基于鲁棒PCA的火电厂湿法烟气脱硫系统的故障诊断

Study on fault diagnosis of wet flue gas desulfurization system in thermal power plant based on robust PCA

针对火电厂湿法烟气脱硫(WFGD)过程建模数据中存在离群点的情况,提出一种基于复合统计量CRS的鲁棒主元分析(PCA)的故障诊断方法.该方法利用含有离群点的样本数据建立PCA模型,求取PCA模型的复合统计量CRS的控制限,剔除样本数据中复合统计量值超过控制限的样本点,使用剩下的样本点重新建立PCA模型.经华能福州电厂的湿法烟气脱硫过程的故障仿真实验,结果表明,该鲁棒PCA比传统PCA具有更好的故障检测能力.

In order to analyze the model data with outliers in wet flue gas desulfurization（WFGD） processes of power plant,a robust principal component analysis（PCA） using compound statistics of residual and score（CRS） is proposed.The control limits of compound statistics is calculated on the basis of PCA model containing outliers.And the sample data whose compound statistics exceeds the control limits is eliminated from the modeling data of PCA model.Then new robust PCA model is built by employing the residual sample data.The WFGD fault simulation results of Huaneng Fuzhou power plant show the new robust PCA approach has better fault detection ability than traditional PCA approach.