Prediction of VLF propagation signal JJI Japan disturbed by lightning using artificial neural networks: Preliminary results

This paper discusses the prediction of very low frequency (VLF) propagation signal that is perturbed by lightning using artificial neural networks (ANN). In this study, the VLF signal is from the transmitter station at Ebino, Japan. Direct interference from lightning discharge is one of the disturbances where the broad spectrum of electromagnetic radiation generated will be trapped in the wave guide formed by the ground and the lower ionosphere. This will trigger the change in the amplitude of the VLF wave propagation that spreads in the sub-ionosphere. This problem illustrates the need to predict the disturbance to the propagation of VLF signals caused by lightning discharge that could result in communication disorders. In this work, a prediction method using ANN with multilayer back propagation is proposed. Other techniques used by the previous researches include theoretical models, statistical models, empirical model and wave hop field strength method. All these techniques require complex mathematical function and are less significant. The proposed method showed better performance in comparison to the conventional approaches in that for the forecasting results of the observations obtained using statistics, the mean percentage error is only 1.822 while for the results of the prediction using the neural network the mean percentage error is 0.1381.