Active area physical parameter and geometric feature fusion-based solar flare forecasting method and system
By integrating the magnetic field physical parameters and geometric characteristics of solar active regions and using neural network models for flare prediction, the problems of single prediction features and insufficient utilization of dynamic information in existing technologies have been solved, achieving higher prediction accuracy and automated prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-12
- Publication Date
- 2026-04-14
AI Technical Summary
Existing solar flare prediction methods lack in-depth integration of the physical parameters and geometric morphology of active regions and fail to effectively utilize dynamic evolution information, resulting in insufficient prediction accuracy and reliability.
By extracting the magnetic field physical parameters and geometric features of solar active regions, constructing time series data, and combining these with a neural network model for feature fusion, the probability of future solar flares can be predicted.
It has improved the accuracy and lead time of flare forecasts, reduced the missed detection rate and false alarm rate, and achieved more comprehensive information fusion and automated forecasting.
Smart Images
Figure CN121859226A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of space weather monitoring and forecasting technology, specifically to a solar flare forecasting method and system based on the fusion of physical parameters and geometric features of active regions. Background Technology
[0002] Solar flares are violent energy release processes in the solar atmosphere, accompanied by intense electromagnetic radiation and particle streams. They can cause serious damage to spacecraft operations, astronaut safety, radio communications, navigation and positioning, and even ground power grids. Therefore, achieving accurate and timely forecasting of solar flares, especially large-scale flares (such as M-class and X-class flares), is an important goal and urgent need in space weather research.
[0003] Existing methods for predicting solar flares can be mainly classified into the following categories: 1. Statistical methods based on empirical parameters. These methods rely on empirical parameters extracted from solar active region magnetic field data, such as total magnetic flux, current helicity, and magnetic field gradient, to make predictions by establishing statistical relationships between these parameters and flare occurrence rates. For example, some pre-calculated parameters from the SHARP data series of the SDO satellite are used. The disadvantage of this type of method is that it relies too heavily on a single type of physical parameter, fails to fully utilize the morphological information of active regions, has limited predictive ability for complex active regions, and its accuracy and reliability need to be improved.
[0004] 2. Morphological methods based on optical images. These methods primarily analyze the morphological characteristics of active regions in solar optical images (such as images from a specific band of SDO / AIA), including the area and complexity of sunspot groups. Their limitation lies in the lack of a direct description of the underlying magnetic field physical processes, resulting in insufficient physical support for predictions.
[0005] 3. Single machine learning methods. In recent years, many studies have attempted to use machine learning models (such as support vector machines and neural networks) for flare prediction, but their input features are often limited to physical parameters provided by SHARP data or image features from a single source, failing to effectively integrate the two complementary key types of physical and geometric information.
[0006] Furthermore, most existing methods rely on "static" characteristics at a single point in time for forecasting, failing to adequately consider the dynamic evolution of the physical parameters and geometry of active regions before solar flares erupt. Since solar flares are essentially the product of the continuous accumulation of magnetic field energy reaching a critical state, capturing the evolutionary trends of active regions before eruptions (such as the rapid emergence of magnetic flux, the enhancement of current, and the increasing complexity of sunspot structures) is crucial for improving forecast accuracy and lead time.
[0007] In summary, existing technologies suffer from limitations such as limited forecast features, failure to deeply integrate the physical nature and geometric morphology of active regions, and lack of effective utilization of dynamic evolution information of active regions, leading to a bottleneck in forecast accuracy. Summary of the Invention
[0008] To achieve the objective of this invention, this application provides a method for predicting solar flares based on the fusion of physical parameters and geometric features of active regions, comprising: Step S1: Acquire magnetic field observation data and corresponding optical image data of the target solar active region; Step S2: Extract the magnetic field physical parameters from the magnetic field observation data and extract the geometric features of the active area from the optical image data; Step S3: Obtain magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct a time series of the magnetic field physical parameters and geometric characteristics; Step S4: Extract the dynamic evolution characteristics of the magnetic field physical parameters and geometric features from the time series; Step S5: The current magnetic field physical parameters and geometric features are used as static features and fused with the dynamic evolution features to construct an enhanced fusion feature vector; Step S6: Input the enhanced fusion feature vector into the pre-trained neural network model to obtain the predicted probability of a specific level of flare occurring in the target solar active region within the future forecast time window; Step S7: Generate and publish flare forecast information based on the comparison result between the predicted probability and the preset threshold.
