Forecasting Of Partial Ring Current Index Using LSTM Artificial Neural Network

Geomagnetic storms profoundly affect the production and life of human society,and the establishment and development of forecasting models for geomagnetic storms is one of the important topics in the field of space weather science.With the help of Long Short-Term Memory(LSTM)artificial neural network which has good time series forecasting performance,we establish a forecasting model for local circulation index and try to improve the model for forecasting geomagnetic storms up to 4 days in advance by combining source surface parameters.The local time dependence of the geomagnetic disturbances during magnetic storms indicates the necessity of forecasting the localized magnetic storm indices.For the first time,we construct prediction models for the Super MAG partial ring current indices(LTSMR-),with the advance time increasing from 1 to 12 hours by using LSTM neural network.Generally,the prediction performance decreases with the advance time and is better for the SMR06-index than for the SMR00-,SMR12-,and SMR18-index.For the predictions with 12 hours ahead,the correlation coefficient is 0.738,0.608,0.665,and 0.613,respectively.To avoid the over-represented effect of massive data during geomagnetic quiet periods,only the data during magnetic storms are used to train and test our models,and the improvement in prediction metrics increases with the advance time.For example,when predicting the storm-time SMR06-index with 12 hours ahead,the correlation coefficient and the prediction efficiency increases from 0.674 to 0.691,and from0.349 to 0.455,respectively.We also evaluate the model performance for forecasting the storm intensity,finding that the relative error for intense storms is usually less than that for moderate storms.Forecasts for the geomagnetic index Dst rapidly decline in model effectiveness when the advance reaches 24 hours and beyond.Geomagnetic storms are one of the consequences of the enhanced solar wind-magnetosphere energy coupling,whose physical source lies in solar activity.By incorporating the source surface parameters into an artificial neural network,we build a Dst prediction model with advance times of 1 to 4 days.The results show that the addition of the source surface training parameters improves the forecasting effect of the Dst index in the order of days,and the longer the advance time,the greater the improvement of the forecasting effect.The model has correlation coefficient of 0.352 and prediction efficiency of 0.103 for forecasts with 4 days advance,which are improved by 22.2% and 77.6%,respectively.