摘要
为了解决风机故障预警问题,提出一种基于多元状态估计技术(MSET)和偏离度的方法.利用MSET建立风机正常运行状态下的非参数模型,对观测向量进行最优估计并得到估计向量,观测向量与估计向量之间的差异可以反映风机工作是否异常.引入偏离度定量衡量观测向量与估计向量之间的偏离程度,有利于捕捉故障发展过程,然后利用滑动窗口法确定故障预警阈值.当平均偏离度超过预警阈值时,发出报警信息提醒运行人员处理.以长春某热电厂引风机的某次故障为例进行应用研究.结果表明:该方法可以及时发现风机异常,实现风机实时故障预警.
To realize early fault warning of power plant fans, a method was proposed based on multivariate state estimation technique （MSET） and the deviation degree. Using MSET, a non-parametric model of the fan was constructed under normal operation conditions, based on which observed vectors were optimally evaluated to obtain the estimated vectors, and the difference between estimated and observed vectors is able to reflect the incipient failure. A deviation degree was introduced to quantify the difference between estimated and observed vectors, so as to capture the fault development process, and to send out early warning for relevant operators to deal with the trouble, when the average deviation has exceeded the predefined threshold set by sliding window method. The method was applied to detect the fault of an induced fan in a power plant in Changchun. Application results show that the method can help to discover abnormal status of fans promptly, and achieve the purpose of online early fault warning of fans.
出处
《动力工程学报》
CAS
CSCD
北大核心
2016年第6期454-460,共7页
Journal of Chinese Society of Power Engineering
基金
国家重点基础研究发展计划(973计划)资助项目(2012CB215203)
国家自然科学基金重点资助项目(51036002)
关键词
电厂风机
多元状态估计
故障预警
滑动窗口法
power plant fan
multivariate state estimation technique
early fault warning
sliding window method
作者简介
刘涛（1990-），男，湖北广水人，硕士研究生，研究方向为火电厂设备故障诊断与预警．电话（Tel．）：18901070798 E-mail：ltncepu@gmail．com．
