A space weather numerical prediction method, system, medium and product

Abstract

A method, system, medium, and product for space weather numerical forecasting are disclosed, relating to the field of data processing technology. This method first preprocesses solar observation data, constructing the initial corona background field through magnetic field extrapolation and velocity inversion. Based on this, ensemble numerical simulations are performed to extract solar wind time-series data and interplanetary propagation ensemble dispersion. Subsequently, magnetospheric numerical simulations are driven using solar wind data as boundary conditions to obtain the magnetospheric background field and error field. The electron radiation belt flux distribution is obtained by combining data assimilation forecasts and used as a driving parameter to couple and generate ionospheric forecast results. Finally, a global quality assessment is performed on the results for each sphere, and by locating error sources and performing local iterative corrections, the end-to-end space weather forecast result is output. Implementing the technical solution provided in this application can improve the accuracy of space weather numerical forecasting under complex space weather events.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a space weather numerical forecasting method, system, medium, and product. Background Technology

[0002] With the increasing frequency of human space activities and the deepening dependence on the space environment, space weather forecasting has become a crucial technical means to ensure the safe operation of spacecraft and the stable functioning of communication and navigation systems. Space weather refers to solar activity and the various physical phenomena and effects it causes in the helio-terrestrial space, encompassing a vast area from the solar surface to the Earth's ionosphere. Accurate space weather forecasts can provide critical early warning information for satellite operations, radio communications, power systems, and more. The accuracy and reliability of these forecasts are of great significance for ensuring the safety of space assets and the stable operation of ground infrastructure.

[0003] Currently, space weather numerical forecasting methods mainly employ stratified forecasting techniques, which involve establishing separate numerical models for different regions of the Sun-Earth space. This forecasting method, based on spatial layer division, achieves end-to-end forecasting from the Sun to the Earth by progressively passing forecast parameters layer by layer. Specifically, it first reconstructs the physical field of solar active regions based on solar observation data, then uses a space propagation model to forecast the evolution of disturbance parameters in the Sun-Earth space. Next, the forecast parameters are input into a near-Earth space environment model to obtain forecast results for the corresponding regions. Finally, it drives downstream space regional models level by level to obtain forecast outputs at each level. This method can cover the main physical processes in the Sun-Earth space and has been widely used in operational space weather forecasting.

[0004] However, due to the complexity of space weather systems and the multi-layered nature of the forecasting chain, the entire forecasting process faces technical challenges in error control. In practical applications, methods based on independent forecasting at each layer are insufficient to effectively suppress the cumulative amplification effect of forecast errors during multi-sphere propagation, and the overall reliability of forecast quality is difficult to guarantee. In particular, when there are deviations in the forecasts of upstream spheres, the lack of error feedback mechanisms due to the independent operation of each sphere model, coupled with the simple sequential propagation method used in existing methods, can easily lead to a significant decrease in the accuracy of downstream sphere forecasts or even systematic deviations, thereby reducing the accuracy of space weather numerical forecasts under complex space weather events. Summary of the Invention

[0005] This application provides a space weather numerical forecasting method, system, medium, and product that can improve the accuracy of space weather numerical forecasting under complex space weather events.

[0006] The first aspect of this application provides a space weather numerical prediction method, comprising: Acquire solar observation data and preprocess the solar observation data to obtain the target solar observation dataset; Based on the target solar observation dataset, magnetic field extrapolation and velocity inversion are performed to construct the initial background field of the solar corona; Using the initial background field of the corona as the initial state, ensemble numerical simulations were performed to obtain interplanetary propagation results, and solar wind time series data and interplanetary propagation ensemble discreteness were extracted based on the interplanetary propagation results. Using the solar wind time series data as upstream boundary data, and based on the interplanetary propagation set discreteness, a magnetosphere numerical simulation is performed to obtain the magnetosphere background field and the magnetosphere background error field. An error covariance matrix is ​​constructed based on the magnetosphere background error field, and data assimilation prediction is performed in conjunction with the magnetosphere background field to obtain the global spatiotemporal distribution of electron radiation band flux. The global spatiotemporal distribution of the electron radiation band flux is used as a coupling driving parameter and coupled with the magnetosphere background field to generate ionospheric prediction results. A global quality assessment is performed on the global spatiotemporal distribution of the magnetosphere background field, the electron radiation band flux, and the ionospheric forecast results. If the global quality assessment does not meet the preset quality standard conditions, the source layer of the error is located and local iterative correction is performed until the preset quality standard conditions are met, and the full-link space weather forecast results are output.

[0007] By adopting the above technical solution, an ensemble simulation method is used to obtain the ensemble dispersion of interplanetary propagation during numerical simulation. This dispersion can quantify the uncertainty of upstream layer forecasts. In subsequent magnetospheric numerical simulations, the magnetospheric background error field is obtained based on this dispersion, thus explicitly transmitting the forecast uncertainty of upstream layers to downstream layers in the form of an error field. Based on this, the error covariance matrix constructed based on the magnetospheric background error field can accurately characterize the spatial distribution characteristics and correlation of forecast errors. This allows for the reasonable allocation of weights between background field forecasts and observational data during data assimilation forecasting, effectively suppressing the impact of upstream layer forecast bias on electron radiation band flux forecasts. Furthermore, by conducting a global quality assessment of the global spatiotemporal distribution of the magnetospheric background field, electron radiation band flux, and ionospheric forecast results, the coordination and reliability of forecast results across all layers can be verified from a full-link perspective. When the assessment does not meet the preset quality standard conditions, by locating the specific error source layer and performing local iterative correction, the inefficiency of rerunning the entire forecast link is avoided, achieving targeted error correction and forecast optimization. This forecasting mechanism, which combines uncertainty quantification and transmission, data assimilation correction, and global quality assessment feedback, can effectively suppress the cumulative amplification effect of forecast errors during multi-sphere transmission, and improve the accuracy of space weather numerical forecasts under complex space weather events.

[0008] Optionally, the collected photosphere magnetic field observation data and solar flare radiation observation data are merged into initial solar observation data; instrument jump points in the initial solar observation data are identified and removed, and aligned with spatiotemporal coordinates to output sample solar observation data; the instrument background noise amplitude and the angle between the solar observation line of sight projection are extracted from the sample solar observation data, and the reciprocal of the data error variance is constructed based on the instrument background noise amplitude and the angle between the solar observation line of sight projection, and the reciprocal is used as the corresponding quality assessment weight; the sample solar observation data and the corresponding quality assessment weight are bound to the elements according to spatial grid points to generate the target solar observation dataset.

[0009] Optionally, the three-dimensional coronal magnetic field is obtained; the quality assessment weights corresponding to each observation data in the target solar observation dataset are used as the underlying boundary constraint weights to perform weighted extrapolation on the three-dimensional coronal magnetic field to obtain the three-dimensional coronal magnetic field distribution; the flux tube expansion factor and the angular distance from the magnetic field line foot point to the coronal hole boundary are extracted from the three-dimensional coronal magnetic field distribution at the source surface; based on the physical mapping relationship between the flux tube expansion factor and the angular distance, the initial solar wind velocity field at the source surface is inverted; the distance decay inverse distribution of the quality assessment weights on the three-dimensional spatial grid is calculated, and the distance decay inverse distribution is used as the error estimate to establish the initial background error field; the three-dimensional coronal magnetic field distribution, the initial solar wind velocity field and the initial background error field are merged to generate the initial coronal background field.

[0010] Optionally, based on the initial background field of the corona, random sampling is performed within a preset perturbation range to generate multiple sets of ensemble perturbation initialization parameters; the ensemble perturbation initialization parameters are applied to the initial background field of the corona to form multiple sets of ensemble simulation initial states, and numerical simulations are performed respectively to obtain interplanetary propagation results corresponding to multiple sets of simulations; the ensemble mean of the solar wind time series data in the interplanetary propagation results is used as the solar wind time series data at the target location; the ensemble standard deviation of the solar wind time series data in the interplanetary propagation results is determined as the interplanetary propagation ensemble dispersion.

