Maximum likelihood frequency estimation in smart grid applications

This paper focuses on the estimation of the fundamental frequency in balanced three-phase power systems. Specifically, we propose a Maximum Likelihood Estimator (MLE) that exploits the multidimensional nature of electrical signals. For perfectly sinusoidal signals, we show that the MLE can be expressed according to the periodogram of the instantaneous positive component. For harmonic signals, we demonstrate that the MLE can be approximated by a cumulated periodogram of the zero, positive and negative sequence components. As compared to single-phase estimators, statistical analysis and simulation results prove that the proposed estimator decreases the Mean Square Error by a factor of three, whatever the Signal to Noise Ratio (SNR) or data length. Furthermore, simulations with experimental data show that the proposed technique outperforms classical spectral estimators such as MUSIC.