基于广义S变换和二维卷积神经网络的油纸绝缘局部放电超声信号识别

Patial Discharge Ultrasonic Signal Recognition in Oil-pressboard Insulation Using GST and 2D CNN

超声法是局部放电测量的重要方法。为更有效提取局部放电超声信号中的关键信息,提出一种基于广义S变换和二维卷积神经网络的油纸绝缘局部放电超声信号识别方法。首先,在实验室采集3种典型缺陷的局部放电超声信号,对预处理后的数据进行广义S变换,获得不同放电样本的时频分布图像;然后,构建二维卷积神经网络,将时频图像作为输入,提取超声信号的时频特征;最后,输出对局部放电类型的识别结果。结果表明:该方法对不同局部放电类型的超声信号进行识别,准确率可以达到97.8%,能够更有效地提取超声信号的内在信息并进行局部放电识别。

Ultrasonic method is an important method for partial discharge measurement.In order to extract the key information of partial discharge ultrasonic signal more effectively,a method of partial discharge ultrasonic signal recognition in oil-pressboard insulation based on generalized S transform and two-dimensional convolutional neural network is proposed.Firstly,the partial discharge ultrasonic signals of three typical defects were collected in the laboratory,and the generalized S transform was applied to the preprocessed data to obtain the time-frequency distribution images of different discharge samples.Then,a two-dimensional convolutional neural network is established,and the time-frequency image is used as input to extract the time-frequency characteristics of the ultrasonic signal.Finally,the recognition result output of the partial discharge type show that this method can identify the ultrasonic signals of different partial discharge types with an accuracy of 97.8%.It can extract the intrinsic information of ultrasonic signals and identify partial discharge more effectively.