Method and system for constructing a space particle environment model

Abstract

This invention provides a method for constructing a space particle environment model. The method involves acquiring a first detection dataset from a main detector and second detection datasets from several auxiliary detectors; incorporating corresponding geomagnetic information into the first and second detection datasets based on several space environment data, an external field model, and an internal field model; assimilating the data in the second detection datasets to construct a generalized linear correlation model; performing correlation analysis between the generalized linear correlation model data and the space environment data, and selecting target environmental factors based on the analysis results; integrating the target environmental factors into the first detection dataset to obtain a target detection dataset; and training the model based on the target detection dataset to obtain a target model. This invention can more effectively standardize short-term environmental disturbances, relax constraints on multi-source data, data time span, and spatiotemporal disturbance characteristics, while also offering a faster solution speed.

Description

Technical Field

[0001] This invention relates to the field of space environment technology, and in particular to a method and system for constructing a space particle environment model. Background Technology

[0002] The particle radiation environment in space can cause radiation effects on spacecraft materials and electronic components, leading to degradation or even failure. Charged particles in near-Earth orbit primarily originate from the Earth's radiation belts, solar cosmic rays, and galactic cosmic rays. The Earth's radiation belts are regions around the Earth where large amounts of high-energy charged particles accumulate; they are divided into inner and outer belts. The inner radiation belt is closer to the Earth's surface, while the outer radiation belt is farther away. High-energy protons are mainly located in the inner radiation belt. Under conditions of no solar proton events and minimal geomagnetic disturbances, the spatial distribution and intensity of high-energy protons in the inner radiation belt are relatively stable. However, this stability is not permanent. The distribution of high-energy protons in the inner radiation belt is affected by long-term changes in the geomagnetic field and by short-term changes in the spatial distribution and intensity caused by solar proton events or geomagnetic disturbances.

[0003] High-energy protons in the radiation belts can cause ionization total dose effects, displacement effects, or proton-induced single-event effects on materials and devices on spacecraft, thereby endangering the safety of spacecraft in orbit. To ensure understanding of the high-energy proton environment along the radiation belts during spacecraft design and in-orbit flight, and to obtain the characteristics of the high-energy proton environment during in-orbit flight, a space high-energy proton radiation effect protection design is needed to provide support.

[0004] Based on current technology, the construction of high-energy particle models of radiation belts is mainly based on the following methods:

[0005] Based on physical mechanisms: Considering important mechanisms such as radial diffusion, local acceleration, local loss, magnetosphere shadowing, and electric convection of particles in the radiation belts, for example, the dynamic radiation belt model constructed by GFZ based on the US VERB4D code has the advantage of good assimilation results with actual detection data, but the disadvantages are high computational complexity, long computation time, and difficulty in using it for engineering applications. It is mainly used for scientific research to analyze some specific space weather events.

[0006] Based on statistical theory, these models include traditional AE / AP radiation belt models and CRRES models, constructed using statistically averaged results from long-term on-orbit observations. Their advantages include providing valuable particle environment parameters, such as energy spectrum and flux, for engineering design. However, they cannot describe the changes in protons in the LEO orbit during the solar activity cycle, and the effects of sparse coverage of measured data are difficult to eliminate.

[0007] The above methods have the following problems in the process of constructing radiation belt models:

[0008] For physical mechanisms that can more dynamically describe the radiation belt environment, the computation speed and input parameter dependence are too slow and complex to provide engineering applications. They also require real detection data for data assimilation before they can be used. They cannot provide relatively accurate results without detection data. At the same time, the current theoretical model construction mainly focuses on electrons, and there is no in-depth research on high-energy protons.

[0009] For engineering radiation belt environment models constructed using statistical methods, their description of dynamic environmental fluctuations is insufficient. It is difficult to use existing models to construct more complex model variations in the B / L coordinate system. Long-term and large-scale detection data are required to establish such models. At the same time, for high-energy protons, traditional models that only consider the influence of magnetic field changes on proton flux may amplify the enhancement of local high-energy proton flux in low-altitude radiation belts, making the models inaccurate.

[0010] In summary, the existing methods have many problems in practical use, so it is necessary to improve them. Summary of the Invention

[0011] To address the aforementioned shortcomings, the present invention aims to provide a method and apparatus for constructing a spatial particle environment model, which can more effectively standardize short-term environmental disturbances, relax constraints on multi-source data and data time span, spatiotemporal disturbance characteristics, etc., and has a faster solution speed.

