One of the recent attempts to localize the spectrum of a non-stationary time series has been the wavelet transform. Although wavelet transform is an excellent tool for detecting and localizing power quality disturbance events, it fails to classify them. This paper therefore, presents a newer approach by phase correcting the wavelet transform known as S-Transform. The S-Transform separates the localizing-in-time aspect of the real valued Gaussian window with the modulation (selection of frequency), so that the window translates and not the modulation. By not translating the oscillatory exponential kernel, the S-Transform localizes the real and imaginary components of the spectrum independently, localizing both the phase and amplitude spectrum. This aspect of the S-Transform is an improvement on the wavelet transform in that the average of all the local spectra does indeed give the same result as the Fourier Transform. Further S-Transform is generalized and used to detect, localize and classify the power quality disturbance events using the extracted features and a simple rule base.