A method, system, computer device and storage medium for predicting the middle and upper atmosphere

Abstract

The application relates to the technical field of weather prediction, in particular to a middle and high layer atmosphere prediction method and system, computer equipment and a storage medium, the method comprises the following steps: acquiring observation data of a multi-source space environment, calling a corresponding prediction model through a preset model scheduling strategy; based on the category of the prediction model, matching corresponding target input data in the observation data, inputting the target input data into the prediction model to obtain a prediction result; according to the prediction result, constructing a visual middle and high layer atmosphere prediction product view, and presenting the visual middle and high layer atmosphere prediction product view through a display interface of the middle and high layer atmosphere prediction platform. The application has the effects of improving the accuracy of middle and high layer atmosphere prediction and the display effect of a prediction result.

Description

Technical Field

[0001] This application relates to the field of weather forecasting technology, and in particular to a method, system, computer equipment, and storage medium for forecasting the middle and upper atmosphere. Background Technology

[0002] Currently, mid-to-upper atmosphere forecasting is a technology that forecasts various environmental conditions originating from solar activity and affecting near-Earth space and Earth's technological systems. Its main objectives are to provide early warning and quantify phenomena such as geomagnetic storms, ionospheric disturbances, and changes in upper atmospheric density, so as to ensure the safe operation of satellites in orbit, the accuracy and reliability of global navigation and positioning services, and the stability of transoceanic communication and power grid systems.

[0003] Existing mid- and upper-level atmospheric forecasting technologies still have certain limitations when providing panoramic, multi-layered forecasts of complex space environments. Most forecasting methods focus on predicting single physical phenomena or specific atmospheric regions; for example, they may only provide forecasts of ionospheric disturbances or only focus on density changes in the thermosphere. These point- or linear forecasting models lack an effective description of the interactions and energy coupling processes between different atmospheric layers, such as the mesosphere and thermosphere, and the thermosphere and ionosphere. Therefore, they struggle to fully represent the entire physical chain of mid- and upper-level atmospheric events from their source to Earth's response. The forecast products provided are often independent, requiring users to integrate and interpret fragmented information from different systems regarding different physical phenomena, resulting in an incomplete understanding of the space environment situation. Therefore, there is room for improvement. Summary of the Invention

[0004] To improve the accuracy of mid- and upper-level atmospheric forecasting and the presentation of forecast results, this application provides a mid- and upper-level atmospheric forecasting method, system, computer equipment, and storage medium.

[0005] The above-mentioned objective of this application is achieved through the following technical solution:

[0006] A method for predicting the middle and upper atmosphere, applied to a middle and upper atmosphere prediction platform, the method comprising:

[0007] Acquire observational data of the multi-source space environment, and call the corresponding prediction model through a preset model scheduling strategy;

[0008] Based on the category of the prediction model, the corresponding target input data is matched in the observation data, and the target input data is input into the prediction model to obtain the prediction result;

[0009] Based on the prediction results, a visualized view of the middle and upper atmosphere prediction product is constructed and presented through the display interface of the middle and upper atmosphere prediction platform.

[0010] By adopting the above technical solutions, and by acquiring observational data of the multi-source space environment and calling various prediction models according to strategies, data from different sources can be systematically integrated within a unified framework. The most suitable analysis tools can be matched for different prediction tasks, thus overcoming the limitations of traditional methods, such as single data sources and poor model versatility. This lays the foundation for comprehensive and accurate forecasting. By matching the model category with corresponding target input data, it ensures that each called specialized model can obtain the corresponding input data, thereby guaranteeing the effectiveness and accuracy of model calculations and avoiding prediction failures or accuracy reductions due to mismatched or incomplete input data. By constructing and presenting a unified visualized view of the middle and upper atmosphere prediction products, complex numerical results with different physical meanings from multiple models can be integrated into an intuitive and easy-to-understand comprehensive situation map. This greatly reduces the threshold for users to interpret and analyze information, significantly improving the efficiency of middle and upper atmosphere situational awareness and the scientific nature of decision-making.

[0011] In a preferred embodiment, this application can be further configured as follows: based on the category of the prediction model, matching the corresponding target input data in the observation data, and inputting the target input data into the prediction model to obtain a prediction result, specifically including:

[0012] When the prediction model is classified as an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the resulting prediction result is the ionospheric storm prediction result.

[0013] When the prediction model is classified as a thermosphere dynamics simulation model, the corresponding target input data is matched as the third data, and the resulting prediction result is the thermosphere state prediction result.

[0014] When the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the prediction result obtained is the irregular body feature recognition prediction result.

[0015] When the prediction model is classified as a mid-to-upper atmosphere density and temperature model, the corresponding target input data is matched as the fourth data, and the resulting prediction result is the atmospheric density and temperature prediction result.

[0016] By adopting the above technical solutions, and by specifically dividing the prediction tasks into ionospheric storm prediction, irregular body feature identification and forecasting, thermospheric state simulation, and middle and upper atmospheric density and temperature modeling, it is possible to provide targeted and professional processing for phenomena at different levels and scales in the middle and upper atmosphere. This results in higher accuracy and reliability in the prediction of macroscopic overall disturbances, local fine risks, physical background fields, and near-space environments, and enables more refined domain-based and layered management and prediction of complex space environments.

[0017] In a preferred embodiment, this application can be further configured such that: when the prediction model is classified as an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the obtained prediction result is the ionospheric storm prediction result, specifically including:

[0018] The first data includes at least the total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index;

[0019] The total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index are input into the ionospheric storm prediction model that has been trained using the LSTM-CNN hybrid model, so as to perform time series prediction.

[0020] The ionospheric storm prediction results are obtained from the ionospheric storm prediction model.

[0021] By adopting the above technical solution and using the LSTM-CNN hybrid model to perform time series prediction on spatiotemporal data including TEC, AE, and F10.7, it is possible to simultaneously utilize the CNN network to extract spatial morphological features from the TEC map and the LSTM network to capture the evolutionary patterns and long-range dependencies of each physical quantity over time. This allows for a deeper understanding of the complete physical process of ionospheric storms compared to a single network structure, significantly improving the prediction accuracy for the entire process of storm occurrence, development, and dissipation.

