On the use of time-frequency distributions for the power quality problem of harmonics

The presence of harmonics in the electric voltage or current waveforms constitutes a steady-state type of Power Quality event Harmonics are sinusoidal components at frequencies that are integer multiples of the fundamental frequency (50 Hz or 60 Hz, in the electric power systems). These are produced by the non-linear characteristics of the devices and loads, causing Power Quality disturbances. In this work signal processing methods are developed or appropriately adapted in order to detect harmonics and to estimate their frequency and power relative to that of the fundamental frequency. A practical problem is that neither are harmonics continuously present in the waveforms nor is their relative powers constant. For the spectral analysis of signals whose statistical properties vary in time (non-stationary signals), time-frequency distribution methods, rather than Fourier analysis based methods, are pertinent. Indeed, time-frequency distributions allow us to observe the evolution of the signal frequency content in time. The most popular time-frequency distribution is the Fourier magnitude spectrogram (squared magnitude of the Short-Time Fourier Transform, STFT). However, for applications such as signal variations detection, where increased time and frequency resolutions are required, the Choi-Willians Distribution (CWD), a member of the Cohen-class distributions family, is preferable. Its efficiency for harmonics detection and estimation is shown here by simulations on synthetic signals with satisfactory results.