Earth space radiation band environment modeling and calculating method based on deep neural network

Abstract

The invention discloses an earth space radiation zone environment modeling and calculating method based on a deep neural network, and belongs to the field of earth space radiation environment simulation modeling. Flux intensity training data and flux distribution training data are constructed based on the earth radiation band environment data; building a regression prediction network and a dichotomy network, optimizing the regression prediction network based on the flux intensity training data, and optimizing the dichotomy network based on the flux distribution training data to obtain a radiation band ion flux intensity model and a radiation band flux distribution model; a user inputs task emission information, and an input tensor is constructed based on the task emission information; inputting the input tensor into the radiation band ion flux intensity model and the radiation band flux distribution model to obtain two prediction tensors; and obtaining the environmental particle energy spectrum data of the earth radiation band based on the two prediction tensors.

Description

Technical Field

[0001] The present invention belongs to the technical field of geospace radiation environment simulation modeling, and in particular to a geospace radiation belt environment modeling and calculation method based on a deep neural network. Background Art

[0002] The Earth's radiation belts are regions of high-energy particles captured and maintained by the Earth's magnetic field. They are mainly divided into the inner radiation belt and the outer radiation belt, with the former dominated by high-energy protons and the latter by high-energy electrons. Since these high-energy particles have a significant impact on human activities and technological facilities, especially artificial satellites and spacecraft, modeling research on the radiation belt environment is particularly critical. Radiation belt environment modeling technology simulates the particle distribution, energy and dynamic changes in the radiation belt through mathematical and physical models. These models usually rely on data from satellite observations (such as NASA's Van Allen Probes and ESA's Cluster mission) and ground-based observations (such as magnetometer networks and radio telescopes). Modeling methods include empirical models based on observational data (such as NASA's AE8 / AP8 model and AE9 / AP9 model), physical models based on physical theories (such as the Fokker-Planck equation, Salammbo model and CRRESPRO / CRRESELE model), and numerical simulations using high-performance computing and numerical methods (such as the Particle-in-Cell method and the Monte Carlo method).

[0003] Although the existing technology has made certain progress, it still has some significant shortcomings. First, many empirical models are based on relatively old data, which makes it difficult to reflect the latest changes in the current radiation belt environment, and the timeliness of the model is poor. Secondly, although the physical model provides a deeper understanding, it still has limitations in dealing with complex nonlinear processes and multi-scale coupling, resulting in limited prediction accuracy. In addition, numerical simulation methods often require high-performance computing resources, with high computational costs and long time consumption, which limits their popularity in real-time prediction and application. The spatiotemporal variability, nonlinear complexity and multi-scale coupling of data increase the difficulty of modeling, making it difficult for existing models to accurately and quickly respond to sudden space weather events. Therefore, the present invention provides a modeling and calculation method for the Earth's space radiation belt environment based on a deep neural network. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method for modeling and calculating the Earth space radiation belt environment based on a deep neural network to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a method for modeling and calculating the earth space radiation belt environment based on a deep neural network, comprising the following steps:

[0006] Loading the earth radiation belt environment data, and constructing flux intensity training data and flux distribution training data based on the earth radiation belt environment data;

[0007] Building a regression prediction network and a binary classification network, optimizing the regression prediction network based on the flux intensity training data and optimizing the binary classification network based on the flux distribution training data to obtain a radiation belt ion flux intensity model and a radiation belt flux distribution model;

[0008] Input task emission information, and construct an input tensor based on the task emission information;

[0009] Inputting the input tensor into the radiation belt ion flux intensity model and the radiation belt flux distribution model to obtain two prediction tensors;

[0010] Based on the two prediction tensors, the particle energy spectrum data of the earth radiation belt environment is obtained.

[0011] Optionally, the process of constructing flux intensity training data and flux distribution training data based on the earth radiation belt environment data includes:

[0012] Preprocessing the earth radiation belt environment data to obtain consistency data;

[0013] Performing coordinate transformation on the consistency data to obtain the earth radiation belt environment data in the BL magnetosphere coordinate system;

[0014] Combining the earth radiation belt environment data in the BL magnetosphere coordinate system into a two-dimensional matrix consisting of magnetic field intensity, drift surface parameters, energy and flux to obtain flux intensity training data;

[0015] The earth radiation belt environment data in the BL magnetosphere coordinate system is combined into a two-dimensional matrix consisting of magnetic field intensity, drift surface parameters, energy and Boolean values ​​to obtain flux distribution training data.

