Study On Anomaly Detection And Location Method Based On PMU Data In Distribution Network

With the rapid development of China’s economy,the demand for power system stability and power quality is getting higher and higher.The slight abnormal disturbance in the distribution network may have a huge impact on social and economic development.As results,the rapid detection and accurate location of the network anomaly signal is crucial.In this paper,the linear shrinkage estimation technique of the covariance matrix,the principal component analysis method and the random matrix theory are used as tools to detect and locate the anomaly in the mass PMU data of the distribution network.The research work of this paper was funded by the State Grid Corporation’s science and technology project“Research on the monitoring and location technology of distribution network anomaly based on large-dimensional dynamic matrix spectrum distribution”.The main work of this paper is as follows:Firstly,by deeply studying the theory of linear shrinkage and random matrix theory,we combine M-P law,LS theory and PMU measurement data for large array distribution network PMU measurement matrix system.A fast detection method for the anomaly in distribution network based on LSRMT is proposed.Secondly,this paper combines linear shrinkage theory and principal component analysis method to propose an anomaly location method based on LS-PCA.The PMU matrix performs a series of matrix transformations,performs matrix dimensionality reduction and extracts the principal components.This paper establishes a score system based on the principal component coefficient scores of each PMU and realizes the accurate location of the abnormal network signal according to the score.Finally,there may be a variety of anomalies occur at the same time.This paper combines the LS theory with the standard condition numbers and proposes a multi-anomaly detection method based on sequential LS-SCN.The method solves the optimization matrix of the PMU sample covariance matrix.According to the SCN rule,the method analyzes the relationship between the matrix statistic and the threshold to realize the multiple anomaly detection.In addition,a multi-anomaly location method based on regression coefficients is also proposed.Finally,the simulation model is built in PSCAD.The simulation results show that the proposed method can detect and locate the anomaly in the distribution network quickly and effectively without knowing the system noise variance.It is also applicable to the large array PMU measurement system in low SNR environment.At the end of the paper,the application of the distribution network PMU device developed by the team in Jinan DJ substation is briefly introduced.