电力变压器励磁涌流判别的自适应小波神经网络方法

励磁涌流识别一直是电力变压器差动保护中比较关注的问题。文中提出了一种基于自适应小波神经网络实现变压器励磁涌流判别的新方法。结合励磁涌流和内部故障电流的特点，构建了一个四层的自适应小波神经网络模型，并对其具体的实现方法进行了详细的分析；利用ATP—EMTP程序进行仿真计算生成训练样本和测试样本，对所构建的网络进行了训练和测试，结果表明自适应小波神经网络能准确、可靠地识别出变压器的励磁涌流状态。

DISTINGUISH BETWEEN INRUSH AND INTERNAL FAULT OF TRANSFORMER BASED ON ADAPTIVE WAVELET NEURAL NETWORK

Inrush identification is always a key problem in power transformer current differential protection. A novel technique using adaptive wavelet neural network is presented to distinguish between inrush and internal fault currents. Based on the characteristics of inrush and internal faults, the architecture of a four-layer adaptive wavelet neural network is given and the implementation details are discussed. ATP-EMTP is used to generate the training samples and testing samples. The test results show the presented technique can make reliable discrimination between inrush and fault currents.