[0022] In a preferred embodiment, this application can be further configured such that: when the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the obtained prediction result is an irregular body feature recognition prediction result, specifically including:

[0023] The second data includes at least GNSS observation data, optical observation data, ionospheric detection parameters, and radar detection data;

[0024] Based on the pre-acquired edge extraction algorithm, phase coherence analysis algorithm, and machine learning algorithm, the irregular body feature identification and prediction model is constructed. The GNSS observation data, the optical observation data, the ionospheric detection parameters, and the radar detection data are input into the irregular body feature identification and prediction model to identify irregular bodies and make short, medium, and long-term predictions.

[0025] The irregular body feature recognition and prediction results are obtained from the irregular body feature recognition and prediction model. The irregular body feature recognition and prediction results include at least the irregular body feature recognition results and the irregular body prediction results.

[0026] By adopting the above technical solution, and using a prediction model constructed by combining edge extraction algorithm, phase coherence analysis algorithm and machine learning algorithm, irregularities can be identified in multi-source heterogeneous diagnostic data including GNSS, optical, and radar. This model can quickly extract complex nonlinear patterns related to ionospheric irregularities from massive, high-dimensional features with extremely high computational efficiency. This enables rapid and accurate probabilistic prediction of local and short-term signal flicker risks, providing tactical-level avoidance guidance for applications with stringent signal quality requirements such as UAVs and high-precision positioning.

[0027] In a preferred embodiment, this application can be further configured such that: when the prediction model is classified as a thermosphere dynamics simulation model, the corresponding target input data is matched as third data, and the resulting prediction result is a thermosphere state prediction result, specifically including:

[0028] The third data includes at least solar activity parameters, geomagnetic disturbance parameters, and time parameters;

[0029] The solar activity parameters, geomagnetic disturbance parameters, and time parameters are input into the thermosphere dynamics simulation model based on TIEGCM, so as to predict the thermosphere electric field and potential through the thermosphere dynamics simulation model.

[0030] The predicted state of the thermal layer is obtained from the thermal layer dynamics simulation model.

[0031] By adopting the above technical solution, using a TIEGCM-based physical model, and inputting external driving parameters such as solar activity and geomagnetic disturbances to predict the electric field and potential of the thermosphere, it is possible to simulate the global electrodynamic process driven by external energy by solving the first-principles physical equations, while ensuring the self-consistency of the physical mechanism. This provides a key upstream physical driving field for understanding and predicting the transport and distribution of ionospheric plasma, and improves the physical completeness of the entire prediction system.

[0032] In a preferred embodiment, this application can be further configured such that: when the prediction model is classified as a mid-to-upper atmosphere density-temperature model, the corresponding target input data is matched as the fourth data, and the obtained prediction result is the atmospheric density-temperature prediction result, specifically including:

[0033] The fourth set of data includes at least solar activity parameters, geomagnetic activity index, occultation atmospheric temperature, and occultation atmospheric density;

[0034] The solar activity parameters, geomagnetic activity index, occultation atmospheric temperature and occultation atmospheric density are input into the trained Light-GBM middle and upper atmospheric density and temperature model to predict the near-space environment.

[0035] The atmospheric density and temperature prediction results are obtained from the aforementioned middle and upper atmospheric density and temperature model.

[0036] By adopting the above technical solution, and by using a density and temperature model of the middle and upper atmosphere, and integrating background driving parameters such as the sun and geomagnetism with real-time observation data such as the temperature and density of occultation atmospheres, the global stability of empirical models or background models can be combined with the accuracy of real-time high-precision observation data to achieve complementary advantages. This generates a more reliable and accurate near-space environment forecast than a single type of model, providing more valuable decision support for applications such as hypersonic vehicles.

[0037] In a preferred embodiment, this application can be further configured such that: constructing a visualized upper-middle atmospheric prediction product view based on the prediction results specifically includes:

[0038] The prediction results of the ionospheric storm, the prediction results of the irregular body disturbance, and the prediction results of the thermosphere state are divided and organized according to spatial regions and time scales to obtain the organized prediction layer data.

[0039] The predicted layer data is mapped to a preset product layer, which includes an ionospheric layer, an irregularity disturbance layer, a thermosphere layer, and a middle and upper atmosphere layer.

[0040] Based on preset layer display rules, the product layers are composited to generate the visualized mid-to-high-altitude atmospheric prediction product view.

[0041] By adopting the above technical solution, and organizing the prediction results into partitions according to spatiotemporal scales, massive continuous data fields can be preprocessed into discrete data units that are easy to index and retrieve quickly. This greatly improves the system's data query performance when responding to user interaction operations, laying the foundation for a smooth visualization experience. By mapping the data to preset product layers and performing layer synthesis processing based on preset layer display rules, a visualized upper-middle atmospheric prediction product view is obtained. This transforms the system from a static data display tool into a powerful platform that can support proactive and exploratory data analysis.

[0042] The second objective of this invention is achieved through the following technical solution:

[0043] A middle and upper atmosphere prediction system, the middle and upper atmosphere prediction system comprising:

[0044] The data acquisition and management module is used to acquire observation data of the multi-source space environment and call the corresponding prediction model through a preset model scheduling strategy.

[0045] The prediction result generation module is used to match the corresponding target input data in the observation data based on the category of the prediction model, input the target input data into the prediction model, and obtain the prediction result;

[0046] The visualization and presentation module is used to construct a visualized view of the upper and middle atmosphere forecast product based on the forecast results, and to present the visualized view of the upper and middle atmosphere forecast product through the display interface of the upper and middle atmosphere forecast platform.

[0047] By adopting the above technical solutions, and by acquiring observational data of the multi-source space environment and calling various prediction models according to strategies, data from different sources can be systematically integrated within a unified framework. The most suitable analysis tools can be matched for different prediction tasks, thus overcoming the limitations of traditional methods, such as single data sources and poor model versatility. This lays the foundation for comprehensive and accurate forecasting. By matching the model category with corresponding target input data, it ensures that each called specialized model can obtain the corresponding input data, thereby guaranteeing the effectiveness and accuracy of model calculations and avoiding prediction failures or accuracy reductions due to mismatched or incomplete input data. By constructing and presenting a unified visualized view of the middle and upper atmosphere prediction products, complex numerical results with different physical meanings from multiple models can be integrated into an intuitive and easy-to-understand comprehensive situation map. This greatly reduces the threshold for users to interpret and analyze information, significantly improving the efficiency of middle and upper atmosphere situational awareness and the scientific nature of decision-making.

[0048] The above-mentioned objective three of this application is achieved through the following technical solution:

[0049] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for predicting the middle and upper atmosphere.

[0050] The fourth objective of this application is achieved through the following technical solution:

[0051] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for predicting the middle and upper atmosphere.

