scatter plots of the analysis field obtained by assimilating the SO2 concentration using three-dimensional variational assimilation and the method proposed in this application respectively and the observed concentration at the ground station are shown;
[0044] Figure 7 It is the internal structure diagram of a computer device in an embodiment. Specific implementation manners
[0045] In order to make the objectives, technical solutions and advantages of this application clearer, the following further elaborates on this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.
[0046] In an embodiment, as Figure 1 and Figure 2 shown, a pollutant assimilation method integrating three-dimensional variational and deep learning is provided, including the following steps:
[0047] Step 1, collect pollutant observation data and background fields in the study area.
[0048] Specifically, Step 1 includes: collecting pollutant observation data at ground stations in the study area and performing quality control; among them, the quality control includes: performing extreme value test and consistency test on the pollutant observation data, filling in the missing data values with the average values before and after the collection time, removing the data with too large or too small pollutant concentrations, and performing grid processing on the pollutant observation data. Collect the background fields of hourly three-dimensional pollutant concentrations provided by the operation results of the WRF-Chem numerical model.
[0049] Step 2, input the pollutant observation data and background fields into the three-dimensional variational assimilation system to generate an hourly three-dimensional variational assimilation analysis field of the pollutants.
[0050] Specifically, Step 2 includes: constructing a three-dimensional variational assimilation system based on the three-dimensional variational assimilation theory, inputting the pollutant observation data and background fields into the three-dimensional variational assimilation system, where the pollutant observation data is used as a constraint and the background field is used as a prior field, constructing a three-dimensional variational assimilation loss function and solving the objective functional to generate an hourly three-dimensional variational assimilation analysis field of the pollutants.
[0051] Step 3: Construct the HATNet assimilation model with a Unet structure integrating the Hadamard attention module and the Transformer module, and input the pollutant observation data, the background field, and the three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained. Among them, the Hadamard attention module extracts the feature information of the model input data by highlighting key features through weighted summation, and the Transformer module further integrates and transforms the features by capturing long-range dependencies.
[0052] The structure of the HATNet assimilation model is as Figure 3 shown. The HATNet assimilation model consists of a symmetric encoder and decoder, and the encoder and decoder are connected by a Transformer module.
[0053] Among them, the encoder contains a convolutional block and multiple Hadamard attention modules. The convolutional block is used to extract the preliminary features of the pollutant concentration in the input pollutant observation data and the background field through convolutional operations, and the Hadamard attention modules are used to perform weighted processing on the preliminary features of the pollutant concentration output by the convolutional block at different levels.
[0054] The Transformer module is used to further integrate and transform the encoded features finally output by the encoder, and output the processed features to the decoder. This module can further enhance the ability of the assimilation model to capture and express important feature information.
[0055] The decoder also contains a convolutional block and multiple Hadamard attention modules, which are used to gradually decode and restore the features output by the Transformer module to generate the pollutant concentration analysis field. Among them, the number of Hadamard attention modules in the encoder and decoder is the same and corresponds one by one, and the encoder and decoder are fused by the DSB module to fuse the outputs of each layer of the encoder and the masks generated by each layer of the decoder to achieve multi-scale feature extraction.
[0056] Specifically, the training of the HATNet assimilation model in Step 3 includes: interpolating the pollutant observation data into a grid form that is consistent with the research area and resolution of the background field, and then inputting it together with the background field and the three-dimensional variational assimilation analysis field into the HATNet assimilation model. Taking the three-dimensional variational assimilation analysis field as the label, the HATNet assimilation model is trained in a supervised learning manner until a trained HATNet assimilation model is obtained. The loss function during the training of the HATNet assimilation model is defined as
[0057] ;
[0058] Among them, represents the three-dimensional variational assimilation analysis field. represents the prediction result of the HATNet assimilation model, and MSE represents the mean square error.
[0059] Step 4: Input the pollutant observation data and background field to be predicted into the trained HATNet assimilation model, predict and generate hourly pollutant concentration analysis fields, and apply the pollutant concentration analysis fields to numerical forecasting.
[0060] Through the above four steps, this method can integrate three-dimensional variational assimilation and deep learning, and use pollutant observation data, background field and three-dimensional variational assimilation analysis field to train a HATNet assimilation model. The HATNet assimilation model can directly assimilate the pollutant concentration analysis field based on the input pollutant observation data and background field data, avoiding the complex calculation process of three-dimensional variational assimilation, improving the efficiency of pollutant assimilation calculation, and based on the Hadamard attention module and Transformer module in the model, it can more effectively capture the spatial distribution characteristics of pollutant concentration and improve the accuracy of pollutant concentration prediction. And experiments have confirmed that the HATNet assimilation model constructed in this application can assimilate the output pollutant concentration analysis field at a speed 34 times that of the three-dimensional variational assimilation system.
