Location Technique for Multiple Partial Discharge Sources using Independent Component Analysis and Direction of Arrival Method of Electromagnetic Waves based on Bayesian Network
Hirokazu Ishimaru, Masatake Kawada · IEEJ Transactions on Power and Energy · 2006
We propose the new locating method for multiple partial discharge sources using independent component analysis (ICA) and direction of arrival method of electromagnetic (EM) waves based on Bayesian Network. The proposed method consists of the following steps. First, the observed signals are converted from the time domain to the frequency domain through the short time Fourier transform (STFT). As the result of this process, ICA for convolved mixture turns into ICA for instantaneous mixture. In order to separate mixed signals, we use the fast ICA algorithm (FastICA) that is based on negentropy as a measure of nongaussianity. Next, we apply the direction of arrival method of EM waves using Bayesian Network to the separated source signals by ICA. Signal sources are estimated by calculating intersection point from the arrival angle. Simulation results showed that our method located multiple signal sources when the observed signals were the convolved mixture of original sources.