On the correlation among partial discharge patterns and degradation in large power transformers

In this paper the implementation of a pattern recognition system for the condition monitoring of power transformers is discussed. Pattern recognition is used in the interpretation of partial discharge measurements in HV plant variously and this paper discusses its application in Power Transformers. The implementation of such techniques may use various signal processing methods but it is absolutely essential that the patterns on which the system is `trained´ reflect the possible effects in practice. This must include both possible geometries and any other parameters which could influence the pattern. Examples are given of metal/insulator configurations used, that emulate degradation in transformers, that give different partial discharge patterns depending upon the field stress and extent of degradation