[0052] In summary, this application includes at least one of the following beneficial technical effects:

[0053] 1. By acquiring observational data of the multi-source space environment and calling various prediction models according to strategies, it is possible to systematically integrate data from different sources within a unified framework and match the most suitable analysis tools for different prediction tasks. This overcomes the limitations of traditional methods, such as single data sources and poor model versatility, laying the foundation for comprehensive and accurate forecasts. By matching the model category and inputting the corresponding target input data, it is possible to ensure that each called specialized model can obtain the corresponding input data, thereby guaranteeing the effectiveness and accuracy of model calculations and avoiding prediction failures or accuracy reductions due to mismatched or incomplete input data. By constructing and presenting a unified visualized view of the middle and upper atmosphere prediction products, it is possible to integrate complex numerical results with different physical meanings from multiple models into an intuitive and easy-to-understand comprehensive situation map, thereby greatly reducing the threshold for users to interpret and analyze information and significantly improving the efficiency of middle and upper atmosphere situation awareness and the scientific nature of decision-making.

[0054] 2. By partitioning and organizing the prediction results according to spatiotemporal scales, massive continuous data fields can be preprocessed into discrete data units that are easy to index and retrieve quickly. This greatly improves the system's data query performance when responding to user interaction operations, laying the foundation for a smooth visualization experience. By mapping the data to preset product layers and performing layer synthesis processing based on preset layer display rules, a visualized upper and middle atmospheric prediction product view is obtained. This transforms the system from a static data display tool into a powerful platform that can support proactive and exploratory data analysis. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the implementation of a method for predicting the middle and upper atmosphere in one embodiment of this application.

[0056] Figure 2This is a flowchart illustrating the implementation of step S20 in a middle and upper atmosphere prediction method according to an embodiment of this application.

[0057] Figure 3 This is a flowchart illustrating the implementation of step S21 in a method for predicting the middle and upper atmosphere according to an embodiment of this application.

[0058] Figure 4 This is a flowchart illustrating the implementation of step S22 in a middle and upper atmosphere prediction method according to an embodiment of this application.

[0059] Figure 5 This is a flowchart illustrating the implementation of step S23 in a middle and upper atmosphere prediction method according to an embodiment of this application.

[0060] Figure 6 This is a flowchart illustrating the implementation of step S24 in a middle and upper atmosphere prediction method according to an embodiment of this application.

[0061] Figure 7 This is a flowchart illustrating the implementation of step S30 in a middle and upper atmosphere prediction method according to an embodiment of this application.

[0062] Figure 8 This is a principle block diagram of a middle and upper atmosphere prediction system according to an embodiment of this application;

[0063] Figure 9 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application. Detailed Implementation

[0064] The following embodiments will help those skilled in the art to further understand the function of this application, but do not limit this application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application. These all fall within the protection scope of this application.

[0065] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0066] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0067] The present application will be further described in detail below with reference to the accompanying drawings.

[0068] In one embodiment, such as Figure 1 As shown, this application discloses a method for predicting the middle and upper atmosphere, applied to a middle and upper atmosphere prediction platform, specifically including the following steps:

[0069] S10: Acquire observation data of the multi-source space environment and call the corresponding prediction model through a preset model scheduling strategy.

[0070] Specifically, acquiring observational data of the multi-source space environment refers to retrieving timely and representative data from multiple space environment detection platforms or data centers. This includes, but is not limited to, ionospheric TEC data acquired through ground GNSS station networks, ionospheric state data collected by electron density monitors on satellite orbits, geomagnetic disturbance parameters such as Kp, Ap, and Sym-H indices collected in real time by geomagnetic arrays, solar activity data such as F10.7 solar radio flux, sunspot number, solar proton event reports, and EUV / FUV flux obtained through public services such as NOAA, NASA, or SEPC, and atmospheric physical quantities such as atmospheric temperature, density, wind speed, water vapor, ozone, and OH measured by satellites such as TIMED, COSMIC, and FY. In practice, these observational data are first jointly normalized using timestamps and spatial coordinates, and then classified according to the physical type, measurement spatial range, and time window of the data so that subsequent models can schedule and call the required subset of data as needed to carry out mid-to-upper atmosphere prediction and analysis.

[0071] S20: Based on the category of the prediction model, match the corresponding target input data in the observation data, input the target input data into the prediction model, and obtain the prediction result.

[0072] Specifically, the model scheduling strategy-based invocation method refers to obtaining the input data required by the existing prediction model based on the category of the existing prediction model, guided by a set of data-model mapping relationship tables or scheduling rules built into the platform. Assuming the model scheduling strategy simultaneously triggers the invocation of the ionospheric storm prediction model and the irregular body feature identification prediction model, then for the ionospheric storm prediction model, based on its category characteristics, it will accurately match and extract multivariate time-series data such as the F10.7 index, Kp index, and global TEC grid from massive amounts of observation data—that is, the target input data—and construct them into a time-series prediction model. The input data is a time-dimensional tensor required for the model. However, for the irregular body feature recognition and prediction model, it matches and extracts a completely different set of feature data, such as S4 scintillation index, airglow image features, and radar echo power spectrum, based on the model's category characteristics. These feature data are then used to construct a feature vector that is suitable for the classification model and does not contain the time dimension. Through this mechanism of on-demand matching based on model category, highly differentiated and specialized data utilization of the same comprehensive data pool is achieved. The prepared target input data is then fed into the corresponding model to drive it to complete a prediction and output the result.

[0073] S30: Based on the forecast results, construct a visualized view of the mid-to-upper atmosphere forecast product and present it through the display interface of the mid-to-upper atmosphere forecast platform.

[0074] Specifically, the prediction results output by each model in the previous step are processed. First, all the gridded prediction results based on geographic coordinates, such as the TEC two-dimensional matrix and the thermosphere density three-dimensional field, are cut according to the preset geographic tile specifications, such as latitude and longitude 2°×2° and a uniform time step, such as 1 hour, and encapsulated into a structured data object containing a geographic location index, a time index and a prediction numerical matrix. Then, these structured data objects are classified and mapped to different product layers according to their physical properties, such as ionospheric TEC, thermosphere density and irregularity probability. For example, all TEC prediction structured data objects are organized into the ionospheric layer, and a set of color mapping tables from blue to red is predefined for them to represent the level of TEC value. Ultimately, based on user interactions such as dragging maps, adjusting time sliders, or system-preset display rules such as a rendering rule that includes layer stacking order, transparency, and color blending mode, different layers are superimposed in a meaningful way, such as superimposing a highlighted risk area layer on top of a physical background parameter layer. This generates a rich, logically clear, and interactive visualization of the mid-to-high atmospheric forecast product, allowing users to zoom, pan, and time relax. This visualization is then pushed to the display interface of the mid-to-high atmospheric forecast platform via real-time communication protocols such as WebSocket.

