基于BRBP神经网络的转子绕组匝间短路故障诊断方法

为了能更准确、容易地诊断出同步发电机转子绕组匝间短路故障, 提出了一种基于BRBP神经网络的故障诊断方法。该方法利用正常运行时不同工况下的机端电压、有功功率、无功功率和转子励磁电流来建立励磁电流的BRBP神经网络预测模型；利用该模型预测正常运行时所需励磁电流，与实测的励磁电流进行比较，相对误差超过阈值就诊断为发生匝间短路故障。通过微型同步发电机动模实验表明，该方法的精度优于BP神经网络法，并且参数设置简单、易于移植和训练速度快，对同步发电机转子绕组匝间短路故障的监测与诊断是有效的。

A Rotor Winding Inter-turn Short-circuit Fault Diagnosis Method Based on BRBP Neural Networks

In order to more accurate and easy diagnosis of synchronous generator rotor winding inter-turn short-circuit fault, a novel fault diagnosis method is put forward, which based on Bayesian regularization back-propagation (BRBP) neural network. Measure and collect sample data in different fault-free operating conditions, including terminal parameters (voltage, active power, reactive power) and field current, then train a BRBP neural network model to predict field current. Input to the model with measured terminal parameters, and a predicted field current is obtained. Finally, compare the predicted field current with the corresponding measured field current, and a synchronous generator rotor winding inter-turn short-circuit fault is diagnosed when relative error exceeds a specific threshold. The micro-synchronous generator dynamic simulation results show that, the method is better accuracy than BP neural network method, only needs to simply set the BRBP neural network structures and parameters, is fast trained and easily applied to other synchronous generator, and is a effective rotor winding inter-turn short-circuit fault diagnosis method for synchronous generator.