Insulator Fouling Monitoring Based on Acoustic Signal and One-Dimensional Convolutional Neural Network

Abstract

Aiming at the problem of pollution insulator discharge mode monitoring in high voltage line, a new one-dimensional convolutional neural network structure (1D-CNN) was designed, and a pollution insulator discharge mode monitoring method based on acoustic emission signal and 1D-CNN was proposed. Firstly, the data was collected in laboratory of acoustic emission signal under different discharge after sliding access way to expand the sample quantity. Thereafter, the sample time and frequency domain was used along with a third octave data as input, using convolution neural network to discharge signal samples adaptive feature extraction and feature dimension reduction. Then, appropriate stride convolution alternative pooling layer was used in order to reduce the training model parameters and the amount of calculation. Finally, Softmax function was used to classify the predicted results. The identified results show that the model can achieve a recognition rate of more than 99.84%, which effectively solves the process of manual data preprocessing in the traditional insulator pollution degree monitoring method. Moreover, at the same time it can be effectively applied to the pollution insulator discharge mode monitoring task.