Skip to Main Content
A new technique for estimating the frequency contents in a signal is applied to power systems. The technique uses two sets of measured samples: samples from the signal itself and samples from a filtered signal obtained by passing the original signal through a continuous-time filter. The estimates of the frequencies contained in the signal are obtained from the eigenvalues of the generalized eigenproblem of two matrices formed from the two sets of samples. The technique assumes estimating multiple sinusoids which makes it suitable in cases with large harmonic distortion. Results show that by the proper choice of the sampling time, the eigenproblem condition is improved such that accurate estimates become attainable in case of noise. Further improvement in the estimate accuracy is achievable by utilizing correlated samples. The technique is also applied to estimating varying frequency.