Abstract
Abstract
Solar activities lead to Sun variation with an 11 yr periodicity. The periodic variation affects space weather and heliophysics research. So it is important to accurately predict solar cycle variations. In this paper, we predicted the ongoing Solar Cycle 25 using neural basis expansion analysis for the interpretable time series deep learning method. 13 months of smoothed monthly total sunspot numbers taken by sunspot Index and Long-term Solar Observations are selected to train and evaluate our model. We used root mean square error (RMSE) and mean absolute time lag (MATL) to evaluate our model performance. RMSE and MATL measure the difference between our predicted values and the actual values along the Y- and X-axis, respectively. The RMSE value is 26.62 ± 1.56 and the MATL value is 1.34 ± 0.35, demonstrating that our model is able to better predict sunspot number variation. Finally, we predicted the variation of the sunspot numbers for Solar Cycle 25 using the model. The sunspot number of Solar Cycle 25 will peak around 2024 February with an amplitude of 133.9 ± 7.2. This means that Solar Cycle 25 will be slightly more intense than Solar Cycle 24.
Publisher
American Astronomical Society
Subject
Space and Planetary Science,Astronomy and Astrophysics
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. Solar cycle prediction using a combinatorial deep learning model;Monthly Notices of the Royal Astronomical Society;2023-11-08