[0075] In one embodiment, such as Figure 2 As shown, in step S20, based on the category of the prediction model, the corresponding target input data is matched from the observation data, and the target input data is input into the prediction model to obtain the prediction result. Specifically, this includes:

[0076] S21: When the prediction model is classified as an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the resulting prediction result is the ionospheric storm prediction result.

[0077] Specifically, when the scheduled model category is identified as an ionospheric storm prediction model, the model is a deep learning network based on a bidirectional long short-term memory network (Bi-directionalLSTM) that performs a targeted data matching and combination operation. It obtains the first data from the observation data. The first data is a multivariate time series dataset that integrates the F10.7 solar activity index and the Kp and AE geomagnetic disturbance indices as external drivers, as well as global TEC grid data and satellite occultation electron density profiles as a representation of the current state of the ionosphere. After the first data is input into the model, the ionospheric storm prediction result generated by the model is a comprehensive product that includes the TEC and three-dimensional electron density evolution forecast for the next few days, and is accompanied by key event markers such as storm onset and peak time.

[0078] S22: When the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the prediction result obtained is the irregular body feature recognition prediction result.

[0079] Specifically, when the scheduled model category is identified as an irregular body feature recognition prediction model, a data matching action targeting multiple diagnostic features is performed to obtain second data from the observation data. This second data specifically includes GNSS carrier phase data and S4 scintillation index from multiple sources such as WHU, CETC, CMPII, and COSMIC; airglow image features processed by the CMPII all-sky imager; F2 layer critical frequency and corresponding altitude from the GIRO global ionospheric altimeter observation network; and plasma parameters and echo characteristics detected by coherent / incoherent scattering radars from institutions such as IGGCAS. By inputting the second data into the trained irregular body feature recognition prediction model, leveraging its ability to process massive amounts of data at high speed and perform high-precision classification, a probability value characterizing the future severe disturbance of the geographic grid is quickly output. Combined with radar velocity data, a disturbance drift vector field is given, which together constitute the irregular body disturbance prediction result.

[0080] S23: When the prediction model is classified as a thermosphere dynamics simulation model, the corresponding target input data is matched as the third data, and the resulting prediction result is the thermosphere state prediction result.

[0081] Specifically, when the scheduled model category is identified as a thermospheric dynamics simulation model, a comprehensive data matching action required to drive a complex physical model is executed. Third data is obtained from the observation data. This third data specifically includes geomagnetic and solar activity indices such as F10.7, Ap, and Dst. During the simulation, a massive amount of in-situ observation data from multiple on-orbit satellites such as NOAA and CSES at different altitudes, such as 507km and 800-850km, is continuously fused. This in-situ observation data specifically covers physical parameters such as in-situ particles, magnetic fields, electric fields, electric potential, electrical conductivity, sedimented particles, FUV / EUV radiation, Joule heating, neutral component density, and ion flux. After inputting this third data into the model, the predicted thermospheric state results covering the entire globe and three-dimensional dynamics are obtained.

[0082] S24: When the prediction model is classified as a mid-to-upper atmosphere density and temperature model, the corresponding target input data is matched as the fourth data, and the prediction result is the atmospheric density and temperature prediction result.

[0083] Specifically, when the scheduled model category is a mid-to-upper atmospheric density and temperature model for refined near-space environment forecasting, a composite data matching action is executed to drive the hybrid forecast model. A fourth data point is obtained from the observation data. This fourth data point specifically includes temperature, water vapor, and ozone profile data from multiple sources such as TIMED / SABER, CMA, and NOAA, as well as wind field data from 20-80km and occultation bending angles or refractive index data from multiple sources such as COSMIC and CMA. After inputting this fourth data point into the model, the resulting atmospheric density and temperature prediction result is a set of probabilistic forecast products containing uncertainty information. It not only provides the most likely estimates of the future near-space environment, such as temperature, density, and ozone concentration, but also gives the reliability of the forecast in the form of ensemble dispersion.

[0084] In one embodiment, such as Figure 3 As shown, in step S21, when the prediction model category is an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the obtained prediction result is the ionospheric storm prediction result, which specifically includes:

[0085] S211: The first data should include at least the total electron content of the ionosphere, the AE index, and the F10.7 solar activity index.

[0086] Specifically, the first data can be the total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index. The F10.7 solar activity index, as an effective proxy indicator of solar extreme ultraviolet radiation, defines the basic state or background density of the ionosphere. The AE index, or auroral photocurrent index, characterizes the direct driving force of energy injection in high-latitude regions. The total electron content data of the ionosphere provides real-time observation of the ionosphere's response to the above-mentioned driving forces, and intuitively shows where the ionosphere thickens (positive phase storm) or thins (negative phase storm).

[0087] S212: Using a trained LSTM-CNN hybrid ionospheric storm prediction model, the total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index are input into the ionospheric storm prediction model to perform time series prediction.

[0088] Specifically, the LSTM-CNN hybrid ionospheric burst prediction model is a prediction unit that integrates the advantages of two different neural networks and has been trained on a large amount of historical data. It can simultaneously and efficiently process the spatial and temporal features contained in the input data, thereby achieving better prediction performance than a single network structure. First, for the total electron content data of the ionosphere, which is input at each time step and exists as a two-dimensional image or grid, it is processed by the convolutional neural network (CNN) part of the model. Utilizing the powerful spatial feature extraction capability of CNN, it automatically learns and identifies key spatial morphological features such as the positive-phase burst region and ionospheric groove in the TEC map, and compresses this complex spatial information into a low-dimensional feature vector. Then, a series of feature vectors extracted by CNN at multiple consecutive time steps are constructed into a time series, and this series is input into the long short-term memory (LSTM) part of the model. Utilizing the ability of LSTM to remember and understand long-range temporal dependencies, it learns and predicts the dynamic evolution of these spatial features. Through this collaborative working mode where CNN is responsible for understanding space and LSTM is responsible for understanding time, a high-quality time series prediction is finally completed.

[0089] S213: Obtain ionospheric storm prediction results from the ionospheric storm prediction model.

