基于波形不对称和神经网络的变压器励磁涌流识别

Transformer magnetizing inrush current discrimination based on waveform asymmetry and neural network

为了实现对变压器励磁涌流和故障电流的识别,将传统二次谐波制动原理与基于波形上下不对称判别励磁涌流的新原理结合起来,建立了一个三层前向神经网络模型。利用Matlab软件对一个双端供电系统进行了仿真运算以得到神经网络的训练样本,用经训练的神经网络实现了对励磁涌流的识别。仿真测试结果验证了该方案是切实可行的,能够准确识别各种变压器状态下的励磁涌流和故障电流。

Aiming at distinguishing magnetizing inrush current and internal fault current,a new principle based on waveform asymmetry and traditional second harmonic brake principle were combined together for building a feedforward three-layered neural network model.Matlab was used for the simulation calculation about a two-source power system,and the trained neural network was used to achieve regonition of the magnetizing inrush current.The result of simulation shows that it is effective for the accurate discrimination of magnetizing inrush current and internal fault current.