Operating state monitoring and fault diagnosis on traction power supply system

How to extract fault feature and divide pattern space is very important for the fault diagnosis. The traveling wave signal characteristic of over-head contact system is analyzed on the base of electromagnetic field theory; the eigenvector is constructed by the wavelet packet energy spectrum. The tests prove that the method in which the traveling wave feature is extracted in terms of wavelet packet energy spectrum is very available. The arithmetic of the pattern recognition is simplified using the method and the accuracy of the fault diagnosis is improved obviously.