Forecasting GICs and Geoelectric Fields From Solar Wind Data Using LSTMs: Application in Austria

The forecasting of local GIC effects has largely relied on the dB/dt as a proxy and, to date, little attention been paid directly geoelectric field or GICs themselves. We approach this problem with machine learning tools, specifically recurrent neural networks LSTMs by taking solar wind observations input and training models predict two different kinds output: first, components Ex Ey; second, in specific substations Austria. is carried out modeled from 26 years one-minute geomagnetic measurements, results are compared measurements recent years. generally predicted better an LSTM trained values substation, but only fraction largest correctly predicted. This model correlation around 0.6, root-mean-square error 0.7 A. probability detecting mild activity 50%, 15% for larger GICs.