Page 377 - Design of Simple and Robust Process Plants

364 Chapter 9 Operation Optimization
nificant effect on the operation of the process, they include feedstocks as well as
reactor conditions also environmental effects can be included.
In general, empirical optimization is limited to processes with:
. four to five DOFs for optimization;
. constant quality feed streams;
. fundamental process model not available; and
. not subject to any fouling, aging or catalyst activity effects, which would have
a negative effect on the plant performance.
Processes with one reactor and a few separations based on pure components as reac-
tants and limited conversions of one of the reactants are candidates for this
approach.
For the development of the model must:
. Define the operating window of the process for optimization.
. Assign the degrees of freedom relevant for the optimization.
. Develop an experimental design program to determine the impact of the
DOFs on plant performance over the whole operating window of the plant.
. Develop and implement a mass balance around the process with sufficient
redundancy to enable a mass balance reconciliation (this can utilized later for
the installation of a performance meter over the process; see Section 9.4). If
the mass balance cannot be closed, it may be best to stop considering optimi-
zation.
. Make a heat balance over the plant to confirm the utility consumption meas-
urements.
. Observe if the process is stable and, if necessary, adapt the control in such a
way that the process is sufficiently stable to achieve a mass balance that is
stable over the planned time horizon of a test. As part of this step, determine
the noise on the mass balance measurement at different operating condi-
tions.
. Start the test program and determine the impact of the DOFs at the plant
mass balance and its utility consumption.
. Develop the empirical model.
. Identify the constraints under different conditions, which set the operational
window.
Based on the empirical model, an optimization needs to be built. The next step will
be to develop the optimization model by adding a price section and economic sec-
tion, and complementing it with an optimizer. For the implementation of empirical
optimization, the best approach is to follow the methodology described later in this
chapter. The optimization might be applied off-line, but it still requires adequate
control with quality loops closed to achieve its maximal savings. The application of
empirical optimization as a stand alone application is limited but empirical models
can play a significant role as a subset of a fundamental model for optimization.