基于WPA优化神经网络的扼流适配变压器故障诊断研究

Fault diagnosis research for choke adaptor transformer based on WPA optimizing neural network

针对传统铁路扼流适配变压器故障诊断模型结构复杂和精度不高的问题,运用狼群算法(WPA)、粗糙集(RS)理论和神经网络(NN)相融合的方法对其进行故障诊断研究。用粗糙集理论对故障样本数据进行约简处理,减少样本数据的监测及关键特征量的输入个数;利用约简后的数据对神经网络训练。利用狼群算法优化BP神经网络参数,提出WPA-BPNN故障诊断模型,以侯马电务段扼流适配变压器故障数据为例进行验证。研究结果表明:WPA-BPNN故障诊断模型相比传统方法,简化了网络结构,缩短了训练所需时间,提高了故障诊断精度,保证了列车行车安全及线路的高效运行。

In order to solve the problem that the fault diagnosis model of the traditional railway Choke adaptor has the complex structure and the low precision, this paper used the WPA (Wolf Pack Algorithm), RS (Rough Sets) theory and neural network (NN) methods to do research on the fault diagnosis. The RS theory was used to reduce the fault sample data, the sample data monitoring and the number of key feature inputs;then the neural network was trained by using the reduced data. Finally, the parameters of BP neural network were optimized by wolves algorithm, and the WPA-BPNN fault diagnosis model was proposed. Fault data of Choke adaptor in Houma Railway Service Section was used as an example for verification. The results show that the WPA-BPNN fault diagnosis model simplifies the network structure compared to the traditional methods, shortens the training time, improves the fault diagnosis accuracy, and ensures train safety and efficient operation of the line.