[0011] Optionally, the spatial gradient features of the interplanetary propagation set discreteness are extracted, and the magnetospheric computational domain is partitioned based on the spatial gradient features to generate a spatial grid. The solar wind time series data is injected into the spatial grid as an upstream boundary driving parameter, and the electrodynamic parameters between the magnetosphere and the spherical ionosphere are bidirectionally mapped and iterated along the three-dimensional magnetic field line path to evolve and generate the magnetospheric background field. The first-order sensitivity coefficient of the magnetospheric background field relative to the solar wind time series data is extracted, and the interplanetary propagation set discreteness is projected onto each node of the spatial grid in combination with the first-order sensitivity coefficient to construct the magnetospheric background error field.

[0012] Optionally, multi-satellite observation data is acquired and cross-consistency calibration is performed to form a multi-satellite fusion dataset in a unified coordinate system; based on the magnetospheric background error field, the error covariance matrix is ​​constructed; diffusion evolution calculation is performed with the magnetospheric background field as the driving force to obtain the basic forecast state of the spatiotemporal evolution of the electron radiation belt; using the error covariance matrix as the background error covariance and the multi-satellite fusion dataset as the observation input, the basic forecast state is filtered and calculated to complete data assimilation and fusion, and the global spatiotemporal distribution of the electron radiation belt flux is output.

[0013] Optionally, the continuity deviations of physical quantities at the sphere interface are extracted from the global spatiotemporal distribution of the magnetospheric background field, the electron radiation band flux, and the ionospheric forecast results to construct a multi-sphere residual matrix, and the trace of the multi-sphere residual matrix is ​​extracted as a global quality metric. When the global quality metric does not reach the threshold requirement corresponding to the preset quality standard condition, singular value decomposition is performed on the multi-sphere residual matrix to extract the spatial mode corresponding to the maximum singular value, and the physical region mapped by the spatial mode is identified as the error source sphere. For the error source sphere, the interplanetary propagation set discreteness and the magnetospheric background error field are extracted as parameters to be adjusted, and incremental compensation is applied to the parameters to be adjusted along the physical evolution direction to generate a corrected driving boundary field. Using the corrected driving boundary field as the input source, the global spatiotemporal distribution of the magnetospheric background field, the electron radiation band flux, and the ionospheric forecast results are updated until the global quality metric calculated after the update reaches the threshold corresponding to the preset quality standard condition, and the end-to-end space weather forecast result is output.

[0014] In a second aspect, embodiments of this application provide a space weather numerical prediction system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the space weather numerical prediction system to perform the methods described in the first aspect and any possible implementation thereof.

[0015] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a space weather numerical prediction system, cause the space weather numerical prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0016] Fourthly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a space weather numerical prediction system, cause the space weather numerical prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0017] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By adopting the above technical solution, an ensemble simulation method is used to obtain the ensemble dispersion of interplanetary propagation during numerical simulation. This dispersion can quantify the uncertainty of upstream layer forecasts. In subsequent magnetospheric numerical simulations, the magnetospheric background error field is obtained based on this dispersion, thus explicitly transmitting the forecast uncertainty of upstream layers to downstream layers in the form of an error field. Based on this, the error covariance matrix constructed based on the magnetospheric background error field can accurately characterize the spatial distribution characteristics and correlation of forecast errors. This allows for the reasonable allocation of weights between background field forecasts and observational data during data assimilation forecasting, effectively suppressing the impact of upstream layer forecast bias on electron radiation band flux forecasts. Furthermore, by conducting a global quality assessment of the global spatiotemporal distribution of the magnetospheric background field, electron radiation band flux, and ionospheric forecast results, the coordination and reliability of forecast results across all layers can be verified from a full-link perspective. When the assessment does not meet the preset quality standard conditions, by locating the specific error source layer and performing local iterative correction, the inefficiency of rerunning the entire forecast link is avoided, achieving targeted error correction and forecast optimization. This forecasting mechanism, which combines uncertainty quantification and transmission, data assimilation correction, and global quality assessment feedback, can effectively suppress the cumulative amplification effect of forecast errors during multi-sphere transmission, and improve the accuracy of space weather numerical forecasts under complex space weather events. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the space weather numerical forecasting method disclosed in the embodiments of this application; Figure 2 This is another schematic flowchart of the space weather numerical forecasting method disclosed in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of a system provided in an embodiment of this application.

[0019] Explanation of reference numerals in the attached drawings: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0021] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0022] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0023] This application provides a space weather numerical prediction method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a space weather numerical prediction method provided in an embodiment of this application. The method is applied to a system, which refers to a hardware and software integrated platform capable of executing a space weather numerical prediction program. The system can execute a space weather numerical prediction program. The method includes steps 101 to 107, as follows: Step 101: Acquire solar observation data and preprocess the solar observation data to obtain the target solar observation dataset.

[0024] In this embodiment of the application, the target solar observation dataset refers to a standardized solar observation data set after quality control and preprocessing, which includes photosphere magnetic field data and solar flare radiation data. Each data point is accompanied by quality assessment weight information. The dataset has eliminated instrument abnormal jump points and completed spatiotemporal coordinate alignment. For example, in a certain solar observation, the preprocessed dataset includes the magnetic field intensity distribution covering the entire solar surface and the corresponding mass weight distribution.

[0025] Specifically, firstly, photosphere magnetic field observation data and solar flare radiation observation data are collected and merged into initial solar observation data. Then, time series analysis is performed on the initial solar observation data. Instrument jump points are identified by calculating the differential gradient between adjacent time points. When the differential gradient exceeds a preset threshold, it is determined as a jump point and removed. Simultaneously, the spatial coordinates of observations from different instruments are uniformly transformed to achieve spatiotemporal coordinate alignment. Next, the data error variance is calculated based on the instrument background noise amplitude and the angle between the solar observation line of sight projection. The quality assessment weight is obtained by the reciprocal of the error variance. Finally, the preprocessed observation data and the corresponding quality assessment weights are bound to spatial grid points to form a target solar observation dataset containing observation values ​​and quality weights.

[0026] In one possible implementation, the method of obtaining the front and rear background light intensity values ​​of the infrared receiver tube of the device under test when the infrared emitter tube is off, and the total light intensity value when the infrared emitter tube is on, specifically includes steps 1011-1013, as follows: Step 1011: Merge the collected photosphere magnetic field observation data and solar flare radiation observation data into initial solar observation data; identify and remove instrument jump points in the initial solar observation data, and perform alignment processing in conjunction with spatiotemporal coordinates to output sample solar observation data.

[0027] Specifically, the data collected from the photosphere magnetic field observation and solar flare radiation observation are first merged according to timestamps to form initial solar observation data. Then, time series analysis is performed on the initial solar observation data to calculate the differential gradient between adjacent observation values. The differential gradient is compared with a preset threshold; when the absolute value of the differential gradient exceeds the preset threshold, it is identified as an instrument jump point, which is then marked and removed. Next, the time reference base and spatial coordinate reference system of different observation instruments are extracted, and the timestamps of all observation data are uniformly converted to the Coordinate System (UTC), and the spatial coordinates of all observation data are uniformly converted to the heliocentric coordinate system, completing the spatiotemporal coordinate alignment process. Finally, sample solar observation data after jump point removal and spatiotemporal coordinate alignment is output.

[0028] Step 1012: Extract the instrument background noise amplitude and the angle between the solar observation line of sight projection of the sample solar observation data. Construct the reciprocal of the data error variance based on the instrument background noise amplitude and the angle between the solar observation line of sight projection, and use the reciprocal as the corresponding quality assessment weight.

