Repetitive Tuning in Optimizing Control of Steady-state Large Scale Systems under Time-varying Uncertainty

Abstract

In the recent work (Brdyś et al., 1982) a certain me-mechanism for an optimizing control of large scale systems operating in steady-state under slowly varying unknown disturbances was introduced and developed. The mechanism utilizes price coordination with local feedback investigated for stationary systems by Brdyś and Ulanicki . The main problem considered by Brdyś et al., consisted in choosing on-line by a tuning layer (one of the control structure layers) the mechanism parameters in such a way that generated controls were feasible and resulting real system performance values didn't differ, on a whole control time interval, from the optimal ones more than it was allowed by the system user. The repetitive tuning scheme introduced there (called here repetitive short horizon tuning) minimized a cost of information exchanged during control by pushing away as far as it was possible the next model optimization layer intervention instant. At a given moment the mechanism parameters were determined only up to this instant, thus not to the end of the control time interval, in general. Although this feature is in general an advantage of such type of tuning it may appear a drawback sometimes. Namely, owing to physically limited frequency of the layer interventions the short horizon tuning may fail even if the proper parameter values exist. To evercome this drawback in the paper a new type of repetitive tuning, called long horizon tuning is proposed and thoroughly investigsated. Comparision of those two repetitive tuning schemes is done. The obtained reults are widely illustrated by numerical example simulations.