[0016] Optionally, the regression prediction network includes a plurality of fully connected hidden layers, ReLU, Leaky ReLU and RReLU activation functions and a plurality of Dropout layers;

[0017] The binary classification network includes several fully connected hidden layers, ReLU, Sigmoid activation functions and a Dropout layer.

[0018] Optionally, the process of optimizing the regression prediction network and the binary classification network includes:

[0019] The first three columns of data in the flux intensity training data and the flux distribution training data are used as feature values, the first three columns of data including: magnetic field intensity, drift surface parameters, and energy; and the fourth column of data is used as a label value;

[0020] Initializing a regression prediction network and a binary classification network and determining a loss function for the regression prediction network and the binary classification network; wherein the loss function of the regression prediction network is a mean square error, and the loss function of the binary classification network is a cross entropy;

[0021] Inputting the first three columns of data of the flux intensity training data into a regression prediction network to obtain a first prediction value; calculating the mean square error between the first prediction value and the label value of the flux intensity training data, and performing cyclic training on the regression prediction network based on the mean square error to obtain a radiation belt ion flux intensity model;

[0022] The first three columns of data of the flux distribution training data are input into a binary classification network to obtain a second prediction value; the cross entropy of the second prediction value and the label value of the flux distribution training data is calculated, and the binary classification network is cyclically trained based on the cross entropy to obtain a radiation belt flux distribution model.

[0023] Optionally, the process of optimizing the regression prediction network and the binary classification network further includes adding regularization to the network training;

[0024] Among them, the regularized expression is:

[0025]

[0026] Where L is the original loss function, λ is the regularization strength parameter, m is the number of samples, and w i are the model weights.

[0027] Optionally, the process of constructing an input tensor based on the task emission information includes:

[0028] Calculating the flight trajectory data of the aircraft using a J2 perturbation model based on the mission launch information;

[0029] Performing coordinate transformation on the flight trajectory data to obtain transformed flight trajectory data;

[0030] The transformed flight trajectory data is combined to obtain a two-dimensional tensor, and the two-dimensional tensor is normalized to obtain an input tensor.

[0031] Optionally, the process of obtaining the earth radiation belt environment particle spectrum data based on the two prediction tensors also includes post-processing the two prediction tensors; wherein the data post-processing process includes:

[0032] Convert the two prediction tensors into two-dimensional matrices respectively;

[0033] The two-dimensional matrix of the radiation belt flux distribution model is transformed to obtain the flux distribution output;

[0034] Performing inverse normalization processing on the two-dimensional matrix of the radiation belt ion flux intensity model to obtain a flux intensity output;

[0035] Multiplying the flux distribution output and the flux intensity output by corresponding terms to obtain an original radiation belt particle flux matrix;

[0036] Based on the original radiation belt particle flux matrix, the average value of flux is calculated for different energy intervals, and a two-dimensional matrix of energy and average flux is constructed.

[0037] Optionally, the process of converting the two-dimensional matrix of the radiation belt flux distribution model includes: setting the output probability of the radiation belt flux distribution model to 0 when it is less than 0.5, and setting it to 1 when it is greater than or equal to 0.5.

[0038] The present invention also provides a computer terminal device, comprising:

[0039] one or more processors;

[0040] A memory, coupled to the processor, for storing one or more programs;

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement a method for modeling and calculating the earth space radiation belt environment based on a deep neural network.

[0042] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a method for modeling and calculating the earth space radiation belt environment based on a deep neural network is implemented.

[0043] Compared with the prior art, the present invention has the following advantages and technical effects:

[0044] This application provides a new method for modeling and calculating the Earth's space radiation belt environment. This method uses the flexibility of deep neural networks and significantly improves the accuracy and effectiveness of model predictions by updating training data sets in real time. It solves the problem that traditional empirical models are difficult to update and maintain in a timely manner, ensuring that the model is always consistent with the dynamic changes of the current Earth's space radiation belt environment.

[0045] The present invention adopts a dual-model approach of "flux intensity" and "flux distribution" to carry out energy spectrum simulation and modeling. By respectively learning the data characteristics of the radiation belt flux intensity and flux distribution, the radiation belt environment conditions (various charged particle flux values ​​in a wide energy range) at each point on the flight trajectory are accurately predicted.

