Identifying linear models from time domain simulations

Abstract

Traditionally, the linear analysis of a power system is performed by first computing a state space model, followed by the application of a suitable analysis method. However, the applicability of this approach is restricted by several practical limitations: typical power systems are represented by very large state matrices that require specialized large-scale eigen analysis programs; the program where the power system models reside does not include linearization capabilities; only time domain data is available. This article illustrates the applicability of a new system identification software program, SYSID, developed to circumvent these limitations. The program incorporates several identification techniques that allow for the computation of equivalent linear models, suitable for control design applications, from time domain simulation data. The program has been used to calculate equivalent models from simulations performed with standard stability programs, as well as from the Electromagnetic Transient Program (EMTP) simulations of detailed three-phase power systems. The equivalent systems are small and retain the modal characteristics of the original system. The identification techniques are based on various algorithms including Prony's method, transfer function fitting, and Hankel matrices. In addition to the identification algorithms, the program also includes signal processing, plotting, modal analysis functions, and basic control design capabilities to assess closed loop system performance.