一种基于潮流指纹识别的电网故障实时诊断方法  

A Real Time Fault Diagnosis Method of Power Gird Based on the Fingerprints Identify Algorithm of Power Flow

作　　者：刘延乐 刘文颖 林玉东[2] 刘海林[2] 

机构地区：[1]新能源电力系统国家重点实验室(华北电力大学),北京102206 [2]东方电子股份有限公司,山东烟台264001

出　　处：《现代电力》2013年第2期1-6,共6页Modern Electric Power

摘　　要：开关保护动作的误报、漏报、误动、拒动问题以及电网拓扑的频繁变化导致传统的基于开关保护动作信号对电网进行故障诊断的方法难以实用化。提出一种利用电网各支路潮流的变化量作为表征该故障的潮流指纹,并采用模式识别算法进行电网故障诊断的新方法。设计了基于节点阻抗矩阵的潮流指纹提取方法,并提出了电网故障模式自适应修正算法以及基于遗传算法的(等值)发电机的出力分配因子优化辨识算法,解决了电网拓扑频繁变化对构建故障模式集的影响及故障样本集的样本特征向量的计算精度问题。在此基础上,按照最近近邻判别准则设计了归一化故障判别函数,实现对电网故障的实时准确识别。利用某地区电网实际数据验证了本方法的有效性和实用性。Because of such problems as misreporting, fail re- porting, misacting and fail acting, and frequent variation of power grid topology, it is very difficult to diagnose fault of power grid based on the switch and protection information in practice. In this paper, the fault diagnosis method of power grid by using of pattern recognition method is pro- posed, which can represent the power flow fingerprints of power grid fault by using of the power flow variation quanti- ty of each branch. Then, the extraction method of power flow fingerprint is proposed based on node impedance ma- trix, and the self- adaptive modification algorithm of fault pattern and the optimization algorithm of （equivalent） gen- erator output allocation factor based on GA are presented, which solve such issues as the effect of frequent variation of the power grid topology on the construction of fault mode set and the calculation accuracy of sample feature vectors. Furthermore, the normalization fault discriminator function is designed according to the nearest neighbor decision rule,which can accurately recognize the power grid failures in real time. In the end, the effectiveness and practicality of this method are verified by the actual data of a regional power grid.

关 键 词：电网故障诊断 故障模式 潮流指纹 模式识别 遗传算法 

分 类 号：TM76[电气工程—电力系统及自动化]