Distribution Cable Partial Discharge Hybrid Noise Suppression Method based on Adaptive Dual Complementary Convolutional Network

Due to the complex interference environment, partial discharge (PD) monitoring signal filtering of noise and interference signals is a necessary step to determine the insulation status of equipment. This paper proposes a method for optimizing filtering based on time-frequency analysis. Since pure partial discharge signals and periodic narrowband noise have different intermediate frequency positions and partial discharge signals usually have burst characteristics, time-frequency diagrams can be used to extract burst intermediate frequency characteristic signals to filter narrowband interference, and innovatively decompose wavelet It is implanted into the convolutional neural network to expand its receptive field, and at the same time, a double complementary convolutional network is used to restore the details and structure of the time-frequency diagram as much as possible, so as to realize the widely distributed white noise data part and the effective signal data in the sampled data. Partially separated, the effectiveness of this method was verified through numerical simulation experiments.