摘要
针对高压断路器的故障诊断,通过分析断路器的合闸电流波形,提取相应的特征量,作为径向基神经网络的输入,经训练后的网络作为断路器的故障诊断模型。由于BP神经网络存在的训练收敛速度慢且容易陷入局部极小等缺陷,提出了一种基于正交算法的RBF网络用于高压断路器的故障诊断方法。仿真结果表明,基于正交算法的RBF网络具有训练速度快、分类性能良好的优点,有很好的实用性。
Aiming at the fault diagnosis of HV circuit breakers and studying coil closing current, some characteristic parameters were abstracted and input into RBF neural network. Network trained can diagnose faults of HV circuit breakers well. For limitation of BP neural network, low learning convergence speed and easily-appearing local minimum, a method of radial basis function network based on orthogonal learning algorithm was proposed in this paper. The simulation result showed that RBF networks has very high learning convergence speed, better classifying performance and good practicality in the field of HV circuit breakers fault diagnosis.
出处
《电力科学与工程》
2008年第3期13-15,共3页
Electric Power Science and Engineering
关键词
高压断路器
故障诊断
正交算法
RBF网络
HV circuit breakers
fault diagnosis
orthogonal learning algorithm
RBF networks