Page 387 - Design of Simple and Robust Process Plants

374 Chapter 9 Operation Optimization
the constraints that have an impact on the overall operation are identified at differ-
ent operational conditions. During model building these factors must be properly
reflected in the model. The model might be very detailed for one unit, with a de-
tailed description of all the constraints to be encountered at different conditions. On
the other hand, it may be very elementary for another unit because it only has mar-
ginal impact on the economics and does not encounter any constraint. In this last
case, one might think about a distillation column or stripping column that could be
formulated in split factors with an energy term connected. The validation of the
model is discussed in Section 9.6.12.

Parameter estimation Typical parameters to be determined include: compressor
and column efficiencies; fouling factors or overall transfer coefficients of exchan-
gers; and catalyst aging factors. These parameters are often only calculated as there
is not enough redundancy built into the flowsheet for them to be estimated. The
technique for the determination of the parameters is essentially the same as applied
for data reconciliation, the Lagrange multiplier method generally being applied (see
data reconciliation).
During the DR /PE the gross errors of a selected set of measurements are deter-
mined by specifying operational limits. One might decide to validate all measure-
ments of the plant with this methodology. It should be realized however, that it
would mean writing very detailed models. In addition, one would also need to be
able to describe the product streams and its physical properties in detail. For exam-
ple, the composition of a tar stream which might change at different feed or reactor
conditions is often not known, but it will have an effect on the vapor pressure and as
such at the bottom temperature or pressure of a tar splitter. Similar comments can
be made about the impact of inertia, if the quantity of inertia is not accurately
known, then the calculated dew point of a stream will not be reliable.

Decider 3 The decisions to be made at this point are:
. Is decider 1 true?
. Is decider 2 true?
. Are the estimated parameters available and within pre-defined range?
. Are the gross errors removed from the selected measurement set?
. Are the DOFs for the optimization still applicable (are the control loops for
the implementation of the set-points of the optimizer still closed)?

9.5.2.4 Optimization
Updated model Three activities can be recognized:
1. The optimization step starts with updated model parameters.
2. The DOFs for the optimization need to be available for implementation..
3. The constraint are set/updated as far as these are not calculated within the
model.