Basic Frequency and Time Domain Least Squares Methods for System Identification

This paper investigates the feasibility of applying least squares (LS) techniques for estimating the low-frequency electromechanical modes of a power system. The performance of time domain least squares (TDLS) and frequency domain least squares (FDLS) system identification algorithms is studied. These algorithms use measured input and output signals to estimate system models. Simulation studies are performed. These simulations include using a test system model developed based on a series of tests performed on the Western North American Power Grid (WNAPG) in June, 2000.

I. Introduction

Accurate knowledge of the low-frequency electromechanical modes is critical for safe low-frequency operation a power system. Monitoring these modes by close examination of measured power system data provides valuable information. Much recent work has focused on estimating the electromechanical modes of oscillation from measured data [1]–[7]. Many of these techniques use only measured output data while others use both measured input and output data. Most of these algorithms focus on time domain data for system identification.