[0012] To achieve the above objectives, the present invention provides a method for constructing a space particle environment model, comprising the following steps:

[0013] Obtain the first detection dataset from the main detector and the second detection datasets from several auxiliary detectors;

[0014] Based on several space environment data, external field models, and internal field models, corresponding geomagnetic information is added to the first detection dataset and the second detection dataset;

[0015] The data in the second detection dataset are assimilated to construct a generalized linear correlation model.

[0016] The generalized linear correlation model data and the spatial environment data are subjected to correlation analysis, and target environmental factors are selected based on the analysis results.

[0017] The target environmental factors are integrated into the first detection dataset to obtain a target detection dataset;

[0018] The model is trained based on the target detection dataset to obtain the target model.

[0019] Optionally, obtaining the first detection dataset of the main detector and the second detection dataset of several sub-detectors includes:

[0020] The first detection data of the main on-orbit high-energy proton detector is acquired, and the first detection data is processed by the geometric factor of the detector to form a first detection data set;

[0021] The second detection data of the secondary on-orbit high-energy proton detector is acquired, and the second detection data is processed by the geometric factor of the detector to form a second detection data set.

[0022] Optionally, the step of adding corresponding geomagnetic information to the first and second detection datasets based on several space environment data, an external field model, and an internal field model includes:

[0023] Based on several space environment data, external field models, and internal field models, corresponding first geomagnetic information is added to the first detection data in the first detection dataset;

[0024] Based on several space environment data, external field models, and internal field models, corresponding second geomagnetic information is added to the second detection data in the second detection dataset.

[0025] Optionally, the step of adding corresponding first geomagnetic information to the first detection data of the first detection dataset based on several space environment data, external field models, and internal field models includes:

[0026] Acquire several spatial environment data, and construct a spatial environment dataset containing timestamps based on the spatial environment data;

[0027] Based on the spatial coordinate information in the first detection dataset and the space environment dataset, the first external magnetic field data corresponding to the orbital position of the main detector is obtained in the external field model;

[0028] Based on the time information and spatial coordinate information in the first detection dataset, the first internal magnetic field data corresponding to the detection position of the main detector is calculated in the internal field model.

[0029] The first external magnetic field data and the first internal magnetic field data are integrated to obtain the first geomagnetic information, and the first geomagnetic information is added to and the first detection dataset is updated.

[0030] Optionally, the step of adding corresponding second geomagnetic information to the second detection data of the second detection dataset based on several space environment data, external field models, and internal field models includes:

[0031] Acquire several spatial environment data, and construct a spatial environment dataset containing timestamps based on the spatial environment data;

[0032] Based on the spatial coordinate information in the second detection dataset and the space environment dataset, the second external magnetic field data corresponding to the orbital position of the sub-detector is obtained in the external field model;

[0033] Based on the time information and spatial coordinate information in the second detection dataset, the second internal magnetic field data corresponding to the detection position of the sub-detector is calculated in the internal source field model.

[0034] The second external magnetic field data and the second internal magnetic field data are integrated to obtain the second geomagnetic information, and the second geomagnetic information is added to and the second detection dataset is updated.

[0035] Optionally, the step of assimilating the data in the second probe dataset to construct data for a generalized linear correlation model includes:

[0036] Select a reference detection data from the second detection dataset;

[0037] The other detection data in the second detection dataset, excluding the reference detection data, are subjected to linear regression processing with the reference detection data, and then merged with the reference detection data to form a generalized linear correlation model data.

[0038] Optionally, the step of performing correlation analysis between the generalized linear correlation model data and the spatial environment data, and selecting target environmental factors based on the analysis results, includes:

[0039] Select the target attribute from the generalized linear correlation model data;

[0040] The count values ​​of the target attributes are correlated with the spatial environment data to obtain the corresponding weighting factors;

[0041] Target environmental factors are selected based on the weighting factors.

[0042] Optionally, training the model based on the target detection dataset to obtain the target model includes:

[0043] The target detection dataset is subjected to balanced sampling to form a training set and a test set;

[0044] The model is trained and tested using the training set and the test set, respectively, to complete the construction of the target model.

