| dc.description.abstract | The selection of controlled variables and measurements is an important aspect of
the control structure design. Using the self-optimizing control framework makes
it possible to select controlled variables as combinations of measurements which
minimizes the steady state loss. However, in a broader perspective, one is not
only interested in the steady state loss, but also in the chemical plant s economy.
This work has combined the steady state loss with prices of measurements using a
Mixed Integer Quadratic Programming (MIQP)-formulation, so that the optimal
subset of measurements also results in the overall best economics for the process
plant. In this way the control structure design could be implemented as part of the
process design phase, to make the plant as profitable as possible. This work also
utilized the power of the MIQP-formulations to include wider selection criteria,
which made it possible to select different measurement devices with a variety of
prices and uncertainty. From this it was possible to calculate the best trade-off
between prices and losses due to measurement uncertainty when different measuring
devices are available to a project. Normal process plants also handles constraints,
which also need to be controlled. The constraints also have a corresponding loss -
also called back-off (due to measurement uncertainty) related them. This has also
been included in the total cost calculation, and evaluated both within an active
constraint region and an unconstrained region. The developed methods have been
tested and evaluated on a Dummy problem and a Subsea separation system. | |