[0009] In some specific embodiments, in step S2, the magnetic field physical parameters include at least one of the total unsigned vertical current, total unsigned current helicity, and total unsigned magnetic flux.
[0010] In some specific embodiments, in step S2, the geometric features include at least one of the following: area, perimeter, compactness, aspect ratio, sunspot distribution range, and morphological moment features of the active area; wherein, the morphological moment features are obtained by calculating the central moment, normalized central moment, and invariant moment based on the binarization result of the active area image.
[0011] In some specific embodiments, in step S4, the dynamic evolution characteristics include at least one of the mean, standard deviation, trend slope, and volatility intensity of the time series within a historical time window.
[0012] In some specific embodiments, the process of extracting the fluctuation intensity includes: performing linear fitting on the time series to obtain a fitted sequence, and determining the fluctuation intensity based on the standard deviation of the residuals between the time series and the fitted sequence.
[0013] In some specific embodiments, in step S6, the neural network model is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolutionary features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
[0014] In some specific embodiments, the processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.
[0015] To achieve the same inventive objective, this application also provides a solar flare prediction system based on the fusion of physical parameters and geometric features of active regions, comprising: Data acquisition module: used to acquire magnetic field observation data and corresponding optical image data of the target solar active region; Feature extraction module: used to extract magnetic field physical parameters from the magnetic field observation data and extract geometric features of the active area from the optical image data; Time series construction module: used to acquire magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct time series of the magnetic field physical parameters and geometric features; Dynamic feature calculation module: used to extract the dynamic evolution features of the magnetic field physical parameters and geometric features from the time series; Feature fusion module: used to fuse the current magnetic field physical parameters and geometric features as static features with the dynamic evolution features to construct an enhanced fusion feature vector; Forecasting model module: used to input the enhanced fusion feature vector into a pre-trained neural network model to obtain the forecast probability of a specific level of flare occurring in the target solar active region within a future forecast time window; Forecast output module: used to generate and publish flare forecast information based on the comparison result between the forecast probability and the preset threshold.
[0016] In some specific embodiments, the neural network model in the prediction model module is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolutionary features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
[0017] In some specific embodiments, the processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.
[0018] The beneficial effects of the above technical solution are as follows: 1. Deep fusion of multi-source information. This invention is the first to effectively fuse magnetic field parameters describing the physical nature of an active region with geometric features describing its macroscopic morphology, overcoming the limitations of single-type features and providing a more comprehensive and complementary information foundation for forecasting models.
[0019] 2. Clear physical mechanisms. The selected physical parameters (vertical current, current helicity, total unsigned magnetic flux, etc.) are directly related to the magnetic energy accumulation and nonpotentiality of flare eruptions, while the geometric features reflect the complexity of magnetic flux emergence and interaction. This integration makes the prediction model not only a "black box" but also provides a clear physical interpretation.
[0020] 3. Introducing Dynamic Evolution Information. This invention overcomes the limitations of traditional static features by extracting the characteristics of the physical parameters and geometric morphology of active regions as they evolve over time, enabling the model to "sense" the development stage and changing trends of active regions. This not only better aligns with the physical processes of solar flares (energy accumulation and release), but also effectively identifies active regions that appear to have insignificant static parameters but are undergoing rapid evolution and are extremely unstable. This is expected to further reduce the false negative rate and potentially enable longer-term advance predictions.
[0021] 4. Improved forecasting performance. Neural networks can effectively learn the complex nonlinear relationship between high-dimensional static and dynamic fusion features and solar flares. Compared with traditional statistical methods or single-feature models, they are expected to significantly improve the accuracy of solar flare forecasts, especially the accuracy rate, and reduce the false alarm rate.
