Geostationary orbital proton energy spectrum inversion based on machine learning
-
摘要: 根据地面中子探测与宇宙线环境之间的关联性，在太阳活动平静期以地球静止环境业务卫星及全球各个中子探测站的探测数据构建数据集。基于极端梯度提升决策树（XGBoost）和人工神经网络建立了由地面中子探测数据反演宇宙线质子环境的模型。模型采用遗传算法求解模型的最优超参数并对神经网络的各个神经元参数进行训练，实现了宇宙线质子环境在太阳活动平静期的反演，模型训练的均方差MSE为0.499，对测试集的平均反演误差分别为26.9%，对比航天常用的辐射环境模型误差通常在200%以内，提高显著。同时使用包括支持向量回归、误差反向传播算法、长短期记忆在内的多种其他机器学习算法进行了对比，结果表明本文所建立的模型具有训练时间短、计算速度快、占用资源小的优点。Abstract: Based on the correlation between ground neutron detection and the cosmic ray environment, a dataset was constructed using the detection data of geostationary operational environmental satellites and various neutron detection stations worldwide for the solar activity quiet period. Models for inverting the cosmic ray proton environment from ground neutron detection data were established based on the extreme gradient boost decision tree (XGBoost) and artificial neural network. They use genetic algorithm to solve the optimal hyperparameter and train the parameters of each neuron of the neural network to realize the inversion of the cosmic ray proton environment. The mean square error of the model training is 0.499, and the average inversion error of the test set is 26.9% respectively. Compared with the radiation environment model commonly used in aerospace, the error is usually within 200%, which is significantly improved. Multiple other machine learning algorithms, including support vector regression, error back propagation training, long short-term memory network, were compared and the results show that the model established in this paper has the advantages of short training time, fast computation speed, and low resource consumption.
-
图 1 所有站点的磁纬度与其截止刚度的分布
Figure 1. Distribution of magnetic latitude and cutoffrigidity of all stations
图 3 太阳活动极小年模型的预测值与探测值及其他算法的对比
Figure 3. Fluxes calculated by our model and SVR, BP, LSTM models, in comparison with flux data detected by GOES10 detector in solar minimum
图 4 太阳活动极大年模型的预测值与探测值及其他算法的对比
Figure 4. Fluxes calculated by our model and SVR, BP, LSTM models, in comparison with flux data detected by GOES13 detector in solar maximum
图 5 太阳活动极小年模型的预测值与CREME96模型、AP8模型的对比
Figure 5. Calculated fluxes comparison among our model, the CREME96 model, and AP8 model in solar minimum
图 6 太阳活动极大年模型的预测值与CREME96模型、AP8模型的对比
Figure 6. Calculated fluxes comparison among our model, the CREME96 model, and AP8 model in solar maximum
表 1 数据集概况
Table 1. Dataset overview
total data volume number of data used for the training set number of data used for the test set 8 442 700 7 598 430 844 270 
下载: 导出CSV
表 2 不同参数方案的各项指标
Table 2. Various indicators of different parameter schemes
indicators MSE RMSE MRE/% default parameter scheme 0.605 0.778 19.6 initial parameter scheme 0.635 0.797 25.5 optimal parameter scheme 0.499 0.706 17.6
下载: 导出CSV
表 3 四种机器学习模型的指标对比
Table 3. Comparison of indicators for four machine learning models
model MSE RMSE MRE/% training time/s GA-XGBoost 0.499 0.706 12.5 135 SVR 2.632 1.622 264.6 620 BP 0.641 0.801 84.2 3 625 LSTM 0.131 0.361 17.5 6 390
下载: 导出CSV
-
