Generative Model-based of Flare Hierarchic Recognition and Forecast of Extreme Ultraviolet Images in Solar Active Region
-
摘要: 近年来，不断发射的空基观测台持续传送回海量日面图像及日地间气象数据，为采用人工智能技术对太阳活动进行预报预警提供了数据基础。但是，极端天气爆发少，样本量较少；中等程度爆发稍多，样本量较多；常规无爆发天气常见，样本较为集中，样本不均衡状况严重影响机器学习方法在空间天气领域的广泛应用。本文面向多源多通道多尺度日面图像信息，构建了来自SOHO和SDO的1996－2015年日面活动区图像数据集；针对数据分布的不平衡，对太阳活动区图像作耀斑分级与预报。在对比分析元学习算法的基础上，设计了结合分类头设计和卷积核初始化的生成式模型；在使网络轻量化的基础上，能够将M和X级耀斑预报的检测率指标相较于普通的深度学习模型和无监督度量式模型分别提升10%和7%。Abstract: In the past 20 years, massive solar images and space meteorological data that have been transmitted back continuously and constantly with increasing space-based observatories launched, provide a promising material basis for the application of artificial intelligence technology to forecast and early warning solar activities. However, due to the less the extreme solar eruption and therefore the smaller relevant sample size, a slightly more the moderate solar activities outbreak and a little more the sample data set size, and the common routine non-outbreak space-weather always occurring and thus the samples become concentrated, thereby these condition and phenomenon result in sample imbalance and unlabeled data and so on which seriously affects the wide application of machine learning methods in the field of space weather forecasting and early warning. To handle the imbalance disturbance of sample data set for flare hierarchic recognition and forecast, this paper designs artificial intelligence algorithms for extreme ultraviolet images of solar active regions. Firstly, a dataset of extreme ultraviolet images of solar active regions from SOHO and SDO from 1996 to 2015 was constructed. Then the generated models combined classification head and initialization of convolution kernel are well-designed, and better index of accuracy for M and X flare are experimentally achieved and proved. Simultaneously, in terms of lightweight networks for deep learning, some comparison and analysis of multi algorithms on Meta learning were also discussed, this proposed method achieves finally 10% and 7% increments in POD accuracy compared with ordinary deep learning based method and unsupervised metric learning method, respectively.
-
Key words:
- EUV images /
- Flare /
- Hierarchic recognition /
- Generative model /
- Artificial intelligence
-
图 2 包含卷积核初始化的生成式度量学习模型
Figure 2. Generative metric learning model with kernel initialization
表 1 耀斑级别数据集分布
Table 1. Flare dataset of different levels
数据集 B C M+X 合计 训练集 669 958 837+98 2562 测试集 194 234 247+28 703 总计 863 1192 1084+126 3265 
下载: 导出CSV
表 2 耀斑分类混淆矩阵
Table 2. Confusion matrix for flare classification
i 级耀斑（True） 非i 级耀斑（False） 预报为i 级（Positive） TP FP 预报非i 级（Negative） FN TN
下载: 导出CSV
表 3 不同类别的分类结果
Table 3. Classification results of different categories
模型 Class POD TSS I3 D[14] B 0.52 0.33 C 1 0.95 MX 0.60 0.40 Unsup_1 shot B 0.44 0.26 C 0.99 0.99 MX 0.63 0.35 Unsup_15 shot B 0.55 0.40 C 1 0.99 MX 0.76 0.48 Generate_1 shot B 0.69 0.54 C 0.70 0.55 MX 0.70 0.55
下载: 导出CSV
表 4 两种模型在4-way情况下的TSS结果
Table 4. TSS index in the 4-way case under different models
模型 Class TSS Unsup_10 shot B 0.413 C 0.976 M 0.166 X 0.209 Generate_1 shot B 0.345 C 0.366 M 0.362 X 0.349
下载: 导出CSV
-