[0022] 5. Automation and Real-time Performance. This method enables a fully automated process from data download to forecast output, meeting the timeliness requirements of operational space weather forecasts. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 A flowchart illustrating a solar flare prediction method based on the fusion of physical parameters and geometric features of active regions, provided as an embodiment of the present invention; Figure 2 This is a schematic diagram of a solar flare prediction system based on the fusion of physical parameters and geometric features of active regions, provided as an embodiment of the present invention. Detailed Implementation
[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0026] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.
[0027] Example 1 One embodiment of the present invention provides a solar flare prediction method based on the fusion of physical parameters and geometric features of active regions, referring to... Figure 1 As shown, it includes: Step S1: Acquire magnetic field observation data and corresponding optical image data of the target solar active region; This embodiment takes the prediction of whether a ≥M-class flare will occur within the next 24 hours as an example, but the scope of application of the present invention is not limited to this.
[0028] Using NASA's SDO satellite data, HMI instruments provided SHARP data (such as the hmi.sharp_720s series), and AIA instruments provided images in the 1700 Å band. The historical time window was set to 24 hours, and the sampling interval was 1 hour, meaning data was extracted for each active region at time points T=24.
[0029] Physical parameters pre-calculated directly from SHARP data, such as TOTUSJZ (total unsigned vertical current), TOTUSJH (total unsigned current helicity), and TOTUSJZ (total unsigned magnetic flux).
[0030] The active region image is preprocessed with flat field correction and edge shading correction, and then active region segmentation is performed to extract the sunspot region.
[0031] Step S2: Extract the magnetic field physical parameters from the magnetic field observation data and extract the geometric features of the active area from the optical image data; In a specific embodiment of the present invention, in step S2, the magnetic field physical parameters include at least one of the following: total unsigned vertical current, total unsigned current helicity, and total unsigned magnetic flux.
[0032] In a specific embodiment of the present invention, in step S2, the geometric features include at least one of the following: area, perimeter, compactness, aspect ratio, sunspot distribution range, and morphological moment features of the active area; wherein, the morphological moment features are obtained by calculating the central moment, normalized central moment, and invariant moment based on the binarization result of the active area image.
[0033] Extract the following physical parameters from the SHARP data: Total unsigned vertical current. Total unsigned current helicity Total unsigned magnetic flux .
[0034] Extracting geometric features includes area, perimeter, and compactness. Aspect ratio; morphological characteristics, including central moments. Normalized central moments and the first two invariant moments .
[0035] The physical parameters selected in this invention (vertical current, current helicity, total unsigned magnetic flux, etc.) are directly related to the magnetic energy accumulation and nonpotentiality of flare eruptions, while geometric features reflect the complexity of magnetic flux emergence and interaction. This fusion makes the prediction model not only a "black box" but also has clear physical interpretability.
[0036] Step S3: Obtain magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct a time series of the magnetic field physical parameters and geometric characteristics; Dynamic characteristics of physical parameters: Calculate the mean, standard deviation, and linear trend slope of the 24-hour time series for each physical parameter. (Using the least squares method), fluctuation intensity Calculate the mean, standard deviation, and slope of each geometric feature time series.
[0037] Step S4: Extract the dynamic evolution characteristics of the magnetic field physical parameters and geometric features from the time series; In a specific embodiment of the present invention, in step S4, the dynamic evolution characteristics include at least one of the mean, standard deviation, trend slope, and fluctuation intensity of the time series within a historical time window.
[0038] In one specific embodiment of the present invention, the process of extracting the fluctuation intensity includes: performing linear fitting on the time series to obtain a fitted sequence, and determining the fluctuation intensity based on the standard deviation of the residuals between the time series and the fitted sequence.