[0029] Specifically, firstly, the instrument background noise amplitude corresponding to each observation point is extracted from the sample solar observation data. This noise amplitude is read from the instrument calibration parameters. Simultaneously, the projection angle from the spatial position of each observation point on the solar surface to the observation line of sight is calculated. This angle is obtained by calculating the heliocentric latitude and longitude coordinates of the observation point and the pointing vector of the observation instrument. Then, the data error variance is calculated by multiplying the square of the instrument background noise amplitude by the secant of the projection angle of the observation line of sight on the solar surface. Next, the reciprocal of the data error variance is calculated, and this reciprocal is used as the quality assessment weight for the corresponding observation point. A one-to-one correspondence is established between the quality assessment weight and the observation data.

[0030] Step 1013: Bind the sample solar observation data and the corresponding quality assessment weights to the spatial grid points to generate the target solar observation dataset.

[0031] Specifically, firstly, a regular spatial grid covering the solar observation area is constructed. This grid uses equally spaced latitude and longitude divisions to spatially partition the solar surface. Then, each observation point in the sample solar observation data is traversed, and its corresponding spatial grid point is determined based on its spatial coordinates. The observed value and corresponding quality assessment weight of the observation point are jointly assigned to that grid point. For cases where the same grid point contains multiple observation points, a weighted average is performed using the quality assessment weights to calculate the comprehensive observed value and comprehensive quality assessment weight of the grid point. For spatial grid points not covered by observation points, spatial interpolation methods are used to obtain the observed values ​​and quality assessment weights from adjacent grid points. Finally, all spatial grid points and their associated observed values ​​and quality assessment weights are organized into a structured data format to generate the target solar observation dataset.

[0032] Step 102: Based on the target solar observation dataset, perform magnetic field extrapolation and velocity inversion to construct the initial background field of the corona.

[0033] In this embodiment of the application, the initial background field of the corona refers to the physical field distribution that describes the initial state of the corona region, including the three-dimensional magnetic field distribution of the corona, the initial velocity field of the solar wind at the source surface, and the corresponding initial background error field. This background field is obtained by extrapolation of physical models based on photosphere observation data, such as the magnetic field intensity distribution at a corona height of 1.5 times the solar radius and the solar wind velocity distribution at a source surface height of 2.5 times the solar radius.

[0034] Specifically, firstly, the photosphere magnetic field data in the target solar observation dataset is extrapolated using a potential source surface model to obtain preliminary extrapolation results of the three-dimensional coronal magnetic field. Then, the quality assessment weights in the target solar observation dataset are used as bottom-level boundary constraint weights to weight and optimize the preliminary extrapolation results, resulting in a three-dimensional coronal magnetic field distribution considering observation quality. Next, based on the three-dimensional coronal magnetic field distribution, the flux tube expansion factor and the angular distance from the magnetic field line footpoint to the coronal hole boundary are extracted. The extracted expansion factor and angular distance are substituted into the empirical velocity model for velocity inversion calculation to obtain the initial solar wind velocity field at the source surface. Finally, the inverse distribution of the distance decay of the quality assessment weights on the three-dimensional spatial grid is calculated. This distribution is used as an error estimate to establish the initial background error field. The three-dimensional coronal magnetic field distribution, the initial solar wind velocity field, and the initial background error field together constitute the initial coronal background field.

[0035] In one possible implementation, the initial background field of the solar corona is constructed by extrapolating the magnetic field and inverting the velocity based on the target solar observation dataset, specifically including steps 1021-1023, as follows: Step 1021: Obtain the three-dimensional magnetic field of the corona; use the quality assessment weights corresponding to each observation data in the target solar observation dataset as the underlying boundary constraint weights, and perform weighted extrapolation on the three-dimensional magnetic field of the corona to obtain the distribution of the three-dimensional magnetic field of the corona.

[0036] Specifically, firstly, based on the photosphere magnetic field data in the target solar observation dataset, a preliminary magnetic field extrapolation calculation is performed using a potential source surface model. The observed magnetic field value is set at the bottom boundary of the photosphere, and the radial magnetic field condition is set at the source surface height. The preliminary extrapolation result of the three-dimensional coronal magnetic field is obtained by solving the Laplace equation. Then, the quality assessment weights corresponding to each observation data in the target solar observation dataset are extracted, and these weights are assigned as bottom boundary constraint weights to the grid nodes of the bottom boundary of the photosphere. Next, a weighted residual function is constructed to calculate the deviation between the preliminary extrapolation result and the observed value at the bottom boundary. The deviation is multiplied by the quality assessment weight and then summed at all boundary nodes to obtain the total weighted residual. Then, an iterative optimization method is used to adjust the three-dimensional coronal magnetic field. In each iteration, the magnetic field distribution is updated to reduce the total weighted residual until the total weighted residual converges to below a preset threshold. Finally, the converged three-dimensional coronal magnetic field distribution is output.

[0037] Step 1022: Extract the flux tube expansion factor and the angular distance from the magnetic field line foot point to the coronal hole boundary of the three-dimensional magnetic field distribution of the corona at the source surface. Based on the physical mapping relationship between the flux tube expansion factor and the angular distance, the initial velocity field of the solar wind at the source surface is obtained by inversion.

[0038] Specifically, firstly, magnetic field data at the source surface height is extracted from the three-dimensional magnetic field distribution of the corona. Regularly distributed starting points are established on the source surface, and magnetic field lines are traced downwards from each starting point to the bottom boundary of the photosphere to determine the location of the magnetic field line foot points. Then, the fluxtube expansion factor is calculated. Magnetic flux per unit area is extracted at the magnetic field line foot points, and the corresponding magnetic field line bundle is traced back to the source surface height. The cross-sectional area of ​​the magnetic field line bundle at the source surface is calculated, and the ratio of the cross-sectional area to the foot point area is used as the fluxtube expansion factor. Next, coronal hole regions in the three-dimensional magnetic field distribution of the corona are identified. Open magnetic field line regions are determined to be coronal holes based on the magnetic field topology. The coronal hole boundary lines are extracted, and the spherical angular distance from each magnetic field line foot point to the nearest coronal hole boundary is calculated. Then, the fluxtube expansion factor and angular distance are substituted into an empirical velocity model. This model uses the product of the negative exponential function of the expansion factor and the exponential function of the angular distance to construct a velocity mapping relationship, and the initial solar wind velocity field at each grid point on the source surface is calculated.

[0039] Step 1023: Calculate the inverse distribution of distance decay of the quality assessment weights on the three-dimensional spatial grid, and use the inverse distribution of distance decay as the error estimate to establish the initial background error field; combine the three-dimensional magnetic field distribution of the corona, the initial velocity field of the solar wind and the initial background error field to generate the initial background field of the corona.

[0040] Specifically, firstly, the quality assessment weights of each grid point at the bottom boundary of the target solar observation dataset are read, and these weights are used as the inverse of the initial error of the bottom boundary. Then, a three-dimensional spatial grid is constructed, which covers the corona region from the bottom boundary of the photosphere to the height of the source surface. Next, the distance attenuation inverse distribution on the three-dimensional spatial grid is calculated. For each three-dimensional spatial grid point, the spatial distance from that grid point to all grid points on the bottom boundary is calculated. The negative exponential function of the distance is used as the attenuation function. The quality assessment weights of each point on the bottom boundary are multiplied by the corresponding attenuation function and then summed over all bottom boundary points to obtain the distance attenuation inverse at that three-dimensional grid point. Then, the distance attenuation inverse distribution is used as the error estimate to establish the initial background error field. Finally, the three-dimensional magnetic field distribution of the corona, the initial velocity field of the solar wind, and the initial background error field are merged according to the spatial grid point positions. Each grid point contains three elements: magnetic field vector, velocity value, and error estimate, to generate the initial background field of the corona.