[0046] This application realizes the end-to-end automated calculation process from data processing to prediction output, greatly improving the efficiency and accuracy of simulation calculations, and providing solid basic data support for the stable operation of spacecraft in complex and changeable radiation environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0048] Figure 1 is a flow chart of the modeling phase of an embodiment of the present invention;

[0049] Figure 2 A radiation belt flux intensity regression prediction network according to an embodiment of the present invention;

[0050] Figure 3 A radiation band flux distribution binary classification network according to an embodiment of the present invention;

[0051] Figure 4 This is a flow chart of calculating the earth radiation belt environment energy spectrum according to an embodiment of the present invention;

[0052] Figure 5 The output result of the radiation band electron environment spectrum of the embodiment of the present invention;

[0053] Figure 6 The output result of the radiation belt proton environment energy spectrum of an embodiment of the present invention;

[0054] Figure 7 The output result of the electron environment spectrum of the radiation belt plasma of the embodiment of the present invention;

[0055] Figure 8 This is the output result of the radiation belt plasma proton environment energy spectrum of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0057] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0058] Embodiment 1

[0059] Radiation belt environment models have important applications in many fields, such as space mission planning, spacecraft design, and space weather forecasting, despite the challenges of spatiotemporal variability of data, nonlinear complexity, and multiscale coupling. In the future, with the advancement of observation technology and the improvement of computing power, radiation belt environment modeling will develop in the direction of higher accuracy, wider application, and stronger prediction capabilities. Through more sophisticated data integration, advanced numerical methods, and interdisciplinary cooperation, it will continue to break through bottlenecks and provide solid guarantees for humans to explore space more deeply. Through continuous technological innovation and interdisciplinary cooperation, radiation belt environment modeling will continue to improve its role in the safe operation of spacecraft and space science research, and ultimately promote the comprehensive exploration and utilization of space by humans.

[0060] This embodiment aims at the problems of insufficient model timeliness, limited prediction accuracy and high computational cost existing in the current technology. The present invention proposes a solution, namely, using deep neural networks to optimize and improve the modeling process of the Earth's radiation belt environment, so as to improve the timeliness and prediction accuracy of the model, while reducing the computational cost. By introducing the deep neural network model, the present invention can easily integrate the latest radiation belt environment data and realize timely updating of the model, thereby improving the timeliness and accuracy of the prediction. At the same time, the computational efficiency of the deep neural network is significantly improved, and complex computing tasks can be completed in a shorter time, occupying fewer computing resources, and significantly reducing the computing cost. In this way, the present invention can not only provide more accurate and real-time radiation belt environment forecasts, but also can be widely used in the fields of space mission planning, spacecraft design and space weather forecasting, so as to solve the deficiencies of the existing technology and promote the development of space science research and safe operation of spacecraft.

[0061] This embodiment provides a method for modeling and calculating the earth space radiation belt environment based on a deep neural network, including the following steps:

[0062] The steps of the first stage are as follows Figure 1 As shown, it includes: loading the earth radiation belt environment data, constructing flux intensity training data and flux distribution training data based on the earth radiation belt environment data; building a regression prediction network and a binary classification network, optimizing the regression prediction network based on the flux intensity training data and optimizing the binary classification network based on the flux distribution training data to obtain the radiation belt ion flux intensity model and the radiation belt flux distribution model.

[0063] (1) Loading data: Reading the Earth radiation belt environment data generated by satellites or other modeling software, which contains time, geographic location coordinate information, particle energy, particle type and ion flux information.

[0064] As a specific implementation of the present embodiment, the radiation belt environment data can be the historical data of the detection satellite or other simulation model output data, wherein the altitude range needs to be controlled at 100-65000km, the latitude range is controlled at 0-88 °, and the data composition at least includes: particle type, time, position coordinates relative to the earth, energy and flux. Different time resolutions need to be limited for data at different orbital altitudes, and the low earth orbit (Low Earth Orbit, LEO) with an altitude of less than 2000km is sampled once every 10s, the middle earth orbit (Middle Earth Orbit, MEO) with an altitude of 2000-35786km is sampled once every 5min, and the high orbit above 35786km is sampled once every 1h. The amount of data is not limited, but it is necessary to cover the altitude and latitude ranges mentioned above as completely as possible. If the original data can cover a complete solar activity cycle, better prediction results will be obtained.

[0065] (2) Data preprocessing: Preprocess the Earth radiation belt environment data to obtain consistent data. In order to ensure the consistency of subsequent training data generation, it is necessary to perform a detailed type check on the loaded data set. Based on these check results, the data is uniformly converted to a uniform data type to ensure the standardization and compatibility of the data set. In addition, the data type conversion process will follow a unified standard to maintain the integrity and accuracy of the data and avoid introducing unnecessary biases or errors during the model training process. This standardized data preprocessing process is critical to the performance and generalization ability of machine learning models.