[0039] All features are Z-score normalized. Static feature dimensions: Assuming 3-dimensional physical parameters and 6-dimensional geometric features, for a total of 9 dimensions. Each feature is evaluated using 3 dynamic metrics (mean, standard deviation, slope), resulting in 9 features × 3 = 27 dimensions. Total dimension of the fused feature vector: 9 + 27 = 36 dimensions.
[0040] Label definition: A flare of ≥M level occurring within the next 24 hours is recorded as 1, otherwise it is recorded as 0.
[0041] This invention overcomes the limitations of traditional static features by extracting the characteristics of the physical parameters and geometric morphology of active regions over time, enabling the model to "sense" the development stage and changing trends of active regions. This not only better aligns with the physical processes of solar flares (energy accumulation and release), but also effectively identifies active regions that appear to have insignificant static parameters but are undergoing rapid evolution and are extremely unstable. This is expected to further reduce the false negative rate and potentially enable longer-term advance predictions.
[0042] Step S5: The current magnetic field physical parameters and geometric features are used as static features and fused with the dynamic evolution features to construct an enhanced fusion feature vector; Specifically, the LSTM layer in the dynamic branch has 64 hidden units, and the output context vector is weighted by the Attention mechanism. The static branch uses a two-layer fully connected network with 32 hidden nodes and ReLU activation function.
[0043] The fusion layer concatenates the outputs of c and the static branch, and then outputs the flare probability through a fully connected layer, using the Sigmoid activation function.
[0044] Using the binary cross-entropy loss function, Adam optimizer, learning rate 0.001, batch size 32, and training epochs 100.
[0045] This invention is the first to effectively integrate magnetic field parameters describing the physical nature of an active region with geometric features describing its macroscopic morphology, overcoming the limitations of single-type features and providing a more comprehensive and complementary information foundation for forecasting models.
[0046] Step S6: Input the enhanced fusion feature vector into the pre-trained neural network model to obtain the predicted probability of a specific level of flare occurring in the target solar active region within the future forecast time window; In a specific embodiment of the present invention, in step S6, the neural network model is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolutionary features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
[0047] In a specific embodiment of the present invention, the processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.
[0048] Obtain the latest 24-hour data of the target activity area, extract the fused feature vector, input it into the trained model, and if the output probability is ≥0.5, issue an alert that "a flare of level M or above may occur within the next 24 hours".
[0049] Step S7: Generate and publish flare forecast information based on the comparison result between the predicted probability and the preset threshold.
[0050] The neural network of this invention can effectively learn the complex nonlinear relationship between high-dimensional static and dynamic fusion features and solar flare eruptions. Compared with traditional statistical methods or single feature models, it is expected to significantly improve the accuracy of solar flare prediction, especially the accuracy rate, and reduce the false alarm rate.
[0051] Example 2 One embodiment of the present invention provides a solar flare prediction system based on the fusion of physical parameters and geometric features of active regions, referring to... Figure 2 As shown, it includes: Data acquisition module 10: used to acquire magnetic field observation data and corresponding optical image data of the target solar active region; Feature extraction module 20: used to extract magnetic field physical parameters from the magnetic field observation data and extract geometric features of the active area from the optical image data; Time series construction module 30: used to acquire magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct a time series of the magnetic field physical parameters and geometric features; Dynamic feature calculation module 40: used to extract the dynamic evolution features of the magnetic field physical parameters and geometric features from the time series; Feature fusion module 50: used to fuse the current magnetic field physical parameters and geometric features as static features with the dynamic evolution features to construct an enhanced fusion feature vector; Forecast model module 60: used to input the enhanced fusion feature vector into a pre-trained neural network model to obtain the forecast probability of a specific level of flare occurring in the target solar active region within a future forecast time window; Forecast output module 70: used to generate and publish flare forecast information based on the comparison result between the forecast probability and the preset threshold.
[0052] In a specific embodiment of the present invention, in the prediction model module 60, the neural network model is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolution features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
[0053] In a specific embodiment of the present invention, the processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.
[0054] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
[0055] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the invention. Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.