[0045] Optionally, the target model can be constructed based on any one of the following methods: decision tree, random forest, neural network, and support vector machine.

[0046] A system for constructing a space particle environment model is also provided, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that the processor implements the method for constructing the space particle environment model described above when executing the computer program.

[0047] The method and system for constructing a space particle environment model described in this invention utilizes machine learning technology to introduce the disturbance factors of the space environment on the geomagnetic field. At the same time, it introduces multiple space environment factors for correlation analysis and constructs nonlinear mapping relationships for multi-dimensional parameters to replace the traditional environment model construction method. This can more effectively standardize short-term environmental disturbances and relax constraints such as multi-source data, data time span, and spatiotemporal disturbance characteristics. In addition, the model has a faster solution speed compared to pure physical mechanism calculations. Attached Figure Description

[0048] Figure 1 A flowchart illustrating the steps of a method for constructing a space particle environment model according to an embodiment of the present invention;

[0049] Figure 2 A flowchart illustrating the steps of a specific implementation of the method for constructing a spatial particle environment model according to an embodiment of the present invention;

[0050] Figure 2 This includes: 1- On-orbit high-energy particle detection data; 2- Multiple other on-orbit high-energy proton detection data; 3- High-energy proton detection data with geomagnetic information; 4- Multiple other high-energy proton detection data with geomagnetic information; 5- Other space environment data with timestamps; 6- External magnetic field data for corresponding orbital positions obtained based on the external field model and space environment data; 7- External magnetic field data for corresponding orbital positions of other satellites obtained based on the external field model and space environment data; 8- Internal field data corresponding to the detection position calculated based on the internal field model; 9- Internal field data corresponding to the positions of multiple other detectors on satellites calculated based on the internal field model; 10- Multiple high-energy proton detection data with geomagnetic information. The high-energy proton detection dataset after assimilation of proton detection data; 11- data of other generalized linear correlation models constructed based on other high-energy proton detection data with geomagnetic information; 12- target detection dataset with timestamps, geomagnetic information, spatial location information, and highly correlated space environment information; 13- environmental factors with high impact factors after correlation analysis between other space environments and high-energy proton detection data with geomagnetic information; 14- training set obtained by balanced sampling in the target detection dataset; 15- test set obtained by balanced sampling in the target detection dataset; 16- model evaluation results obtained based on machine learning training; 17- test results of the test set; 18- the final environmental model. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0052] It should be noted that references to "an embodiment," "embodiment," "example embodiment," etc., in this specification refer to the described embodiment including specific features, structures, or characteristics, but not every embodiment must include these specific features, structures, or characteristics. Furthermore, such expressions do not refer to the same embodiment. Moreover, when describing specific features, structures, or characteristics in conjunction with embodiments, whether or not explicitly described, it is indicated that incorporating such features, structures, or characteristics into other embodiments is within the knowledge of those skilled in the art.

[0053] Furthermore, certain terms are used in the specification and subsequent claims to refer to specific components or parts. Those skilled in the art will understand that manufacturers may use different names or terms to refer to the same component or part. This specification and subsequent claims do not distinguish components or parts by differences in name, but rather by differences in function. The terms "comprising" and "including" used throughout the specification and subsequent claims are open-ended and should be interpreted as "including but not limited to." Additionally, the term "connection" here includes any direct and indirect electrical connection means. Indirect electrical connection means include connections made through other means.

[0054] For constructing a high-energy proton environment model in the radiation belt, traditional methods mainly integrate on-orbit detection data from multiple satellites using statistical methods and transform it to the B / L coordinate system. The model is then constructed using the detection data in the following form: I(>E,B,L,τ,T)=N(>E,L)Φ(>E,L,τ)G(B,L); where: I is the omnidirectional integral flux of particles; N is the flux of particles with energy greater than E at a distance L; Φ is the longitude function; G is the distribution function; E is the particle energy; B is the magnetic field strength; L is the magnetospheric parameter; τ is the local time; and T is the time period.

[0055] Using particle counts with energies greater than E from probe data and three-dimensional spatial coordinates (X, Y, Z), a geomagnetic field model G(B, L) is constructed. Based on the obtained B and L values, functions N and Φ are then constructed, ultimately yielding model I. The geomagnetic field model is an internal source field model of the geomagnetic field, determined by higher-order spherical harmonic functions. Functions N and Φ are simple functions. The model is based on the assumption that the geomagnetic field constrains high-energy particles; therefore, high-energy particles of different energies exhibit a simple distribution pattern under the constraints of B and L values, which can be expressed using simple functions such as polynomials, logarithmic / exponential functions, or power functions.

