Research Of Partial Discharge Detection Method Based On Date Fusion

Partial discharge is not only an important factor in transformer life, but also an important factor of power system stability. However, due to the occurrence of location and time of partial discharge is very random, timely and accurately find the point of partial discharge has become an important index of the transformer detection, and it is also an important guarantee to the normal operation of power system. For partial discharge, the industry has developed a variety of detection methods, some of these methods have been successfully applied to partial discharge detection, and some will not put into practice only after further study. In this paper, based on comparison of various detection methods, the author proposed a method for partial discharge location, which combined ultrasonic detection and ultra-high frequency detection. Ultrasonic detection is easy to locate, but ultrasonic attenuates severely during its spread; ultra-high frequency detection method can avoid power system disturbances which are difficult to identify by conventional electrical testing methods, and this detection can significantly improve signal to noise ratio. This article used the advantages of these two methods to greatly improve the accuracy of detection. As to the identification of PD type, this article promoted a simple discrimination rules based on the chromatographic analysis of transformer oil.This paper analyzed the radiation mechanism of ultrasonic and ultra-high frequency electromagnetic wave in partial discharge, discussed the propagation characteristics of these two signals, described the possibility and criterion of joint detection, and designed the hardware of the ultrasonic detection subsystem and ultra-PD high-frequency detection subsystem. The former included sensor selection, pre-amplifier circuit design, filter design and the optical isolation module to guarantee the safety of user; The latter included UHF antenna selection, mixing unit circuit design and filter circuit design.Data acquisition system was shared by ultra-high frequency and ultrasonic detection subsystem. In order to improve the sampling rate and ensure the detection be of real-time, this design used a highly efficient and reliable high-speed data acquisition card for data acquisition, analog PD signals transmitted directly to the main control room through optical fiber or cable, in the main control room, the PCI-based acquisition cards will complete the signal acquisition. As to the system software, we designed it using virtual instruments.For the partial discharge signals from multiple sensors, we gave a location method based on Time Difference of Arrival (TDOA), in order to improve the accuracy, the paper also used the BP (back propagation) neural network for the data fusion; and in the end of this paper, we gave a partial discharge type recognition rules useing rough set theory.