A method to evaluate harmonic model-based estimations under non-white measured noise

Automatic extracting information from power-system event recordings requires applications of signal-processing estimation techniques whose performance has been verified under white noise. This paper proposes a method to test these techniques under real power-system noise, which is very different from white noise, to evaluate their application feasibility. The first part of the paper describes the evaluation method used to evaluate the techniques in a statistical sense and a method to extract noise from measured power-system recordings. The second part of the paper focuses on the evaluation of a number of harmonic model-based techniques under non-white noise, including: Kalman filter, MUSIC, ESPRIT, and segmentation algorithms. The paper shows that for the Kalman filter, a very high order with high computational burden is necessary only if high frequency components are of interest. The application of MUSIC, ESPRIT, and the segmentation algorithms under natural power-system noise is shown to be feasible.