GA analysis of discharge currents

A time series has been analyzed by several methods such as Fourier analysis, wavelet analysis and so on. However, these methods are quite unsatisfactory for prediction. In this paper, we describe a prediction method of a time series and attempt to predict discharge currents accompanied by tree development. In order to pick up characteristics of discharge currents, we implemented the wavelet transformation. A prediction method integrates the GA (genetic algorithm) into the GMDH (Group Method of Data Handling). The result shows that a network generated by GMDH can estimate a system generating the time series. This suggests a prediction possibility of discharge breakdown, generally dielectric breakdown and so on