中国电力 ›› 2021, Vol. 54 ›› Issue (2): 58-65.doi: 10.11930/j.issn.1004-9649.202004012

• 国家“十三五”智能电网重大专项专栏:(五)电力传感技术及应用专栏 • 上一篇    

基于振动信号区间特征快速提取的断路器储能状态辨识方法

夏小飞, 芦宇峰, 苏毅, 杨健   

  1. 广西电网有限责任公司电力科学研究院,广西 南宁 530023
  • 收稿日期:2020-04-06 修回日期:2020-12-14 发布日期:2021-02-06
  • 作者简介:夏小飞(1981-),男,硕士研究生,高级工程师,从事开关技术研究,E-mail:23459133@qq.com;芦宇峰(1982-),男,博士研究生,高级工程师,从事开关技术研究,E-mail:luyufenghust@163.com
  • 基金资助:
    中国南方电网有限责任公司科技项目(GXKJXM20180905)

Circuit Breaker Energy Storage State Identification Based on Quick Extraction of Vibration Signal Interval Features

XIA Xiaofei, LU Yufeng, SU Yi, YANG Jian   

  1. Electric Power Research Institute of Guangxi Power Grid Co., Ltd., Nanning 530023, China
  • Received:2020-04-06 Revised:2020-12-14 Published:2021-02-06
  • Supported by:
    This work is supported by Science and Technology Project of CSG (No.GXKJXM20180905)

摘要: 针对断路器伴随振动信号分析故障的特征提取费时、实时性差无法用于在线监测问题,提出一种基于快速提取区间特征的断路器储能状态辨识方法。首先由峭度-小波模极大值检测断路器储能状态起始点,将振动信号通过KS检验标记包络幅值差异明显区间,然后提取信号包络和作为特征向量,采用ReliefF-SFS方法对特征进行筛选降维得到最优特征子集。最后通过模糊C均值聚类(KFCM)对特征进行预分类获得风险最小的最优超平面,由支持向量机(SVM)建立训练模型进行状态辨识。实验结果表明:所提出振动信号区间特征快速提取的储能状态辨识方法,在保证准确率的前提下,提取特征仅需0.2 s,在断路器状态监测领域具有重要的应用价值。

关键词: 区间特征, KS检验, ReliefF-SFS, KFCM-SVM, 状态辨识

Abstract: The vibration signal based circuit breaker faults diagnosis has the problem of time-consuming in feature extraction and poor real-time, which makes the method inapplicable to on-line monitoring. We therefore proposed a circuit breaker energy storage state identification method based on fast extraction of interval features. Firstly, the starting point of the energy storage state of the circuit breaker was detected by the kurtosis-wavelet modulus maximum value, and the vibration signals were marked through KS test to indicate the significant difference in the envelope amplitude. Then the signal envelope was extracted and used as the feature vector, and the ReliefF-SFS method was used to reduce the dimensionality of features to obtain the optimal feature subset. Finally, the fuzzy C-means clustering (KFCM) was used to pre-classify the features to obtain the optimal hyperplane with the least risk, and a training model was established with support vector machine (SVM) for state identification. The experimental results show that the proposed state identification method only takes 0.2 s to extract features with reliable recognition accuracy, which has important application value in the field of circuit breaker state monitoring.

Key words: interval feature, KS test, ReliefF-SFS, KFCM-SVM, state identification