Using artificial neural networks in the investigation of four moderate geomagnetic storms (mGSs) that occurred in 2015

Abstract

This essay aims to investigate the four moderate Geomagnetic Storms (mGSs) (storms of March 02, April 11, April 16, and May 13) that occurred in 2015. When examining GSs, Zonal Geomagnetic Indice (ZGi) values are predicted with an Artificial Neural Network model (ANNm). GSs are evaluated with the Solar Wind Parameters (SWp) (Bz, E, P, N, v, T) and the ZGi (Dst, ap, AE) obtained from the OMNI web. The study displays the ZGi estimations of 2015 with their errors, specifies the dual relationships of the variables with the covariance matrix, and exhibits the classified characteristics of the data with the Cluster Analysis. The physical principles govern the ANNm whose inputs are the SWp and outputs are the ZGi. For each event data; training, validation, and testing set ratio are regarded as 70%, 10%, and 20%, respectively. The backpropagation algorithm is determined as Scaled Conjugate Gradient (trainscg). The network performance is evaluated by the Correlation Constant (R) and Root Mean Square Error (RMSE) (nT) scales for each geomagnetic activity. Hereby, the study argues that the model is reliable, with an R-value of up to 89% and an RMSE value of down to 1.34 nT.