基于图像处理及降噪的局部放电图谱智能识别方法

局部放电检测是目前电力设备状态评价的主要手段,得到广泛应用推广。由于缺陷图谱的复杂性及现场干扰的多样性,传统的局部放电模式识别方法正确率低,训练时间长。针对上述问题,本文提出了一种基于图像处理技术及数据深度稀疏降噪的电力设备局部放电图谱智能识别方法。首先,运用图像处理技术对检测得到的图谱进行预处理；然后利用深度稀疏降噪自编码器进行数据稀疏降噪；最后对得到的有效去噪的数学模型,利用极限学习机(Extreme Learning Machine, ELM)网络,实现对局部放电的智能分类和识别。利用在变电站现场实测数据对本方法进行验证,证明本方法对含有多样干扰的局部放电信号有更好的识别效果,能很好适用于目前的电力设备图像信息模式识别应用当中。

Partial Discharge Pattern Intelligent Recognition Algorithm Based on image processing and noise reduction

Partial discharge (PD) detection is the main means of power equipment status evaluation at present. Due to the complexity of defect map and the diversity of field interference, the traditional local discharge pattern recognition method has low recognition accuracy and long training time. Based on these problems, an intelligent recognition method of partial discharge map of power equipment based on image processing technology and sparse data depth de-noising is proposed in this paper. Firstly, image processing technology is used to preprocess the detected image data. Then the active de-noising is carried out by using the deep sparse de-noising self-encoder. Finally, extreme learning machine is used as classifier to realize intelligent classification and identification of partial discharge. Using this method to verify the measured data at the scene of the substation, prove that the method of partial discharge signals containing various interference has better recognition effect, can apply to the current electric power equipment images information pattern recognition applications.