REAL-TIME ESTIMATION OF THE PARAMETERS OF POWER-SYSTEM SMALL-SIGNAL OSCILLATIONS

It is shown that finite records of measured power system signals can be exactly modelled or generated by a linear transfer function excited by a unit impulse input. The coefficients of the numerator polynomial and denominator polynomial of the transfer function completely characterize the signal and are called the linear predictive code (LPC). The signal model using the LPC can then be used to compute various signal attributes such as frequency, the presence or absence of critical modes, the degree of damping, the amplitude of various signal components and most importantly to predict the future evolution of the signal. A scheme for on-line monitoring of the LPC's of short time records of power system signals is developed such that the signal attributes may be displayed in decreasing order of importance but increasing precision. The proposed scheme is evaluated using real-time simulations of various power system oscillations.