基于径向基函数网络的高压断路器在线监测和故障诊断

ON-LINE MONITORING AND FAULT DIAGNOSIS OF HIGH VOLTAGE CIRCUIT BREAKERS BASED ON RADIAL BASIS FUNCTION NETWORKS

介绍了一种高压断路器在线检测和故障诊断的方法。由于高压断路器出现机械故障不仅会引起振动冲击事件的时间漂移，还会引起时域波形中一些波峰幅值的变化。基于径向基函数（ＲＢＦ）网络理论，将健康振动信号和断路器实际振动信号波峰幅值之差形成的残差以及波峰幅值发生的时间区段作为断路器故障诊断的特征参数，来判别断路器是否发生故障及其故障类型。在实际工程应用中，可以将断路器正常工作时的动静触头接触时产生的振动信号纪录下来，存入用于巡回检测的微机系统。用ＲＢＦ网络预测器的输出与预先给定的阈值进行比较，实现故障的自动诊断。对模拟故障信号进行了仿真实验，仿真结果表明：ＲＢＦ网络在线学习只需１组样本，因而其收敛速度比ＢＰ网络的收敛速度快，更适合于断路器在线检测。该方法还具有精度高和鲁棒性的特点，是一种比较有效的方法。

The occurrence of mechanical failure of high voltage circuit breakers not only can cause the time shift of vi- bration impulse events, but also can bring the change of some wave crest shape in time domain. In this paper on the basis of Radial Basis Function (RBF) Network theory a new method of on-line monitoring and fault diagnosis for high voltage cir- cuit breakers is put forward, in which two parameters, i. e., the residual error between natural vibration signal and practi- cal vibration signal and the time interval in which the wave crest appears, are taken as eigen parameters to determine whether the circuit is faulty or not and what kind of the fault should be if it occurs. In practical utility when the movable contacts make contact with fixed ones under normal operation of circuit breaker the vibration signal can be recorded and saved into the microcomputer system for data logging. Com- paring the given threshold with the output of the predictor of RBF network the automatic fault diagnosis can be implement- ed. The simulation results of analogue fault signals show that for the on-line training of RBF network one group of samples is enough, therefore, it can converge quickly and the require- ment of real time fault identification can be satisfied. The pre- sented method is of robustness and high accuracy, so it is suit- able for on-line monitoring of circuit breakers.