This paper addresses the subject of power quality data mining and the role of signal transforms played in the above knowledge discovery process. We explore the performance of both Discrete Wavelet Transform (DWT) and S-transform in the feature extraction and classification stages of this mining process In our work, wavelet transform and S-transform(ST) were applied to transient power system data and pertinent features were extracted to further train a Learning Vector Quantization network (LVQ) for power disturbance classification. It was found that the number of features and hence the size and the training time of the LVQ network were considerably reduced in case of ST features. Also, the classification accuracy of the LVQ classifier was increased in the case of time -domain featured disturbances, such as sags, swell, etc when trained with ST features. Moreover, unlike Wavelet transform -based recognition system which is highly sensitive to the presence of noise, in the case of S-transform based system results are found to be quite satisfactory up to a noise level of 3.5%.