[0041] Step 103: Perform ensemble numerical simulation with the initial background field of the corona as the initial state to obtain interplanetary propagation results, and extract solar wind time series data and interplanetary propagation ensemble discreteness based on the interplanetary propagation results.

[0042] In this embodiment, the interplanetary propagation ensemble dispersion refers to the degree of dispersion of the prediction results of multiple ensemble simulation members in the process of simulating the interplanetary propagation of solar disturbances. It is quantified by calculating the standard deviation of the ensemble members, reflecting the magnitude of uncertainty in the propagation process of disturbances. For example, for a certain coronal mass ejection event, the ensemble standard deviation of the solar wind speed at the Earth orbit position of thirty ensemble simulations is fifty kilometers per second, which indicates the prediction uncertainty at that position.

[0043] Specifically, firstly, based on the initial background error field of the initial background field of the corona, multiple sets of ensemble perturbation initialization parameters are generated using a random sampling method within a preset perturbation parameter range. These parameters include the velocity perturbation amplitude, azimuth angle perturbation amplitude, and magnetic flux rope strength perturbation amplitude of the coronal mass ejection. Subsequently, based on the three-dimensional magnetic field distribution of the corona and the initial velocity field of the solar wind in the initial background field of the corona, multiple sets of parallel magnetohydrodynamic numerical simulations are performed after superimposing each set of ensemble perturbation initialization parameters to simulate the propagation and evolution of solar perturbations in interplanetary space, obtaining interplanetary propagation results. Then, the time series of solar wind physical parameters at a specified spatial location is extracted from the interplanetary propagation results, and the mean of all ensemble members at each time step is calculated to obtain solar wind time series data. At the same time, the standard deviation of each physical quantity of all ensemble members at each time step is calculated to obtain the interplanetary propagation ensemble dispersion.

[0044] Step 104: Using solar wind time series data as upstream boundary data, and based on the interplanetary propagation set discreteness, perform magnetosphere numerical simulation to obtain the magnetosphere background field and magnetosphere background error field.

[0045] In this embodiment of the application, the magnetospheric background error field refers to the error distribution of each physical quantity in the magnetospheric simulation results. This error distribution is obtained by projecting the interplanetary propagation set discreteness into the magnetospheric computational domain, reflecting the influence of the uncertainty of the upstream solar wind on the accuracy of magnetospheric prediction. For example, the magnetic field strength error near the magnetopause is 20 nanoteslas, which indicates the uncertainty range of the magnetic field strength prediction at that location.

[0046] Specifically, firstly, the spatial gradient characteristics of the interplanetary propagation set discreteness are extracted. A strategy of fine-grid partitioning is used in high-gradient regions and coarse-grid partitioning in low-gradient regions to adaptively partition the magnetospheric computational domain. Then, solar wind time-series data is injected as upstream boundary conditions into the boundary nodes of the magnetospheric computational domain. Multiple Riemannian algorithms are used to solve the magnetohydrodynamic equations, with the Riemannian algorithm type adaptively selected for different plasma characteristic regions. Simultaneously, a spherical shell ionospheric approximation model is constructed, and the magnetospheric electric field, convection velocity, and field-directed current are bidirectionally mapped to the ionospheric along the three-dimensional magnetic field lines, completing the coupled calculation of the magnetospheric and ionospheric layers. The magnetospheric background field is obtained through time-progressive iterative evolution. Next, the first-order sensitivity coefficient of the magnetospheric background field to the solar wind boundary conditions is calculated. The interplanetary propagation set discreteness is then projected onto the magnetospheric grid nodes using this sensitivity coefficient to obtain the magnetospheric background error field.

[0047] Step 105: Construct an error covariance matrix based on the magnetospheric background error field, and combine it with the magnetospheric background field to perform data assimilation prediction, thereby obtaining the global spatiotemporal distribution of electron radiation band flux.

[0048] In this embodiment, the error covariance matrix refers to a matrix that describes the statistical characteristics of the background error of the electron radiation zone phase spatial density. The matrix elements represent the covariance of the error between different spatial locations or different energy channels. This matrix is ​​constructed based on the background error field of the magnetosphere through coordinate transformation and statistical modeling. For example, the error covariance matrix in the first adiabatic invariant coordinate system can describe the correlation of the electron phase spatial density error at different magnetic shell locations.

[0049] Specifically, the process begins by integrating electron flux observation data from multiple satellite series. Physical conjugate calibration is performed on the observation data from different satellites to quantify flux adjustment factors between satellites and remove outliers, thus completing multi-satellite data fusion. Next, the magnetospheric background field and satellite electron flux observation data are read in, and the magnetic field model is used to calculate the three adiabatic invariants, converting the electron flux into phase space density with adiabatic invariants as coordinates. Then, based on the magnetospheric background error field, the error distribution in magnetospheric space is mapped to the phase space density coordinate system through coordinate transformation, establishing an error covariance matrix. Then, driven by the magnetospheric background field, the radial diffusion coefficient and local diffusion coefficient of magnetospheric waves are quantified, and the multidimensional diffusion equation of the electron radiation belt is solved to obtain the basic forecast state. Finally, the Kalman filter algorithm is used, with the error covariance matrix as the background error covariance input. Multi-satellite observation data and the basic forecast state are fused, and data assimilation forecasting is completed through a prediction update iteration process to obtain the global spatiotemporal distribution of electron radiation belt flux.

[0050] In one possible implementation, an error covariance matrix is ​​constructed based on the magnetospheric background error field, and data assimilation prediction is performed in conjunction with the magnetospheric background field to obtain the global spatiotemporal distribution of electron radiation band flux. Specifically, this includes steps 1051-1053, as follows: Step 1051: Acquire multi-satellite observation data and perform cross-consistency calibration to form a multi-satellite fusion dataset in a unified coordinate system.

[0051] Specifically, the process begins by acquiring electron radiation belt observation data from multiple in-orbit satellites. This data includes electron flux, energy spectrum, and timestamp information measured by different satellites at their respective orbital positions. Next, cross-consistency calibration is performed, identifying spatiotemporally overlapping observation points in the satellite orbital intersection region. Differences in observation values ​​from different satellites at the same location and time are calculated, and a systematic bias correction factor is constructed. The observation values ​​of each satellite are multiplied by the corresponding correction factor to eliminate instrument calibration differences. Then, a unified spatial coordinate system is implemented, transforming the different coordinate systems used by each satellite, such as geocentric inertial coordinates and geomagnetic coordinates, to a unified magnetospheric coordinate system. A magnetic field model is used to calculate the magnetic shell parameters and magnetic local time parameters of each observation point as standardized spatial coordinates. Finally, a unified time reference is established, converting the timestamps of all observation data to Coordinated Universal Time (UTC). Finally, the calibrated and coordinate-unified multi-satellite observation data is integrated according to a spatiotemporal grid to form a multi-satellite fusion dataset.

[0052] Step 1052: Based on the magnetospheric background error field, construct the error covariance matrix; drive the diffusion evolution calculation with the magnetospheric background field to obtain the basic prediction state of the spatiotemporal evolution of the electron radiation belt.

