Forecasting Solar Cycle 25 Using Deep Neural Networks

Benson, Bernard; Pan, W. David; Prasad, A.; Gary, G. A.; Hu, Qiang

doi:10.1007/s11207-020-01634-y

Cited by 50 publications

(28 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pala and Atici (2019) used the LSTM model and predicted a maximum sunspot number of 167.3 in July 2022. Benson et al (2020) estimated the peak sunspot number in Solar Cycle 25 to be 106 ± 19.75 around March 2025 ± 1 year by using a combination of the LSTM and WaveNet methods. Xiong et al (2021) predicted with multiple regression a peak of 140.2 in March 2024.…”

Section: Resultsmentioning

confidence: 99%

“…Although there have been many studies on predicting the sunspot number by using non-deep learning (Xu et al 2008;Hiremath 2008;Chattopadhyay, Jhajharia, and Chattopadhyay 2011;Tabassum, Rabbani, and Omar 2020) or deep learning forecasting methods (Pala and Atici 2019;Benson et al 2020;Arfianti et al 2021;Prasad et al 2022), most are based on ARMA models or deep learning methods like LSTM or GRU. Little work has been done on the more recent time-series models Prophet, Transformer, and Informer for the sunspot number prediction.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A comparative study of non-deep learning, deep learning, and ensemble learning methods for sunspot number prediction

Dang,

Chen,

et al. 2022

Preprint

View full text Add to dashboard Cite

Solar activity has significant impacts on human activities and health. One most commonly used measure of solar activity is the sunspot number. This paper compares three important non-deep learning models, four popular deep learning models, and their five ensemble models in forecasting sunspot numbers. Our proposed ensemble model XGBoost-DL, which uses XGBoost as a two-level nonlinear ensemble method to combine the deep learning models, achieves the best forecasting performance among all considered models and the NASA's forecast. Our XGBoost-DL forecasts a peak sunspot number of 133.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A comparative study of non-deep learning, deep learning, and ensemble learning methods for sunspot number prediction

Dang,

Chen,

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In order to determine which of the hidden units 15 and 19 would provide the most accurate forecast for solar cycle 25, we put our model through a battery of experiments at a variety of different time points utilising those hidden units. Benson et al (2020) suggested that the window for the dataset should include at least four solar cycles in order to accurately capture the long-term trend present within the data and provide a projection for the future. As a result, we have a database that spans over four solar cycles, with a particular focus on the maximum phase.…”

Section: Preparing the Datasetmentioning

confidence: 99%

Modulation of Cosmic Ray with Solar activities During Solar Cycles 19-24 to forecast Solar Cycle 25

Kumar

Pal

Rani

et al. 2022

Preprint

View full text Add to dashboard Cite

As of currently, a spotless event solar activity precursor technique is being employed to estimate the intensity and onset periods of future 11-year cycles. In the past, it was utilised to make predictions concerning the maximum amplitudes and rising times of cycles ranging from 19 to 24. The reassuring outcomes that have been obtained from all of these iterations can be summarised as follows: In the past, an updated variation of the proposed method was utilised in order to make a prediction concerning the cycle 24's defining parameters. The observed values of spotless events are utilised in the validation of the preliminary forecasts for cycle 24's characteristics. Additionally, the maximum amplitude and ascension period of the 25th solar cycle can be predicted with the assistance of this method. We have made projections on the forthcoming cycle 25 by utilising the stacking LSTM forecasting model. According to the findings of our study, the model that we devised is able to accurately predict both long-term dependencies and trends in the data. The maximum strength of solar cycle 25 will be 47 percent higher than that of cycle 24 thanks to the sunspot amplitude of 171.9 3.4.

show abstract

“…Only six were done for SC-24 (Prasad et al, 2022). Long Short-Term Memory (LSTM) neural network was used in combination with other models for the prediction of SC-25 (Pala and Atici, 2019;Benson et al, 2020;Lee, 2020;Prasad et al, 2022). Several machine learning methods were used by Dani and Sulistiani (2019) to compare the predicted peak amplitude of SSN for SC-25, and the obtained results were different among these methods, namely: 159.4 ± 22.3, 95.5 ± 21.9, 110.2 ± 12.8, and 93.7 ± 23.2 respectively for Linear Regression (LR), Radial Basis Function (RBF), Random Forest (RF) and Support Vector Machine (SVM), and peak occurring times of SC-25 would be September 2023, December 2024, December 2024 and July 2024.…”

Section: Introductionmentioning

confidence: 99%

Solar Cycle 25 Prediction Using an Optimized Long Short-Term Memory Mode with F10.7

Zhu

2022

Sol Phys

View full text Add to dashboard Cite

In this paper, an optimized long short-term memory (LSTM) model is proposed to deal with the smoothed monthly $F_{10.7}$ F 10.7 data, aiming to predict the peak amplitude of $F_{10.7}$ F 10.7 and the occurring time for Solar Cycle 25 (SC-25) to obtain the maximum amplitude of sunspot number (SSN) and the reaching time. The “re-prediction” process in the model uses the latest prediction results obtained from the previous prediction as the input for the next prediction calculation. The prediction errors between the predicted and observed peak amplitude of $F_{10.7}$ F 10.7 for SC-23 and SC-24 are 2.87% and 1.09%, respectively. The predicted peak amplitude of $F_{10.7}$ F 10.7 for SC-25 is 156.3, and the maximum value of SSN is calculated as 147.9, which implies that SC-25 will be stronger than SC-24. SC-25 will reach its peak in July 2025.

show abstract

Forecasting Solar Cycle 25 Using Deep Neural Networks

Cited by 50 publications

References 31 publications

A comparative study of non-deep learning, deep learning, and ensemble learning methods for sunspot number prediction

A comparative study of non-deep learning, deep learning, and ensemble learning methods for sunspot number prediction

Modulation of Cosmic Ray with Solar activities During Solar Cycles 19-24 to forecast Solar Cycle 25

Solar Cycle 25 Prediction Using an Optimized Long Short-Term Memory Mode with F10.7

Contact Info

Product

Resources

About