The Signal Analysis And Interference Suppression Of Power Transformer Partial Discharge

Large power transformer is one of the most important and expensive equipments inpower industry. Its operation reliability involves the security of power supply and hassignificant impact on national economy development and people’s lives. Oil-paperinsulation is the most common insulation method of power transformers, As insulationmedia of the power transformer have to withstand heat, electricity impact long-term, whichresults in partial discharge at run-time. the event of partial discharge would affectedseriously of the transformer safe operation. Therefore, carry the on-line monitoring ofoperation Transformer’s Partial Discharge is a very good response to internal insulation oflatent defects to determine the extent of transformer inside of insulation deterioration, Andhas very important significance to the safe and reliable operation for transformer. Based onthe simulation experiment system of transformer partial discharge in the laboratory,thispaper had an in-depth study on the two vital areas of signal detection and patternrecognition, the main research works of this dissertation are shown as follows:①3kinds of experimental models simulating discharges were designed and modelexperiments were performed. with the constant pressure method to study the partialdischarge characteristics. With time growth, we collected the partial discharge signals atdifferent time points②Through the analysis of the common interfering signals about partial dischargedetection, the method using wavelet thresholds estimation is applied in white-noisesuppression and the cascade digital filter is applied in narrow-band noise suppression.Meanwhile,a de-noising method based on wavelet thresholds combined with the cascadedigital filter is applied in mixture noise suppression. By comparing the simulatedPD-signals with real PD-signals measured by online measuring system, it shows that theproposed method is an effective tool to remove noise.③Aim at the problem of neural network input choice, a new way which usestatistical operator to generate input vector was brought forward. the analysis andsimulation results show that three-layers neural network based on modified algorithm canbuild clear mapping between partial discharge data and models of power transformers.