Practical applications for the machine intelligent partial discharge disturbing pulse suppression system NeuroTEK II

Partial discharge (PD) measurements, especially for on-site applications, have the problem that PD information is superimposed with electromagnetic disturbances. To solve this problem, a fast machine intelligent signal recognition system for the suppression of disturbing pulses was developed. This system distinguishes between PD and pulse shaped noise by referring to the time-resolved signal shape in real-time using a neural network hardware in a PC host. In this paper, practical applications for the disturbance pulse suppression are presented. Therefore measurements with a laboratory test set-up and artificial PD sources as well as results from an on-site measurement in gas insulated switchgear (GIS) were performed. Another application is the disturbance free determination of a PD fault location on-site.