Analyzing Signals from a Primary Winding of a Locomotive Transformer with Three Types of Wavelet-Based Transforms

Abstract

The paper is concerned with the analysis of in-stantaneous values and evaluation of certain power quality indices corresponding to phase currents and voltages ac-quired from the primary winding of a locomotive trans-former. Representative data sets acquired during all operat-ing regimes (acceleration, constant speed, normal and re-spectively regenerative braking) were analyzed. Firstly the background noise was estimated by using a wavelet thrash-ing tree considering a wavelet mother with a short filter. Specific computational aspects related to the use of the Sta-tionary Wavelet Transform were presented and it was used to evaluate the instantaneous values of the fundamental harmonic, respectively the distorting residue. Discrete Wavelet Transform was used for 3 situations to deal with time-frequency localization of deviations from stationarity. A nodal analysis was also made with an original implemen-tation of the Wavelet Packet Transform in a special situa-tion, when the deviation from stationarity nearby the right edge prevents the use of DWT. Excepting the thrashing tree, the wavelet mother used in all transformations relied on a filter of length 40 due to the good properties of selectivity relative to the harmonics’ energies. All methods were used to evaluate the 3 major root mean square values (for fun-damental frequency, for distorting residue and total). The results yielded by different methods were compared. A good convergence of methods was noticed. Explanations for the small differences are suggested. images