Prediction of Kp index using NARMAX models with a robust model structure selection method

Abstract

Abstract-The severity of global magnetic disturbances in Near-Earth space can crucially affect human life. These geomagnetic disturbances are often indicated by a Kp index, which is derived from magnetic field data from ground stations, and is known to be correlated with solar wind observations. Forecasting of Kp index is important for understanding the dynamic relationship between the magnetosphere and solar wind. This study presents 3 hours ahead prediction for Kp index using the NARMAX model identified by a novel robust model structure detection method. The identified models are evaluated using 4 years of Kp data. Overall, the models with robust structure can produce very good Kp forecast results and provide transparent and compact representations of the relationship between Kp index and solar wind variables. The robustness and conciseness of the models can highly benefit the space weather forecast tasks.

Kp forecasting with a recurrent neural network

“…A technique that has proven successful in modeling highly nonlinear systems known as NARMAX (nonlinear autoregressive moving average model with exogenous inputs) has been studied as a means of predicting Kp (Balikhin et al, 2011;Wei et al, 2011;Rasttter et al, 2013;Gu et al, 2017). Similar to the network proposed here, NARMAX models are fit to a lagged time series and make predictions for some time t in the future.…”