基于Kohonen网络的高压断路器机械故障识别方法

Detecting Method for Mechanical Fault in High Voltage Circuit Breaker Based on Kohonen Network

引起高压断路器故障的原因大部分都是机械故障,因此对高压断路器进行故障诊断,使其可靠、高效地工作对电力系统的运行具有很大的意义。实验室模拟断路器基座支架上的合闸振动信号,采用小波包变换对取得的3种状态下的振动信号进行分解,计算各频段能量并做归一化处理,构造特征向量作为Kohonen网络的输入,进而进行机械故障识别。最后,将基于Kohonen网络的识别方法与BP网络以及RBF网络识别方法对高压断路器机械故障的识别效果进行比较。结果表明提出的基于Kohonen网络的高压断路器机械故障识别方法优于BP网络以及RBF网络,具有较高的准确性。

The causes of the failure of high-voltage circuit breaker are mostly mechanical faults. The high-voltage circuit breaker fault diagnosis makes its operation more reliable and efficient and is of great significance for power system. The paper makes the simulation of closing vibration signal on the circuit breaker base in the Laboratory, decomposes the vibration signal under three states by wavelet packet transform, calculates and normalizes each frequency band energy, and constructs eigenveetor as the input of Kohonen network to detect the mechanical fault. Finally, the mechanical fault detection for the high voltage circuit breaker is done with the Kohonen network, BP network and RBF network based detection method, and the detection effects are compared. The results show that the detecting method based on the Kohonen network for the mechanical fault in the high voltage circuit breaker has high accuracy, surpassing the BP network and RBF network.