[0066] As a specific implementation method of this embodiment, since the loaded data comes from a wide range of sources, the data type will be more complex. For the quality of subsequent training data, the original data needs to be uniformly converted. Specifically, this method makes a judgment through the unit system of the read-in data, and realizes the conversion of the following common data types: 1) Conversion of simplified Julian day to decimal year; 2) Conversion of Cartesian coordinate system to WGS-84 coordinate system; 3) Conversion of integral energy spectrum data to differential energy spectrum data. Through the above transformations, the data is unified into the form of {decimal year, longitude, latitude, altitude, energy, differential flux}. Note that the energy unit in the data is unified into MeV, and the differential flux unit is unified into MeV. -1 cm -2 ·s -1 , the following is the calculation formula for converting the integral energy spectrum to the differential energy spectrum:

[0067]

[0068] In the formula, E i (n) represents energy, F d (n) and F i(n) are the differential and integrated flux of the nth energy point, respectively. Note that a maximum energy point will be lost when the integral is converted to a differential spectrum.

[0069] (3) Coordinate transformation: The consistent data is transformed to obtain the Earth radiation belt environment data in the BL magnetospheric coordinate system. The time and geographic location coordinates in the data are converted into BL coordinates (a magnetospheric coordinate system introduced to describe the intensity distribution of radiation belt particles). On the one hand, according to the formation mechanism of the radiation belt, the particle flux distribution is described by the magnetic field intensity B and the parameter L marking each drift surface. On the other hand, this operation reduces the strong dependence of the label data (particle flux data) on space and time, which is more conducive to the convergence of the model in the subsequent training process.

[0070] As a specific implementation of this embodiment, in order to more accurately describe the distribution of radiation belt particle intensity, the BL magnetospheric coordinate system is introduced to replace the traditional time-space coordinate system, where B represents the magnetic field intensity and L is a parameter marking each drift surface. The magnetic field intensity B is obtained by the latest International Geomagnetic Reference Field (IGRF13) model, which is constructed based on geomagnetic observation data on a global scale. Since the Earth's magnetic field changes over time (including long-term and short-term changes), the IGRF model needs to be updated regularly to maintain its accuracy. The IGRF model is currently updated every 5 years to reflect changes in the Earth's magnetic field. The calculation formula for the L value is as follows:

[0071]

[0072] Where M is the dipole moment constant of the Earth's magnetic field, which is 7.9×10 30 nT·cm 3 , V represents the calculation parameters, B0 is the magnetic field intensity at the intersection of the magnetic field lines at the calculation point and the equator line, which can all be obtained through the IGRF13 model calculation.

[0073] (4) Constructing flux intensity training data: Combine the data into a two-dimensional matrix composed of B, L, energy, and particle differential flux, remove meaningless values ​​and outliers in the flux data, normalize each column, and save it as flux intensity training data.

[0074] As a specific implementation of this embodiment, the data is combined into a two-dimensional matrix such as {B, L, energy, flux}. If the flux value is meaningless or an erroneous value, the entire row of the matrix is ​​deleted to avoid contamination of the training data. After completing the above processing, each column of the matrix is ​​normalized and the value is scaled to between 0 and 1 to speed up the convergence of the model during training. The normalization formula adopted is as follows:

[0075]

[0076] Where, Xtrain is the normalized output value, x is the input data, X o Represents the original data matrix.

[0077] (5) Constructing flux distribution training data: Combine the data into a two-dimensional matrix composed of B, L, energy, and particle differential flux. Set the normal data in the flux data to 1 and the abnormal data to 0. Normalize each column and save it as flux distribution training data.

[0078] As a specific implementation of this embodiment, the data is combined into a two-dimensional matrix such as {B, L, energy, Boolean value}, and the Boolean value 0 or 1 is determined by the flux value. If the flux is meaningless or an erroneous value, the flux value is set to 0, and the normal value is set to 1. After completing all row operations, the first three columns are normalized in the same way as in the previous step.

[0079] (6) Building a deep neural network model: Based on the characteristics of the training data, two deep neural network models were built. For the flux intensity training data, a regression prediction network was built. The structure is as follows: Figure 2 As shown in Figure 1, a deep neural network with three fully connected hidden layers, ReLU, Leaky ReLU and RReLU activation functions and two Dropout layers has a total of 166401 parameters. For the flux distribution training data, a binary classification network is built with the following structure: Figure 3 As shown in the figure, it consists of two fully connected hidden layers, ReLU and Sigmoid activation functions, and a Dropout layer, with a total of 34049 parameters.