[0090] Specifically, the original, high-dimensional numerical matrix output by the model is transformed into a more user-friendly and information-dense application product. The obtained prediction results can include the following parts: First, quantitative physical quantity forecasts, such as hourly evolution maps of global total electron content (TEC) over the next 120 hours, gridded data of three-dimensional global electron density, and distribution maps of F2 layer critical frequencies (foF2) calculated from peak electron density. Second, event-type forecasts generated by algorithm post-processing, such as automatically identifying and marking key warning information such as the expected onset time, peak time, expected intensity level (e.g., moderate, severe), and storm phase type (e.g., positive or negative phase storm) by analyzing the shape of the TEC forecast curve. This provides intuitive and clear decision-making basis for applications such as satellite navigation and shortwave communication.

[0091] In one embodiment, such as Figure 4 As shown, in step S22, when the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the resulting prediction result is the irregular body feature recognition prediction result, which specifically includes:

[0092] S221: The second data shall include at least GNSS observation data, optical observation data, ionospheric detection parameters, and radar detection data.

[0093] Specifically, GNSS observation data refers to carrier phase and S4 index, optical observation data refers to brightness distribution of all-weather imager, ionospheric detection parameters refer to critical frequency and peak height of F2 layer, and radar detection data refers to coherent or incoherent scattering power spectrum characteristics. This combination can comprehensively reflect the intensity and dynamics of the disturbance in space.

[0094] S222: Based on pre-acquired edge extraction algorithms, phase coherence analysis algorithms, and machine learning algorithms, an irregular body feature recognition and forecasting model is constructed. GNSS observation data, optical observation data, ionospheric detection parameters, and radar detection data are input into the irregular body feature recognition and forecasting model to identify irregular bodies and make short, medium, and long-term forecasts.

[0095] Specifically, the irregular body feature recognition and prediction model does not use a single algorithm, but rather constructs a multi-stage composite processing flow that integrates the advantages of multiple algorithms. It first uses traditional algorithms with clear physical meaning to enhance the features of the raw data, extracting key, interpretable intermediate features. Then, it leverages the powerful nonlinear fitting capabilities of machine learning algorithms to perform deep pattern recognition and prediction on these enhanced features. First, edge extraction algorithms, such as the Canny operator, are applied to optical observation data, i.e., airglow images, to automatically identify and quantify the morphological features of dark fringes in plasma bubbles. Simultaneously, phase coherence analysis algorithms are applied to GNSS observation data to accurately quantify the severity of signal phase jitter. Then, these features extracted through physical algorithms, along with other direct observation parameters such as F2 layer height and radar echo power, are fed into a machine learning algorithm, such as a gradient boosting decision tree or a deep neural network, for final recognition and prediction.

[0096] S223: Obtain irregular body feature recognition forecast results from the irregular body feature recognition forecast model. The irregular body feature recognition forecast results shall include at least the irregular body feature recognition results and the irregular body forecast results.

[0097] Specifically, the irregular body feature recognition and forecasting model identifies and forecasts irregular bodies in the short, medium, and long term, resulting in a rich set of irregular body feature recognition and forecasting results that includes both current situation identification and future forecasting. This provides users with a complete information loop from current situation awareness to future risk prediction. The irregular body feature recognition result can be a highlighted analysis map showing the spatial extent of existing irregular bodies, such as plasma bubbles, along with a parameter list that includes a quantitative description of the irregular body, such as its maximum intensity, coverage area, and internal structural complexity. The irregular body forecasting result can be a future-oriented prediction product, such as a probability distribution map showing the occurrence of significant signal flickering in various regions within the next 0-3 hours, and a dynamic drift vector field predicting how the identified irregular body will move in the future.

[0098] In one embodiment, such as Figure 5 As shown, in step S23, when the prediction model category is a thermosphere dynamics simulation model, the corresponding target input data is matched as the third data, and the obtained prediction result is the thermosphere state prediction result, which specifically includes:

[0099] S231: The third data should include at least solar activity parameters, geomagnetic disturbance parameters, and time parameters.

[0100] Specifically, solar activity parameters in the third set of data, such as the F10.7 index or EUV spectral flux, serve to set the basic energy source for the entire simulation. They determine the background temperature and ionization level of the upper atmosphere. Geomagnetic disturbance parameters, such as the Kp index or Ap index, serve to set the disturbance energy source during the simulation. They quantify the energy injection intensity from the magnetosphere and are key to driving violent phenomena such as thermospheric bursts. Time parameters, including year, month, day, and UTC, serve to set the spatiotemporal reference for the simulation. They determine the Earth's solar radiation, seasonal effects, and the duration of disturbance events. These three factors together constitute the necessary conditions for driving the physical model to evolve reasonably.

[0101] S232: Obtain the thermospheric dynamics simulation model based on TIEGCM, input solar activity parameters, geomagnetic disturbance parameters and time parameters into the thermospheric dynamics simulation model, so as to predict the thermospheric electric field and potential through the thermospheric dynamics simulation model.

[0102] Specifically, the TIEGCM-based thermospheric dynamics simulation model is a complex physical model unit that has been fully validated and parameter optimized. It performs calculations by solving a set of partial differential equations describing the hydrodynamic, thermodynamic, and electrodynamic processes of the upper atmosphere. Through direct simulation of physical processes, it can generate a spatiotemporally continuous three-dimensional atmospheric state in which the physical quantities are coupled and the internal logic is completely self-consistent. After receiving input parameters, the model calculates the global circulation, i.e., the wind field, of neutral atmosphere such as oxygen atoms and nitrogen molecules under the influence of solar radiation heating and collisions with ions (i.e., ion drag) on ​​a global three-dimensional grid through numerical integration. At the same time, the model also calculates the thermospheric electric field and corresponding potential distribution generated by the atmospheric wind turbine effect when the neutral wind field cuts the geomagnetic field. This predicted electric field and potential are key physical quantities that determine how ionospheric plasma is transported and distributed, thus constituting an important upstream link in the entire forecast of the middle and upper atmosphere system.

[0103] S233: Obtain the predicted state of the thermosphere from the thermosphere dynamics simulation model.

[0104] Specifically, the thermospheric state prediction result is a high-resolution and high-fidelity digital twin of the upper atmosphere. The obtained results can be specifically included in the following parts: First, a three-dimensional thermospheric atmospheric density field, which is the core input with the highest accuracy requirements for calculating atmospheric drag force and orbit prediction of low-Earth orbit spacecraft; second, a three-dimensional thermospheric temperature field and wind field, including zonal, meridional and vertical wind speeds; third, key energy budget parameters, such as the Joule heating power distribution map, which can be used to diagnose the energy deposition process during geomagnetic storms; and fourth, application-oriented derivative products, such as simplified low, medium and high level total density disturbance level maps at specific altitudes of 250km, 400km, etc., to facilitate users to quickly assess risks.

