Research On Power Quality Disturbance Identification And Treatment Method Based On Distributed Compressive Sensing

A distribution network is an important part of the power system.With the increasing diversification of new energy power generation and the trend of load equipment towards power electronics,more and more distributed generations（DGs）and power electronic equipment are connected to the distribution network,which not only makes the amount of data in the power grid increasingly large,but also makes the problem of power quality more prominent.Therefore,in order to realize the efficient collection,accurate identification and comprehensive treatment of power quality disturbance data,and improve the means of power quality monitoring and control,this paper carries out the research of power quality disturbance identification and management method based on distributed compressed sensing.Firstly,aiming at the problem that power quality data in the distribution network is increasingly large and difficult to collect and store,this paper proposes a power quality data compression storage method based on distributed compressed sensing and edge computing.An adaptive joint reconstruction algorithm is proposed by combining synchronous orthogonal match tracing algorithm and K-SVD（K-Singular Value Decomposition）dictionary learning algorithm,which is applied to the cloud edge cooperation framework with distributed compressive sensing as the edge algorithm.The compressed partition storage of power quality data is realized by analyzing the dictionary atoms and measured values uploaded on the edge in the cloud,which lays a foundation for the subsequent research on power quality disturbance identification and treatment methods.Secondly,aiming at the problems of difficult feature extraction and vulnerable to harmonic interference of power quality disturbance data,a power system disturbance data acquisition and classification algorithm based on compressed sensing is proposed.The typical single and composite disturbance data are modeled,and the sparse characteristics of disturbance data,the standard deviation,kurtosis,margin factor and the number of dominant frequencies of adaptive dictionary atoms are combined as the classification characteristics of disturbance data.BP（back propagation）neural network is used to realize sample learning and classification.It has the advantages of high classification and identification accuracy and strong anti-interference,which is conducive to the targeted treatment of power quality disturbance.Finally,aiming at the problems of high penetration and decentralized treatment of harmonic pollution in the distribution network,this paper proposes a zoning treatment method of harmonic pollution based on distributed compressed sensing.The voltage data of each node in the distribution network are decomposed under the same dictionary atom to obtain the corresponding sparse distribution characteristics of harmonic voltage.Then the fitness function of the genetic algorithm is designed by comprehensively considering the relationship between the sum of Euclidean distances of sparse characteristics between nodes in each partition and the number of total partitions,so as to realize the fast partition governance of distribution network harmonics.The simulation results show that the proposed method is easier to perceive the harmonic pollution in the distribution network,has the advantages of a small amount of calculation and strong noise resistance,which provides an important reference for the localization and management of harmonic pollution in the distribution network.