Abstract
Abstract This paper is concerned with developing an algorithm for gross error detection and data reconciliation that is based on Dynamic Bayesian Networks (DBNs). The proposed method replaces each traditional step in gross error detection (data reconciliation, gathering of residual error data, and hypothesis testing) with an appropriate DBN representation. In addition to detecting gross errors, the proposed method also estimates gross errors in real time and validates them against future data. Upon a successful validation, estimation is terminated and the reference point change is recorded and applied to subsequent data reconciliation steps. Issues such as data reconciliation through Kalman filtering, observability of augmented state space models, and Bayesian inference based on statistical test p values are discussed. The real time gross error detection performance is demonstrated through a simulation example with industrial application potential.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call