Study On On-line Monitoring And Pattern Recognition Of Partial Discharge Based On Ultra-high-frequency Sub-band Energy

Partial discharge (PD), which does not completely bridge the insulation betweenthe terminals, damages the insulation of equipment and indicates the attenuation ofcomplete insulation. According to that, it is important method to detecting partialdischarge of electrical equipment for evaluating the insulation condition of apparatus.Since partial discharge is a paroxysmal and latent phenomenon, it can hardly be foundin time based maintenance. So, on-line monitoring, which can report delitescentinsulation fault timely, is a more effective method to avoid insulation fault. After takingcondition monitoring and fault diagnosis technology, it allows the transition frompreventive maintenance to predictive maintenance which means that state maintenance.From the “maturity required repairing transition” to “necessity required repairing”. Ifthe analysis of the offline preventive test results has accumulated a wealth of experienceand has been worked out the appropriate procedures to popularize to implement, thenfor online diagnosis, it is still under the stage of research, trial operation andaccumulation of experience. To develop insulation online diagnostic technology, notonly the deeply understanding of the mechanism of insulation structure aging, but alsothe application of sensing, microelectronics and other high-tech are needed. It is anemerging technology with interdisciplinary nature, and it has significant academic andeconomic worth.Among these PD detection techniques, ultra-high-frequency (UHF) method iswidely used because of its high sensibility and good interference rejection. In this paper,the application of UHF method in PD on-line monitoring is analyzed and a new type ofcoin-shaped fractal antenna is designed. The antenna has a size of250mm×250mm×1mm, and a bandwidth of400-1000MHz according to VoltageStanding Wave Ratio(VSWR) less than2, which means that the bandwidth is2.5:1.Meanwhile, a mean gain of1.7dB and an omnidirectional radiation property are gotten,which allows it to effectively receive the weak UHF signals from every direction.Through testing, the antenna with a high signal to noise ratio (SNR) is a set ofsmall-sized, high-gain, ultra-wideband, and omni-directional UHF sensor. However,The frequency of UHF signal received by antenna is pretty high(300-3000MHz), whichmeans a great challenge for sampling and collecting, as well as increases the complexityfor post process. This shortcoming decreases the performance-price-ratio of UHF employment. In this paper, after a comprehensive analysis of the partial discharge UHFsignal characteristics, the author proposes to make use of mixing Proposed mixingtechniques and the use of an envelope detector technology, design a mixer detectorcircuit and envelope detection technology, and designs a mixer-detector circuit. Thecircuit can transform UHF signal to a lower frequency under the condition ofunchanging PDs’ peak value and phase. Therefore, a common ADC will work to satisfythe sampling law.According to experience, the UHF energy spectrum of different types of dischargedifferences distinctly. This paper attempts to the method of using sub-band energy ratiofor pattern recognition. The author introduces of four typical types of partial dischargeas research subjects, and divides the full400-1000MHz band into three consecutivesections evenly, as the lower band of400-600MHz, the middle band of600-800MHz,the higher band of800-1000MHz. Then by means of extraction the percentage of thethree sub-band energy of the full-band energy, which works as fingerprint patternrecognition feature amount, the paper depicts a triangular plot. In the plot, the energydistributions of different PD models are significantly different; on the contrary, theenergy distributions of same PD models are largely similar. They present apparently afeature of “regional gathering” in the triangular plot. Furthermore, a cluster analysis ofthis phenomenon based on the Mahalanobis distance is launched in this paper.Meanwhile, this paper quantifies the fingerprint feature amount of the patternidentification method mentioned above. And a scientific evaluation criteria is putforward to assess the pattern recognition method, the evaluation results show that themethod has great accuracy and feasibility.