[0053] Specifically, firstly, error estimates for each spatial grid point are extracted from the magnetospheric background error field to construct an error covariance matrix. The error correlation between any two spatial grid points is calculated, and the correlation coefficient is determined using a distance decay function. This function sets the correlation length scale to twice the distance between the magnetospheric shell parameters. The error estimates for each grid point are multiplied by the correlation coefficient to form the error covariance matrix. Subsequently, the magnetic field configuration, plasma density distribution, and geomagnetic activity parameters in the magnetospheric background field are extracted and input into the electron radiation belt diffusion evolution model as driving conditions. Next, diffusion evolution calculations are performed, and an evolution equation for the electron phase spatial density is established in the magnetospheric shell coordinate system. This equation includes radial diffusion terms and local acceleration terms. The radial diffusion coefficient is determined based on the geomagnetic activity parameters in the magnetospheric background field, and the local acceleration term is calculated based on plasma wave conditions. Then, the evolution equation is integrated over time using the finite difference method, advancing from the current time to the predicted time, to calculate the temporal evolution of the electron phase spatial density at each magnetospheric shell location. Finally, the electron phase spatial density is converted into electron flux, and the basic predicted state of the spatiotemporal evolution of the electron radiation belt is output.

[0054] Step 1053: Using the error covariance matrix as the background error covariance and the multi-satellite fusion dataset as the observation input, filter the basic forecast state to complete the data assimilation and fusion, and output the global spatiotemporal distribution of electron radiation band flux.

[0055] Specifically, firstly, the error covariance matrix is ​​used as the background error covariance matrix, which quantifies the uncertainty and spatial correlation of the basic forecast state at each spatial grid point. Then, an observation operator is constructed to map the electron flux in the model space to the observation space corresponding to the multi-satellite fusion dataset, establishing a mapping relationship based on satellite orbital positions and instrument response functions. Next, the observation error covariance matrix is ​​calculated, extracting instrument errors and representative errors for each observation point from the multi-satellite fusion dataset to construct a diagonal matrix form of the observation error covariance. Then, an ensemble Kalman filter is used for filtering calculations, generating ensemble members of multiple basic forecast states. The ensemble mean and ensemble spread are calculated, and the ensemble members are projected onto the observation space using the observation operator. The innovation in the observation space, i.e., the difference between the observed and forecast values, is calculated. The Kalman gain matrix is ​​determined based on the background error covariance and the observation error covariance. The innovation is multiplied by the Kalman gain and then superimposed onto the basic forecast state to complete the state update. Finally, the state after assimilation analysis is interpolated and extended on the global magnetospheric spatial grid to output the global spatiotemporal distribution of electron radiation band flux.

[0056] Step 106: Using the global spatiotemporal distribution of electron radiation band flux as a coupling driving parameter, and combining it with the magnetosphere background field for coupling calculation, ionospheric prediction results are generated.

[0057] In the embodiments of this application, the coupling driving parameter refers to the input physical quantity that drives the operation of the downstream layer numerical model. Here, it specifically refers to the global spatiotemporal distribution of electron radiation band flux. This parameter provides information on the particle source of electrons falling into the ionosphere. For example, the electron flux of one megaelectron volt at the position of the 3.5 magnetic shell in a certain local time sector is 10,000 per square centimeter per second per steradian per megaelectron volt. This flux will drive the calculation of the ionospheric electron density.

[0058] Specifically, the system first integrates observation data from both ground-based and space-based Global Navigation Satellite Systems (GNSS), performing standardization preprocessing on both. Then, a multi-parameter coupled physical model of the thermospheric ionospheric dynamo is constructed. The global spatiotemporal distribution of electron radiation band flux is used as the particle deposition source term in the ionospheric electron density equation, while the electric field and field current in the magnetosphere background field are used as driving terms in the ionospheric dynamo equation. The thermospheric neutral parameter equation, the ionospheric parameter equation, and the ionospheric dynamo parameter equation are solved simultaneously to achieve multi-parameter bidirectional coupled calculation. Next, a three-dimensional variational assimilation algorithm and a Kalman filter algorithm are integrated to assimilate the preprocessed observation data into the coupled physical model's calculation results. High-resolution grids are used for regional assimilation in China, while low-resolution grids are used for global assimilation, constructing a dual-scale digital ionosphere. Finally, multi-time-series forecasts of ionospheric electron density, temperature, wind speed, electric field, and potential are output to generate ionospheric forecast results.

[0059] Step 107: Perform a global quality assessment on the global spatiotemporal distribution of the magnetosphere background field, the electron radiation band flux, and the ionospheric forecast results. If the global quality assessment does not meet the preset quality standard conditions, locate the error source layer and perform local iterative correction until the preset quality standard conditions are met, and output the full-link space weather forecast results.

[0060] In this embodiment of the application, global quality assessment refers to the process of verifying the consistency of multi-sphere forecast results. The global rationality of the forecast results is measured by quantifying the continuity deviation of physical quantities at the interface of the spheres. The assessment uses the trace of the multi-sphere residual matrix as the quality metric. For example, when the magnetic field strength deviation at the interface between the magnetosphere and the radiation belt is five nanotesla and the electron flux deviation at the interface between the radiation belt and the ionosphere is ten percent, the global quality metric obtained by comprehensive calculation is used to determine whether the forecast quality meets the standard.

[0061] Specifically, firstly, the global spatiotemporal distribution of the magnetospheric background field and electron radiation band flux, as well as the physical quantities at the interfaces of the ionospheric forecast results, are extracted. The differences between the corresponding physical quantities at the interfaces of the upstream and downstream spheres are calculated to construct a multi-sphere residual matrix. Then, the trace of the multi-sphere residual matrix is ​​calculated, and the trace value is used as the global quality metric. It is then determined whether the metric meets the preset quality standard conditions. If the conditions are met, the full-link space weather forecast result is directly output. If the conditions are not met, singular value decomposition is performed on the multi-sphere residual matrix, and the left singular vector corresponding to the largest singular value is extracted. The source sphere of the error is located based on the position of the element with the largest component in the left singular vector. Next, the parameters to be adjusted are extracted from the forecast results corresponding to the source sphere of the error, and incremental compensation is applied to the parameters to be adjusted along the physical evolution direction. The numerical simulation and data assimilation process of the source sphere of the error and its downstream spheres are re-executed to complete the local iterative correction. The global quality assessment and local iterative correction process are repeated until the global quality metric meets the preset quality standard conditions, and the full-link space weather forecast result is output.

[0062] In one possible implementation, a global quality assessment is performed on the global spatiotemporal distribution of the magnetospheric background field, the electron radiation band flux, and the ionospheric forecast results. If the global quality assessment does not meet the preset quality standard conditions, the source layer of the error is located and local iterative correction is performed until the preset quality standard conditions are met. Finally, the full-link space weather forecast result is output. This specifically includes steps 1071-1074, as follows: Step 1071: Extract the global spatiotemporal distribution of the magnetosphere background field and electron radiation band flux, and the continuity deviation of physical quantities at the sphere interface of the ionospheric prediction results. Construct a multi-sphere residual matrix and extract the trace of the multi-sphere residual matrix as a global quality metric.

[0063] Specifically, the process first identifies the interfaces between the global spatiotemporal distribution of the magnetospheric background field, electron radiation belt flux, and ionospheric prediction results, including the interfaces between the magnetosphere and the ionosphere, and between the electron radiation belt and the ionosphere. Then, physical quantities at the interfaces of each layer are extracted, including parameters such as magnetic field strength, electric field strength, plasma density, and electron flux. Corresponding physical quantities are extracted from the upper and lower layers at the same spatial location at the interface. Next, the continuity deviation of the physical quantities is calculated by subtracting the physical quantities of the upper and lower layers. This difference is then normalized by dividing the difference by the characteristic scale of the physical quantity to obtain a dimensionless continuity deviation. Then, a multi-layer residual matrix is ​​constructed, arranging the continuity deviations of various physical quantities at all interface grid points into rows and columns, with the matrix rows corresponding to the spatial locations of the interfaces and the columns corresponding to different physical quantity types. Finally, the trace of the multi-layer residual matrix is ​​extracted, and the diagonal elements of the matrix are summed to obtain the global mass metric.

[0064] Step 1072: When the global quality metric does not meet the threshold requirement corresponding to the preset quality standard condition, perform singular value decomposition on the multi-concentric residual matrix, extract the spatial mode corresponding to the maximum singular value, and identify the physical region mapped by the spatial mode as the error source concentricity.

