In this study, an algorithm based on pattern recognition approach is proposed for classification of power quality disturbance types. For feature extraction which is an important part of the pattern recognition, a method based on entropy which uses the decomposition coefficients of wavelet transform is presented. The most important advantage of the method is the reduction of data size without losing main distinguishing characteristics of signal. Support vector machines based on statistical learning theory is used as a classifier. The performance of the proposed algorithm is evaluated by using real and synthetic power quality disturbance data. Real power quality disturbance data are obtained from our national power system. Besides, the synthetic power quality disturbance data are obtained from ATP/EMTP and mathematical models. The analyses and results obtained in this study show that proposed algorithm has an efficient, feasible and practical structure.