Invariant subspace method for eigenvalue computation

Methods which first subdivide a large power system into subsystems, and which then study how the interactions between subsystems cause the eigenvalues and eigenvectors to vary between the subsystems and the total system are presented. The invariant subspace method, which allows eigenvalues that are difficult to study individually to be grouped into a subspace which can be handled quite efficiently, is described. The way in which the method allows the local and global subsystem eigenvalues to be identified is also described. The greatest change occurs in the global eigenvalues, those which influence the response of more than one of the subsystems. These eigenvalues are of particular interest as they are the type that could cause interarea oscillations