Research On Clustering Analysis And Visualization Based On K-means Algorithm In High-dimensional Power Data

With the rapid development of smart grid technology,the country’s comprehensive implementation of renewable resources policy,the demand for electricity in all sectors of society is growing,and the power energy system is gradually changing to be more intelligent,more flexible,and more interactive.In each link of smart grid operation,high-dimensional data with large volume,complex structure and complex correlation are generated.It is of great significance to mine the valuable information in the high-dimensional electricity data and carry out visual analysis.Based on the research of domestic and foreign scholars,this thesis conducts research on k-means clustering algorithm and visualization of high-dimensional electricity consumption data.The main conducted research work in this dissertation are expounded as follows:1.Regarding the two problems of how to determine the K value and how to choose the initial clustering center in the improved k-means algorithm,an improved K-means algorithm is proposed in view of the characteristics of the existing high-dimensional power consumption data set.This improved algorithm uses the dimensionality reduction algorithm to reduce the data set dimension in the high-dimensional feature space in the 3D visual space,combined with the Canopy algorithm to determine the K value of the improved k-means algorithm.The sample density parameter is introduced into the Canopy algorithm to select the initial clustering center.Finally,the experiment is carried out based on the high-dimensional data set of student users collected from smart electricity meters in smart campus.Experimental results show that the improved k-means algorithm has higher clustering accuracy,stronger anti-noise interference ability and better clustering effect.2.Because the high dimension affects the traditional Radviz visualization display,a Improved visualization technology is proposed,which uses the dimensionality reduction algorithm to optimize the data display and reduce the problem of covering overlap between the high-dimensional power consumption data;Then,K-means clustering analysis is used to find out the relationship between the data,improve the influence of high-dimensional data,and better help people to find the laws among the data.Finally,based on the high-dimensional data set of student users collected from smart electricity meters in smart campus,the experiment proves that the improved Radviz visualization technology improves the deficiency of traditional Radviz visualization technology in data display and enhances the effect of data visualization.3.Aiming at the high-dimensional electricity consumption data of student users,a Tableau-based visualization platform for student users electricity clustering results and analysis was established.The platform makes use of latitude and longitude information to make visual maps,and draws the electricity information of student users into bar charts,line charts,bubble charts and other intuitive and clear forms,and presents the geographic location and corresponding electricity information of student users to the logistics management staff in a visual interactive interface.The platform not only brings convenience to managers,but also contributes to the lean management of power consumption information.