Online dynamic optimization studies of catalyzed esterfication of propionic anhydride with 2-butanol in the presence of disturbance and uncertainty
Loading...
Date
2015-09-01
Authors
Fakhrony Sholahudin Rohman
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In the presence of significant disturbance and uncertainty, a plant can suffer with
the off-spec products which lead to significant loss in profit. To overcome those
problems, the online dynamic optimization is the best strategy to be implemented.
Consequently, in this study, the simple online dynamic optimization strategy namely
cascaded-conditional based optimization has been proposed. In the cascaded-conditional
based optimization framework proposed, the re-optimization and control problems are
solved separately in two layers. The control problem is solved by using an adaptive PID
controller. Meanwhile, the re-optimization problem solving includes conditional update
and dynamic re-optimization mechanism. A simple trigger to activate the dynamic reoptimizer
is introduced when a deviation of ±5% conversion occurred in the plant which
acts as a conditional update. The dynamic re-optimization mechanism proposed also
consists of estimator and dynamic re-optimizer. The embedded estimator is to provide
the full information of current states and the control variable input as required in
dynamic re-optimizer for computing the new optimal trajectory.
To determine the most effective dynamic optimization technique, an offline
dynamic optimization study has been carried out. Five dynamic optimization techniques
namely control vector parameterization (CVP), orthogonal collocation (OC), multiple
shooting (MS), differential evolution (DE) and hybrid strategy (HS) have been
implemented to optimize three different objective functions, i.e. maximize conversion,
minimize final time and maximize operation profit. The most effective optimal
trajectories achieved are those obtained from OC in maximizing profit. The optimum
value of conversion, final time and profit obtained from the OC are 99.9%, 60 min and
RM12.84/min. Meanwhile, in order to determine the most effective estimator, derivative-free estimators, i.e. the scaled unscented Kalman filter (sUKF), the Divided Difference Kalman Filter (DDKF) and the Cubature Kalman Filter (CKF) have been evaluated in six different cases. The results showed that the CKF is the most effective estimators since it can give estimation with the highest degree accuracy with lowest RMSE for most cases tested. Finally, the proposed online dynamic optimization strategy has been implemented and evaluated in a Catalyzed Esterification of Propionic Anhydride with 2- Butanol. When the significant disturbance and uncertainty occurred, the proposed online dynamic optimization strategy manages to generate a new optimal trajectory to maintain the conversion within desired range (on-spec).