Overstiffening of low frequency synchronizing modes in reduced order power system models

The construction of a reduced order power system model by coherency-based techniques requires first the identification of coherent groups, and second the aggregation of the generators belonging to the same coherent group. For an r-area partition of the original N-machine system, two aggregation methods, the inertial averaging and the singular perturbation, are applied to a linearized model of the power system.

Several properties of the constituent matrices of a power system model are developed in this thesis. The constituent matrix concept appears to be a powerful tool analyzing the whole aggregation problem.

The two aggregation procedures, the inertial averaging and the singular perturbation, are compared with respect to how accurately they preserve the dominant modes of the unreduced order model in the equivalent system. The analysis presented in this thesis proves that while both aggregation methods exhibit the overstiffening phenomenon, the effects are less pronounced when the equivalent model is obtained by the singular perturbation aggregation method. Based on this fact, the singular perturbation aggregation method is superior to inertial averaging aggregation method in preserving the dominant modes of the unreduced order model in the equivalent model.