Page 384 - Design of Simple and Robust Process Plants

9.5 Closed Loop Steady-state Optimization 371
The output of the reconciliation module is a set of reconciled data to be used as
actual input for the simulation.

Decider 2 Before the next step is activated, the following decisions need to be
made:
1. Is decider 1 true?
2. Are the reconciled data available and within a certain pre-defined range?
3. Are gross errors removed from the data set?

9.5.2.3 Data reconciliation and parameter estimation (DR&PE)
The DR&PE module is built to achieve a close match between the operational plant
and the simulated process plant. The objective of the step is to provide an updated
model for optimization by estimation of the assigned parameters. The detailed activ-
ities are:

± all involved measurements are evaluated on gross errors;
± parameters estimated /calculated at actual process conditions;
± actual plant performance; and
± simulated performance, available for comparison with actual performance.
Note: Parameters are determined at actual process conditions. If the optimization
were to force the process into a different operational point, it would require an addi-
tional optimization cycle before updated parameters could be incorporated into the
optimization.
If gross errors are detected they might be eliminated from the measured set, and
the DR&PE might be repeated as described under DR in the previous paragraph.
At this point we exclude the model validation, which is discussed during the
methodology description (see Section on Validation 9.6.12). In the optimization
cycle it is assumed, that the model has been validated.
Simultaneous data reconciliation and parameter estimation will provide the best
fit between actual and simulated performances, as the reconciliation emphasizes an
extended set of measured data covering the rigorous model. The precondition is that
enough redundancy is available in the plant measurement not limited to DR but
also for PE.

The model The entire empirical or fundamental model form the bases for the esti-
mation. The factors to be addressed during model building are:
. Selection of commercial flowsheeter equation-based, provided with model
library, physical property data bank, optimization routines and economic sec-
tion.
. Overall process model, empirical versus fundamental.
. Reactor modeling ± are there kinetics available at sufficient detail?
. Projected accuracy of the model. The level of detail of the model, partly deter-
mined at the feasibility stage (see Section 9.6.1)but also during model build-
ing, to comply with model accuracy requirements.