[0061] In a specific embodiment, taking the optimization of SO2 concentration at 18:00 on November 19, 2023 as an example, the longitude range is from 76.72° east longitude to 151.28° east longitude, the latitude range is from 11.78° north latitude to 57.16° north latitude, the resolution is 12km, and the study area includes most of China. Specifically, the following steps are included:
[0062] S1: Collect SO2 observation data from 1641 ground stations at 18:00 on November 19, 2023, perform quality control on the observation data, fill in missing values and remove outliers. Collect the forecast field of WRF-Chem model at 18:00 on November 19, 2023 as the SO2 concentration background field.
[0063] S2: Based on the 3Dvar theory, a 3Dvar assimilation system is constructed. The SO2 observation data and the SO2 concentration background field are input into the three-dimensional variational assimilation system. The SO2 observation data are used as constraints and the SO2 concentration background field is used as a priori field. After constructing the three-dimensional variational assimilation loss function and solving the target functional, the three-dimensional variational assimilation field of SO2 concentration hour by hour is obtained.
[0064] S3: Construct the HATNet assimilation model, interpolate the SO2 observation data into a grid form consistent with the corresponding background field research area and resolution, then use the SO2 observation data, the three-dimensional SO2 concentration background field and the three-dimensional variational assimilation field output in step S2 as training data, and use supervised learning to train the HATNet assimilation model. The loss function in the HATNet assimilation model is defined as follows:
[0065] ;
[0066] After the training is completed, a trained HATNet assimilation model is obtained;
[0067] S4: Input the SO2 observation data and background field to be predicted into the trained HATNet assimilation model to generate an hourly SO2 concentration analysis field, and the optimized result of the SO2 concentration analysis field can be used for subsequent numerical weather prediction.
[0068] Furthermore, the SO2 data in the study area at 18:00 on November 19, 2023 was assimilated using the method proposed in this application, and the three-dimensional variational assimilation results were compared and verified. The comparison results are as Figures 4 - 6 shown. As Figures 4 - 6 can be seen, compared with the three-dimensional variational assimilation, the assimilation analysis result of the method proposed in this application for the SO2 concentration is better. Based on this method, the accuracy of the initial field of pollutant concentration prediction can be effectively improved, which has important scientific significance and popularization and application value for improving the prediction accuracy of pollutants; it is an effective supplement to traditional meteorological data assimilation, avoiding large matrices such as background error and covariance and complex solution calculations, and also avoiding additional errors introduced by matrix design. It can directly perform end-to-end data assimilation, which is simple and easy to implement, with a small amount of calculation and less financial investment, and is of great significance for meteorological data assimilation and pollutant prediction.
[0069] In one embodiment, a pollutant assimilation device integrating three-dimensional variational and deep learning is provided, including:
[0070] A data acquisition module for collecting pollutant observation data and background fields in the study area;
[0071] A variational assimilation module for inputting pollutant observation data and background fields into a three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants hourly;
[0072] A deep learning module for constructing a HATNet assimilation model with a Unet structure integrating a Hadamard attention module and a Transformer module, and inputting pollutant observation data, background fields, and three-dimensional variational assimilation analysis fields into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained; among them, the Hadamard attention module extracts the feature information of the model input data by highlighting key features through weighted summation, and the Transformer module further integrates and transforms the features by capturing long-range dependencies;
[0073] A prediction output module is used to input the pollutant observation data to be predicted and the background field into the trained HATNet assimilation model, predict and generate an hourly pollutant concentration analysis field, and apply the pollutant concentration analysis field to numerical weather prediction.
[0074] For the specific limitations of the pollutant assimilation device integrating 3D variational and deep learning, reference can be made to the limitations of the pollutant assimilation method integrating 3D variational and deep learning in the above text, which will not be elaborated here. Each module in the above pollutant assimilation device integrating 3D variational and deep learning can be implemented in whole or in part by software, hardware, and their combinations. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.
[0075] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a pollutant assimilation method integrating 3D variational and deep learning. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.
[0076] Those skilled in the art can understand that Figure 7 the structure shown in
[0077] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.
[0078] Step 1, collect pollutant observation data and background fields in the study area;
[0079] Step 2: Input the pollutant observation data and the background field into the three-dimensional variational assimilation system to generate the hourly three-dimensional variational assimilation analysis field of pollutants;
[0080] Step 3: Construct a HATNet assimilation model with a Unet structure integrating a Hadamard attention module and a Transformer module, and input the pollutant observation data, the background field, and the three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained; among them, the Hadamard attention module extracts the feature information of the model input data by highlighting key features through weighted summation, and the Transformer module further integrates and transforms the features by capturing long-range dependencies;
[0081] Step 4: Input the pollutant observation data and the background field to be predicted into the trained HATNet assimilation model, predict and generate the hourly pollutant concentration analysis field, and apply the pollutant concentration analysis field to numerical weather prediction.
