基于PCA和多元状态估计的引风机故障预警

Induced Draft Fan Fault Warning based on PCA and Multivariate State Estimation Technique

针对电厂引风机运行条件恶劣、故障种类多的特点,为了提高机组安全和设备利用率,提出了一种基于主元分析(Principal Component Analysis,PCA)和多元状态估计(Multivariate State Estimation Technique,MSET)的引风机故障预警方法:首先采用PCA对原始数据进行简化,消除冗余和噪音,选取主要状态参数,同时采用小波变换(Wavelet Transform,WT)进一步优化数据质量;然后通过MSET对引风机正常运行工况下的历史数据进行建模,包括历史记忆矩阵的构建、观测向量的估计和残差的计算等。以华能上海某电厂的引风机为例进行故障预警分析,验证该方法可以有效实现引风机早期的故障预警。

In order to improve the safety of the unit and the utilization rate of the equipment,a Principal Component Analysis(PCA)and Multivariate State Estimation Technique(MSET)are proposed to improve the safety of the unit^induced draft fan and the variety of faults.The early warning method of induced draft fan failure:firstly use PCA to simplify the original data,eliminate redundancy and noise,select the main state parameters,and further optimize the data quality by Wavelet Transform(WT);then run the normal operation of the induced draft fan through MSET.The historical data under the working conditions are modeled,including the construction of the historical memory matrix,the estimation of the observation vector and the calculation of the residual.With the induced draft fan of a power plant in Huaneng Shanghai as an example,the fault early warning analysis and verification is earned out.This method can effectively realize the early fault warning of the induced draft fan.