基于DPCA-KD的污水处理过程故障诊断  

Fault Diagnosis in the Wastewater Treatment Process Based on DPCA-KD

作　　者：徐宝昌[1] 庄朋 李巨峰[2] 唐智和 栾辉 何为 XU Bao-chang;ZHUANG Peng;LI Ju-feng;TANG Zhi-he;LUAN Hui;HE Wei(College of Information Science and Engineering,China University of Petroleum(Beijing);CNPC Safety and Environmental Protection Technology Research Institute)

机构地区：[1]中国石油大学(北京)信息科学与工程学院 [2]中国石油天然气股份有限公司安全环保技术研究院

出　　处：《化工自动化及仪表》2023年第5期700-706,719,共8页Control and Instruments in Chemical Industry

基　　金：中国石油天然气股份有限公司科学研究与技术开发项目(2022DJ6904)。

摘　　要：污水生化处理过程是一类强非线性、变量耦合、工况复杂的过程。由于环境恶劣,污水生化处理过程传感器故障频发,导致传统基于动态主元分析的故障检测方法漏报率较高、检测率较低。提出了一种基于Kantorovich距离的动态主元分析故障检测方法。首先,通过动态主元分析构建增广矩阵,对多维数据进行降维,降低数据的自相关性。其次,通过Kantorovich距离对过程的数据进行故障检测。最后,基于国际水协会的基准仿真模型BSM1对所提方法进行验证。结果表明,所提出的方法相较于传统的动态主元分析方法,降低了故障误报率、提高了检测率。The biochemical process of wastewater treatment is a process with strong nonlinearity,variable coupling and complex working conditions.The bad environment,frequent sensor failure in the biochemical process can result in high false positive rate and false positive rate of the traditional fault detection methods based on dynamic principal component analysis.In this paper,a fault detection method based on Kantorovich distance dynamic principal component analysis was proposed.Firstly,having an augmented matrix constructed through the dynamic principal component analysis to reduce the dimension of multi-dimensional data and reduce the autocorrelation of data;secondly,having the process data detected by the Kantorovich distance and the method proposed validated based on the International Water Association’s BSM1(benchmark simulation model).The results show that,as compared to the traditional dynamic principal component analysis,the method proposed here can reduce false fault alarm rate and improve failure detection rate.

关 键 词：污水处理 故障诊断 动态主元分析 Kantorovich距离 BSM1 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]