| With the proposal of the development concept of strong smart grid and energy Internet,and the promotion of power Io T sensing technology,the operation of the power grid is developing in the direction of intelligence,flatness and openness,requiring the establishment of a sound and comprehensive online monitoring system for electrical equipment.However,the continuous development and expansion of the power system has brought a large number of power distribution equipment,and the need t o monitor all power distribution equipment has brought a lot of challenges to the power big data platform.Massive monitoring signals need to be compressed and stored,but the traditional compression sampling method based on Nyquist sampling theorem has a low compression rate and requires a high sampling frequency,which cannot effectively reduce the pressure.In order to reduce storage capacity and improve transmission efficiency,this paper studies the spars e dictionary training method and observation matrix optimization method of compressed sensing,and further studies its specific application in power distribution equipment monitoring signal processing,which has important theoretical significance and practical value..The main work and results of this paper are as follows:1)An atomic adaptive singular value decomposition algorithm based on dynamic threshold is proposed.Based on the problem of dictionary redundancy in the traditional KSVD algorithm,a dy namic threshold judgment link is added in the dictionary training process to adaptively delete atoms in the dictionary that contribute less to the reconstruction,reducing the redundancy of the sparse dictionary.The simulation experiment verifies that the sparse dictionary obtained by the proposed algorithm has fewer atoms,faster operation speed,larger peak signal-to-noise ratio and smaller root mean square error of the reconstruction result,and has higher reconstruction accuracy.2)An observation matrix optimization method based on sparse representation error and QR decomposition is proposed.Based on the relevant conditions of the observation matrix and the traditional optimization method,the sparse representation error is added to the observation matrix optimization model as a penalty term,and the observation matrix obtained by the algorithm is decomposed by QR to improve its column independence.Simulation experiments verify the effectiveness and superiority of the proposed ob servation matrix optimization method.3)An abnormal identification algorithm of power distribution equipment monitoring signal based on compressive sensing theory is proposed.Using the non-zero value distribution of the sparse coefficient change matrix i n the compressed sensing process,the health threshold of the change o f the sparse coefficient matrix is obtained,so as to judge the abnormal monitoring signal of the power distribution equipment.The simulation experiment verifies the effectiveness of the algorithm for abnormal identification of monitoring signals of power distribution equipment. |
