基于自编码网络的局部放电信号特征提取与识别

气体绝缘金属封闭开关设备(GIS)的状态影响电力系统运行的可靠性,而局部放电是设备潜伏性绝缘故障的重要表现之一。传统局部放电模式识别方法依赖专家经验选取局部放电特征,主观性强且不确定度高。针对这一问题,文中提出将深度学习技术引入局部放电模式识别领域,运用卷积神经网络及其扩展自编码网络提取局部放电信号特征,充分发挥自编码网络的特征抽取能力。同时,将所提取的特征与经典分类器进行衔接,有机结合传统机器学习方法与深度学习方法,实现局部放电信号的基本参数提取、统计特征计算与放电类型识别。实验结果表明,文中所提方法提取的特征相较传统的人工特征可明显提高局部放电的分类准确率和分类效率,具有广阔的工程应用前景。

Feature extraction and recognition of partial discharge signal based on self-encoding network

The status of gas insulated fully-insulated switchgear (GIS) determines the reliability of power equipment operation. Partial discharge (PD) is one of manifestation for various early-stage latent insulation failures. The traditional partial discharge pattern recognition method relies on expert experience to select the feature which have the disadvantages of strong subjectivity and high uncertainty. To solve this problem, this paper introduces the encoding-decoding network structure to fully extract the features of partial discharge signals adaptively and we interface these features with classical classifiers to combine traditional machine learning methods with deep learning methods conveniently. Features are connected with classical classifiers, realizing the organic combination of traditional machine learning method and deep learning method. The basic parameters extraction, statistical feature calculation and discharge type identification of PD signal are realized. Experimental result shows that the features extracted by our method can improve the classification accuracy of partial discharge patterns significantly.