[0080] As a specific implementation of this embodiment, a nonlinear regression deep neural network and a binary classification deep neural network are built according to the data distribution characteristics.

[0081] In the regression network, a linear layer (func1) with an input dimension of 3 and an output dimension of 512 is first created, and then a Dropout layer is added with a retention rate of 0.5 to reduce the overfitting of the model. Next is a linear layer (func2) with an input dimension of 512 and an output dimension of 256, followed by another Dropout layer with a retention rate of 0.2. Finally, a linear layer (func3) with an input dimension of 256 and an output dimension of 128 is added, and a linear regression layer (regression) with an output dimension of 1 is connected for the regression task. During the forward propagation process, the input data is first passed to func1 through the ReLU activation function, and then the previously defined Dropout layer is applied. Next, it is passed to func2 through the Leaky ReLU activation function, and the Dropout layer is applied again. Finally, it is passed to func3 using the Randomized Leaky ReLU activation function, and finally predicted through the linear regression layer. The returned output is a one-dimensional tensor. The first element is taken out by slicing as the final output result. The structure is as shown in the attached figure. Figure 2 shown.

[0082] In the binary classification network, the structure is as shown in the following Figure 3 As shown in the figure, it contains three fully connected layers, namely func1, func2 and func3. During the initialization process, func1 with an input dimension of 3 and an output dimension of 256 is followed by a Dropout layer with a retention rate of 0.2. Next is a func2 with an input dimension of 256 and an output dimension of 128. Finally, there is a func3 with an input dimension of 128 and an output dimension of 1. The output result is processed by the sigmoid function for the binary classification task. During the forward propagation process, the data is first passed to func1 through the ReLU activation function, and then passes through the Dropout layer. Then, it is passed to func2 through the ReLU activation function, and finally passed to func3 through the sigmoid function to obtain the classification result.

[0083] (7) Model training: The first three columns of the training data (B, L, and energy) are used as feature values, and the fourth column of particle differential flux is used as the label value, and are brought into the built deep neural network for training.

[0084] Furthermore, the process of optimizing the regression prediction network and the binary classification network includes: taking the first three columns of data in the flux intensity training data and the flux distribution training data as feature values, the first three columns of data including: magnetic field intensity, drift surface parameters, and energy; taking the fourth column of data as label values; initializing the regression prediction network and the binary classification network and determining loss functions for the regression prediction network and the binary classification network; wherein the loss function of the regression prediction network is the mean square error, and the loss function of the binary classification network is the cross entropy; inputting the first three columns of data of the flux intensity training data into the regression prediction network to obtain a first prediction value; calculating the mean square error between the first prediction value and the label value of the flux intensity training data, and performing cyclic training on the regression prediction network based on the mean square error to obtain a radiation belt ion flux intensity model; inputting the first three columns of data of the flux distribution training data into the binary classification network to obtain a second prediction value; calculating the cross entropy between the second prediction value and the label value of the flux distribution training data, and performing cyclic training on the binary classification network based on the cross entropy to obtain a radiation belt flux distribution model.

[0085] As a specific implementation of this embodiment, this application selects the most popular PyTorch deep learning framework for model training. The training process is mainly divided into: 1) data preparation; 2) network loading; 3) damage function and optimizer selection; 4) training cycle; 5) verification. The specific description is as follows:

[0086] 1) In the data preparation process, the first three columns of the training data {B, L, energy} are first used as feature values, and the fourth column {flux} or {Boolean value} is used as the label value. Then, the train_test_split function is used to divide the data into a training set and a test set, which account for 99% and 1% respectively. Finally, the DataLoader function is used to randomly shuffle the training set and test set data, and process them in batches into Tensor format. After many experiments, the model training effect reaches the balance point of speed and effect when the batch_size is set to 512.

[0087] 2) Load the deep neural network built in step 6 and set the initialization parameters of the network model. This application adopts the uniform distribution method.

[0088] 3) The two networks are not the same in terms of the choice of damage function. For the regression network, this application chooses to use the mean square error loss function, and the formula is as follows:

[0089]

[0090] Among them, y i is the true value of the validation set label, a i is the predicted value of the model, and n is the number of samples. For the binary classification network, the cross entropy loss function is used, and the formula is as follows:

[0091] L=-[y i *ln(a i )+(1-y i )*ln(1-a i )] (5)

[0092] In the formula, y i is the validation set label (0 or 1), a i is the probability that the model predicts a positive class. In terms of the choice of optimizer, this application uses the Adam (Adaptive Moment Estimation) optimizer, which is an adaptive optimization algorithm that can adjust the learning rate based on historical gradient information.

