This thesis deals with insulation diagnosis technology by analysis of partial discharges(PDs) for a 22.9kV cast resin transformer.
Patch-type capacitive probes were proposed as sensor to detection PD, which were installed on the vessel and completely insulated from high voltage of the transformer. A low-noise, high-gain amplifier with frequency ranges from 500kHz to 35MHz at -3dB was fabricated, including the capacitive probes. Two kinds of the cast resin transformer defects were simulated using needle electrodes with a radius of 5μm and 10μm. To analyze the phase(Φ), magnitude(q) and numbers(n) of PD pulses as the defects, virtual instrument system(VIS) consisted of a digitizer(NI PXI-5114, 125MHz, 8bit, 250MS/s) and an analysis program was fabricated. Also, a denoise algorithm using the wavelet transform was designed to separate the internal and external noise from the PD signals.
Experimental results, proved that the proposed insulation diagnosis system could detect the PD of 10pC for a cast resin transformer. In the results of analysis on the Φ-q-n pattern, the qmax and numbers of the PD pulses showed a tendency to decrease from 26.3pC to 23.9pC and from 2,219 times to 1,422 times respectively as the radius of the defects decreased from 10μm to 5μm. Also, the phase distribution of PD pulses shifted in range from phase 30°～60°(+) and 210°～240°(-) to phase 60°～90°(+) and 240°～270°(-) as the acutance of the defects increased.
In conclusion, it is possible to presume strange defects by analysis on the Φ-q-n pattern and the approximate location of the PD source to install several capacitive probes on the vessel of the cast resin transformer.