Page 401 - Design of Simple and Robust Process Plants

388 Chapter 9 Operation Optimization
In the example of the total plant control (Figure 9.13 and Table 9.4), a reactor and
three finishing towers are shown. The pressure control planned with the overhead
condensers is not shown on the flowsheet; neither are the level controllers on the
overhead drums and the bottoms. The products A and B have a hard quality specifi-
cation.
The following set-points are controlled by MBC: The quality controllers of top and
bottom streams of the distillation towers; and feed maximization for the reactant X.
OO manipulates the set-points for the feed (optional)and the ratio of the reactant
Y versus X, the recycle flow of product B, the reactor temperature and pressure, the
quality of recycle Y, the quality of product C.
The choices of who is controlling what has an impact on the modeling activities
for MBC, as well as for OO, particularly in relation to the description of the con-
straints. The objectives for BC are stable operation with hard specification quality
loops closed, and the soft quality loops open in case of interaction. This is based on
the control design philosophy of Luyben et al. (1998), who phrased it as:

It is always the best to utilize the simplest control system that will achieve the desired
objective.

To differentiate BC from MBC, the author introduced the following concept:

Basic control design has conceptual to be simple and easily operable in case of failure of
a higher control layer.

The above statement includes a preference for the self-optimizing control approach
at the basic control level, as is advocated by Skogestad.
The objectives for MBC in steady-state optimization are: self-optimization of
units, decoupling of interaction, constraint control like capacity maximization and
model predictive controllers to achieve better control particular for transient opera-
tions like batch reactors.
The objectives for OO are: optimize operation by downloading optimal set-points
or transients operations trajectories from process-wide optimization simulations.
It is the executive who, depending on the situation, sets the decision as to which
controller is active for implementation. The communication lines are illustrated for
the different layers in Figure 9.15, which shows that OO may send set-points to
MBC as well as to the BC layer, and also receive information from both systems. In
case of an outage of a specific MBC function (as in the example of the bottom qual-
ity controller), the OO may still function. It can perform its function with a DOF
less or, it can determine optimal bottom quality and select the corresponding set-
point of the steam reboiler flow with an additional margin, to avoid off-spec situa-
tions as a result of dynamics. All the above situations need to be considered and
implemented through the executive.