State Estimation in the Control of Integrated Batch Reaction with Distillation Processes

Abstract

Batch processes involving integrated reaction and fractionation steps call for a higher than normal degree of co-ordination in their control. Ideally, active feedback control action would operate on component concentrations within the reactor/still system but high capital and operating costs, associated with suitable on-line composition sensors, usually preclude this approach. With vapour/liquid equilibrium involved, an alternative is to use temperature measurements to infer underlying component concentrations. As a potential basis for this, the performance of extended Luenberger observer (ELO) and extended Kalman filter (EKF) state estimation strategies is compared in a case study based on an industrial process, for which detailed mathematical models were available. The EKF scheme proves to have superior capability in delivering acceptable estimate accuracy, even in the face of process/model mismatch. This mismatch arose in using a reduced order model in the estimators which cuts computational overheads dramatically but neglects column fluid hold-up and assumes binary rather than true multicomponent VLE.