This paper describes the first partial eigensolution algorithm to effi
ciently and simultaneously compute the set of dominant closed-loop pol
es in any high-order scalar transfer function. The proposed algorithm
is called the Dominant Pole Spectrum Eigensolver (DPSE). The DPSE is a
subspace iteration method, which operates on both left and right subs
paces to produce better estimates for the dominant poles of the transf
er function F(s). The results presented are related to the study of lo
w-frequency oscillations in a large power system dynamic model, but th
e proposed algorithm is completely general. Reduced-order models are e
ffectively obtained with the use of DPSE, as shown in the payer.