In this paper, we investigate the potential of support vector machines (SVMs) for power quality data mining in electrical power systems. Modified wavelet transform, known as S-transform, has been used to extract unique features of the various power quality disturbances. Feature vectors from S-transform analysis are used to train the SVM classifier. Various multi-class SVM algorithms have been applied on the power quality data under study and the directed acyclic graph (DAGSVM) algorithm is found to be performing well. A comparison between the DAGSVM method and the one based on artificial neural network demonstrates the efficiency of the SVM method in classifying PQ disturbances