[0082] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.
[0083] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims
1. A pollutant assimilation method integrating three-dimensional variational and deep learning, characterized in that: The method comprises: Step 1: Collect pollutant observation data and background fields in the study area; Step 2, inputting the pollutant observation data and background field into a three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants on an hourly basis; Step 3, constructing a HATNet assimilation model of a Unet structure integrating a Hadamard attention module and a Transformer module, and inputting the pollutant observation data, background field, and three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained; wherein the Hadamard attention module extracts feature information of the model input data by weighting and highlighting key features, and the Transformer module further integrates and transforms features by capturing long-distance dependencies; Step 4, inputting the pollutant observation data and background field to be predicted into the trained HATNet assimilation model, predicting and generating hourly pollutant concentration analysis fields, and applying the pollutant concentration analysis fields to numerical forecasting; The pollutant observation data, background field and three-dimensional variational assimilation analysis field are input into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained, including: After the pollutant observation data is interpolated into a grid form that is consistent with the study area and resolution of the background field, it is input into the HATNet assimilation model together with the background field and the three-dimensional variational assimilation analysis field. The three-dimensional variational assimilation analysis field is used as a label, and the HATNet assimilation model is trained using supervised learning until a trained HATNet assimilation model is obtained.
2. The method according to claim 1, characterized in that Collect pollutant observation data and background fields in the study area, including: Collect pollutant observation data from ground stations in the study area and perform quality control; wherein the quality control includes: performing extreme value test and consistency test on pollutant observation data, filling missing values, removing outliers, and gridding pollutant observation data; The WRF-Chem numerical model running results provide a background field of hourly three-dimensional pollutant concentrations.
3. The method according to claim 2, characterized in that The pollutant observation data and background field are input into the three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants every hour, including: A three-dimensional variational assimilation system is constructed based on the three-dimensional variational assimilation theory, and the pollutant observation data and the background field are input into the three-dimensional variational assimilation system, wherein the pollutant observation data are used as constraints and the background field is used as a priori field. A three-dimensional variational assimilation loss function is constructed and the target functional is solved to generate a three-dimensional variational assimilation analysis field of pollutants hour by hour.
4. The method according to claim 1, characterized in that The HATNet assimilation model consists of a symmetrical encoder and decoder, and the encoder and decoder are connected through a Transformer module; The encoder includes a convolution block and multiple Hadamard attention modules, wherein the convolution block is used to extract the input pollutant observation data and the preliminary features of the pollutant concentration in the background field through convolution operation, and the Hadamard attention module is used to perform weighted processing on the preliminary features of the pollutant concentration output by the convolution block at different levels; The Transformer module is used to further integrate and transform the encoding features finally output by the encoder, and output the processed features to the decoder; The decoder also includes a convolution block and multiple Hadamard attention modules, which are used to gradually decode and restore the features output by the Transformer module to generate a pollutant concentration analysis field; wherein, the number of Hadamard attention modules in the encoder and decoder is consistent and one-to-one corresponding, and the output of each layer of the encoder is fused with the mask generated by each layer of the decoder through the DSB module between the encoder and the decoder to achieve multi-scale feature extraction.
5. The method according to claim 1, characterized in that The loss function of the HATNet assimilation model training is defined as ； in, represents the three-dimensional variational assimilation analysis field, represents the prediction result of the HATNet assimilation model, and MSE represents the mean square error.
6. A pollutant assimilation device integrating three-dimensional variation and deep learning, characterized in that: The device comprises: Data acquisition module, used to collect pollutant observation data and background fields in the study area; A variational assimilation module, used for inputting the pollutant observation data and background field into a three-dimensional variational assimilation system to generate a three-dimensional variational assimilation analysis field of pollutants on an hourly basis; A deep learning module, used to construct a HATNet assimilation model of a Unet structure integrating a Hadamard attention module and a Transformer module, and input the pollutant observation data, background field, and three-dimensional variational assimilation analysis field into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained; wherein the Hadamard attention module extracts feature information of the model input data by weighting and highlighting key features, and the Transformer module further integrates and transforms features by capturing long-distance dependencies; A prediction output module, used to input the pollutant observation data and background field to be predicted into the trained HATNet assimilation model, predict and generate hourly pollutant concentration analysis fields, and apply the pollutant concentration analysis fields to numerical forecasting; The pollutant observation data, background field and three-dimensional variational assimilation analysis field are input into the HATNet assimilation model for training until a trained HATNet assimilation model is obtained, including: After the pollutant observation data is interpolated into a grid form that is consistent with the study area and resolution of the background field, it is input into the HATNet assimilation model together with the background field and the three-dimensional variational assimilation analysis field. The three-dimensional variational assimilation analysis field is used as a label, and the HATNet assimilation model is trained using supervised learning until a trained HATNet assimilation model is obtained.
7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.