This thesis presents methods for power quality event detection and classification using selective 2D-wavelet subspaces. Distinctive features of signals obtained from different disturbance events are extracted using unique characteristics of corresponding wavelet coefficients presented in different 2D subspaces. Comparison between the proposed method with existing methods which are based on 1D wavelet and 2D wavelet with non-selective subspaces have been carried out. In addition, the thesis proposes a method for disturbance classification using neural network. The detection and classification algorithms were developed and implemented using MATLAB program and verified using both synthesized and actual signals captured from real power systems. The results show the method based on 2D-wavelet subspaces suitable for power quality event detection and has superior performance in classification than the traditional method using 1D wavelet