[0093] In terms of parameter configuration, the initial learning rate is set to 0.001 to prevent the network from failing to converge during training. At the same time, to prevent overfitting in training, L2 regularization is added. The formula is as follows:

[0094]

[0095] Among them, L is the original loss function, λ is the regularization strength parameter, m is the number of samples, and w i is the model weight. In the process of gradient descent, L2 regularization affects the update of weights by adding the sum of squares of weights as a penalty term, so that when the weights are updated, not only the gradient of the loss function is considered, but also the contribution of the regularization term to the gradient is considered.

[0096] 4) In the training cycle, first forward propagation is performed to input data into the network and calculate the network output; then the defined loss function is used to calculate the loss value between the network output and the true label; then, according to the calculated loss, back propagation is performed to calculate the gradient; then the optimizer is used to update the network weights and biases according to the calculated gradient; finally, during the training process, loss, accuracy and other indicators are recorded for subsequent analysis. This application performs 100 cycles of cyclic training on the two networks respectively, and reduces the learning rate by 50% after every 20 cycles to obtain the optimal training results.

[0097] 5) The test validation process is to evaluate the model using the validation set after each training cycle (epoch), which can help understand the performance of the model on unseen data.

[0098] (8) Output model: After successful training, two deep neural network models with corresponding weight values ​​will be output, namely the radiation belt ion flux intensity model and the radiation belt flux distribution model.

[0099] As a specific implementation of this embodiment, the torch.save function is used to save the parameter weight values ​​of the two types of model networks respectively, and the files are stored in the "xxx.pkl" file format. Since different charged particles in the radiation belt environment need to be trained and saved separately, a total of 8 models are trained, namely: radiation belt proton flux intensity model, radiation belt proton flux distribution model, radiation belt electron flux intensity model, radiation belt electron flux distribution model, radiation belt plasma proton flux intensity model, radiation belt plasma proton flux distribution model, radiation belt plasma electron flux intensity model and radiation belt plasma electron flux distribution model.

[0100] The calculation process of the second stage is as follows Figure 4 As shown, it includes: the user inputs the mission launch information, and constructs the input tensor based on the mission launch information; the input tensor is input into the radiation belt ion flux intensity model and the radiation belt flux distribution model to obtain two prediction tensors; based on the two prediction tensors, the earth radiation belt environment particle energy spectrum data is obtained.

[0101] (1) User input: To calculate the radiation belt environment data, the user needs to input the launch information of the spacecraft (launch date, apogee, perigee, inclination, right ascension of the ascending node, argument of perigee, mean anomaly, mission period and time interval), the charged particle types and energy range of the calculated radiation belt environment.

[0102] As a specific implementation of this embodiment, in order to drive the algorithm to implement the calculation, the user needs to input detailed mission launch information, which includes the launch date of the spacecraft, the expected apogee and perigee distances, orbital inclination, ascending node right ascension, perigee argument, mean anomaly, and the specific period and time interval of the mission. In addition, we also need the user to specify the type of charged particles in the radiation belt environment to be calculated and the corresponding energy range, so as to more accurately evaluate the radiation environment that the spacecraft may encounter during its on-orbit operation. These input data are crucial to ensure that the algorithm can accurately predict and plan the trajectory of the spacecraft, the efficiency and safety of mission execution. For example: the launch date is the early morning of January 1, 2004, the apogee is 23000km, the perigee is 3000km, the inclination is 5 degrees, the other orbital elements are 0 degrees, the flight is 7 days, and the time interval is 60s. Calculate the electron and proton environment (including plasma) in the radiation belt, the proton energy range is 1.15keV-2GeV, and the electron energy range is 1keV-10MeV.

[0103] (2) Flight trajectory calculation: Based on the aircraft launch information input by the user, the J2 perturbation model is used to calculate the time, longitude, latitude, and altitude data sets of the aircraft during the mission cycle.

