Analysis and Forecasting of Geomagnetic Field Signal in Active Period

摘要

Geomagnetic variation is divided into quiet period and active period, while active period is non-periodic and random. This paper utilizes analyze the magnetic variation field signals in active period in time-frequency domain, then models and forecasts the signals by using Artificial Neural Networks. The results show that the 4-hour-forecasting error of the three methods is MAE ≤ 2.3nT. In multi-step forecasting, the method of GRNN is smooth, while LNN causes error increasing apparently. RBFNN has the best performance as its MAE is the smallest one for each time.