Power System Disturbance Detection Based On Improved Canonical Variable Analysis

Disturbance detection of power system is of great significance to ensure the stability and security of power system operation,the safety of people’s electricity consumption and social and economic development.With the increasing abundance of power system data,multivariate statistical detection method has become a research hotspot in the field of power system security monitoring.As an effective multivariate statistical method,canonical variable analysis can better mine the dynamic correlation of data and provide an effective means for disturbance detection of power system.Aiming at the dynamic characteristics of power system data,this paper studies the power system disturbance detection method based on the improved canonical variable analysis,in order to improve the sensitivity of small disturbance and wide area disturbance detection.First,in view of the problem that the existing multivariate statistical methods ignore the dynamic correlation characteristics of power system data,this paper proposes a disturbance detection method of power system based on Canonical Variate Analysis-K Nearest Neighbors(CVAKNN).This method uses the canonical variable analysis method(CVA)to extract typical variables to maximize the correlation between past information data and future information data,and further combines the K-Nearest Neighbor(KNN)method to construct a disturbance monitoring index based on the CVA statistic KNN distance.Verification by numerical example simulation data and actual field-collected data shows that the CVAKNN method has better disturbance detection ability than the traditional CVA method and the existing K-nearest neighbor principal component analysis method.Secondly,to solve the problem that the CVAKNN method does not consider the local information of the data,which leads to the low detection rate of small disturbances,a power system small disturbance detection method based on the CVAKNN based on Double Locality Preserving(DLP-CVAKNN)method is proposed.On the one hand,this method integrates the idea of Locality Preserving Projections(LPP)into the CVA optimization process,and designs an optimization objective function that maintains the local structure.On the other hand,this method introduces the Statistical Local Approach(SLA)method to improve the monitoring statistics and increase the sensitivity to small disturbances.Verification by numerical examples,Simulink simulation data,and actual field-collected data shows that the DLP-CVAKNN method has better detection capabilities for small disturbances.Finally,in view of the problem that the overall modeling in wide-area power system monitoring can easily conceal local disturbance information,this paper proposes a Bayesian fusion based dual local preserving K-Nearest Neighbor canonical variable analysis(Bayesian fusion based DLP-CVAKNN,BDLP-CVAKNN)wide-area disturbance detection method of power system.The method first uses the principal component space difference of the prior data to divide the data of each area of the power system,and establishes a regional local monitoring model for each data block.Then,the method uses the Bayesian integration method to fuse each local sub-model to construct a global statistic,and construct a contribution graph to locate the disturbance source.The simulation results verify the effectiveness of the BDLP-CVAKNN method for power system wide-area detection and positioning.