Comparison of the first four weak and moderate geomagnetic storms of the 2022 using artificial neural networks

Abstract

Solar wind parameters (SWP) and geomagnetic indices act as defining factors between the Earth's magnetosphere-ionosphere and the Sun. Models between SWP and indices should be examined seriously to better interpret the dynamics of geomagnetic storms (GSs). This work touches on the first two weak (25 January and 22 February) and moderate (14 January and 13 March) GSs of the 2022 year with an artificial neural network (ANN) model. The models are based on SWP (E, v, P, T, N, Bz) and geomagnetic indices (Dst, Kp, ap). The ANN employs SWP as inputs and indices as outputs, predicting the indices via SWP. The Scaled Conjugate Gradient (trainscg) algorithm is used for the back-propagation iteration. Four different geomagnetic storms, which are weak; Dst = -35 nT (25 January), Dst = –32 nT (22 February), and moderate storms; Dst = -91 (14 January) nT, Dst = -85 nT (13 March), are considered as case studies. The results of the estimations have acceptable accuracy. While the model is trained with mean square error (MSE) loss function, the accuracy of the model is evaluated by both mean square error (MSE) and mean error (ME).