Abstract: Power system disturbance data is of great significance for monitoring power system operation state and regulating its working mode. A new algorithm for power system disturbance data acquisition and classification is proposed based on compressed sensing. Firstly, an AtOMP ( Atom adaptive Matching Pursuit) algorithm is designed based on compressed sensing and K-SVD ( K-Singular Value Decomposition ) dictionary learning algorithm to compress and collect various disturbed data. Then, the sparse feature of the disturbed data, the standard deviation of the adaptive dictionary atom, kurtosis, margin factor and the number of principal frequencies are extracted as the training samples, and the BP neural network is used to realize the sample learning and classification. The experimental results show that the proposed algorithm can achieve highly compressed data collection of disturbed data, small data amount, high classification recognition, strong anti-interference and other advantages.

Key words: compressed sensing, sparsely dictionary, K-singular value decomposition ( K-SVD) algorithm, disturbance signal classification, back propagation (BP) neural network