Efficient data reconciliation and estimation for dynamic processes using nonlinear programming techniques

Abstract

Through the use of data reconciliation techniques, the level of process variable corruption due to measurement noise can be reduced and both process knowledge and control system performance can be improved. Process data from systems governed by dynamic equations are typically reconciled using the Kalman filter or the extended Kalman filter. Unfortunately, chemical engineering systems often operate dynamically in highly nonlinear regions where the extended Kalman filter may be inaccurate. In addition, the Kalman filter may not be adequate in the presence of inequality constraint Thus, a more robust means for reronciling process measurements for nonlinear dynamic systems is desirable. In this paper, a new method for nonlinear dynamic data reconciliation (NDDR) using nonlinear programming is proposed. Through the use of enhanced simultaneous optimization and solution techniques the algorithm provides a general framework within which efficient state and parameter estimation can be performed. Extensions for the treatment of biased measurements are also discussed. We demonstrate the use of NDDR and its extensions on a reactor example.