Application of LSTM Neural Network in F10.7 Solar Radio Flux Mid-term Forecast
-
摘要: F10.7指数作为大气密度经验模型的重要输入参量,其预报精度直接影响航天器轨道预报精度.研究发现,太阳活动表现出长时间尺度上平均11年和中短时间尺度平均27天的周期性变化特征.依据这一观测事实,基于长短期记忆单元(Long Short-term Memory,LSTM)递归神经网络方法进行F10.7指数未来27天的中期预报.利用一个连续长时段F10.7数据作为训练数据,构建LSTM神经网络训练和预测模型,分别预测太阳活动高低年未来27天的F10.7指数.结果表明,太阳活动高年的第27天F10.7指数预报平均相对误差最优可达10%以内,低年最优可达2%以内.Abstract: The F10.7 index is an important input parameter for the empirical models of atmospheric density, and its prediction accuracy directly affects the accuracy of spacecraft orbit prediction. The solar activity exhibited an average of 11 years on a long-term scale and a 27-day periodic variation on a short-term scale. Based on this observational fact, a l Long and Short Term Memory (LSTM) recurrent neural network method is proposed to conduct the mid-term forecast of F10.7 index for the next 27 days. Using a continuous long period of F10.7 data as training data, the LSTM neural network training is constructed, and the upper and lower bounds of model parameters based on empirical formula are determined. The method of trial and error is used to select the optimal model parameters, and the prediction models to predict solar activity of high and low years F10.7 index in the next 27 days are constructed. The results show that the average relative error of the 27th day F10.7 index forecast for solar activity in the high year can reach about 10%, and can reach 2% or less in the low year. In 1998, the correlation coefficient between the predicted value of the F10.7 index on the 27th day and the measured value was 0.60.
-
Key words:
- F10.7 /
- LSTM neural network /
- Medium-term forecast
-
[1] YIN F, MA S Y, LI J, et al. Simulation of orbit decay for LEO satellites caused by atmospheric drag[J]. Chin. J. Geophys., 2013, 56(12):3980-3987(尹凡, 马淑英, 李晶, 等. 大气阻力引起卫星轨道衰减的数值模拟[J]. 地球物理学报, 2013, 56(12):3980-3987) [2] LIU Siqing, ZHONG Qiuzhen, WEN Jing, et al. Modeling research of 10.7cm solar flux 27-day forecast (I)[J]. Chin. J. Space Sci., 2010, 30(1):1-8(刘四清, 钟秋珍, 温靖, 等. 太阳10.7cm射电流量中期预报模型研究(I)[J]. 空间科学学报, 2010, 30(1):1-8) [3] WANG Xin. Deep learning for mid-term forecast of daily index of solar 10.7cm radio flux[J]. J. Spacecraft TT&C Technol., 2017, 36(2):118-122(王歆. 基于深度学习的太阳10.7cm射电流量日值的中期预报[J]. 飞行器测控学报, 2017, 36(2):118-122) [4] MA Ruiping, JI Qiao, XU Jiyao. Wavelet analysis of quasi-27-day oscillations in the solar index F10.7[J]. Chin. J. Space Sci., 2007, 27(2):89-95(马瑞平, 纪巧, 徐寄遥. 太阳F10.7指数准27天振荡的小波分析[J]. 空间科学学报, 2007, 27(2):89-95) [5] ZHONG Qiuzhen, LIU Siqing, HE Juanxiong, et al. Application of singular spectrum analysis to solar 10.7cm radio flux 27-day forecast[J]. Chin. J. Space Sci., 2005, 25(3):199-204(钟秋珍, 刘四清, 何卷雄, 等. 奇异谱分析在太阳10.7cm射电流量中期预测中的应用[J]. 空间科学学报, 2005, 25(3):199-204) [6] WEN Jing, ZHONG Qiuzhen, LIU Siqing. Model research of 10.7cm solar radio flux 27-day forecast (I!I)[J]. Chin. J. Space Sci., 2010, 30(3):198-204(温靖, 钟秋珍, 刘四清. 太阳10.7cm射电流量中期预报模型研究(I!I)[J]. 空间科学学报, 2010, 30(3):198-204) [7] WANG Hongbo, XIONG Jianning, ZHAO Changyin. The Mid-term forecast method of solar radiation[J]. Acta Astron. Sin., 2014, 55(4):302-312(汪宏波, 熊建宁, 赵长印. 太阳辐射指数F10.7的中期预报方法[J]. 天文学报, 2014, 55(4):302-312) [8] MIAO Juan, LIU Siqing, XUE Bingsen, et al. Primary research on prediction method of 10.7cm solar radio flux[J]. Chin. J. Space Sci., 2003, 23(1):50-54(苗娟, 刘四清, 薛炳森, 等. 太阳10.7cm射电辐射流量预报方法初探[J]. 空间科学学报, 2003, 23(1):50-54) [9] XU Yiaoqiang, FANG Yueheng, ZHAO Donghua, et al. Electricity consumption prediction based on LSTM neural networks[J]. Power Syst. Technol., 2017, 20(8):25-29(徐尧强, 方乐恒, 赵冬华, 等. 基于LSTM神经网络的用电量预测[J]. 电网技术, 2017, 20(8):25-29) [10] REN Jun, WANG Jianhua, WANG Chuanmei, et al. Stock index forecast based on regularized LSTM model[J]. Comput. Appl. Software, 2018, 35(4):44-48, 108(任君, 王建华, 王传美, 等. 基于正则化LSTM模型的股票指数预测[J]. 计算机应用与软件, 2018, 35(4):44-48, 108) [11] ZHAO Zhenyu, XU Yongmao. Introduction to Fuzzy Theory, Neural Networks and Their Application[M]. Beijing:Tsinghua University Press, 1996(赵振宇, 徐用懋. 模糊理论和神经网络的基础与应用[M]. 北京:清华大学出版社, 1996) [12] HEATON J. Ian Goodfellow, Yoshua Bengio, and Aaron Courville:deep learning[J]. Genet. Program. Evol. Mach., 2017, 19(1/2):1-3 [13] NASSER J M, FAIRBAIRN D R. The application of neural network techniques to structural analysis by implementing an adaptive finite-element mesh generation[J]. Artif. Intell. Eng. Design Anal. Manuf., 1994, 8(3):177 [14] ZHANG Qingliang, LI Xianming. A new method to determine hidden note number in neural network[J]. J. Jishou Univ.:Natl. Sci. Ed., 2002, 23(1):89-91(张清良, 李先明. 一种确定神经网络隐层节点数的新方法[J]. 吉首大学学报(自然科学版), 2002, 23(1):89-91) [15] GAO Daqi. On structures of supervised linear basis function feed forward three-layered neural networks[J]. J. Circuits Syst., 1997, 21(3):31-37(高大启. 有教师的线性基本函数前向三层神经网络结构研究[J]. 电路与系统学报, 1997, 21(3):31-37) [16] DING Yongsheng. Computational Intelligence:Theory, Technology and Application[M]. Beijing:Science Press, 2004(丁永生. 计算智能:理论, 技术与应用[M]. 北京:科学出版社, 2004)
点击查看大图
计量
- 文章访问数: 1147
- HTML全文浏览量: 201
- PDF下载量: 121
- 被引次数: 0