Introducing adaptive neurofuzzy modeling with online learning method for prediction of time-varying solar and geomagnetic activity indices

Research in space weather has in recent years become an active field of research requiring international cooperation because of its importance in hazard warning especially for satellite technology and power utility systems. The time-varying sun as the main source of space weather impacts the Earth’s magnetosphere by emitting hot magnetized plasma called solar wind into interplanetary space. The emission of Solar Energetic Particles (SEPs) and consequently the magnitude of Interplanetary Magnetic Field (IMF) vary almost periodically with an approximate life cycle of 11 years. It is shown that the solar and geomagnetic activity indices have complex behavior often characterizable as quasi-periodic or even chaotic, which causes the long-term prediction to be a conundrum. Moreover, solar and geomagnetic activity indices and their chaotic characteristics vary abruptly during solar and geomagnetic storms. This variation depicts the difficulties in modeling and long-term prediction of solar and geomagnetic storms. On the other hand, neural networks and related neurofuzzy tools as general function approximators have been the subjects of interest due to their many practical applications in modeling and predicting complex phenomena. However, most of these systems are trained by algorithms that need to be carried out by an off-line data set which influence their performance in prediction of time-varying solar and geomagnetic activity indices. This paper proposes an adaptive neurofuzzy approach with a recursive learning algorithm for modeling and prediction of space weather indices which fulfill requirements of prediction of time-varying solar and geomagnetic activities for long time spans. The obtained results depict the power of the proposed method in online prediction of time-varying solar and geomagnetic activity indices.