This paper presents statistical methods and wavelet based effective feature extraction method for power quality (PQ) disturbance classification problem. The PQ signals used in this study are two common types named as swell and sag. First, the signals consisting of sag and swell are determined by using statistical methods. In the previous studies, validation of PQ disturbances for obtaining skewness and kurtosis coefficients were created at the zero crossing points of the voltage signal. In practice, occurrence of disturbances at these points is not guaranteed. So in this paper, disturbances are constituted in eight different points (0° ,45°, 90°, 135°, 180°, 225°, 270°, 315°) having different characteristics. Skewness and kurtosis coefficients of the constituted signals are calculated in local frames. These coefficients are obtained during one period long sliding frame. It has been observed that in swell and sag events this statistical method gives different results depending on moment occurrence of disturbances. So another method is needed. Multi-resolution analysis (MRA) technique of discrete wavelet technique (DWT) and Parseval's theorem are employed to extract the energy distribution features of sag and swell signals constituted in eight different points (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°).