[0105] In one embodiment, such as Figure 6 As shown, in step S24, when the prediction model category is the middle and upper atmosphere density and temperature model, the corresponding target input data is matched as the fourth data, and the obtained prediction result is the atmospheric density and temperature prediction result, which specifically includes:

[0106] S241: The fourth set of data includes at least solar activity parameters, geomagnetic activity index, occultation atmospheric temperature, and occultation atmospheric density.

[0107] Specifically, solar activity parameters and geomagnetic activity indices are mainly used as background driving parameters to drive the empirical or climatological part of the model, so as to quickly generate a globally covered and stable baseline atmospheric state. Occultation atmospheric temperature and occultation atmospheric density are used as real-world constraint data, which come from GNSS occultation detection technology with high vertical resolution and high precision. They can provide ground-based real information about the current atmospheric thermodynamic structure, so as to correct and pull back theoretical results generated purely by background driving, and ensure that the prediction can reflect real weather changes.

[0108] S242: Obtain the trained Light-GBM middle and upper atmospheric density and temperature model, input solar activity parameters, geomagnetic activity index, occultation atmospheric temperature and occultation atmospheric density into the middle and upper atmospheric density and temperature model, and use the middle and upper atmospheric density and temperature model to predict the near-space environment.

[0109] Specifically, the middle and upper atmosphere density and temperature model is a composite prediction unit that may contain multiple computational components and has been trained or optimized with a large amount of data. It can leverage the strengths of both empirical models and real-time observation data, thus generating more reliable prediction results than single-type models. After receiving input data, an empirical model component within the model first uses solar activity parameters and geomagnetic activity indices to quickly calculate a globally covered background field. At the same time, another data-driven correction component within the model, such as a pre-trained neural network, compares the high-resolution measured data of occultation atmospheric temperature and density with the values ​​of the aforementioned background field at the corresponding spatiotemporal points, calculates the deviation between the two, and finally, this deviation calculated based on real observations is superimposed on the background field as a correction field. This results in a final near-space environment prediction result, namely the atmospheric density and temperature prediction result, that has both a global physical morphology and a high degree of consistency with real observations in local details.

[0110] S243: Obtain atmospheric density and temperature prediction results from the middle and upper atmospheric density and temperature model.

[0111] Specifically, atmospheric density and temperature prediction results are a set of probabilistic forecast products that provide users with the risk assessment information needed for decision-making. The results can specifically include the following parts: First, the ensemble average forecast field, which is the average of the results of all ensemble members, as an estimate of future atmospheric conditions such as temperature, density, air pressure, wind speed, and major components such as OH and O atoms; second, the ensemble dispersion or standard deviation field, which is the degree of dispersion of the results of all ensemble members. The larger this value, the higher the uncertainty and the lower the reliability of the forecast; and third, a probability map of specific events, for example, calculating the percentage of members whose density exceeds a certain threshold, thereby obtaining a probability map of atmospheric density exceeding the danger threshold at a certain time and place in the future.

[0112] In one embodiment, such as Figure 7 As shown, in step S30, which involves constructing a visualized upper-middle atmospheric forecast product view based on the forecast results, the specific steps include:

[0113] S31: The prediction results of ionospheric storms, irregular body features, thermosphere state, and atmospheric density and temperature are divided and organized according to spatial regions and time scales to obtain the organized prediction layer data.

[0114] Specifically, the prediction results output by each model in the previous step are processed. First, the prediction results are preprocessed into a series of discretized, miniaturized, and structured data objects with standardized spatiotemporal indexes, i.e., prediction layer data. For example, a three-dimensional thermosphere density prediction result covering the globe and with a forecast duration of 72 hours would result in a delay of several seconds or even longer if the entire file were read and segmented for each user request. However, by pre-partitioning and organizing the data according to spatial regions and time scales, it is preprocessed into tens of thousands of independent small files or cached objects. Each object corresponds precisely to a specific geographic grid, such as 2°×2°, and a specific time point, such as 1 hour. When a user requests to view the situation of a certain local area, the backend can directly locate and read the dozens of corresponding data objects in parallel through the spatiotemporal index, thereby reducing the data retrieval latency to the millisecond level.

[0115] S32: Map the predicted layer data to the preset product layer, which includes the ionosphere layer, irregularity disturbance layer, thermosphere layer, and middle and upper atmosphere layer.

[0116] Specifically, the product layer is a logical container that is business-oriented and associated with specific physical quantities or forecast products. It decouples the underlying physical data storage from the upper-level business logic calls, making the entire system highly modular and scalable. All the forecast layer data containing the total electron content (TEC) prediction value generated in the above steps, regardless of which specific model it comes from, will be logically classified and mapped to the ionospheric layer. All the forecast layer data containing the scintillation probability will be mapped to the irregular body perturbation layer. At the same time, when defining each product layer, some default rendering attributes will be pre-bound to it. For example, a color mapping table from cool to warm tones is preset for the thermal layer to intuitively show the change of density from low to high.

[0117] S33: Based on preset layer display rules, perform layer compositing on product layers to generate a visualized mid-to-high-altitude atmospheric prediction product view.

[0118] Specifically, the layer display rules predefine the spatial range, time window, layer priority, transparency, and overlay mode. The application first reads the latitude and longitude boundaries of the user view and the selected prediction time as dynamic parameters. Then, it retrieves the ionospheric layer, irregular body disturbance layer, thermosphere layer, and middle and upper atmosphere layer that match the spatial range and time window from the layer repository. Each layer is sent to the WebGL rendering pipeline in order of priority from low to high, and alpha blending is performed according to transparency and color fusion is controlled according to overlay mode. After rendering, a single bitmap containing the superposition effect of multiple physical quantities is output and pushed to the front-end canvas. At the same time, a time slider and layer switch control are loaded for user interaction, thereby generating a visualized middle and upper atmosphere prediction product view.

[0119] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0120] In one embodiment, a mid-to-upper atmosphere prediction system is provided, which corresponds one-to-one with the mid-to-upper atmosphere prediction methods described in the above embodiments. For example... Figure 8 As shown, this mid-to-upper atmosphere prediction system includes a data acquisition and management module, a prediction result generation module, and a visualization and presentation module. Detailed descriptions of each functional module are as follows:

[0121] The data acquisition and management module is used to acquire observation data of the multi-source space environment and call the corresponding prediction model through a preset model scheduling strategy.