[1] GOLUB L, PASACHOFF J M. Nearest Star: The Surprising Science of Our Sun[M]. Cambridge: Harvard University Press, 2001 [2] DUDA R O, HART P E, STORK D G. Pattern Classification[M]. Beijing: China Machine Press, 2003 [3] CHEN Yunji, LI Ling, LI Wei, et al. AI Computing Systems[M]. Beijing: China Machine Press, 2020 [4] PARK E, MOON Y J, SHIN S, et al. Application of the deep convolutional neural network to the forecast of solar flare occurrence using full-disk solar magnetograms[J]. The Astrophysical Journal, 2018, 869(2): 91 doi: 10.3847/1538-4357/aaed40 [5] NISHIZUKA N, SUGIURA K, KUBO Y, et al. Deep flare net (DeFN) model for solar flare prediction[J]. The Astrophysical Journal, 2018, 858(2): 113 doi: 10.3847/1538-4357/aab9a7 [6] HUANG X, WANG H N, XU L, et al. Deep learning based solar flare forecasting model. I. Results for line-of-sight magnetograms[J]. The Astrophysical Journal, 2018, 856(1): 7 doi: 10.3847/1538-4357/aaae00 [7] SHEN F, YANG Z C, ZHANG J, et al. Three-dimensional MHD simulation of solar wind using a new boundary treatment: comparison with in situ data at Earth[J]. The Astrophysical Journal, 2018, 866(1): 18 doi: 10.3847/1538-4357/aad806 [8] 杨易, 沈芳, 杨子才. 多种观测数据驱动的三维行星际太阳风MHD模拟[J]. 空间科学学报, 2020, 40(3): 305-314 doi: 10.11728/cjss2020.03.305YANG Yi, SHEN Fang, YANG Zicai. Simulation of interplanetary solar wind with three-dimensional MHD model driven by multiple observations[J]. Chinese Journal of Space Science, 2020, 40(3): 305-314 doi: 10.11728/cjss2020.03.305 [9] GALVEZ R, FOUHEY D F, JIN M, et al. A machine-learning data set prepared from the NASA solar dynamics observatory mission[J]. The Astrophysical Journal Supplement Series, 2019, 242(1): 7 doi: 10.3847/1538-4365/ab1005 [10] HE K M, ZHANG X Y, REN S Q, et al. Delving deep into rectifiers: Surpassing human-level performance on ImageNet classification[C]//2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile: IEEE, 2015: 1026-1034 [11] ALIPOUR N, MOHAMMADI F, SAFARI H. Prediction of flares within 10 days before they occur on the sun[J]. The Astrophysical Journal Supplement Series, 2019, 243(2): 20 doi: 10.3847/1538-4365/ab289b [12] The SunPy Community, MUMFORD S J, CHRISTE S, et al. SunPy-python for solar physics[J]. Computational Science & Discovery, 2015, 8(1): 014009 doi: 10.1088/1749-4699/8/1/014009 [13] 叶茜, 宋乔, 薛炳森. 基于活动区面积统计的F10.7指数预报方法[J]. 空间科学学报, 2019, 39(5): 582-590 doi: 10.11728/cjss2019.05.582YE Qian, SONG Qiao, XUE Bingsen. F10.7 index forecasting method based on area statistics of solar active regions[J]. Chinese Journal of Space Science, 2019, 39(5): 582-590 doi: 10.11728/cjss2019.05.582 [14] 朱凌锋. 多尺度日面信息的生成式度量模型空间天气预报[D]. 北京: 中国科学院大学, 2021ZHU Lingfeng. Generative Model Space Weather Forecasting with Multi-Scale Solar Information[D]. Beijing: University of Chinese Academy of Sciences, 2021 [15] LI H Y, DONG W M, MEI X, et al. LGM-Net: Learning to generate matching networks for few-shot learning[C]//Proceedings of the 36 th International Conference on Machine Learning. Long Beach, USA: ICML, 2019: 3825-3834 [16] BLOOMFIELD D S, HIGGINS P A, MCATEER R T J, et al. Toward reliable benchmarking of solar flare forecasting methods[J]. The Astrophysical Journal Letters, 2012, 747(2): L41 doi: 10.1088/2041-8205/747/2/L41 -
下载:
点击查看大图
图(3) / 表(4)
计量
- 文章访问数: 44
- HTML全文浏览量: 20
- PDF下载量: 19
- 被引次数: 0
