Dynamic Data Reconciliation in Process Simulation, 1992 Engineering Conference Proceedings
As with steady-state data reconciliation, methods for dynamic data reconciliation can be built into an equation-based process flowsheeting system, in this case a dynamic simulator (MASSBAL 3.0). Advantage can be taken of the equation-solving and derivative generating capabilities of such a system.
We have found the most suitable method for dynamic data reconciliation to be an extended, hybrid Kalman filter. A brief, qualitative description is presented of the Kalman filter as implemented in MASSBAL, with an explanation of the terminology employed.
Dynamic data reconciliation will be illustrated by examples showing the filtering of level measurements in a tank of non-constant cross section, and a small flowsheet with recycle and a first order chemical reaction. The importance of tuning the Kalman filter parameters is discussed, including the role of assumptions about process noise in enabling a poor model to be detected.
Dynamic Data Reconciliation in Process Simulation, 1992 Engineering Conference Proceedings