摘要
弓网电弧已成为电力机车安全运行的隐患,及时识别弓网电弧对于评价受流质量、调控弓网电弧、指导线路检修具有重要意义。该文开展了不同条件的弓网系统受流特性实验,将系统受流分为正常受流和电弧受流两种状态。提出一种基于回路电流和支持向量机(SVM)的弓网电弧在线识别方法及其工程实现方案。采用改进的F-score算法选择回路电流的平均值、标准差和相关系数作为弓网电弧的典型特征,利用svmtrain函数创建SVM模型,利用网格搜索算法优化SVM的径向基核函数。实验表明,该方法能够有效识别弓网电弧。接触压力、滑动速度和接触电流以及燃弧时间、电流采样频率均会影响弓网电弧的识别准确率。燃弧时间对识别准确率的影响较大,相同条件下燃弧时间越长,识别准确率越高。
Pantograph arc has become a serious hidden trouble to the safety operation of train. It’s of great significance to recognize timely pantograph arc for evaluating current collection quality, controlling arc and guiding maintenance. Current collection experiments under different conditions were carried out. The current collection status was divided into normal collection status and arc collection status. A kind of on-time recognition method of pantograph arc based on current signal and support vector machine （SVM） was proposed. An engineering implementation method was briefly introduced. The mean value, standard deviation and correlation coefficient of current were selected by the improved F-score （I-F-score） algorithm as typical characteristics of pantograph arc. The SVM model was established by using svmtrain function with Matlab software and the parameters of radial basis function were optimized by using grid search method. Lots of testing results showed that the suggested method can identify pantograph arc effectively. Some factors such as contact pressure, sliding speed, contact current, arc duration and sample rate have certain effects on recognition accuracy. Arc duration has larger effect on the recognition accuracy. The longer the arc duration, the higher the recognition accuracy is.
出处
《电工技术学报》
EI
CSCD
北大核心
2018年第1期82-91,共10页
Transactions of China Electrotechnical Society
基金
国家自然科学基金(51277090)
辽宁省教育厅重点实验室基础研究(LZ2014024)
辽宁省教育厅基金(LJYL015)
辽宁工大第五批生产技术问题创新研究基金(20160054T)资助项目
