Study On PD Detection And Pattern Recognition For Power Cables

The cross-linked polyethylene (XLPE) power cable has been employed greatly for more than 40 years. It has substituted for the oiled paper insulating power cable in the field of the middle and low voltage and is being used more and more widely in the field of the high and extra high voltage. At present, there are many XLPE cable lines of 110kV voltage class in service. So many kinds of faults appear which can result in partial discharge (PD) and some accidents when the XLPE cables are in service. Partial discharge detection of XLPE cables is an effective method for diagnosing faults of XLPE cables.Based on summarizing the present technology on XLPE faults detection, this paper uses the theory of Rogowski coil to make the current transducer and detect the PD of XLPE cables. Considering the effects of all the parameters of Rogowski coil, the current transducer is produced firstly. Then the experiment of the XLPE cables PD detection and the four different cable faulty patterns are developed.In this paper, a de-noise method based on PD signals from electrical sensors is introduced. Three kinds of sensors are employed to obtain PD signals for the time domain characteristics of XLPE cables partial discharge pulses. The starting time and ending time of discharge can be confirmed by multiplying the signals from the three different sensors. A shielding filter could be constructed through the two thresholds. Then the original PD signals are processed and de-noised by the shielding filter and improved the signal-to-noise. The result shows that the proposed method has simple configuration and possesses well de-noise effect and application prospect.Finally, this paper presents a method that extracting the PD features from the time domain oscillogram. The back propagation neural network is set up and then the extracted features are put into it for recognition. The consequence indicates that the features extracted by the way mentioned before can totally represent the PD characters.