[0104] As a specific implementation of this embodiment, according to the user input in the previous step, the J2 perturbation model is used to calculate the trajectory of the aircraft. The J2 term in the model represents the main influence of the elliptical flattening of the earth's equator on the satellite orbit. When calculating, it is first necessary to determine the initial orbital elements of the satellite, and then apply the perturbation effect caused by the J2 term to these orbital elements. This includes periodic corrections to parameters such as the right ascension of the ascending node, the perigee angular distance, and the orbital inclination of the satellite orbit. By solving the orbital equation after the perturbation correction, the actual flight trajectory of the satellite under the influence of the non-uniform mass distribution of the earth can be predicted. This method usually uses numerical integration technology to process the perturbation equation, so as to accurately simulate the orbital changes of the satellite over a long period of time.

[0105] (3) Coordinate conversion: Same as the coordinate conversion in the first stage; convert the traditional time-space coordinate system (time, longitude, latitude and altitude) into the BL magnetosphere coordinate system.

[0106] (4) Construct input tensor: Combine the data into a two-dimensional tensor consisting of B, L, and energy, and normalize each column separately. The method is the same as step 4 in the modeling phase. Use the torch.tensor function to convert the matrix into the Tensor form required by the model.

[0107] (5) Data prediction: Load the flux intensity model and flux distribution model generated in the previous stage, set them to evaluation mode, inject the input tensors into the two models in batches, and obtain two prediction tensors.

[0108] (6) Data post-processing: First, the tensors output by the two models are converted into two-dimensional matrices. Secondly, the probability values ​​output by the flux distribution model are converted. Those less than 0.5 are set to 0, and those less than 0.5 are set to 1. Then, the corresponding items of the two matrices are multiplied to obtain the radiation belt flux intensity distribution matrix. Finally, the average flux value of each energy point is calculated.

[0109] Furthermore, the data post-processing process includes: converting the two prediction tensors into two-dimensional matrices respectively; converting the two-dimensional matrix of the radiation belt flux distribution model to obtain the flux distribution output; inverse normalizing the two-dimensional matrix of the radiation belt ion flux intensity model to obtain the flux intensity output; multiplying the flux distribution output and the flux intensity output by corresponding terms to obtain the original radiation belt particle flux matrix; based on the original radiation belt particle flux matrix, calculating the average value of the flux for different energy intervals, and constructing a two-dimensional matrix of energy and average flux.

[0110] As a specific implementation of this embodiment, first convert the two tensors output in the previous step into matrix form, and then further process the output of the flux distribution model, setting the probability value less than or equal to 0.5 to 0, and the probability value greater than 0.5 to 1. Then, the output of the flux intensity model is inversely normalized and multiplied with the corresponding item output by the flux distribution model to obtain the original radiation belt particle flux matrix. Finally, the average value of the flux is calculated for different energy intervals, and a two-dimensional matrix of energy and average flux is constructed.

[0111] (7) Output energy spectrum: Sort by energy from small to large, and output the energy spectrum data of particles in the Earth's radiation belt environment.

[0112] As a specific implementation of this embodiment, the two-dimensional matrix constructed in the previous step is written into a file, named according to the computing environment category input by the user in step 1, and saved in the form of "xxx.dat". Figure 5-Figure 8 The output results are plotted and compared with the latest radiation belt environment model (AE9 / AP9 / SPM) of the US Naval Laboratory.

[0113] in, Figure 5 Output results for the radiation belt electron environment spectrum; Figure 6 Output results for the radiation belt proton environment spectrum; Figure 7 Output results for the electron environment spectrum of the radiation belt plasma; Figure 8 Output results for the proton environment spectrum of the radiation belt plasma.

[0114] This method makes full use of GPU acceleration technology, greatly improving the efficiency of model calculation. In previous long-term flight missions, the calculation of the radiation belt energy spectrum took a long time, which brought many inconveniences to the simulation calculation of the radiation environment of the spacecraft. However, this embodiment uses the powerful computing power of GPU to fundamentally solve this problem, and the calculation speed has been significantly improved, providing strong support for real-time decision-making and adjustment of spacecraft.

[0115] Embodiment 2

[0116] This embodiment also provides a computer terminal device, including:

[0117] one or more processors;

[0118] A memory, coupled to the processor, for storing one or more programs;

[0119] When one or more programs are executed by one or more processors, the one or more processors implement a method for modeling and calculating the earth space radiation belt environment based on a deep neural network.

[0120] Embodiment 3

[0121] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, a method for modeling and calculating the earth space radiation belt environment based on a deep neural network is implemented.

