摘要
SF6电气设备存在放电故障时,内部的SF6气体会分解成诸多衍生物,对设备的安全运行造成隐患.因此,通过SF6衍生物的状态可以推断设备的放电故障.在已有实验数据的基础上,将SF6衍生物的状态作为神经网络的输入,放电故障作为神经网络的输出,构建了基于概率神经网络的SF6电气设备故障诊断模型.实验表明,构建的模型对放电故障的预测达到88.23%,并与BP神经网络模型的预测结果进行了比较,证实了在SF6电气设备故障诊断的研究中,概率神经网络要优于BP神经网络.
The SF6 gas will break down into many derivatives as the SF6 electrical equipments include discharge faults, which may cause hidden dangers. Therefore, the discharge faults can be inferred with the state of SF6 derivatives. With the existing experimental data, the probabilistic Neural network based faults diagnosis model of SF6 electrical equipments is constructed with inputs of the states of SF6 derivatives, and outputs of discharge faults. Experiments show that the discharge faults prediction rate of constructed model reaches 88.23%, and compared with the predictions of BP neural network model, which confirmed that probabilistic neural network is better than BP neural network on the research of SF6 electrical equipment fault diagnosis.
出处
《湖北文理学院学报》
2013年第8期10-14,共5页
Journal of Hubei University of Arts and Science
基金
湖北省自然科学基金项目(2010CDZ051)
湖北省教育厅项目(B20122503)
关键词
概率神经网络
六氟化硫电气设备
故障诊断
Probabilistic neural network
SF6 electrical equipments
Fault diagnosis
