发电机励磁控制系统故障诊断的神经网络模型

鉴于发电机励磁控制系统是电网进行无功和电压调节的关键控制设备,为确保发电机安全运行,采用先进的人工神经网络的理论和方法完成了励磁系统故障诊断,并建立了一种新型的励磁系统故障诊断4层(含输入层)神经网络模型。采用神经网络方法对励磁系统进行故障诊断的关键在于网络的学习,网络学习又依赖于正确的输入输出样本。因此还收集了一定数量的励磁系统现场运行故障实例和大量的专家知识样本,对于正确调整神经网络中的各参数具有重要的实用价值。

Generator Excitation Control System Fault Diagnosis with ANN

Excitation Control System (ECS) of the generator is a key control unit that can adjust the reactive power and bus voltage for power network. If the faults in ECS could not be detected,the security,quality and economy of the power system will be affected. Therefore,fast and precise fault diagnosis for ECS is very important. This thesis makes a conclusion that artificial neural networks (ANN) possess special advantages for the excitation control system fault diagnosis (ECSFD) and has set up a new four-layer (including the input layer) model for ECSFD. The key of ECSFD is learning of ANN,and the learning of ANN is depended on the correct input/output (I/O) patterns. For this reason,this paper collects a number of real fault patterns and plentiful expert knowledge patterns. Through analyzing the model effect causing from different ANN parameters,this paper discovered one group of parameters that can diagnose the fault from the Generator Excitation Control System exactly and speedily. Since the research can be applied to practice directly,this paper will show the important and useful value in the application.