J.A. Jordaan and B.J. van Wyk (South Africa)
Frequency estimation, Newton method, Separable Least Squares, Power systems.
This paper presents the Separable Least Squares (SLS) method for estimating the frequency and other model pa rameters of signals recorded in power systems, and com pares it to other algorithms. These algorithms take into account an exponentially decaying DC component, fun damental frequency and components until the M-th har monic. The other algorithms tested are: Linear Least Squares method using a second order Taylor series approx imation of an exponentially decaying DC component, and the Newton method. The parameters estimated are: ampli tudes of the DC component and harmonics, time constant of the decaying DC component, the frequency of the funda mental component and the phase angle of each harmonic. The SLS method appears to be more robust than the other methods since more accurate results were obtained when the assumed fundamental frequency was far away from the actual frequency.
Important Links:
Go Back
