Partial Discharge Inference by an Advanced System. Analysis of Online Measurements Performed on Hydrogenerator

A method for discriminating original pulses in online partial discharge measurement

“…Fortunately, the shape of PD pulses has useful information about the location and type of PD source. Till now, several parameters are introduced and evaluated for PD signal inference [11][12][13][14][15]22], i.e., gravity center and deviation of the normalized signal in time and frequency domain (equivalent time length r T and equivalent bandwidth r F ), pulse height and rise time. For a digitized pulse shape p(n), the gravity center in time l T [23] and frequency domain n F are measured as follows:…”

“…According to Heisenberg uncertainty, it has been proven that r 2 T r 2 F P 1 4 [23] and therefore these two features are almost dependent variables. Several efforts have been done to use these two conventional features for pulse identification and classification [13][14][15]. But the main questions are: ''are these two features optimal for PD discrimination especially for online measurement in noisy environments?"…”

“…The proposed features use first moment for calculating the averages and deviations as seen in (13)(14)(15)(16). Unlike the common standard deviations these two introduced features are more robust in online measurements and they are less sensitive to the interferences and they produce more concentrated clusters than conventional features.The next proposed feature is total variation that measures the total oscillation of a continuous signal [23] and is defined by:…”

“…Refs. [11][12][13][14][15] have used an ultra wideband measuring system in identification and localization of PD pulses and reduction of interferences based on the pulse shape features. Finally, in [16] a novel wavelet based technique has been proposed for extracting PD signals embedded in high noise level.…”

Classification and separation of partial discharge signals by means of their auto-correlation function evaluation

“…Thus, they were merged to form the PRPD pattern reported in Figure 10. The visual inspection, [1,11], and the identification system, [67], applied to the pattern of Figure 10 provided the indication of "insulation-conductor delamination".…”