基于VMD-TCN的太阳活动指数F_10.7的短期预报

Short-Term Prediction of Solar Activity Factor F_10.7 Based on VMD-TCN Model

摘要: F_10.7是波长为10.7 cm太阳辐射通量，它经常作为不同空间天气模型、大气密度经验模型中一个重要的参数输入，同时也是最广泛使用的衡量太阳活动水平强弱的参数之一，因此对其进行准确的预报是空间天气研究的一个重点关注的方向。本文利用1957~2019年观测的太阳辐射通量F_10.7数据建立了基于VMD-TCN的F_10.7预测模型，其中将1957~2008年作为模型训练集，2009~2019年作为模型的测试集，并使用留一法对数据集进行分组，进行多次交叉验证。实验结果表明：VMD-TCN模型预报F_10.7的均方根误差1~4 sfu，平均绝对误差百分比为0~2 sfu，相关系数高达0.99。与基于相同数据集建模的其他模型如TCN，VMD-ELM模型相比，VMD-TCN模型预测F_10.7的均方根误差更低，相关系数更高，模型综合预测效果最佳。

Abstract: F_10.7 is the solar radiation flux with a wavelength of 10.7 cm, which is often used as an important parameter input in different space weather models, empirical models of atmospheric density, and is also one of the most widely used parameters to measure the strength of solar activity level, so its accurate forecasting is a key concern in space weather research. In this paper, a F_10.7 prediction model based on VMD-TCN is established using the observed solar radiation flux F_10.7 data from 1957~2019, in which 1957~2008 is taken as the training set of the model, and 2009~2019 as the test set of the model, and the datasets are grouped by using the leave-one-out method for multiple cross-validations. The experimental results show that the VMD-TCN model predicts F_10.7 with the root mean square error of 1~4 sfu, the average absolute error percentage of 0~2 sfu, and the correlation coefficient is as high as 0.99. Compared with other models based on the same dataset, such as the TCN, VMD-ELM models, the VMD-TCN model predicts F_10.7 with a lower root mean square error and a higher correlation coefficient, which gives the best overall prediction effect of the model.