New multi-level algorithm for linear dynamical systems

Abstract

A new approach is developed for the solution of linear interconnected dynamical systems. The algorithm is within the class of non-feasible methods. Its main feature is that it allows the transfer of information between the subsystems with each iteration. The convergence properties of this technique are given in detail. It is shown that the proposed methodology may be faster than the well-known prediction algorithm given in the literature since in each iteration better interconnection information is used within the period of the optimization horizon.