[0065] Specifically, firstly, it is determined whether the global quality metric value meets the threshold requirement corresponding to the preset quality standard condition, and the global quality metric value is compared with the preset threshold. When the global quality metric value does not meet the threshold, singular value decomposition is performed on the multi-concentric residual matrix, decomposing the matrix into the product of the left singular vector matrix, the singular value diagonal matrix, and the right singular vector matrix. Then, the maximum singular value in the singular value diagonal matrix is ​​extracted, which corresponds to the error mode that contributes the most in the residual matrix. Next, the left and right singular vectors corresponding to the maximum singular value are extracted. The left singular vector represents the spatial modal distribution, and the right singular vector represents the physical quantity type weight. Then, the left singular vector is mapped back to the spatial coordinate grid of the concentric interface, and the spatial position with the largest amplitude of the left singular vector is identified. The corresponding physical concentric layer is determined according to the attribution of this position in the concentric structure. Finally, this physical concentric layer is identified as the error source layer, and the layer is marked as needing parameter adjustment.

[0066] Step 1073: For the error source layer, extract the interplanetary propagation set discreteness and magnetospheric background error field of the preceding correlation as parameters to be tuned, apply incremental compensation to the parameters to be tuned along the physical evolution direction, and generate the corrected driving boundary field.

[0067] Specifically, firstly, based on the identified error source spheres, the preceding associated driving boundary conditions are traced. For magnetospheric error sources, the interplanetary propagation set discreteness is extracted as a parameter to be tuned; for electron radiation belt or ionospheric error sources, the magnetospheric background error field is extracted as a parameter to be tuned. Then, the sensitivity coefficients of the parameters to be tuned are calculated, and the adjoint method is used to trace back along the physical evolution direction to calculate the partial derivatives of the sphere interface residuals with respect to the parameters to be tuned. Next, parameter increments are constructed based on the sensitivity coefficients and the multi-sphere residual matrix. The gradient direction is obtained by multiplying the residual vector with the transpose of the sensitivity matrix, and the magnitude of the parameter increment is determined along the negative gradient direction. Then, increment compensation is applied to the parameters to be tuned by superimposing the calculated parameter increments onto the original parameters to be tuned, updating the values ​​of the interplanetary propagation set discreteness or the magnetospheric background error field. Finally, the updated parameters are re-inputted into the sphere evolution model to generate the corrected driving boundary field.

[0068] Step 1074: Using the corrected driving boundary field as the input source, update the global spatiotemporal distribution of the magnetospheric background field, the electron radiation band flux, and the ionospheric forecast results until the global mass metric value calculated after the update reaches the threshold corresponding to the preset mass standard conditions, and output the full-link space weather forecast results.

[0069] Specifically, firstly, the modified driving boundary field is used as the input source to update the initial or boundary conditions of the interplanetary propagation model; then, the magnetosphere evolution model is rerun, driven by the updated interplanetary parameters, to calculate the new magnetosphere background field, including the magnetosphere magnetic field, plasma distribution, and geomagnetic activity index; next, using the updated magnetosphere background field as the driving condition, the electron radiation belt diffusion evolution model and data assimilation system are rerun to calculate the new global spatiotemporal distribution of electron radiation belt flux; finally, using the updated magnetosphere parameters and electron deposition flux as input, the ionospheric model is rerun. The new ionospheric forecast results are calculated, including electron density, conductivity, and electric field distribution. Then, the continuity deviation of physical quantities at the interface of each layer after the update is extracted, the multi-layer residual matrix is ​​reconstructed, and the new global mass metric is calculated. Next, it is determined whether the new global mass metric meets the threshold corresponding to the preset mass standard conditions. If it does not meet the threshold, the process returns to step 1072 to continue iterative optimization. If it does meet the threshold, the iteration is terminated. Finally, the global spatiotemporal distribution of the magnetospheric background field and electron radiation band flux, as well as the ionospheric forecast results that meet the mass standards, are output and merged into the full-link space weather forecast results.

[0070] In the above embodiments, the solar wind propagation forecasting function from the corona to interplanetary space was realized through a single deterministic numerical simulation. To further quantify the uncertainty propagation characteristics during propagation, improve the reliability assessment capability of magnetosphere-driven boundary data, and reduce the impact of initial condition errors and model parameter uncertainties on the accuracy of downstream magnetosphere forecasts, this application also provides a space weather numerical forecasting method. This method generates multiple sets of simulation initial states by applying ensemble perturbation parameters to the initial background field of the corona, performs statistical analysis on the interplanetary propagation results to extract the ensemble mean and dispersion characteristics of the solar wind time series data, and projects the propagation uncertainty onto the magnetosphere spatial grid using a sensitivity coefficient to construct a magnetosphere background field containing error field information. This enables the system to more accurately characterize the sources of uncertainty in space weather forecasting and their propagation patterns across multiple spheres, achieving a quantitative assessment of forecast reliability. The following section combines... Figure 2 The space weather numerical prediction method in the embodiments of this application is described as follows: Please see Figure 2 This is another flowchart illustrating a space weather numerical forecasting method in an embodiment of this application.

[0071] Step 201: Based on the initial background field of the corona, randomly sample within the preset perturbation range to generate multiple sets of set perturbation initialization parameters.

[0072] Specifically, firstly, the types of physical quantities that need to be perturbed in the initial background field of the corona are determined, including parameters such as the coronal magnetic field, plasma density, temperature, and solar wind speed. Then, a preset perturbation range is determined based on the initial background error field, and the error estimates of each physical quantity are extracted as the upper limit of the perturbation amplitude. Next, random sampling is performed within the preset perturbation range, and a Gaussian random number generator is used to generate perturbation values ​​that follow a normal distribution for each physical quantity, with a perturbation mean of zero and a perturbation standard deviation as the error estimate. Then, multiple sets of ensemble perturbation initialization parameters are generated, with the number of ensemble members set to a preset value, such as fifty or one hundred. Perturbation parameters are generated independently for each ensemble member at all spatial grid points, forming multiple complete sets of perturbation parameters. Finally, the generated multiple sets of ensemble perturbation initialization parameters are stored as parameter matrices, with the matrix rows corresponding to the ensemble member numbers and the matrix columns corresponding to the spatial grid points and physical quantity types.

[0073] Step 202: Apply ensemble perturbation initialization parameters to the initial background field of the corona to form multiple sets of ensemble simulation initial states, and perform numerical simulations respectively to obtain the interplanetary propagation results corresponding to the multiple sets of simulations.

[0074] Specifically, firstly, ensemble perturbation initialization parameters are applied to the initial background field of the corona. The perturbation parameters of each ensemble member are iterated, and the perturbation values ​​are superimposed on the physical quantities of the corresponding grid points of the initial background field of the corona to generate multiple sets of initial states for ensemble simulation. Subsequently, numerical simulations are performed on each set of initial states. A set of governing equations is established using a magnetohydrodynamic numerical model, which includes equations for mass conservation, momentum conservation, energy conservation, and magnetic field evolution. Next, the simulation time step and spatial grid are set. The time step is determined according to the Courant condition to ensure numerical stability, and the spatial grid extends from the corona floor boundary to the target location, such as Earth's orbit. Then, time-progression calculations are performed in parallel for each ensemble member. The governing equations are solved in each time step, and spatial discretization is performed using the finite volume method. Time integration is performed using the Runge-Kutta method, progressively advancing from the initial moment to the simulation termination moment. Finally, the simulation outputs of all ensemble members are collected, and the solar wind parameters at each moment at the target location are extracted to form multiple sets of interplanetary propagation results corresponding to the simulation.

