Assessment of utility of the corona audible signal in the diagnostic processof the technical condition of UHV transmission lines

The paper deals with application of neural network techniques to recognition of typical damages of the transmission circuit elements in UHV power lines. The damage indicator is the acoustic signal generated by the corona effect, the intensity of which increases as a result of damage or contamination of the conductor surface. The primary difficulty in the diagnostic process is the necessity to distinguish between the signals generated by the surface damages and contaminations. The problem can not be solved neither by the analysis of RF interference signal nor by the classical methods of the acoustic signal analysis. Particular attention has been focused on the parametrization of the acoustic signal of corona effect and selection of distinctive features, which could be useful in distinguishment of the origins of the increased corona effect in dry conductor surface conditions. For that task the aggregation analysis has been applied, in particular the possibilities offered by "Statistica"[ampersand]reg; software package. The data used in the analysis has been obtained from laboratory studies in Institute of Power Engineering in Warsaw, where typical surface damages and contaminations of the transmission line elements have been simulated. Keywords: corona noise, diagnosis, acoustics, cluster analysis, neural network.