Short Term Forecast of the F10.7 Solar Flux Index based on Neural Networks

The F10.7 solar flux index reflects the intensity of solar activity and serves as a significant input parameter for the MSIS model of the atmosphere. Accurate forecasting of the F10.7 index aids in enhancing the accuracy of atmospheric density predictions, which is of paramount importance for space missions such as spacecraft collision warning. The forecasting methods for the F10.7 index mainly involve physical models, statistical models and machine learning models. Specifically, physical models can predict the F10.7 index based on the physical laws, while statistical models and machine learning models utilize historical data and algorithms for prediction. In this paper, a neural network model for short term forecasting F10.7 index is studied. The model is composed of a long short term memory(LSTM) network and a feed forward network(FFN). Observational data from 1957 to 2019 is used to optimized the model. Simulation results show that the trained model exhibits an average absolute percentage error of less than 3.70% and 5.73% in forecasting the F10.7 indices three days in advance for year 2021 and 2022, respectively. The predicted F10.7 solar flux indices present good correlation with the actual observational data, proving efficacy of the model.