[0056] However, for spatially sparse data with short detection periods, these data contain disturbances from short-term solar activity and other complex factors. Therefore, when using this modeling approach, these disturbances become noise signals without independent variables to represent them. For models based on long-term average data, these disturbances can be removed statistically, but at the same time, the model loses its ability to represent short-term environmental disturbances to the radiation belts.

[0057] Figure 1 This invention illustrates a method for constructing a space particle environment model according to an embodiment of the present invention. Based on this method, a space particle radiation environment model can be obtained. The space particles can be charged particles with a certain energy, such as high-energy protons, medium-to-high-energy electrons, and space thermal plasma. The constructed model is used to characterize the space environment distribution characteristics of the aforementioned charged particles. This embodiment will use the construction of a high-energy proton environment model as an example for illustration. The construction method includes the following steps:

[0058] S101: Obtain the first detection dataset of the main detector and the second detection dataset of several auxiliary detectors. That is, the data in the first detection dataset comes from the data detected by the main detector, and the data in the second detection dataset comes from the data detected by several auxiliary detectors; the main detector is relative to the auxiliary detectors, the main difference being their different on-orbit locations, specifically the different locations of the satellites they ride on. The main detector and the auxiliary detectors can be detectors of the same type or different types.

[0059] In one embodiment, step S101 includes: acquiring first detection data from the primary on-orbit high-energy proton detector and processing the first detection data using the detector's geometric factor to form a first detection data set; acquiring second detection data from the secondary on-orbit high-energy proton detector and processing the second detection data using the detector's geometric factor to form a second detection data set.

[0060] See Figure 2The first and second detection data correspond to boxes 1 and 2 in the figure, respectively. In specific implementation, the first detection data is obtained based on the main on-orbit high-energy proton detector, and the data is processed using the detector's geometric factors to form a first dataset DF = Data(UT(time),>Energy,Counts,{X... Positions}); where, {X Positions} represents spatial location coordinates, including longitude, latitude, and altitude; similarly, the second detection data obtained by other secondary on-orbit high-energy proton detectors are processed using the detector's geometric factors to form a second dataset denoted as {DF}. i}

[0061] S102: Based on several space environment data, an external field model, and an internal field model, corresponding geomagnetic information is added to the first and second detection datasets. This means using a model that includes an external field to replace the original model that only constructs an internal field. The space environment data is used to characterize the dynamic changes of the external magnetic field, such as the ring current index of geomagnetic disturbances, solar wind speed, and solar wind dynamic pressure. Specifically, step S102 includes: based on several space environment data, an external field model, and an internal field model, adding corresponding first geomagnetic information to the first detection data of the first detection dataset; and based on several space environment data, an external field model, and an internal field model, adding corresponding second geomagnetic information to the second detection data of the second detection dataset.

[0062] In one optional implementation, the step of adding corresponding first geomagnetic information to the first detection data of the first detection dataset based on several spatial environment data, an external field model, and an internal field model includes: acquiring several spatial environment data and constructing a spatial environment dataset containing timestamps based on the spatial environment data; obtaining first external magnetic field data corresponding to the orbital position of the main detector in the external field model according to the spatial coordinate information in the first detection dataset and the spatial environment dataset; calculating first internal magnetic field data corresponding to the detection position of the main detector in the internal field model according to the time information in the first detection dataset and the spatial coordinate information; integrating the first external magnetic field data and the first internal magnetic field data to obtain first geomagnetic information, and adding and updating the first detection dataset with the first geomagnetic information.

[0063] Specifically, a spatial environment dataset containing timestamps is constructed using various spatial environment data: DE = Data(UT(time), {X Environments}).

[0064] Using {X} in DE and DF Positions Information, using an external field model to obtain the B value, for example, when using the T96 external field model, {XEnvironments This includes the Dst index, solar wind velocity, and solar wind dynamic pressure at each UT (time). For missing or low-temporal-resolution space environment data, interpolation can be used to obtain the values ​​at the corresponding time. Positions It needs to include three-dimensional spatial coordinates, i.e., DM1 = G. outer (B)=T96({X Positions},Dst,SW speed ,SW preasure ); where T96 is the tsyganenko 96 external field model function; Dst is the geomagnetic loop current exponent; SW speed Solar wind speed; SW preasure This represents the solar wind dynamic pressure; DM1 represents the first external magnetic field data.

