基于热点位置分类的电流互感器发热故障判别方法

针对变电设备故障发热特征选择、提取和故障类型判别问题,提出一种基于热点位置分类的电流互感器发热故障判别方法。首先运用CNN神经网络和YOLO融合算法完成电流互感器零部件检测为基础,建立二维坐标参考基准；然后采用PCNN分层聚类迭代方法分割出故障区域,并利用灰度质心法获取故障区域的热点等效中心；最后选择并提取热点等效中心到坐标原点的距离、角度等参数,应用这些参数判别热点位置的故障类别属性,从而实现电流互感器故障类型判别。结果表明,该判别方法准确率高达92,具有较好的实用性和推广性。

Thermal fault identification method of current transformerbased on hotspot position classification

Aiming at the problems of fault heating feature selection,extraction and fault type identification of substation equipment,a thermal fault identification method of current transformer based on hot spot position classification was proposed.Firstly,based on the detection of current transformer parts by using CNN neural network and YOLO fusion algorithm,a two dimensional coordinate reference benchmark was established.Then,the fault region was segmented by PCNN hierarchical clustering iteration method,and the hot spot equivalent center of the fault region was obtained by gray center method.Finally,parameters such as the distance and angle between the equivalent center of the hot spot and the origin of coordinates are selected and extracted,and these parameters are applied to identify the fault category attributes of the hot spot position,so as to realize the fault type identification of the CT.The results show that the accuracy of this method is as high as 92%,and it has good practicability and generalization.