Predicting the Solar Activity Cycle Based on A LSTM-ARIMA Hybrid Model

Yusen Zhou

doi:10.54097/r4r4rt85

Authors

Yusen Zhou

DOI:

https://doi.org/10.54097/r4r4rt85

Keywords:

solar activity, LSTM, ARIMA, Fusion Model.

Abstract

Solar activity has a profound impact on the Earth, and the development of effective models to accurately predict solar activity will help us to further our understanding of space weather changes caused by solar activity and their possible consequences.In this paper, data on the area of the solar magnetic field, the number of sunspots and the area of sunspot regions were collected and processed, and histograms and box plots of the distributions of the three types of data were obtained, and it was found that the former showed a normal distribution and the latter two showed a skewed distribution. Subsequently, the data were further processed using the sliding average method to obtain the smoothed trend time series and fluctuation time series of the above three groups of data.The use of a single common time series forecasting model to predict solar activity proves to be unreliable due to the uncertainty and non-linearity of the solar activity cycle and intensity. Therefore, in this paper, separate ARIMA and LSTM forecasting models are first constructed to predict the general and fluctuating trends of solar activity indicators, respectively. Then, a hybrid prediction model is built to integrate the prediction results of the two models through denormalization, weight adjustment and noise removal to obtain the final prediction results.This paper focuses on predicting the changes in the solar magnetic field area, the number of sunspots, and the sunspot area to further speculate on the beginning and end of solar activity and its intensity. Finally, this paper provides a comparative analysis, reliability validation, and parameterization of the established models. From the perspectives of sliding and rolling validation, this paper compares the stability and accuracy of the independent ARIMA prediction model and LSTM prediction model as well as the hybrid model in processing raw data. In terms of model parameter determination, this paper analyzes and determines the number of cycles, hidden layer grid size in the LSTM model and the autoregressive term, moving average term and difference order in the ARIMA model, respectively.

Downloads

Download data is not yet available.

References

E. W. Cliver, “Solar activity and geomagnetic storms: The first 40 years,” Eos, Transactions American Geophysical Union, vol. 75, no. 49, pp. 569–575, 1994.

S. W. Samwel, E. A. El-Aziz, H. B. Garrett, A. A. Hady, M. Ibrahim, and M. Y. Amin, “Space radiation impact on smallsats during maximum and minimum solar activity,” Advances in Space Research, vol. 64, no. 1, pp. 239–251, Jul. 2019.

Wei Sun, G. Li, J. Wang, Shuting Xu, Shiaoxiao Wei, and F. Liu, "Construction of magnetic polarity index for sunspot magnetic field and its periodicity analysis," Geophysical Progress, vol. 37, no. 4, pp. 1475-1483, Aug. 2022.

S. B. Waykar, A. W. Kale, and N. P. Karlekar, “Deep Learning-Based Solar Activity Prediction Using Sunspot Number and Solar Radio Flux,” in Advances in Parallel Computing, D. J. Hemanth, T. N. Nguyen, J. Indumathi, and S. Lakshmanan, Eds., IOS Press, 2022.

I. Krasheninnikov and S. Chumakov, “Forecasting the Sunspots Number Function in the Cycle of Solar Activity Based on the Method of Applying Artificial Neural Network,” in Artificial Intelligence Application in Networks and Systems, vol. 724, R. Silhavy and P. Silhavy, Eds., in Lecture Notes in Networks and Systems, vol. 724. Cham: Springer International Publishing, 2023, pp. 824–835.

C. J. Schrijver and C. Zwaan, Solar and Stellar Magnetic Activity. in Cambridge Astrophysics. Cambridge: Cambridge University Press, 2000.

G. Jensen, “Applied forecasting with an autoregressive integrated moving average (ARIMA) model,” 1977.

P. R. Low and E. Sakk, “Comparison between autoregressive integrated moving average and long short-term memory models for stock price prediction,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 12, no. 4, Art. no. 4, Dec. 2023.

A. Sherstinsky, “Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) network,” Physica D: Nonlinear Phenomena, vol. 404, p. 132306, Mar. 2020.

D. S. Mohanty, “An International Study of Application of Long Short-Term Memory (LSTM) Neural Networks for the prediction of stock and forex markets,” IJFMR - International Journal for Multidisciplinary Research, vol. 5, no. 3.

X.-H. Le, H. V. Ho, G. Lee, and S. Jung, “Application of Long Short-Term Memory (LSTM) Neural Network for Flood Forecasting,” Water, vol. 11, no. 7, Art. no. 7, Jul. 2019.

J. Gonzalez and W. Yu, “Non-linear system modeling using LSTM neural networks,” IFAC-PapersOnLine, vol. 51, no. 13, pp. 485–489, Jan. 2018.