[0065] Using UT(time) and {X} in DF Positions Information is obtained by using an internal source field model to obtain the B value. For example, when using the IGRF13 geomagnetic field model, DM2 = G. inner (B) = IGRF({X) Positions},UT(time));DM2 is the first internal magnetic field data.

[0066] Integrate DM1 and DM2 to form a dataset of magnetic field values ​​B: DM(B) = DM1(B) + DM2(B). Based on the constructed global geomagnetic field vector B data of DM, magnetic field lines are traced to obtain the distance of the magnetic field lines from the Earth's center at the magnetic equator, denoted as L. The ratio of |B| to the magnetic field at that location, |B0|, yields B / B0. Therefore, DM(B,L) = {B,L,B / B0} can be constructed; DM(B,L) represents the first geomagnetic information. Adding the DM data updates the first detection dataset to DF = Data(UT(time),>Energy,Counts,B,L,{X Positions For specific processing procedures, please refer to [link / reference]. Figure 2 As shown, the spatial environment dataset corresponds to Figure 2 Box 5 shown corresponds to the first external magnetic field data. Figure 2 Box 6 shown corresponds to the first intrinsic magnetic field data. Figure 2 Box 8 is shown.

[0067] The step of adding corresponding second geomagnetic information to the second detection data of the second detection dataset based on several space environment data, an external field model, and an internal field model includes: acquiring several space environment data and constructing a space environment dataset containing timestamps based on the space environment data; obtaining second external magnetic field data corresponding to the orbital position of the sub-detector in the external field model according to the spatial coordinate information in the second detection dataset and the space environment dataset; calculating second internal magnetic field data corresponding to the detection position of the sub-detector in the internal field model according to the time information in the second detection dataset and the spatial coordinate information; integrating the second external magnetic field data and the second internal magnetic field data to obtain second geomagnetic information, and adding the second geomagnetic information to update the second detection dataset. The specific implementation method for updating the second detection dataset in this example can adopt the specific implementation method for the first detection dataset described above, and will not be repeated here.

[0068] S103: Assimilate the data in the second probe dataset to construct a generalized linear correlation model. When multi-source probe data is introduced as the dataset, each dataset should be converted into a dataset with the same spatial environment, location parameters, and BL coordinates. For each dataset, a generalized linear correlation model should be constructed with B, L, and other highly correlated factors as independent variables, and one of them should be designated as the primary data source. The other data sources should be assimilated, and the model should be constructed after assimilation.

[0069] In an optional implementation, step S103 includes: selecting a reference detector data from the second detector dataset; performing linear regression processing on other detector data in the second detector dataset besides the reference detector data and the reference detector data, and merging them with the reference detector data to form generalized linear correlation model data. Specifically, based on the updated dataset described above, firstly, let DF1 in DF be the primary detector data (i.e., the reference detector data), and then perform linear regression processing on {DF1}. i Assimilate the dataset}(i≠1) to construct a generalized linear correlation model Fi, such that DFi(Counts)~Fi(>Energy,B,L), and obtain its mapping characteristics for Energy, B, and L, coefi={WE,WB,WL}i. Taking DF2 as an example, select specific values ​​of >Energy, B, and L, perform linear regression on F1(>Energy,B,L) and F2(>Energy,B,L), and use this method to assimilate DF2. Process all data in this way and merge them to form the overall dataset DF. total =Data(UT(time),>Energy,Counts,B,L,{X Positions For specific processing procedures, please refer to [link / reference]. Figure 2As shown, the dataset DF total and Figure 2 This corresponds to box 11 shown.

[0070] S104: Perform correlation analysis between the generalized linear correlation model data and the spatial environment data, and select target environmental factors based on the analysis results.

