Tracking of instantaneous spectrum components using time-frequency methods

This paper serves the idea of applying joint time-frequency representations for analysis of non-stationary signals in electrical engineering. Main directions of researches are concentrated around Cohenpsilas class of transformation which gives some possibilities of adaptation to analyzed signal by choosing appropriate kernel function. Additionally, novel approach delivered by S-transform is also introduced. In order to investigate the methods several experiments were performed using simulated phenomena of switching on the capacitor banks in distribution system. Obtained results were supplemented by comparison to classical spectrogram. Proposed methods allowed to track instantaneous spectrum components well as its energy with better precision than classical spectrogram. It leads to applications in diagnosis and power quality area.