[0075] Step 203: Use the set mean of the solar wind time series data in the interplanetary propagation results as the solar wind time series data at the target location.

[0076] Specifically, firstly, solar wind parameters at the target location are extracted from interplanetary propagation results. For each ensemble member, time series of physical quantities such as solar wind velocity, density, temperature, and magnetic field components are extracted at the target location. Then, the ensemble mean is calculated for each physical quantity. At each time step, all ensemble members are traversed, and the sum of the values ​​of the same physical quantity for all ensemble members at that time is divided by the total number of ensemble members to obtain the ensemble mean of that physical quantity at that time. Next, the ensemble means at each time step are arranged in chronological order to form a time series of ensemble mean values ​​for each physical quantity of the solar wind. Then, the calculated time series of ensemble mean values ​​is used as the solar wind time series data at the target location. This data represents the optimal prediction result after comprehensively considering the uncertainty of the initial state. Finally, the solar wind time series data at the target location is output in a standard format, including timestamps and corresponding values ​​of solar wind velocity, density, temperature, magnetic field components, and other parameters.

[0077] Step 204: Determine the set standard deviation of the solar wind time series data in the interplanetary propagation results as the interplanetary propagation set dispersion.

[0078] Specifically, firstly, solar wind time-series data for all ensemble members at the target location are extracted from the interplanetary propagation results, obtaining parameter values ​​such as solar wind velocity, density, temperature, and magnetic field components for each ensemble member at each time step. Then, the ensemble standard deviation is calculated for each physical quantity. At each time step, all ensemble members are traversed, and the variance of that physical quantity across all ensemble members is calculated. Specifically, the value of each ensemble member is subtracted from the ensemble mean, the square is calculated, the sum is taken over all members, divided by the total number of ensemble members minus one, and the square root of the result is taken to obtain the ensemble standard deviation. Next, the ensemble standard deviations at each time step are arranged in chronological order to form a time series of ensemble standard deviations for each solar wind physical quantity. Then, the calculated time series of ensemble standard deviations is determined as the interplanetary propagation ensemble dispersion, which has different values ​​for different physical quantities and at different times. Finally, the interplanetary propagation ensemble dispersion is output and stored, serving as input for uncertainty information of the driving boundary in subsequent magnetosphere simulations.

[0079] In one possible implementation, after determining the ensemble standard deviation of the solar wind time series data in the interplanetary propagation results as the ensemble dispersion of the interplanetary propagation, the implementation further includes steps 2041-2043, as follows: Step 2041: Extract the spatial gradient features of the interplanetary propagation set discreteness, divide the magnetosphere computational domain according to the spatial gradient features, and generate a spatial grid.

[0080] Specifically, firstly, the spatial distribution data of the interplanetary propagation set discreteness is extracted, which includes the discreteness values ​​of various spatial locations in the upstream region of the solar wind. Then, the spatial gradient characteristics are calculated by taking the partial derivatives of the discreteness field along the three coordinate directions, and the square root of the sum of the squares of the partial derivatives in the three directions is used to obtain the gradient magnitude. High-gradient regions with gradient magnitudes exceeding a preset threshold are marked. Next, the magnetospheric computational domain is partitioned based on the spatial gradient characteristics. The magnetospheric computational domain is initially divided into a basic grid. In high-gradient regions, the basic grid is recursively refined, dividing the grid cells into eight sub-cells using an octree structure. The refinement level is determined based on the gradient magnitude, while the basic grid scale is maintained in low-gradient regions. Then, a spatial grid topology is generated, recording the spatial coordinates, size, and adjacency relationships of each grid cell, and constructing coordinate arrays for grid nodes and grid centers. Finally, a complete spatial grid data structure is output, covering the entire magnetospheric computational domain from the upstream boundary of the solar wind to the far end of the magnetotail.

[0081] Step 2042: Inject solar wind time series data as upstream boundary driving parameters into the spatial grid, and perform bidirectional mapping and iteration of electrodynamic parameters between the magnetosphere and the spherical shell ionosphere along the three-dimensional magnetic field line path to evolve and generate the magnetosphere background field.

[0082] Specifically, firstly, solar wind time-series data is injected into the spatial grid as upstream boundary driving parameters, and time-varying boundary conditions for parameters such as solar wind velocity, density, temperature, and magnetic field are set at the upstream boundary of the spatial grid. Then, a set of magnetohydrodynamic governing equations for the magnetosphere is established, and the mass, momentum, energy, and magnetic field evolution equations are discretized and solved on each cell of the spatial grid. Next, magnetosphere-ionosphere coupling calculations are performed, constructing a spherical ionosphere model with an ionospheric region having an ionospheric height of 100 km to 500 km, and the conductivity distribution on the ionospheric shell is calculated. Finally, a mapping relationship is established along the three-dimensional magnetic field line path, from the magnetic field... The layered grid nodes trace magnetic field lines to the ionospheric shell. At the foot of the magnetic field lines, the field current and electric field parameters of the magnetosphere are extracted and mapped to the ionosphere to calculate the ionospheric potential distribution. The ionospheric potential is then mapped back to the magnetosphere along the magnetic field lines to update the magnetosphere electric field. Next, bidirectional mapping and iteration are performed, and the parameter mapping process from the magnetosphere to the ionosphere and from the ionosphere to the magnetosphere is repeated. Each iteration updates the magnetosphere background field until the electrodynamic parameters of the magnetosphere and the ionosphere satisfy the continuity condition at the foot of the magnetic field lines and the iteration residual is less than a preset threshold. Finally, the evolved magnetosphere background field is output, which contains the complete physical parameter distribution of all nodes in the spatial grid.

[0083] Step 2043: Extract the first-order sensitivity coefficient of the magnetosphere background field relative to the solar wind time series data, and combine the first-order sensitivity coefficient to project the interplanetary propagation set discreteness onto each node of the spatial grid to construct the magnetosphere background error field.

[0084] Specifically, firstly, the first-order sensitivity coefficients of the magnetosphere background field relative to the solar wind time series data are extracted. A perturbation analysis method is used to apply unit perturbations to the upstream boundary parameters of the solar wind, and the magnetosphere background field is recalculated. The change in magnetosphere parameters after perturbation is divided by the perturbation of the solar wind parameters to obtain the first-order sensitivity coefficients. Sensitivity coefficients are calculated separately for each component of the solar wind velocity, density, temperature, and magnetic field. Then, error projection is performed using the first-order sensitivity coefficients. At each node of the spatial grid, the sensitivity coefficients for each solar wind parameter are extracted. The dispersion values ​​of the corresponding solar wind parameters are read from the interplanetary propagation set discreteness. The sensitivity coefficients are multiplied by the dispersion values ​​to obtain the magnetosphere parameter error components caused by the solar wind parameter. Next, the error components of all solar wind parameters are synthesized. Assuming that the errors of different solar wind parameters are independent, the square root of the sum of the squares of each error component is taken to obtain the total error estimate. Then, the error estimates of all spatial grid nodes are organized into a field distribution to construct the magnetosphere background error field. Finally, the magnetosphere background error field is output. This error field has the same spatial grid structure as the magnetosphere background field, providing corresponding error estimation information for the physical parameters of each node.

[0085] The following describes a space weather numerical prediction system according to an embodiment of the present invention from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of the structure of a space weather numerical prediction system in an embodiment of this application.

[0086] It should be noted that, Figure 3 The structure of a space weather numerical prediction system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0087] like Figure 3 As shown, a space weather numerical forecasting system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on a program stored in a Read-Only Memory (ROM) 302 or a program loaded from a storage section 308 into a Random Access Memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0088] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0089] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0090] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0091] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0092] Specifically, a space weather numerical forecasting system according to this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements a space weather numerical forecasting method provided in the above embodiment.

