Bilinear time-frequency distributions (TFDs) are developed to represent time-varying signal jointly in time and frequency representation (TFR). Since the TFDs offer good time and frequency resolution, they are appropriate to analyze power quality signals that consist of magnitude variation and multiple frequencies. However, the TFD suffer from cross-terms interferences due to their bilinear structures. In this paper, smooth-windowed Wigner-Ville distribution (SWWVD) is used to analyze power quality signals. The power quality signals are swell, sag, interruption, harmonic, interharmonic and transient. To get accurate TFR, the parameters of the separable kernel are estimated from the signal. A set of performance measures is defined and used to compare the TFR for various kernel parameters. The comparison shows that signals with different parameters require different kernel settings to get the optimal TFR.