Abstract: In view of the complexity of power quality disturbance (PQD), a recognition method of PQD feature extraction method based on maximum variance unfolding (MVU) is proposed and PQD recognition is completed by combining classifier algorithm in this paper. The wavelet energy features of PQD are extracted by wavelet transform to construct the original feature set.For the reason that non-convex quadratic programming is transformed into convex semi-definite optimization problem by kernel function, the MVU method is applied to compress the sample features into 3-dimension features which keep low dimensional structure in high dimensional space and make a rough recognition of PQD. General classifiers are used to verify the method. The experimental results show that this MVU method reduces the number of eigenvectors and improves the recognition accuracy of the PQD. It is promising in engineering.

Key words: power quality disturbance, manifold learning, maximum variance unfolding, features extraction, disturbance recognition