Model Reduction and Decentralized Control of Large Scale Systems Using Chained Aggregation.

摘要

This report proposes the model reduction of linear time invariant systems and the syntehsis of decentralized control with parallel computations by utilizing the given system information structure, e.g., structural and output information. We first develop the chained aggregation procedure which leads to the generalized Hessenberg system representation (GHR). The structural features of the GHR are studied. The influences of the Hessenberg Blocks on trajectory and eigenvalue perturbation are examined. Secondly we develop the generalized QL (GQL) and the restricted QL (RQL) algorithms to be used in conjunction with the GHR for construction of reduced order models. The GHR is shown to exhibit as class of plausible reduced order models suitable for further parameter adjustment if necessary. Finally, we exploit the use of GHR structure for the synthesis of decentralized control in the tracking of reference interconnection variables in an interconnected large scale system. A hierarchical structure is proposed which allows decentralized and parallel computations. Model reduction and control synthesis are illustrated using power system and rocket dynamics as examples. (Author)