Functional design
Steady state and dynamic simulation can be
effectively used for checking regulatory control
system behaviour, identifying new instrumentation, checking feasibility of inferential development and estimating potential benefits of APC
applications.
Commissioning
The simulation of the APC strategy using the
dynamic model can be used to train engineers
and operators, increasing their confidence in the
Methodology description
The first step in this approach is that of building a steadystate simulation of the plant or unit. The second step is to
convert the steady state model into a dynamic model and
to use this to generate data that will allow the identifica-
Steady-state simulation
The steady-state simulation is a necessary step in this
approach because it provides the starting values for the
dynamic simulation, but it can also be used for many other
purposes. By looking at the steady state gains of the system, it is possible to detect if the control design is ill conditioned, in which case other manipulated variables can
be selected.
Steady state gains are also helpful in deciding the size
and direction of the moves that will be applied to the
manipulated variables to obtain the dynamic models.
Finally, repeating the steady state analysis in different
operating conditions can also help to identify strong nonlinearities, which can then be accounted for during the
model identification.
Casestudies
This section will show some examples where dynamic
simulations have been used to build models for MPC controllers and also give some closed-loop results.
Propane/propylene splitter
The second example is a propane/propylene splitter,
where the main advantage of the simulation approach is
that performing a step test on a real plant can be
extremely time consuming (the settling time of the unit is
more than 48 hours) and difficult because of the many disturbances that impact the unit.
The process flow diagram used for the steady state
and dynamic simulation of the unit is shown in Figure 5. It
can be observed that the scheme includes compression
of the overhead vapours to provide a heating medium to
the reboiler. This complicates the column operation considerably, as it introduces an energy recycle that otherwise does not exist on conventionally reboiled towers.
The model matrix obtained from the simulated step
test is shown on Figure 6. It can be observed that the
purity of the propylene product has been modelled as a
ramp variable (models in the first column), because the
settling time is much longer than the controller prediction
horizon.
These models have then been used to design the
DMCplus controller and its performance has been evaluated with a series of closed-loop simulations. One example can be seen in Figure 7, where the purity specification
for the propylene product was varied a few times (thick
blue line) and the controller was capable of achieving the
desired product grades without violating the other CV limits and rejecting feed flow disturbances.
Conclusion
Historically, the use of dynamic simulation in APC projects has been limited. The main reasons behind this are