基于人工神经网络的电力变压器励磁涌流判别

Discrimination Research of Power Transformer Magnetizing Inrush Current Based on Artificial Neural Network

研究了励磁涌流的产生机理和特性 ,探讨了影响励磁涌流的因素 ,对多种条件下的电力变压器励磁涌流及故障电流进行了数字仿真 ,比较了两者在物理特性上的区别。同时利用Matlab中的神经网络工具箱 ,建立了径向基函数 (RBF)神经网络模型 ,对励磁涌流和短路电流的样本进行训练 ,并对训练好的神经网络进行测试。结果表明 :RBF神经网络可以正确地区分励磁涌流和短路电流。

In this paper, the originating principles and characters of magnetizing inrush currents, and the influencing factors are analyzed. The physical characters of magnetizing inrush currents and internal fault currents are compared through digital simulation under several conditions. The model of RBF NN is established by means of ANN toolbox in Matlab software. The sample data of magnetizing inrush currents and internal fault currents have been trained, and the performance of the trained ANN has been tested. The results showed that RBF NN can distinguish magnetizing inrush currents from internal fault currents correctly.