[1] Cao Meng, Deng Jiahao, He Weifan, et al. Geant4 simulation of energy deposition ratio and physical processes of CdZnTe crystals irradiated by high energy particles[J]. Applied Physics A, 2023, 129: 595. doi: 10.1007/s00339-023-06862-0 [2] Loffredo F, Vardaci E, Bianco D, et al. Protons interaction with Nomex target: secondary radiation from a Monte Carlo simulation with Geant4[J]. Applied Sciences, 2022, 12: 2643. doi: 10.3390/app12052643 [3] Ye Bing, Cai Li, Wu Zhaoxi, et al. Investigation of radiation response for III-V binary compound semiconductors due to protons using Geant4[J]. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 2022, 529: 38-48. [4] Loto'aniu T M, Singer H J, Rodriguez J V, et al. Space weather conditions during the Galaxy 15 spacecraft anomaly[J]. Space Weather, 2015, 13(8): 484-502. doi: 10.1002/2015SW001239 [5] Davis A J, Mewaldt R A, Binns W R, et al. The evolution of galactic cosmic ray element spectra from solar minimum to solar maximum: ACE measurements[C]//Proceedings of the 27th International Cosmic Ray Conference. 2001: 1. [6] Badhwar G D, O'Neill P M. An improved model of galactic cosmic radiation for space exploration missions[C]//Proceedings of the 22nd International Cosmic Ray Conference. 1991: 643-646. [7] Nymmik R A. Initial conditions for radiation analysis: Models of galactic cosmic rays and solar particle events[J]. Advances in Space Research, 2006, 38(6): 1182-1190. doi: 10.1016/j.asr.2006.07.002 [8] O’Neill P M. Badhwar-O'Neill galactic cosmic ray model update based on advanced composition explorer (ACE) energy spectra from 1997 to present[J]. Advances in Space Research, 2006, 37(9): 1727-1733. doi: 10.1016/j.asr.2005.02.001 [9] Adams J H. Cosmic ray effects on microelectronics. Part I: the near-earth environment[R]. Memorandum: NRL Report, 1986, A506. [10] Belov A. Large scale modulation: view from the earth[J]. Space Science Reviews, 2000, 93(1): 79-105. [11] 都亨, 叶宗海. 低轨道航天器空间环境手册[M]. 北京: 国防工业出版社, 1996Du Heng, Ye Zonghai. Low orbit spacecraft space environment manual[M]. Beijing: National Defense Industry Press, 1996. [12] Tomassetti N, Bertucci B, Donnini F, et al. Data driven analysis of cosmic rays in the heliosphere: diffusion of cosmic protons[J]. Rendiconti Lincei. Scienze Fisiche e Naturali, 2023, 34(2): 333-338. doi: 10.1007/s12210-023-01149-1 [13] Vette J I. The AE-8 trapped electron model environment[R]. Greenbelt: University of Michigan Library, 1991. [14] vakyan S V, Voronin N A, Nikol’sky G A. Response of atmospheric pressure and air temperature to the solar events in October 2003[J]. Geomagnetism and Aeronomy, 2015, 55(8): 1180-1185. doi: 10.1134/S0016793215080034 [15] Kudela K, Storini M, Hofer M Y, et al. Cosmic rays in relation to space weather[J]. Space Science Reviews, 2000, 93(1): 153-174. [16] Solanik M, Bobík P, Genči J. Geliosphere-parallel CPU and GPU based models of cosmic ray modulation in the heliosphere[J]. Computer Physics Communications, 2023, 291: 108847. doi: 10.1016/j.cpc.2023.108847 [17] 罗昊鑫, 姚宇翔, 潘文武, 等. 基于神经网络的消色差旋转双棱镜反向解算[J]. 强激光与粒子束, 2023, 35:071008 doi: 10.11884/HPLPB202335.220332Luo Haoxin, Yao Yuxiang, Pan Wenwu, et al. Inverse calculation of achromatic Risley prism based on neural network[J]. High Power Laser and Particle Beams, 2023, 35: 071008. doi: 10.11884/HPLPB202335.220332 -
点击查看大图
计量
- 文章访问数: 521
- HTML全文浏览量: 188
- PDF下载量: 62
- 被引次数: 0