[0093] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in a space weather numerical prediction system described in the above embodiments; or it may exist independently and not assembled into the space weather numerical prediction system. The storage medium carries one or more computer programs, which, when executed by a processor of the space weather numerical prediction system, cause the space weather numerical prediction system to implement the space weather numerical prediction method based on encrypted data transmission via the Internet of Things provided in the above embodiments.

Claims

1. A method for space weather numerical prediction, characterized by, The method includes: Acquire solar observation data and preprocess the solar observation data to obtain the target solar observation dataset; Based on the target solar observation dataset, magnetic field extrapolation and velocity inversion are performed to construct the initial background field of the solar corona; Using the initial background field of the corona as the initial state, ensemble numerical simulations were performed to obtain interplanetary propagation results, and solar wind time series data and interplanetary propagation ensemble discreteness were extracted based on the interplanetary propagation results. Using the solar wind time series data as upstream boundary data, and based on the interplanetary propagation set discreteness, a magnetosphere numerical simulation is performed to obtain the magnetosphere background field and the magnetosphere background error field. An error covariance matrix is ​​constructed based on the magnetosphere background error field, and data assimilation prediction is performed in conjunction with the magnetosphere background field to obtain the global spatiotemporal distribution of electron radiation band flux. The global spatiotemporal distribution of the electron radiation band flux is used as a coupling driving parameter and coupled with the magnetosphere background field to generate ionospheric prediction results. A global quality assessment is performed on the global spatiotemporal distribution of the magnetosphere background field, the electron radiation band flux, and the ionospheric forecast results. If the global quality assessment does not meet the preset quality standard conditions, the source layer of the error is located and local iterative correction is performed until the preset quality standard conditions are met, and the full-link space weather forecast results are output.

2. The method of claim 1, wherein, The process of acquiring solar observation data and preprocessing the solar observation data to obtain the target solar observation dataset includes: The collected photosphere magnetic field observation data and solar flare radiation observation data were combined into initial solar observation data; The instrument jump points in the initial solar observation data are identified and removed, and then aligned with spatiotemporal coordinates to output sample solar observation data. The instrument background noise amplitude and the angle between the solar observation line projection and the sample solar observation data are extracted. Based on the instrument background noise amplitude and the angle between the solar observation line projection and the angle, the reciprocal of the data error variance is constructed, and the reciprocal is used as the corresponding quality assessment weight. The sample solar observation data and corresponding quality assessment weights are bound to spatial grid points to generate the target solar observation dataset.

3. The method of claim 1, wherein, The process of extrapolating the magnetic field and inverting the velocity based on the target solar observation dataset to construct the initial background field of the corona includes: Obtain the three-dimensional magnetic field of the solar corona; The quality assessment weights corresponding to each observation data in the target solar observation dataset are used as the underlying boundary constraint weights to perform weighted extrapolation of the three-dimensional coronal magnetic field, thereby obtaining the three-dimensional coronal magnetic field distribution. The flux tube expansion factor and the angular distance from the foot of the magnetic field line to the coronal hole boundary are extracted from the three-dimensional magnetic field distribution of the corona at the source surface. Based on the physical mapping relationship between the flux tube expansion factor and the angular distance, the initial velocity field of the solar wind at the source surface is obtained by inversion. Calculate the inverse distance decay distribution of the quality assessment weights on a three-dimensional spatial grid, and use the inverse distance decay distribution as an error estimate to establish an initial background error field. The three-dimensional magnetic field distribution of the corona, the initial velocity field of the solar wind, and the initial background error are combined to generate the initial background field of the corona.

4. The method of claim 1, wherein, The process of performing ensemble numerical simulations using the initial background field of the corona as the initial state to obtain interplanetary propagation results, and extracting solar wind time-series data and interplanetary propagation ensemble discreteness based on the interplanetary propagation results, includes: Based on the initial background field of the corona, random sampling is performed within a preset perturbation range to generate multiple sets of set perturbation initialization parameters; The ensemble perturbation initialization parameters are applied to the initial background field of the corona to form multiple sets of ensemble simulation initial states, and numerical simulations are performed respectively to obtain the interplanetary propagation results corresponding to the multiple sets of simulations; The set mean of the solar wind time series data in the interplanetary propagation results is taken as the solar wind time series data at the target location; The set standard deviation of the solar wind time series data in the interplanetary propagation results is determined as the set discreteness of the interplanetary propagation.

5. The method according to claim 4, characterized in that, The process of using the solar wind time series data as upstream boundary data and performing magnetosphere numerical simulation based on the interplanetary propagation set discreteness to obtain the magnetosphere background field and magnetosphere background error field includes: Extract the spatial gradient features of the interplanetary propagation set discreteness, and divide the magnetosphere computational domain according to the spatial gradient features to generate a spatial grid; The solar wind time series data is injected into the spatial grid as an upstream boundary driving parameter, and the electrodynamic parameters between the magnetosphere and the spherical shell ionosphere are bidirectionally mapped and iterated along the three-dimensional magnetic field line path to generate the magnetosphere background field. The first-order sensitivity coefficient of the magnetosphere background field relative to the solar wind time series data is extracted, and the discretization of the interplanetary propagation set is projected onto each node of the spatial grid in combination with the first-order sensitivity coefficient to construct the magnetosphere background error field.

6. The method according to claim 1, characterized in that, The process of constructing an error covariance matrix based on the magnetosphere background error field, and combining it with the magnetosphere background field for data assimilation prediction, yields the global spatiotemporal distribution of electron radiation band flux, including: Acquire multi-satellite observation data and perform cross-consistency calibration to form a multi-satellite fusion dataset in a unified coordinate system; Based on the background error field of the magnetic layer, the error covariance matrix is ​​constructed; Diffusion evolution calculations were performed using the aforementioned magnetosphere background field as the driving force to obtain the basic predicted state of the spatiotemporal evolution of the electron radiation belt; Using the error covariance matrix as the background error covariance and the multi-satellite fusion dataset as the observation input, the basic forecast state is filtered and calculated to complete data assimilation and fusion, and the global spatiotemporal distribution of the electron radiation band flux is output.

7. The method according to claim 1, characterized in that, The process involves a global quality assessment of the magnetosphere background field, the global spatiotemporal distribution of the electron radiation band flux, and the ionospheric forecast results. If the global quality assessment does not meet preset quality standard conditions, the source layer of the error is located and local iterative corrections are performed until the preset quality standard conditions are met. Finally, a full-link space weather forecast result is output, including: The global spatiotemporal distribution of the magnetosphere background field, the electron radiation band flux, and the continuity deviation of physical quantities at the sphere interface of the ionospheric prediction results are extracted to construct a multi-sphere residual matrix, and the trace of the multi-sphere residual matrix is ​​extracted as a global quality metric. When the global quality metric does not reach the threshold requirement corresponding to the preset quality standard condition, singular value decomposition is performed on the multi-concentric residual matrix to extract the spatial mode corresponding to the maximum singular value, and the physical region mapped by the spatial mode is identified as the error source concentric layer. For the aforementioned error source layer, the interplanetary propagation set discreteness and magnetosphere background error field of the preceding correlation are extracted as parameters to be tuned. Incremental compensation is applied to the parameters to be tuned along the physical evolution direction to generate the corrected driving boundary field. Using the corrected driving boundary field as the input source, the global spatiotemporal distribution of the magnetospheric background field, the electron radiation band flux, and the ionospheric forecast results are updated until the global mass metric value calculated after the update reaches the threshold corresponding to the preset mass standard condition, and the full-link space weather forecast results are output.

8. A space weather numerical prediction system, characterized in that, The space weather numerical prediction system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the space weather numerical prediction system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on a space weather numerical prediction system, the space weather numerical prediction system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on a space weather numerical prediction system, it causes the space weather numerical prediction system to perform the method as described in any one of claims 1-7.

Images