Robust automatic segregation of harmonics in electric drives by means of the Mahalanobis distance

Abstract

This paper discusses the use of the Mahalanobis distance as a solution for a robust and automatic segregation of harmonics in the signals measured in electric drives, as for example the phase currents or the mechanical speed. The Mahalanobis distance is used during real-time operation to clean the results of an on-line frequency analysis, reducing the uncertainties related to frequency resolution, spectral leakage and background noise. At the same time, it is a powerful tool to classify the known harmonics from the unknown ones, thus providing the drive with valuable information that can be used for further decisional control actions. The overall theoretical background is reviewed as well as its implementation in an operational drive system. Experimental results are provided, proving the validity of the concept.