[0056] The methods and apparatus provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
[0057] In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "a specific embodiment" or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0058] Finally, it should be noted that the above 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.
Claims
1. A solar flare prediction method based on the fusion of physical parameters and geometric features of active regions, characterized in that, include: Step S1: Acquire magnetic field observation data and corresponding optical image data of the target solar active region; Step S2: Extract the magnetic field physical parameters from the magnetic field observation data and extract the geometric features of the active area from the optical image data; Step S3: Obtain magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct a time series of the magnetic field physical parameters and geometric characteristics; Step S4: Extract the dynamic evolution characteristics of the magnetic field physical parameters and geometric features from the time series; Step S5: The current magnetic field physical parameters and geometric features are used as static features and fused with the dynamic evolution features to construct an enhanced fusion feature vector; Step S6: Input the enhanced fusion feature vector into the pre-trained neural network model to obtain the predicted probability of a specific level of flare occurring in the target solar active region within the future forecast time window; Step S7: Generate and publish flare forecast information based on the comparison result between the predicted probability and the preset threshold.
2. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 1, characterized in that, In step S2, the magnetic field physical parameters include at least one of the following: total unsigned vertical current, total unsigned current helicity, and total unsigned magnetic flux.
3. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 1, characterized in that, In step S2, the geometric features include at least one of the following: area, perimeter, compactness, aspect ratio, sunspot distribution range, and morphological moment features of the active region; wherein, the morphological moment features are obtained by calculating the central moment, normalized central moment, and invariant moment based on the binarization result of the active region image.
4. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 1, characterized in that, In step S4, the dynamic evolution characteristics include at least one of the mean, standard deviation, trend slope, and volatility intensity of the time series within the historical time window.
5. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 4, characterized in that, The process of extracting the fluctuation intensity includes: performing linear fitting on the time series to obtain a fitted sequence, and determining the fluctuation intensity based on the standard deviation of the residuals between the time series and the fitted sequence.
6. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 1, characterized in that, In step S6, the neural network model is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolutionary features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
7. The solar flare prediction method based on the fusion of physical parameters and geometric features of active regions according to claim 6, characterized in that, The processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.
8. A solar flare prediction system based on the fusion of physical parameters and geometric features of active regions, characterized in that, include: Data acquisition module: used to acquire magnetic field observation data and corresponding optical image data of the target solar active region; Feature extraction module: used to extract magnetic field physical parameters from the magnetic field observation data and extract geometric features of the active area from the optical image data; Time series construction module: used to acquire magnetic field observation data and optical image data of the target solar active region within a historical time window, and construct time series of the magnetic field physical parameters and geometric features; Dynamic feature calculation module: used to extract the dynamic evolution features of the magnetic field physical parameters and geometric features from the time series; Feature fusion module: used to fuse the current magnetic field physical parameters and geometric features as static features with the dynamic evolution features to construct an enhanced fusion feature vector; Forecasting model module: used to input the enhanced fusion feature vector into a pre-trained neural network model to obtain the forecast probability of a specific level of flare occurring in the target solar active region within a future forecast time window; Forecast output module: used to generate and publish flare forecast information based on the comparison result between the forecast probability and the preset threshold.
9. The solar flare prediction system based on the fusion of physical parameters and geometric features of active regions according to claim 8, characterized in that, In the prediction model module, the neural network model is a hybrid neural network structure, including a temporal processing branch for processing dynamic evolution features and a fully connected network branch for processing static features; the temporal processing branch includes at least one of a long short-term memory network layer and a gated recurrent unit layer, and is combined with an attention mechanism.
10. The solar flare prediction system based on the fusion of physical parameters and geometric features of active regions according to claim 9, characterized in that, The processing procedure of the timing processing branch includes: The time series of dynamic evolution features are input into the LSTM layer to obtain the hidden state sequence; The hidden state sequence is weighted using an attention mechanism to generate a context vector; The context vector is concatenated with the output vector of the fully connected network branch, and the predicted probability is obtained after processing by the output layer.