A STRATEGY FOR INCLUSION OF CLOSED LOOP DYNAMICS IN REAL TIME OPTIMIZATION WITH APPLICATION TO AN OXYGEN DELIGNIFICATION REACTOR

Abstract

As driven by increasing energy costs, raw material costs and market competition, it is a necessity that modern chemical plants operate at the optimum operating point and are responsive to changes in product specification. The calculated plant optimum operating point may be at or close to constraint boundaries, which makes the process susceptible to constraint violation, off-specification products and loss of profitability in the presence of disturbances. A new approach has been developed to track the optimum of the chemical process such that violation of constraints can be prevented by inclusion of closed loop dynamics in Real Time Optimization. Constrained model predictive control will be used as the regulatory control. This new approach introduces an additional layer in the process automation hierarchy which determines an appropriate amount of back-off from target set points based on a closed-loop dynamic model of the process. It does not require a large effort in modelling since the dynamic model is that the model used in model predictive control and the steady-state relation is the steady-state process gain of the dynamic model inside the model predictive control. It is assumed that the target set-points from the Real-Time Optimization are available to be used in our approach. The new approach (dynamic real-time optimization) is formulated here as a multilevel program where the upper-level problem has a quadratic objective function with linear con straints and the lower-level optimization problems have quadratic objective functions that are strictly convex with linear constraints. A quadratic dynamic matrix control formulation gives rise to the lower-level optimization problems. The upper-level determines set-points that are as close as possible to set-point targets calculated at the steady-state Real-Time Optimization level, but are such that the closed loop inputs and outputs satisfy specified constraints Oxygen bleaching in pulp mills is an example of chemical plants facing economic and environmental challenges. Improvements in the operation of oxygen delignification reactors could have a potentially significant impact on the controllability of downstream units of the bleaching plant and the overall plant performance. Developing a dynamic model of the oxygen delignification reactor is a necessity toward meeting this objective through the development of model-based control schemes and finding the optimum set-point to the controller. A first-principles nonlinear dynamic model of an oxygen delignification tower is developed, and used in the design and performance evaluation of a model-based control strategy. The proposed dynamic real-time optimization approach was then applied to the oxygen deligni fication reactor model developed to calculate the required optimum set-points by the model predictive controller in face of disturbances.