PREDICTION OF MAJOR STORMS BY USING NEURAL NETWORKS

Good prediction results for geomagnetic indices have been obtained with the use of artificial intelligent(AI)technique,especially by the neural networks.While a common Problem is how to increase the advance time of the prediction.It is true that the earlier of the prediction, the less possibility to predict the detail of an impending event.Therefore threshold prediction should be also a very helpful and useful method for early time prediction when forecasting the occurrence of the events which bring disaster on the cormmunication,power transmission and satellite life etc. Threshold prediction means the predicted events are over a threshold,such as-120nT. In this paper BP model of neural networks is used to predict the major storms 4 hours ahead, with the minimum of Dst less than-120nT. The inputs are 13 solar wind parameters and geomagnetic AE and Dst.The ought is I for major storms or 0 for non major storms. The results show that the threshold prediction by neural networks can give accurate prediction for major storm occurrence.