[0071] Optionally, step S104 includes: selecting a target attribute from the generalized linear correlation model data; performing correlation analysis between the count value of the target attribute and the spatial environment data to obtain the corresponding weight factors; and filtering out target environmental factors based on the weight factors. In this embodiment, it is preferable to filter out target environmental factors from high-weight factors. Specifically, in conjunction with the above assimilation processing results, the dataset DF... total =Data(UT(time),>Energy,Counts,B,L,{X Positions}) Select the Counts count value after determining the Energy, B, and L values, and then use DE = Data(UT(time),{X Environments Correlation analysis was conducted; where the determined Energy, B, and L values ​​refer to the energy resolution division given by the detector's own energy channels, and the B and L values ​​formed after geomagnetic coordinate transformation of the satellite orbit coordinates where the detector is located. Counts is the count value of particles of a specified energy detected by the detector under the above conditions. Based on the correlation analysis, high-weight factors {X} were selected. Environments} makes DF total (Counts(time))~DE({X Environments (time-lag i )}), where {lag i} represents the time lag deviation between each highly correlated environmental factor and the detection data results, derived from correlation analysis; DF total (Counts(time))~DE({X Environments (time-lag i The expression for )}) is used to characterize and filter space environment factors and their time delay values ​​that are significantly correlated with probe data fluctuations. This result is used to introduce more space environment factors into the model to explain and characterize changes in probe data fluctuations. For specific processing procedures, please refer to... Figure 2 As shown, the target environmental factor and Figure 2 It corresponds to box 13 in the text.

[0072] To construct the correlation between input environment parameters, indicators such as correlation coefficient, AIC, BIC, and information entropy can be used to test whether they are correlated.

[0073] S105: Integrate the target environmental factors into the first detection dataset to obtain a target detection dataset. Specifically, combining the target environmental factors selected above, the target environmental factors {X} are integrated into the first detection dataset. Environments (time-lag i )} Integrate datasets to form DF total =Data(UT(time),>Energy,Counts,B,L,{X Positions},{X Environments The target detection dataset and Figure 2 This corresponds to box 12 shown.

[0074] S106: Train the model based on the target detection dataset to obtain the target model. The trained target model can be used to drive the prediction results through various class parameters, and the prediction results are the high-energy particle environment results for this spatiotemporal region. When using this target model, in addition to traditional input parameters such as B, L, energy E, and time, real-time values ​​of space environment factors introduced in the construction of various radiation belt models can be added as driving parameters. Alternatively, the average values ​​of these environmental parameters during high / low solar activity years can be used as driving parameters to obtain only the average high-energy particle environment of the radiation belts.

[0075] In an optional implementation, step S106 includes: performing balanced sampling on the target detection dataset to form a training set and a test set; training and testing the model based on the training set and test set respectively to complete the construction of the target model. The construction of the target model includes, but is not limited to, machine learning methods based on decision trees, random forests, neural networks, and support vector machines. In practical applications, the best-performing method can be selected based on the NFL theory and their fit to the results.

[0076] The target detection dataset was divided into a training set and a test set based on a balanced sampling method, and trained using a machine learning model, such as a random forest algorithm with 100 trees and a feature value M of 3. The factors used to build the model included L, B, B / B0, longitude, latitude, energy, and Dst. The model was then tested using the test set, showing that it also has good adaptability to the test set data.

[0077] The embodiments provided by this invention, for the standard magnetic field, use a model that includes an external source field instead of the original model that only constructs an internal source field, i.e., G. total (B,L)=G outer (B,L,Dst,SW speed ,SW preasure ,{OtherIndexs})+G inner(B,L). Wherein, G total G represents the magnetic coordinates of the total geomagnetic field used in this model. outer The external magnetic field component caused by the solar wind; G inner Let be the intrinsic magnetic field component caused by the Earth's magnetic field; B be the magnetic field strength; and L be the distance from the Earth's center to the magnetic field line at that location when it crosses the Earth's magnetic equatorial plane. This formula adds the contribution of the extrinsic magnetic field component to the overall magnetic field, based on the original intrinsic magnetic field. The resulting geomagnetic coordinate system will more closely approximate reality at locations with larger L values. Introducing the solar wind and interplanetary magnetic field can describe the influencing factors at locations with higher L values. Where Dst is the geomagnetic loop current index, and SW... speed and SW preasure These represent solar wind velocity and dynamic pressure, respectively, while {OtherIndexs} represents other standard space environment index data products required for the external field model. Dst,SW speed ,SW preasure The space environment factors {OtherIndexs} are used to characterize the dynamic changes of exogenous magnetic fields. For example, Dst is the ring current index of geomagnetic disturbance, which characterizes the degree of compression of the Earth's magnetosphere under the influence of the solar wind; SW speed SW preasure The parameters {G} represent solar wind velocity and dynamic pressure, which can also characterize the level of solar wind influence on the Earth's magnetosphere. {OtherIndexs} represents other optional space environment indices or factors, such as the G index, W index, and other artificial indices obtained by moving averages of various geomagnetic or solar wind environments over different time windows. By selecting appropriate characterization parameters, the impact of changes in the exogenous magnetic field on the probe location (B, L) can be more reasonably described. This reduces errors caused by inaccurate geomagnetic coordinate systems or environmental disturbances during model construction. Traditional environmental models do not consider the influence of exogenous magnetic fields and directly remove probe data with large L values ​​that are more likely to be affected by exogenous magnetic fields. Therefore, this method can convert more probe data with large L values ​​from high-latitude regions into effective data, resulting in better data utilization.

