Self-optimizing control.

Thumbnail

Event details

Date 19.11.2010
Hour 10:15
Speaker Prof. S. Skogestad, Department of Chemical Engineering, NTNU Trondheim, Norway.
Location
ME C2405
Category Conferences - Seminars
Self-optimizing control deals with the selection of measurements (y) or measurement combinations as controlled variables (CVs), c = Hy. The issue is to select H. This is an important decision which is usually not view as a decision at all, and certainly not treated systematically. The term "self-optimizing" refers to cases where one can keep constant setpoints for the CVs, without any need to reoptimize when disturbances (d) occur. In the talk, some approaches for selecting self-optimizing variables are reviewed, including the very simple nullspace method which is to select H such that HF=0 where F = dyopt/dd is the optimal sensitivity. More generally, one should select H to minimize the norm of HF. Normally, F is obtained from the model, but it can also be used to find patterns in optimal data. It is interesting that the focus is on the small singular values of the data matrix F,and not on the large singular values as is normally the case with "chemiometric" methods. It is argued that self-optimizing control is not an alternative to real-time optimization (RTO), NCO tracking or model predictive control (MPC), but is to be seen as complementary. In self-optimizing control we determine controlled variables (CV). Preferably, the CV set points are kept constant, but they may also be adjusted using RTO or NCO tracking. In any case, a good choice of CVs will reduce the frequency of setpoint changes by RTO or NCO tracking. When selecting self-optimizing CVs, a set of disturbances has to be assumed, as unexpected disturbances are not rejected in SOC. On the other hand, RTO and NCO tracking adapt the inputs at given sample times without any assumptions on what disturbances occur. By using NCO tracking in the optimization layer and SOC in the control layer below, we demonstrate that the methods are complementary. This combination allows for fast optimal action for the expected disturbances (by SOC), while other disturbances are compensated by NCO tracking on a slower time scale.

Practical information

  • General public
  • Free

Event broadcasted in

Share