| With the proposal of carbon peak and carbon neutralization strategy,clean energy has gradually become an important development demand for building an environmental protection society,and electric energy plays a pivotal role in it,and saving electric energy has become the general trend.At the same time,the modernization of power grid development,the scale of power system and the increasing amount of data,measuring instrument in the direction of the intelligent upgrade,promote the informatization construction and the depth of the grid business integration,improves the data collection process and application level,improves the production efficiency.HPLC smart meters which arises at the historic moment.Outlier diagnosis is indispensable in power data,and traditional statistical methods have gradually lost their original advantages,so the event information contained in the outlier data cannot be effectively mined.At present,it is necessary to introduce new outlier diagnosis methods according to the deeper characteristics of power data.Based on the HPLC smart electricity meter,this paper explored the original data of 3168 observation points in 33 days.The main works were as follow:（1）The data pretreatment,using the nearest neighbor（KNN）fill the missing data interpolation method,inspection by the changing point of Bin Seg algorithm will be divided into apparent power required time period have similar characteristics,and sorted according to the average fitting data again,from a certain extent,eliminate the tedious,users irregular electricity by b-spline basis function transfer electricity data Into in this paper,the functional data,lay the foundation for subsequent research.（2）This parper uses anlysis based on functional principal component analysis（FPCA）to96 points in high dimensional HPLC on dimension reduction characteristics of the electric meter data,then use of principal component score vector K means clustering and hierarchical clustering compared validation,according to the characteristics of different classification,get including three types of power consumption of the user.（3）From the perspective of high,medium and low power consumption types,the diagnostic effect of outlier diagnosis method of functional data on power consumption of HPLC smart meters was discussed,mainly including Bagplot and HDR boxplot based on principal component analysis Outliergram algorithm based on functional depth and MS-plot method based on functional radial outliers were visualized to display the monitored outliers respectively.Finally,Venn diagram was used to summarize the results of different outlier diagnosis methods,and the feasibility of functional data outlier diagnosis method in empirical data was demonstrated by comparison.By diagnosing the abnormal date of the electricity consumption of HPLC smart meter residents,the method based on the functional principal component cannot identify the shape outlier,while the Outliergram method based on the outlier diagnosis of the functional depth can effectively identify the position degree and shape abnormality of the abnormal curve,but it is susceptible to the influence of the outlier ratio,and the first two methods cannot be applied to the multivariate functional data;while the MS-plot diagnostic method based on the radial anomaly of the functional data is more widely used.It can detect radial anomalies and shape outliers,and can also be used for multivariate functional data,when the more constraints,the more demanding the diagnosis of outliers,thereby reducing the false positive rate,the specific method needs to depend on the type of data and the needs of the study.In addition,individual outliers meet the criteria for selecting outliers from all detection methods,and such outliers need to be paid close attention to,strengthen the monitoring of abnormal users’ electricity-related equipment,early detection of abnormalities in electricity consumption data and corresponding processing,to ensure the normal operation of power enterprise services,thereby reducing unnecessary economic losses.In summary,the diagnosis and analysis of outliers in the electricity consumption data of HPLC smart meters has certain theoretical and practical significance. |
