基于波形记忆和模糊极小--极大神经网络的变压器励磁涌流和内部短路的鉴别

传统的区分变压器励磁涌流和内部短路的各种方法存在原理性缺陷，不能满足现代超高压电力系统的要求，此文根据内部故障和单纯涌流这两种情况下波形的不同，提出了波形记忆的原理并采用了一种模糊神经网络模型——模糊极小-极大神经网络来对这两种波形进行记忆和鉴别。运用EMTP程序对变压器各种内部故障或涌流的情况进行较为全面的仿真的以形成网络的训练样本，通过学习和测试，表明该网络所形成的新算法能够正确鉴别变压器各咱运行工况下的励磁涌流和内部短路，所需的鉴别时间小于20ms。

DISTINGUISH TRANSFORMER MAGNETIZING INRUSH FROM ITS INTERNAL FAULTS BASED ON WAVE SHAPE REMEMBRANCE AND FUZZY MINIMUM-MAXIMUM NEURAL NETWORK

Traditional methods to distinguish transformer magnetizing inrush from its internal faults have drawbacks, so that they cannot meet the need of modern EHV power systems. According to the wave shape difference between transformer magnetizing inrush and its internal faults, in this paper a principle of wave shape remembrance is put forward and these two kinds of wave shape are remembered and identified by fuzzy minimum-maximum neural network. To obtain the training sample of neural network the transformer magnetizing inrush and its various internal faults are almost overall simulated by EMTP program. Through training and testing it has shown that under various operating modes the new algorithm obtained by this neural network can correctly distinguish the magnetizing inrushes from internal short circuits, and the needed time for distinguishing is less than 20ms.