Page 414 - Design of Simple and Robust Process Plants

9.6 Project Methodology for Operation Optimization 401
. Development and installation of the performance meter needs to follow the
scope of the project and satisfy the accuracy requirements.
. Development of the control structure is an essential element of an OO project
as it is based on the assumption of steady-state operation, and emphasizes
closed loop operation for BC and MBC. The results of this step are used for
the design of the BC and MBC, and the DOFs for MBC and OO are defined.
. Model development needs to be based on an equation-based simulator to
achieve rapid solutions. Next to the modeling of the performance of the indi-
vidual units (including the reactor model(s)), it is especially the description of
the constraints that requires attention. During optimization, the process will
be pushed against its constraints, and either these constraints are incorpo-
rated into a constraint controller or they are calculated and implemented in
the model. The parameters are selected preliminarily for the model building;
the final selection will be done during model validation. The robustness of
the model need to be extensively tested over the operating window.
. Build executive defines the OO operational architecture with all necessary
empty steps included. The communication between OO, BC and MBC
through a database at BC level must be defined.
. Implementation of data analysis, data reconciliation (including full designed
performance meter and the parameter estimation)takes place step by step.
In effect, the optimization is taken in operation step by step. During the data
analysis step, gross instrument errors are detected, the steady-state situation
is observed, and the decision criteria are defined. If steady-state criteria are
not satisfied, the control needs to be reconsidered. Ultimately, the perfor-
mance of the control is verified.
. Implementation of data reconciliation emphasizes the determination of the
reconciled mass balances, resulting in determination of gross errors, feed
rates and composition, and the reconciled performance measurement. The
output of this step starts the DR&PE module.
. Implementation of parameter estimation is initially done in interaction with
the model validation step. The parameter estimation will be done simulta-
neously with data reconciliation of plant measurements (being selected dis-
tributed over the process), with the overall process model. The output of this
step is summarized in the simulated process performance. A comparison
with the reconciled performance measurement should pass a defined criteria,
to be used as warning. The parameters ultimately selected for optimization
are available for the next processing step.
. Model validation is an essential step in the design of an OO system, but it
does not directly play a role in the execution of OO cycle. The validation is
done by performing an experimental design program of the DOFs over its
operational range. During the tests runs extensive measurements and sam-
ples are taken for detailed analysis and fixed parameters are set free for esti-
mation purposes. Now, an extensive off-line study must be performed to eval-
uate the overall process model over its operational range and search for mod-
eling errors, and update parameters, including those that were fixed. Off-line