摘要
在故障电弧伴生早期弧声频谱特性研究的基础上,提出基于小波包分解的早期弧声频带局部能量特征参数的快速提取方法。早期弧声的功率谱分析表明,故障电弧发生之前,在可听波段产生5 kH z^10 kH z的电弧声,其带宽和中心频率与电极形状、放电距离、放电电压等试验条件有关。利用小波包多分辨率技术对弧声信号进行三层分解,对各子频带进行能量统计,根据不同子频带能量的分布特征建立起"能量—信号"的映射关系。实验研究表明,弧声出现前后第二子频带S31和第三子频带S32能量变化明显,可以作为识别早期弧声的特征参数。通过在线监测信号这两个特征子频带能量的变化即可判断早期故障弧声是否存在,从而实现故障电弧的早期预测预警。
The internal arcing faults in the switch cabinet can cause serious damage to the equipments and fatal injury to personnel if not detected and isolated promptly. The arc sound is one of the signals which often occurred during the early time of arcing faults. In order to reveal the inherent relation and regularity of early time arc sound and arcing faults, an all-purpose test system for detecting faults arc has been set up. The study on spectrum analysis of arc sound signal has showed that a signature band of 5- 10kHz of sound wave would be generated before faults arc take place, which is strong in intensity and variable in both width and center depending on arcing electrodes, space between electrodes, and discharge voltage. On this basis, feature extraction of faults arc sound signal has been detailed by using multi-resolution analysis of wavelet packet. A local energy feature extraction method of arc sounds frequency bands based on three-layer wavelet packet decomposition is presented. The results of research have showed that the local energy of some frequency bands are significantly increased in all experiment conditions, and they can be used as the characteristic parameters of arc sounds recognition. So it is feasible to forecast and early warn faults arc in switchgear by online detecting the energy of incipient arc sounds feature sub-bands.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2008年第4期57-62,共6页
Proceedings of the CSU-EPSA
基金
福建省高新技术研究开发计划重点项目(2005H036)
关键词
故障电弧
早期弧声
特征提取
小波包分析
预测预警
faults arc
incipient arc sound
feature extraction
wavelet packet analysis
forecast and early warn
