摘要
为保障供电系统的安全、可靠运行,对电网环网柜在线故障检测问题进行研究,提出了一种新的基于数据局部特征的环网柜数据建模和在线监控方法。利用邻域保持嵌入(NPE)算法局部特征提取的策略,基于环网柜的多个测量变量信息以及环境变量信息,获取实时数据特征,构建了基于数据特征的环网柜故障检测模型。将构建的NPE模型应用于实际环网柜在线检测,并将原始数据空间划分为不相关的特征空间和数据残差空间。针对这两个空间,分别构造Hotelling T^2和预测误差平方和(SPE)的监控统计量,并基于这两个监控统计量,实现了环网柜的在线实时监控和故障报警。将该故障检测方法应用于实际环网柜的的监控案例研究中,试验结果证明了该方法在环网柜故障检测方面的有效性。通过该数据监控模型,改善了环网柜故障检测的效果,为降低风险、提高环网柜的安全稳定和运行品质提供技术保障。
For ensuring safety and reliable operation of power supply system,the online fault detection of ring main unit of power grid is researched,and a new method of data modeling and online monitoring based on data local features is proposed.By using the local feature extraction strategy of Neighborhood Preserving Embedding （NPE） algorithm,based on the information of multiple variables in ring main unit and the environmental information,the real time data features are obtained,and then the fault detection model of ring main unit is constructed based on these data features.The NPE model established is applied in online detection of actual ring main unit,and the raw data space is divided into two parts： feature space and residual space.In accordance with these two spaces,the monitoring statistics of Hotelling T^2 and squared prediction error （SPE） are constructed,and the online real time monitoring and fault alarm of ring main unit are achieved based on statistics of T^2 and SPE.This fault detection method has been used in research on a case based on real ring main unit,the results verify that the effectiveness of this method in fault detection process ofring main unit.The data monitoring model improves the effects of fault detection for ring main unit,and provides technical support for reducing risks,and enhancing the safety,stability and operation quality of the ring main unit.
出处
《自动化仪表》
CAS
2017年第10期69-73,共5页
Process Automation Instrumentation
基金
国家自然科学基金资助项目(61540070)
云南省教育厅科学研究基金资助项目(2015Y019)
云南省科技计划应用基础研究基金资助项目(2014FB112)
轻工过程先进控制教育部重点实验室开放课题(江南大学)资助项目(APCLI1606)
关键词
能源
电网
供电
故障检测
数据驱动
数据挖掘
Energy
Power grid
Power supply
Fault detection
Data -driven
Data mining
