Dynamic Distributed Monitoring Strategy for Large-Scale Nonstationary Processes Subject to Frequently Varying Conditions Under Closed-Loop Control

• Published in: IEEE transactions on industrial electronics (1982) Vol. 66; no. 6; pp. 4749 - 4758
• Main Authors: Zhao, Chunhui, Sun, He
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.06.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Bayes methods; Bayesian analysis; Closed-loop control; dynamic distributed monitoring; frequent varying conditions; Heuristic algorithms; large-scale nonstationary processes; Monitoring; Principal component analysis; Process control; Process variables; Statistical inference; Steady-state
• ISSN: 0278-0046; 1557-9948
• DOI: 10.1109/TIE.2018.2864703

Large-scale processes under closed-loop control are commonly subjected to frequently varying conditions due to load changes or other causes, resulting in typical nonstationary characteristics. For closed-loop control processes, the normal changes in operation conditions may distort the static and dynamic variations in a different way from real faults. In this paper, a dynamic distributed monitoring strategy is proposed to separate the dynamic variations from the steady states, and concurrently, monitor them to distinguish changes in the normal operating condition and real faults for large-scale nonstationary processes under closed-loop control. First, large-scale nonstationary process variables are decomposed into different blocks to mine the local information. Second, the static and dynamic equilibrium relations are separated by probing into the cointegration analysis solution in each block. Third, the concurrent monitoring models are constructed to supervise both the steady variations and their dynamic counterparts for each block. Finally, the local monitoring results are combined by Bayesian inference to obtain global results, which enable description and monitoring of both static and dynamic equilibrium relations from the global and local viewpoints. The feasibility and performance of the proposed method are illustrated with a real industrial process, which is a 1000-MW ultra-supercritical thermal power unit.