Chemical Engineering November 2010 - 50

Feature Report
Connecting Process
Simulators to the
Control Room
The benefits are no longer restricted to high-end
integrations or large projects. Here is how to do it in
small-to-medium sized projects and why
O
ne of the best ways to realize
value from your process data is
to use those data - both realtime
and historical - in a process
simulator. The many applications
for doing so are generally categorized
into offline and online categories.
Offline applications, where the
simulator is not directly connected to
the control system, use process data to
validate dynamic simulations of the
process, or to improve steady-state
model fidelity. The goals of online applications
include checking current
performance, using the simulator as
a software sensor, and even monitoring
sensors. With online applications,
results of the simulation can be used
by the control system.
In the past, combining control system
data with process simulation models
was restricted to high-end integrations,
such as dynamic matrix control
(DMC) and realtime optimizers. As a
result of improvements in computer
hardware, software, and open standards
for data exchange, it is now possible
for the average engineer (and the
average computer) to make useful connections
between simulation models
and control systems (Figure 1).
This article discusses possible applications,
provides some illustrative examples
and addresses what is needed
to connect the systems.
About process simulators
Process simulators are heat- and massbalance
software programs that model
process unit operations. Traditional
simulators are steady-state models;
they calculate 'snapshot' results based
on the assumption that feedstreams
and specifications are constant, with
no holdup, delay or transient conditions
in the process being modeled.
Dynamic simulators, in contrast, do
account for holdup, delay, and some
transient conditions. Some vendors
build a dynamic simulator on top of
their steady-state programs, while
others have separate standalone programs
for dynamics.
About modern DCS
Modern distributed control systems
(DCS) connect data from sensors,
control systems and operator panels.
Data historians record the information
for future analysis. Computer
network connections allow access to
DCS information from other systems
and software.
Reasons for integration
Improve steady-state models by
reconciling them with data from
the plant historian/DCS. Allow the
optimization package of your simulator
to change specifications for the
model until it agrees (statistically)
better with plant data. An optimization
package can do this more accurately
than trial-and-error manual
adjustments by the user. It is possible
to automate this process so that the
simulator model data are regularly
reconciled against the DCS data.
34 ChemiCal engineering www.Che.Com november 2010
David Hill
Chemstations
Kenichi Matsuoka
Inpex
Validate test-control logic. Validate
test-control logic in a dynamic model
by using the model with historical
data. Specify disturbances at the time
they occurred in real life, and determine
whether the model responds
the same way the real-life equipment
responded. A validated model can be
used to test changes to the control
logic in a virtual plant. It can also be
used to try to forecast the current system's
response to disturbances.
Use a simulation as an online software
sensor to predict values that
are difficult to measure. A simulation
model could, for example, be used
to predict conversion from a reactor or
liquid composition of a stream, based
on more easily measured variables
and unit operation calculations. Depending
on application and confidence
in the model, this approach could be
used as a " for information only " value
by operators, or it could be used as
part of closed-loop control. It is also
possible to perform sensor validation,
to determine whether the values that
your sensors report are consistent
with each other.
Additional possibilities include creating
a simplified operator training
simulator (OTS), startup and shutdown
models based on historical data, data filtering
and statistical sensor validation,
dynamics optimizations, and modeling
predictive control using the simulator
as a model for the controllers.
Example applications
Historical data in dynamic simulations.
A natural gas pipeline had
two looped branches. A third loop was
being added to expand capacity (Figure
2). There was a concern that the controllers
for the three branches would
battle each other, causing oscillations
that would affect the flow to custom
http://www.Che.Com
Chemical Engineering November 2010

Table of Contents for the Digital Edition of Chemical Engineering November 2010
