An improved robust dynamic data reconciliation approach

Jiang Chunyang,Qiu Tong,Chen Bingzhen,Zhao Jinsong

DOI: https://doi.org/10.3969/j.issn.1001-4160.2007.10.001

2007-01-01

Abstract:Because process measurements from modern chemical plants typically contain random errors unavoidably, it is necessary to reconcile the measurements, but the reconciliation is usually affected by outlier errors. Present dynamic data reconciliation methods have limitations to handle them. Based on robust estimation principles, a new robust estimator (RE) is constructed in this paper. Its advantages in physical conceptions and parameter adjustment make it outstanding from its peers reported in literature. An improved robust dynamic data reconciliation (IRDR) methodology is proposed to identify and correct both random and outlier errors by incorporating the new RE into the objective function of dynamic data reconciliation. A dynamic CSTR case study shows that, IRDR can recognize multiple outlier errors with favorable variations. Compared with other methods, IRDR is more powerful in both detection and idenfica-tion of the outliers.