[0122] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for modeling and calculating the earth space radiation belt environment based on deep neural network, characterized in that: The following steps are involved: Loading the earth radiation belt environment data, and constructing flux intensity training data and flux distribution training data based on the earth radiation belt environment data; Building a regression prediction network and a binary classification network, optimizing the regression prediction network based on the flux intensity training data and optimizing the binary classification network based on the flux distribution training data to obtain a radiation belt ion flux intensity model and a radiation belt flux distribution model; Input task emission information, and construct an input tensor based on the task emission information; Inputting the input tensor into the radiation belt ion flux intensity model and the radiation belt flux distribution model to obtain two prediction tensors; Based on the two prediction tensors, the particle energy spectrum data of the earth radiation belt environment is obtained.

2. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 1, characterized in that: The process of constructing flux intensity training data and flux distribution training data based on the earth radiation belt environment data includes: Preprocessing the earth radiation belt environment data to obtain consistency data; Performing coordinate transformation on the consistency data to obtain the earth radiation belt environment data in the BL magnetosphere coordinate system; Combining the earth radiation belt environment data in the BL magnetosphere coordinate system into a two-dimensional matrix consisting of magnetic field intensity, drift surface parameters, energy and flux to obtain flux intensity training data; The earth radiation belt environment data in the BL magnetosphere coordinate system is combined into a two-dimensional matrix consisting of magnetic field intensity, drift surface parameters, energy and Boolean values ​​to obtain flux distribution training data.

3. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 2 is characterized in that: The regression prediction network includes several fully connected hidden layers, ReLU, Leaky ReLU and RReLU activation functions and several Dropout layers; The binary classification network includes several fully connected hidden layers, ReLU, Sigmoid activation functions and a Dropout layer.

4. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 3 is characterized in that: The process of optimizing the regression prediction network and the binary classification network includes: The first three columns of data in the flux intensity training data and the flux distribution training data are used as feature values, the first three columns of data including: magnetic field intensity, drift surface parameters, and energy; and the fourth column of data is used as a label value; Initializing a regression prediction network and a binary classification network and determining a loss function for the regression prediction network and the binary classification network; wherein the loss function of the regression prediction network is a mean square error, and the loss function of the binary classification network is a cross entropy; Inputting the first three columns of data of the flux intensity training data into a regression prediction network to obtain a first prediction value; calculating the mean square error between the first prediction value and the label value of the flux intensity training data, and performing cyclic training on the regression prediction network based on the mean square error to obtain a radiation belt ion flux intensity model; The first three columns of data of the flux distribution training data are input into a binary classification network to obtain a second prediction value; the cross entropy of the second prediction value and the label value of the flux distribution training data is calculated, and the binary classification network is cyclically trained based on the cross entropy to obtain a radiation belt flux distribution model.

5. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 4 is characterized in that: The process of optimizing the regression prediction network and the binary classification network also includes adding regularization to the network training; The regularized expression is: Where L is the original loss function, λ is the regularization strength parameter, m is the number of samples, and w i are the model weights.

6. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 1, characterized in that: The process of constructing the input tensor based on the task emission information includes: Calculating the flight trajectory data of the aircraft using a J2 perturbation model based on the mission launch information; Performing coordinate transformation on the flight trajectory data to obtain transformed flight trajectory data; The transformed flight trajectory data is combined to obtain a two-dimensional tensor, and the two-dimensional tensor is normalized to obtain an input tensor.

7. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 1, characterized in that: The process of obtaining the particle energy spectrum data of the earth radiation belt environment based on the two prediction tensors also includes post-processing the data of the two prediction tensors; wherein the process of post-processing the data includes: Convert the two prediction tensors into two-dimensional matrices respectively; The two-dimensional matrix of the radiation belt flux distribution model is transformed to obtain the flux distribution output; Performing inverse normalization processing on the two-dimensional matrix of the radiation belt ion flux intensity model to obtain a flux intensity output; Multiplying the flux distribution output and the flux intensity output by corresponding terms to obtain an original radiation belt particle flux matrix; Based on the original radiation belt particle flux matrix, the average value of flux is calculated for different energy intervals, and a two-dimensional matrix of energy and average flux is constructed.

8. The method for modeling and calculating the earth space radiation belt environment based on deep neural network according to claim 7, characterized in that: The process of converting the two-dimensional matrix of the radiation belt flux distribution model includes: setting the output probability of the radiation belt flux distribution model to 0 when it is less than 0.5, and setting it to 1 when it is greater than or equal to 0.

5.

9. A computer terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the Earth space radiation belt environment modeling and calculation method based on deep neural network as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the Earth space radiation belt environment modeling and calculation method based on deep neural network as described in any one of claims 1-8.

Images