抄録
The aggregation method is effectively used in the control synthesis of large-scale linear systems to reduce requirement on the computer. It is essentially important for the synthesis of a satisfactory sub-optimal control to construct an appropriate aggregated model. In this paper, we propose a new criterion for selecting a proper aggregation matrix based on the projection error analysis. By introducing a projection operator, the projection error between the model and the original system is expressed by an explicit function of the aggregation matrix and the system dynamics. Then the aggregation matrix may be determined under the minimization of time-integration of mean square projection error weighted by the loss matrix of the system control performance. Our method seems to be more practical than the existing ones from the viewpoint of the computing labor and the model quality, since various establishable parameters are included in this criterion. Furthermore, we also investigate about the model confidence index, which represents the pre-estimate of the performance degradation due to the aggregation control compared with the optimal control. A numerical example for the gas-absorber system is considered to indicate the effectiveness of our method.