[0122] The prediction result generation module is used to match the corresponding target input data in the observation data based on the category of the prediction model, input the target input data into the prediction model, and obtain the prediction result.

[0123] The visualization and presentation module is used to construct a visualized view of the mid-to-upper atmosphere forecast product based on the forecast results, and to present the visualized mid-to-upper atmosphere forecast product view through the display interface of the mid-to-upper atmosphere forecast platform.

[0124] Optional, the prediction result generation module specifically includes:

[0125] The ionospheric storm prediction submodule is used to match the corresponding target input data as the first data when the prediction model is classified as an ionospheric storm prediction model, and the resulting prediction result is the ionospheric storm prediction result.

[0126] The irregular body recognition submodule is used to match the corresponding target input data as the second data when the prediction model is classified as an irregular body feature recognition prediction model, and the prediction result obtained is the irregular body feature recognition prediction result.

[0127] The thermosphere state simulation submodule is used to match the corresponding target input data as the third data when the prediction model is classified as a thermosphere dynamics simulation model, and the resulting prediction result is the thermosphere state prediction result.

[0128] The middle atmosphere modeling submodule is used to match the corresponding target input data as the fourth data when the prediction model is classified as a middle and upper atmosphere density and temperature model, and the resulting prediction result is the atmospheric density and temperature prediction result.

[0129] Optional, the ionospheric storm prediction submodule specifically includes:

[0130] The ionospheric state data unit is used for the first data, which includes at least the total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index.

[0131] The model prediction unit is used to input the total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index into the ionospheric storm prediction model through the trained LSTM-CNN hybrid ionospheric storm prediction model, so as to make time series predictions through the ionospheric storm prediction model.

[0132] The prediction result acquisition unit is used to obtain ionospheric storm prediction results from the ionospheric storm prediction model.

[0133] Optional, irregular body recognition submodule, specifically including:

[0134] The irregular body data unit is used for the second data, which includes at least GNSS observation data, optical observation data, ionospheric detection parameters, and radar detection data;

[0135] The model prediction unit is used to construct an irregular body feature recognition and prediction model based on pre-acquired edge extraction algorithm, phase coherence analysis algorithm and machine learning algorithm. GNSS observation data, optical observation data, ionospheric detection parameters and radar detection data are input into the irregular body feature recognition and prediction model to identify irregular bodies and make short, medium and long-term predictions.

[0136] The prediction result acquisition unit is used to obtain the irregular body feature recognition prediction result from the irregular body feature recognition prediction model. The irregular body feature recognition prediction result includes at least the irregular body feature recognition result and the irregular body prediction result.

[0137] Optional, the thermal state simulation submodule specifically includes:

[0138] The solar activity data unit is used for third-party data that includes at least solar activity parameters, geomagnetic disturbance parameters, and time parameters.

[0139] The model prediction unit is used to obtain the thermospheric dynamics simulation model based on TIEGCM. Solar activity parameters, geomagnetic disturbance parameters and time parameters are input into the thermospheric dynamics simulation model to predict the thermospheric electric field and potential through the thermospheric dynamics simulation model.

[0140] The prediction result acquisition unit is used to obtain the predicted results of the thermosphere state from the thermosphere dynamics simulation model.

[0141] Optional, the mid-atmosphere modeling submodule includes:

[0142] The middle and upper atmospheric parameter data unit, used for the fourth data, includes at least solar activity parameters, geomagnetic activity index, occultation atmospheric temperature and occultation atmospheric density;

[0143] The model prediction unit is used to acquire the trained Light-GBM middle and upper atmospheric density and temperature model. Solar activity parameters, geomagnetic activity index, occultation atmospheric temperature and occultation atmospheric density are input into the middle and upper atmospheric density and temperature model to predict the near-space environment.

[0144] The prediction result acquisition unit is used to obtain atmospheric density and temperature prediction results from the middle and upper atmospheric density and temperature model.

[0145] Optional visualization and presentation modules include:

[0146] The data partitioning and processing submodule is used to partition and organize the ionospheric storm prediction results, irregular body feature prediction results, thermosphere state prediction results, and atmospheric density and temperature prediction results according to spatial regions and time scales to obtain the organized prediction layer data.

[0147] The layer mapping submodule is used to map the predicted layer data to the preset product layers, which include the ionosphere layer, the irregular body disturbance layer, the thermosphere layer, and the middle and upper atmosphere layer.

[0148] The dynamic view compositing submodule, based on preset layer display rules, performs layer compositing on product layers to generate a visualized mid-to-high-altitude atmospheric prediction product view.

[0149] Specific limitations regarding the middle and upper atmosphere prediction system can be found in the limitations of the middle and upper atmosphere prediction methods described above, and will not be repeated here. Each module in the aforementioned middle and upper atmosphere prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0150] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores data such as multi-source space environment observation data, ionospheric storm prediction models, irregular body disturbance identification models, thermospheric dynamics simulation models, and meso-atmospheric density and temperature models. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for predicting the middle and upper atmosphere.

[0151] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0152] Acquire observational data of the multi-source space environment, and call the corresponding prediction model through a preset model scheduling strategy;

[0153] Based on the category of the prediction model, the corresponding target input data is matched in the observation data, and the target input data is input into the prediction model to obtain the prediction result;

[0154] Based on the forecast results, a visualized view of the middle and upper atmosphere forecast products is constructed and presented through the display interface of the middle and upper atmosphere forecast platform.

[0155] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0156] Acquire observational data of the multi-source space environment, and call the corresponding prediction model through a preset model scheduling strategy;

[0157] Based on the category of the prediction model, the corresponding target input data is matched in the observation data, and the target input data is input into the prediction model to obtain the prediction result;

[0158] Based on the forecast results, a visualized view of the middle and upper atmosphere forecast products is constructed and presented through the display interface of the middle and upper atmosphere forecast platform.

