443 / 2022-03-15 22:07:03

Research on online partial discharge recognition methods based on multi-sensor fusion

online monitoring,multi-sensor fusion,type recognition,Partial discharge

In order to monitor and recognize the types of partial discharge (PD) in cable joints, this paper builds an experimental platform for simulating PD in cable joints and proposes a monitoring scheme with the fusion of ultrasonic sensor (US) and high-frequency current transformer (HFCT). For both types of data, firstly, wavelet transform is used for denoising; then empirical mode decomposition (EMD) is used to obtain feature values; finally, convolutional neural network (CNN) is used for type recognition. In addition, online monitoring software for PD is also developed based on the above methods. The results show that the multi-sensor fusion method has a high accuracy of more than 95\% for monitoring discharges and more than 90\% for recognizing discharge types. Compared with single-sensor detection, the accuracy of the fusion method can be improved by more than 10\%. These studies and results can provide support to ensure the safe operation of power cables.