基于多元统计过程监控的锅炉过程故障检测  

Fault Detection of Industrial Processes Based on Multivariate Statistical Process Monitoring

作　　者：牛玉广[1] 王世林[2] 林忠伟[1,2] 李晓明 NIU Yuguang WANG Shilin LIN Zhongwei LI Xiaoming(State Key Laboratory for Alternate Electric Power System with Renewable Energy Sources, North China Electric Power University, Beijing 102206, China School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China School of Automation Engineering, Northeast Dianli University, Jilin 132012, Jilin Province, China)

机构地区：[1]华北电力大学新能源电力系统国家重点实验室,北京102206 [2]华北电力大学控制与计算机工程学院,北京102206 [3]东北电力大学自动化工程学院,吉林省吉林132012

出　　处：《动力工程学报》2017年第10期829-836,共8页Journal of Chinese Society of Power Engineering

基　　金：国家自然科学基金青年基金资助项目(51606033);中央高校基本科研业务专项资金资助项目(JB2015181)

摘　　要：提出了一种新的基于稀疏约束非负矩阵分解(SCNMF)的复杂工业过程故障检测方法.首先在交替约束最小二乘算法(ACLS)求解非负矩阵分解(NMF)问题的基础上对系数矩阵H实施稀疏约束,随后采用非负双奇异值分解(NDSVD)方法对SCNMF算法进行初始化,并将所提算法应用于某火力发电厂1 000 MW机组锅炉过程中.结果表明:SCNMF算法的收敛性和稀疏度明显优于传统的NMF算法,且对故障的检测效率也要优于NMF算法和主元分析(PCA)算法.A novel fault detection method based on sparseness-constrained non-negative matrix factorization（SCNMF）was proposed for complex industrial processes.The specific way is to use alternating constrained least squares（ACLS）with sparseness constraint on coefficient matrix Hto solve the non-negative matrix fracterization（NMF）problems,then to enhance the initialization stage of SCNMF by non-negative double singular value decomposition（NDSVD）,and finally to apply the presented method to the fault detection in various boiler processes of a 1 000 MW unit.Results show that the SCNMF is superior to conventional NMF on both the convergence and the sparsity,and its monitoring performance is also better than NMF and principal component analysis（PCA）.

关 键 词：故障检测 非负矩阵分解 奇异值分解 锅炉过程 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]