[0078] Secondly, machine learning is used to construct multi-source mapping relationships, transforming N(>E,L)Φ(>E,L,τ) into F(>E,L,τ,{X Environments},{X PoesitionsThe mapping object is the particle flux count, and the mapped object is a control variable with specified energy, L value, specific spatial environmental factors, and specific spatial location; N(>E,L)Φ(>E,L,τ) represents the particle count and its distribution weight along geographical longitude. The traditional way to construct N(>E,L)Φ(>E,L,τ) in a model is to divide the space into grids. This is because traditional models need to find an algebraically analytical expression to describe the variation of N(>E,L)Φ(>E,L,τ), such as the relationship between the particle count and L value and energy, the relationship between longitude distribution and L value and energy, etc. However, satellite flight trajectories do not cover the Earth's surface uniformly. Therefore, dividing the space into grids can effectively count the detectors whose trajectories fall within them, and use interpolation functions to obtain the calculated coefficients of each grid and save them in a data file for easy model access. However, this approach has several drawbacks: 1. It results in a loss of spatial resolution due to grid partitioning; 2. The averaging within the grid makes the data lack the ability to reflect short-term dynamic changes; 3. In reality, multiple factors may influence the distribution of N and Φ, and considering only the influence of L and energy E will miss the influence of other factors, reducing the model's ability to describe dynamic environmental changes. Environments {X} represents a set of independent space environmental factors, such as geomagnetic index and solar activity data. Correlation analysis with the probe data is used to screen factors with higher weights and more significant p-values; the p-value is the probability of rejecting the null hypothesis, and a smaller p-value indicates higher significance. Poesitions} represents the spatial location parameters of the probe data, which are used to replace the original Φ function to represent the information of the latitude and longitude region. The probe data C(>E,{X}) is then used to... Environments}) Construct the mapping relationship F to complete the model establishment.

[0079] The method for constructing the space particle environment model provided in this embodiment has good compatibility with in-orbit multi-source high-energy proton detection data with good data quality; it is not sensitive to the sparsity of data in space and can also construct a global environment model; it can add other parameters such as space location and space environment to describe more complex and changing radiation belt environment; the calculation speed is much faster than that of pure physics models, reducing the time complexity of various complex physical calculation processes from O(n2) or O(nlogn) to a linear time complexity level of O(kn), where n is the number of parameters to be calculated and k depends on different machine learning algorithms.

[0080] The present invention also provides a system for constructing a space particle environment model, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, it implements the method for constructing the space particle environment model described above.

[0081] In summary, the method and system for constructing a space particle environment model as described in this invention address the problems of difficulty in reflecting space environment disturbances, long required spatiotemporal coverage of data, and difficulty in reflecting current changes due to early data in the process of constructing a radiation belt proton environment model using high-energy proton detection data. It introduces machine learning technology and space environment data correlation analysis technology to reconstruct the model construction process, resulting in a high-energy proton environment model of the space radiation belt with advantages such as short required data time, ability to reflect short-term and long-term disturbance characteristics of the space environment, and engineering applicability. This improves the ability to construct autonomous radiation belt environment models and the application capability of autonomous on-orbit detection data, and strengthens the space radiation environment protection design and on-orbit response capability for spacecraft under development and in orbit.

