基于改进BP神经网络的牵引变流器故障诊断

Traction converter's fault diagnosis based on improved BP neural network

在变流器故障诊断系统中,通过MATLAB对牵引变流器建立故障仿真模型,提取故障特征,对输入输出数据进行标幺化和模糊化的处理,并基于改进的动量BP神经网络算法,完成对变流器开关管开路的诊断,误差满足要求范围,结果表明:该算法收敛迅速,能避免陷入局部极值,而且准确率很高,是一种快速有效的方法。

In the field of the converter fault diagnosis ,the simulation model of the traction converter＇s fault is established through MATLAB and extract the fault characteristics .Then make the input date per unit and let the output data fuzzy processed. Based on improved BP neural network algorithm, the converter＇s switch open fault diagnosis is completed. The error meets the requirements and the results show that： the rapid convergence of the algorithm can avoid falling into local minima, with high accuracy, is a fast and effective algorithm.