[0159] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0160] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0161] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for predicting the middle and upper atmosphere, characterized in that, Applied to a mid-to-upper atmosphere prediction platform, the mid-to-upper atmosphere prediction method includes: Acquire observational data of the multi-source space environment, and call the corresponding prediction model through a preset model scheduling strategy; Based on the category of the prediction model, the corresponding target input data is matched in the observation data, and the target input data is input into the prediction model to obtain the prediction result; Based on the prediction results, a visualized mid-to-upper atmosphere prediction product view is constructed and presented through the display interface of the mid-to-upper atmosphere prediction platform. Specifically, the step of matching corresponding target input data from the observed data based on the category of the prediction model, and inputting the target input data into the prediction model to obtain the prediction result includes: When the prediction model is classified as an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the resulting prediction result is the ionospheric storm prediction result. When the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the prediction result obtained is the irregular body feature recognition prediction result. When the prediction model is classified as a thermosphere dynamics simulation model, the corresponding target input data is matched as the third data, and the resulting prediction result is the thermosphere state prediction result. When the prediction model is classified as a mid-to-upper atmosphere density-temperature model, the corresponding target input data is matched as the fourth data, and the resulting prediction result is the atmospheric density-temperature prediction result. The step of constructing a visualized mid-to-upper atmosphere forecast product view based on the forecast results specifically includes: The ionospheric storm prediction results, the irregular body feature identification prediction results, the thermosphere state prediction results, and the atmospheric density and temperature prediction results are divided and organized according to spatial regions and time scales to obtain the organized prediction layer data. The predicted layer data is mapped to a preset product layer, which includes an ionospheric layer, an irregularity disturbance layer, a thermosphere layer, and a middle and upper atmosphere layer. Based on preset layer display rules, the product layers are composited to generate the visualized mid-to-high-altitude atmospheric prediction product view.

2. The method for predicting the middle and upper atmosphere according to claim 1, characterized in that, When the prediction model is classified as an ionospheric storm prediction model, the corresponding target input data is matched as the first data, and the resulting prediction result is the ionospheric storm prediction result, specifically including: The first data includes at least the total electron content of the ionosphere, the AE index, and the F10.7 solar activity index; The total electron content data of the ionosphere, the AE index, and the F10.7 solar activity index are input into the ionospheric storm prediction model that has been trained using the LSTM-CNN hybrid model, so as to perform time series prediction. The ionospheric storm prediction results are obtained from the ionospheric storm prediction model.

3. The method for predicting the middle and upper atmosphere according to claim 1, characterized in that, When the prediction model is classified as an irregular body feature recognition prediction model, the corresponding target input data is matched as the second data, and the resulting prediction result is the irregular body feature recognition prediction result, specifically including: The second data includes at least GNSS observation data, optical observation data, ionospheric detection parameters, and radar detection data; Based on the pre-acquired edge extraction algorithm, phase coherence analysis algorithm, and machine learning algorithm, the irregular body feature identification and prediction model is constructed. The GNSS observation data, the optical observation data, the ionospheric detection parameters, and the radar detection data are input into the irregular body feature identification and prediction model to identify irregular bodies and make short, medium, and long-term predictions. The irregular body feature recognition and prediction results are obtained from the irregular body feature recognition and prediction model. The irregular body feature recognition and prediction results include at least the irregular body feature recognition results and the irregular body prediction results.

4. The method for predicting the middle and upper atmosphere according to claim 1, characterized in that, When the prediction model is classified as a thermosphere dynamics simulation model, the corresponding target input data is matched as the third data, and the resulting prediction result is the thermosphere state prediction result, which specifically includes: The third data includes at least solar activity parameters, geomagnetic disturbance parameters, and time parameters; A thermosphere dynamics simulation model based on TIEGCM is obtained, and the solar activity parameters, geomagnetic disturbance parameters and time parameters are input into the thermosphere dynamics simulation model to predict the thermosphere electric field and potential through the thermosphere dynamics simulation model; The predicted state of the thermal layer is obtained from the thermal layer dynamics simulation model.

5. The method for predicting the middle and upper atmosphere according to claim 1, characterized in that, When the prediction model is classified as a mid-to-upper atmosphere density and temperature model, the corresponding target input data is matched as the fourth data, and the resulting prediction result is the atmospheric density and temperature prediction result, which specifically includes: The fourth set of data includes at least solar activity parameters, geomagnetic activity index, occultation atmospheric temperature, and occultation atmospheric density; The trained Light-GBM middle and upper atmosphere density and temperature model is obtained. The solar activity parameters, geomagnetic activity index, occultation atmospheric temperature and occultation atmospheric density are input into the middle and upper atmosphere density and temperature model to predict the near space environment. The atmospheric density and temperature prediction results are obtained from the aforementioned middle and upper atmospheric density and temperature model.

6. A mid-to-upper atmosphere prediction system, characterized in that, The middle and upper atmosphere prediction system includes: The data acquisition and management module is used to acquire observation data of the multi-source space environment and call the corresponding prediction model through a preset model scheduling strategy. The prediction result generation module is used to match the corresponding target input data in the observation data based on the category of the prediction model, input the target input data into the prediction model, and obtain the prediction result; The visualization and presentation module is used to construct a visualized mid-to-upper atmosphere prediction product view based on the prediction results, and to present the visualized mid-to-upper atmosphere prediction product view through the display interface of the mid-to-upper atmosphere prediction platform. The prediction result generation module specifically includes: The ionospheric storm prediction submodule is used to match the corresponding target input data as the first data when the prediction model is classified as an ionospheric storm prediction model, and the resulting prediction result is the ionospheric storm prediction result. The irregular body recognition submodule is used to match the corresponding target input data as the second data when the prediction model is classified as an irregular body feature recognition prediction model, and the prediction result obtained is the irregular body feature recognition prediction result. The thermosphere state simulation submodule is used to match the corresponding target input data as the third data when the prediction model is classified as a thermosphere dynamics simulation model, and the resulting prediction result is the thermosphere state prediction result. The middle atmosphere modeling submodule is used to match the corresponding target input data as the fourth data when the prediction model is classified as a middle and upper atmosphere density and temperature model, and the resulting prediction result is the atmospheric density and temperature prediction result. The visualization and presentation module specifically includes: The data partitioning and processing submodule is used to partition and organize the ionospheric storm prediction results, irregular body feature identification prediction results, thermosphere state prediction results, and atmospheric density and temperature prediction results according to spatial regions and time scales to obtain the organized prediction layer data. The layer mapping submodule is used to map the predicted layer data to the preset product layers, which include the ionosphere layer, the irregular body disturbance layer, the thermosphere layer, and the middle and upper atmosphere layer. The dynamic view compositing submodule, based on preset layer display rules, performs layer compositing on product layers to generate a visualized mid-to-high-altitude atmospheric prediction product view.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the middle and upper atmosphere prediction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the middle and upper atmosphere prediction method as described in any one of claims 1 to 5.

Images