摘要
开关柜内部的故障电弧是一种频发性、灾难性的严重故障,传统的故障电弧保护方案都是在发生燃弧之后开始报警并切断断路器,会造成一定的经济损失。实验研究表明,放电声响能作为故障电弧发生之前的主要表征信息,放电声响经过声道调制和周围障碍物反射后形成复杂的声音混合体,且放电声响湮没在复杂的背景噪声中。用双谱对故障电弧早期放电声响进行了分析,并以B-P神经网络为分类器提取放电声响。结果表明,双谱能有效地识别放电声响,为进一步以放电声响为判据建立故障电弧早期预警系统奠定基础。
In switch tank, fault arcs are constant and disastrous trouble. Traditionally, an alarm was given and the circuit was cut off only after aflame arcing was generated, which caused considerable economic loss. Switch tank and electric facilities would be destructed by aflame arcing. Experiments show that the pre -arc sound serving as the main signal forms complicated compound sound after being adjusted by sound regulator and reflected by surrounding barriers and is submerged in background noise. In this paper, pre -arcing sounds are analyzed by using Bispectrum. Three layers B - P artificial neural network is established. Results indicate that feature of Bispectrum can recognize pre -arcing sounds effectively with arcing sounds generated before aflame arcing. The system of fault arcs early warning system is established, which avoids endangerment brought about by aflame arcing and establishes good foundation for fault arcs protection in switch tank.
出处
《淮阴工学院学报》
CAS
2008年第5期47-50,共4页
Journal of Huaiyin Institute of Technology
基金
福建省科技计划重点项目(2005H036)
关键词
故障电弧
双谱
放电声响
累积量
神经网络
fault arcs
bispectrum
pre - arc sounds
cumulant
neural network
