基于阈值判断和支持向量机的电网实测过电压识别  

Identification Method of Grid Measured Over-voltage Based on Threshold Judgment and Support Vector Machine

作　　者：杨庆[1] 赵洪彬 司马文霞[1] 韩睿[1] 陈勇[1] 

机构地区：[1]重庆大学输配电装备及系统安全与新技术国家重点实验室,重庆400030 [2]国网四川省电力公司乐山供电公司,乐山614000

出　　处：《高电压技术》2016年第10期3188-3198,共11页High Voltage Engineering

基　　金：国家自然科学基金(51177182;51477018);国家创新研究群体基金(51321063)~~

摘　　要：过电压严重危害电网的安全稳定运行。目前,许多变电站已配置了可记录过电压波形的过电压在线监测装置,从而为过电压识别提供了基础条件。快速、准确识别出过电压种类对确定过电压发生原因及在此基础上开展检修工作具有指导意义。为此,提出了一种基于阈值判断和支持向量机的电网过电压识别方法,对暂时、操作以及雷电共11种过电压进行识别。在变电站采集得到共480条实测过电压波形数据,采用时域分析、频域分析和小波分析等数学方法相结合、多角度地提取过电压特征量,并将这些特征量输入由阈值判断和多级支持向量机构成的模式识别结构中进行识别。识别结果证明,所提特征量维数低、具有较大区别性,采用的识别方法准确性高、速度快,该识别方法可以较好地适用于对电网暂时、操作以及雷电过电压的识别。Over-voltage is seriously damaging to the safe and stable operation of the power grid. At present, many substations have configured over-voltage on-line monitoring device for recording the over-voltage waveform, and providing the basis for over-voltage identification. Quickly and accurately identifying the over-voltage category has great effects on determining over-voltage causes and has guiding significance for carrying out the maintenance work. Consequently, we propose an identification method based on threshold judgment and support vector machine to temporary, operating and lightning over-voltage, which have a total of 11 categories. According to 480 measured over-voltage signals,the time domain, frequency domain, and wavelet analysis are adopted to extract over-voltage characteristic values, and identify those over-voltages based on threshold judgment and multistage support vector machine. Identification results show that the proposed characteristic values have a low dimension and big distinctiveness, and the recognition method has high speed and high accuracy. The identification method can well be applied to identify temporary, operation, and lightning over-voltage in power grid.

关 键 词：过电压识别 实测波形 特征量 阈值判断 支持向量机 

分 类 号：TM76[电气工程—电力系统及自动化] TM63