EXTRACTION OF THE VECTOR OF DISTINCTIVE FEATURES FOR THE ACOUSTIC SIGNAL FROM CORONA IN OVERHEAD POWER LINES

摘要

One of possible practical benefits of the corona process is the possibility to evaluate the technical condition of some line elements using the acoustic signal accompanying the corona discharge. That signal can be easily recorded and its analysis, making use of the contemporary signal processing techniques and artificial intelligence methods, may be the source of information concerning the line damages, which are very difficult to detect and assess by other methods. The essence of the problem is the fact that useful information about the forms and locations of the UHV line damages is hidden in the structure of the signal and has to be extracted, in which process the artificial intelligence techniques applied by the authors can be very helpful. Some distinctive features of such phenomena can be extracted in laboratory but in real world conditions, there may be some problems with collecting a properly representative learning set. The paper presents some results gathered in real world conditions and an extraction method of the vector of distinctive features, based on cluster analysis method).