Page 393 - Design of Simple and Robust Process Plants

380 Chapter 9 Operation Optimization
It should be realized that the implementation of an OO project takes between 18
months to 2 years. An approaching improvement project will have an impact on an
OO project, and therefore should be considered during implementation.
The feasibility study should conclude with a list of potential project options to be
explored in more detail. The next steps are explained for a static/dynamic OO pro-
ject, and they all serve as back-up for the cost±benefit analysis. (Scheduling projects
would follow another track, but this is not discussed here.)Most of the following
activities will be worked out in more detail during the development of the OO pro-
ject.

9.6.1.2 Modeling background
It is important to have the modeling background defined. The process modeling
background is important, as most companies have modeling experience and simula-
tions available of the process under study, with some preferred modeling software.
This has an impact on the modeling effort to be performed, and also on the eventual
software license cost.
Reactor modeling is a must to enable accurate OO models for processes with a
reactor. The type of reactor models depends on the available knowledge ± it may be
empirical, a conversion model, or a fundamental model. It is the required accuracy
(which is a function of the overall contribution of the reactor system to the process
operational cost)that dominates the type of reactor model. A reactor that sets the
product distribution of the process is most likely a prime candidate for a rigorous
fundamental model. A reactor that operates at close to 100% conversion (e.g., hydro-
genation of a di-olefin in the presence of olefins)can often be optimized locally by a
self-optimizing controller. In such a case, a conversion model implemented in the
overall process model might be sufficient. For commercially available processes,
reactor models might also be available commercially. Often these are closed ver-
sions, in which case it is better to negotiate excess to the derivatives functions in
order to enable optimization.
Although the cost of a reactor model can be high, is often a prerequisite for an
OO project. It should be noted that the development and validation of a model
requires stable process conditions and accurate measuring techniques that are nor-
mally not available in a process plant. This could be an argument to include these
requirements in a OO project.

9.6.1.3 DOFs for optimization and model-based control
The DOFs for the plant-wide optimization are in general the specification of internal
streams and the set-points for reactor conditions. These set-points will be down-
loaded from the OO executive to the MBC system or the basic control system,
depending on who is controlling what. Sub optimizations such as self-optimizing
control have to be predetermined at this step in relation to plant-wide optimization.
A preliminary split must be made for the DOFs applied at the MBC or OO layer.
Constraints are preferably approached by MBC, as they often take advantage of
direct process measurements with avoidance of calculations. For example, if a con-
denser is subject to fouling, optimization does need to estimate a fouling factor and