An Efficient Power User Classification Method Based On Dimensionality Reduction And Load Clustering

With the continuous promotion of smart meters,the data accumulated in the power industry has gradually become a massive trend.The diversification of the grid business has also promoted the development of data mining research in the power industry.The classification of power users is the basis of many power industry applications.The scientific and rational classification of power users has further developments such as time-of-use electricity price,load forecasting,and peak-shifting.The traditional classification of electric power users has shown some drawbacks in today's diversified power usage habits.The application of massive load data to classify power users can make full use of the power consumption characteristics of power users and make better and reasonable division of users.According to the summary of the classification method of power users,this paper has done the following work:1)Analyzed the background and significance of the research of the subject,and analyzed the current research status and some existing problems of the classification method of power users at home and abroad.And on this basis,put forward the idea of improvement.2)Introduced some relevant theoretical knowledge involved in this paper,and analyzed and compared the big data platforms Hadoop and Spark.The dimension reduction algorithm and clustering algorithm in Spark MLlib are sorted out and analyzed and compared.Finally,according to the specific requirements of power user classification,the principal component analysis algorithm and k-means algorithm are determined as the specific algorithms for realizing power user classification in this paper.3)Make full use of the advantages of Spark platform and Spark MLlib,and design a high-efficiency power user classification method that combines dimensionality reduction and load clustering.The implementation process of this method and the principle and implementation process of the specific algorithm are introduced in detail.4)Using the measured data,the method designed in this paper was verified.The experimental results of the method designed in this paper are compared with the traditional k-means algorithm.The comparison between algorithm efficiency and time highlights the significant advantages of the design method.