The principal objective of this project is to consider the link between optimization and control, that is, how to
implement optimal operation in practice. One approach is self-optimizing control, where one uses a constant setpoint policy for the controlle d
variables, and in this project we will extend and develop this idea further.
This project is related to a larger activity in plantwide control performed in the group of Professor Sigurd Skogestad
at NTNU. Plantwide control deals primarily with the stru ctural issues of the control system, such as what to
measure and control, which inputs to use, and how to pair these sets of variables. This is widely recognized as the
most important issue for chemical process control (Seborg et al 2004), yet its theoret ical basis remains quite weak,
especially when compared with the large amount of work on optimization and control design (with structure fixed).
We also plan to use the same idea to precondition dynamic optimization problems and thus speed-up their solut ion and improve their convergence properties. This idea is already used in the context of robust model predictive control where it known that
the standard open-loop optimization, with feedback indirectly introduced through a moving horizon, does not yield the optimal
solution because it does not account for the fact that feedback actually will be used (e.g., see work by Mayne and coworkers).
A much better approach is to include a reasonable "pre-conditioning" (suboptimal) feedback policy,
which then cha nges the dynamics of the process to be optimized, and then let the open-loop provide a correction to this.
We plan to use the same idea in a different context, namely to improve the speed and convergence of dynamic optimization problems.
This application is for the funding of two Ph.D. students and one postdoc, each for a period of 3 years. We will
encourage females to apply.