[0082] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for constructing a space particle environment model, characterized in that, Including the following steps: Obtain the first detection dataset from the main detector and the second detection datasets from several auxiliary detectors; Based on several space environment data, external field models, and internal field models, corresponding geomagnetic information is added to the first detection dataset and the second detection dataset; The data in the second detection dataset are assimilated to construct a generalized linear correlation model. The generalized linear correlation model data and the spatial environment data are subjected to correlation analysis, and target environmental factors are selected based on the analysis results. The target environmental factors are integrated into the first detection dataset to obtain a target detection dataset; The target detection dataset is used to train a model to obtain a target model. The step of assimilating the data in the second detection dataset to construct the generalized linear correlation model data includes: Select a reference detection data from the second detection dataset; The other detection data in the second detection dataset, excluding the reference detection data, are subjected to linear regression processing with the reference detection data, and then merged with the reference detection data to form a generalized linear correlation model data.

2. The method for constructing a space particle environment model according to claim 1, characterized in that, The acquisition of the first detection dataset of the main detector and the second detection dataset of several sub-detectors includes: The first detection data of the main on-orbit high-energy proton detector is acquired, and the first detection data is processed by the geometric factor of the detector to form a first detection data set; The second detection data of the secondary on-orbit high-energy proton detector is acquired, and the second detection data is processed by the geometric factor of the detector to form a second detection data set.

3. The method for constructing a space particle environment model according to claim 1, characterized in that, The step of adding corresponding geomagnetic information to the first and second detection datasets based on several space environment data, external field models, and internal field models includes: Based on several space environment data, external field models, and internal field models, corresponding first geomagnetic information is added to the first detection data in the first detection dataset; Based on several space environment data, external field models, and internal field models, corresponding second geomagnetic information is added to the second detection data in the second detection dataset.

4. The method for constructing a space particle environment model according to claim 3, characterized in that, The step of adding corresponding first geomagnetic information to the first detection data of the first detection dataset based on several space environment data, external field models, and internal field models includes: Acquire several spatial environment data, and construct a spatial environment dataset containing timestamps based on the spatial environment data; Based on the spatial coordinate information in the first detection dataset and the space environment dataset, the first external magnetic field data corresponding to the orbital position of the main detector is obtained in the external field model; Based on the time information and spatial coordinate information in the first detection dataset, the first internal magnetic field data corresponding to the detection position of the main detector is calculated in the internal field model. The first external magnetic field data and the first internal magnetic field data are integrated to obtain the first geomagnetic information, and the first geomagnetic information is added to and the first detection dataset is updated.

5. The method for constructing a space particle environment model according to claim 3, characterized in that, The step of adding corresponding second geomagnetic information to the second detection data of the second detection dataset based on several space environment data, external field models, and internal field models includes: Acquire several spatial environment data, and construct a spatial environment dataset containing timestamps based on the spatial environment data; Based on the spatial coordinate information in the second detection dataset and the space environment dataset, the second external magnetic field data corresponding to the orbital position of the sub-detector is obtained in the external field model; Based on the time information and spatial coordinate information in the second detection dataset, the second internal magnetic field data corresponding to the detection position of the sub-detector is calculated in the internal source field model. The second external magnetic field data and the second internal magnetic field data are integrated to obtain the second geomagnetic information, and the second geomagnetic information is added to and the second detection dataset is updated.

6. The method for constructing a space particle environment model according to claim 1, characterized in that, The step of performing correlation analysis between the generalized linear correlation model data and the spatial environment data, and selecting target environmental factors based on the analysis results, includes: Select the target attribute from the generalized linear correlation model data; The count values ​​of the target attributes are correlated with the spatial environment data to obtain the corresponding weighting factors; Target environmental factors are selected based on the weighting factors.

7. The method for constructing a space particle environment model according to claim 1, characterized in that, The step of training the model based on the target detection dataset to obtain the target model includes: The target detection dataset is subjected to balanced sampling to form a training set and a test set; The model is trained and tested using the training set and the test set, respectively, to complete the construction of the target model.

8. The method for constructing a space particle environment model according to claim 7, characterized in that, The target model is constructed based on any one of the following methods: decision tree, random forest, neural network, or support vector machine.

9. A system for constructing a space particle environment model, characterized in that, The system includes a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the computer program to implement the method for constructing the space particle environment model according to any one of claims 1 to 8.

Images