Solar activity predicted with artificial intelligence

Abstract

The variability of solar activity has been described as a non-linear chaotic dynamic system. Al methods are therefore especially suitable for modelling and predicting solar activity. Many indicators of the solar activity have been used, such as sunspot numbers, FI0.7 cm solar radio flux, X-ray flux, and magnetic field data. Artificial neural networks have also been used by many authors to predict solar cycle activity. Such predictions will be discussed. A new attempt to predict the solar activity using SOHO/MDI high-time resolution solar magnetic field data is discussed. The purpose of this new attempt is to be able to predict episodic events and to predict occurrence of coronal mass ejections. These predictions will be a part of the Lund Space Weather Model.