抄録
Before starting the design of a multilevel control system, it is necessary to investigate whether or not the decentralized control will really reduce the computing cost of large scale system optimization. In this paper it is examined, through a case study of a linear dynamic system with a quadratic performance index, when a two-level control system is effective with respect to the computer time.
The two-level control algorithm employed here is the successive substitution method of interaction variables. Empirical equations for the computer time are shown and the “maximum allowable iteration number” for an effective two-level control is defined.
As a result, it is pointed out that the two-level control algorithm for the linear quadratic problem considered here is not effective and also that the multilivel control may be effective if the local-optimization cost function will increase much slower than the overall-optimization one. And a new first level algorithm is proposed which meets such a computing cost qualification.
