Suppression Of The Periodic Narrow-band Noise With Mixed Frequencies In Partial Discharge On-line Monitoring And Pd Location Of GIS

Partial Discharge (PD) in Gas Insulated Switchgear (GIS) that is one of the important factors for the fault of GIS must have been monitored. How to accurately obtain PD signal from strong electromagnetic interferences is the key to achieving real-time and on-line monitoring. Only exactly getting PD, can some follow-up researches such as PD location,PD verification,assessment of security status ,etc be developed.At present, when the multiple-frequency narrow-band interferences simultaneously occur in PD detection(narrow-band noise with mixed frequencies),there is no appropriate methods or measures for suppressing the narrow-band noise with mixed frequencies. In order to suppress the periodic narrow-band noise in the partial discharge (PD) on-line monitoring system, the different characteristics between PD signals and the narrow-band noise with respect to wavelet power spectrum were analyzed, it was found that the huge interfering power peak of the narrow-band noise with mixed frequencies appears in the wavelet power spectrum. Thus, a new method to suppress above noise was put forward based on this discovery, and the basic principles and process of applying this de-noise method were expounded.Besides, the K-mean threshold algorithm originated in classic threshold methods was taken to improve the self-adapting ability of this de-noise method with regard to the frequency of narrow-band noise with mixed frequencies. Principles and implementation steps of the K-mean threshold algorithm are explained clearly in this paper. The threshold-selecting formula corresponding to Wavelet power spectrum is given and the superiority of the K-mean threshold algorithm in regard to the classic threshold methods is shown from a theoretical point of view. The simulation researches and the results of the real signals'de-noising show that the proposed method in this paper has a fine self-adapting ability and strong narrow-band noise-suppressing capacity.After suppressing interferences and obtaining real PD signal, the PD source location in GIS can be started .In order to achieve precise location, the propagation characteristic of electromagnetic wave must be researched deeply. In this paper, the relationship between the distance of sensors to PD source and time delay characteristic and wave sharp has been deeply studied by simulation and experiment. It is found that the phenomenon of electromagnetic waves Catadioptric in GIS is very prominent, that when the distances between sensors and PD source are different, not only time delay of each PD signal is different, but also wave sharp of each one has discrepancy.Based on the above researches about the propagation characteristic of electromagnetic wave in GIS, the TDOA (time different of Arrival) that is applied PD source in GIS is improved in this paper. A new method that comparing peak value on power-curve is proposed .It is shown that this location method has more high accuracy, up to centimeter level and when white noise occured, the location results was more stable.At the same time, the relationship of installation position of the inner-ring sensor and the location accuracy has been researched.It is found that the long-to-radius ratio was more large, the theoretical error of location was more small.