Variable Projection Method for Power System Modal Identification

Abstract

Common power system disturbances are characterized by oscillations that either damp out, sustain or grow. This oscillatory response can be measured and analyzed to extract modal information of the system. In this paper we apply a nonlinear least-squares optimization method known as the variable projection method to power system models and data. The variable projection method is compared against several linear methods such as Prony analysis, Matrix-Pencil and ERA, and the results suggest the nonlinear least-squares approach is superior.