Classification based Detection of Geomagnetic Storms using LSTM Neural Network

Solar activity influences near-Earth space phenomena by releasing solar flares, high speed solar wind, and CMEs which tend to give rise to geomagnetic storms. They affect the Earth’s magnetic field and upper atmosphere by interfering with the radio signals sent from space and hinder GNSS technology that relies on the accuracy of these signals. If information about the onset of the storms is available even a few hours prior to the actual storm, it can help in estimating the economic and global financial losses that ensue because of them. A large percentage of our daily work relies on accurate positioning, navigation and timing. Promising results have been achieved in the recent years for predicting Dst by using artificial neural networks. In this paper, a neural network technique called the LSTM is designed which recognises numerical time series data. An algorithm that detects the Dst, an hourly measure of the ring current in the form of time series data, using other geomagnetic indices has been proposed. Preliminary results exhibit a positive response to the classification based detection technique used in this paper.