运用模糊神经网络来区分变压器励磁涌流和内部故障

Use fuzzy neural network to discriminate transformer magnetizing inrush and internal faults

变压器励磁涌流和内部故障的鉴别一直是变压器差动保护中的一个热点问题。在几种传统的识别励磁涌流方法的基础上 ,结合模糊神经网络这一新型的人工智能技术 ,综合利用这几种原理对电气量的采样值分别提取形成网络的特征输入量 ,并采用了Simpson模糊极小 -极大神经网络来形成区分励磁涌流和内部故障的模糊模式分类器。运用EMTP程序通过大量的仿真计算获取网络的训练和测试样本 ,结果表明 ,训练后的网络能快速地区分变压器各种运行工况下的励磁涌流和内部故障 ,对测试样本的正确率达到 10 0 %。

The discrimination between transformer magnetizing inrush and internal faults is a hot point in transformer differential protection. In this paper, based on several traditional methods to identify magnetizing inrush such as second harmonic theory, high harmonic theory, symmetry of wave shape theory, low voltage accelerating criterion and differential power theory, etc, and considering the new artificial intelligence technique of fuzzy neural network, we make full use of these theories and develop several corresponding characteristic inputs from electrical sampling values. Furthermore, we use Simpson's fuzzy min-max neural network to develop a fuzzy pattern classifier to discriminate between magnetizing inrush and internal faults. We also use EMTP simulation to get plenty of training and testing patterns. The training and testing result shows that the trained network can discriminate magnetizing inrush and internal fault quickly under all kinds of transformer operating conditions, the correct rate of testing samples is up to 100%.