Research On Power User Behavior Analysis In Big Data

With the increasing application of smart meters,user-side big data emerges which contains potential information on users' electricity consumption behaviors.Under such a circumstance,how to fast and validly analyze these data and dig out useful information on users' electricity consumption behaviors becomes one of the most important research topics.Mining the similarity of different users' behaviors on electricity consumption,and finding out the relationship between all kinds of factors and different electricity consumption behavior modes,could help us to customize service-oriented applications for electric power companies,users and the government.In this paper,on the basis of existing researches,made an in-depth analysis on users' electricity consumption behaviors,and put forward the idea of digging out the relationship between each affecting factor and different kinds of residential electricity consumption behaviors.The detailed content is as follows:1.Analyzed the load characteristics of all kinds of users and the building methods of the electricity consumption data samples.To improve clustering segmentation performance and guarantee clustering speed,adopted Autoencoder algorithm to conduct dimension reduction processing on the data of each week's load sample.2.Fully studied clustering algorithm,combined the thinking of spectrum clustering,put forward the optimized US-ELM-Kmeans algorithm to improve the existing clustering method and improve its effect.Also,clustered and segmented the residential users.3.In the process of relational analysis,put forward the optimized Apriori algorithm which amended the disadvantages of existing relational analysis methods,reduced memory space and improved algorithm efficiency.Made relational analysis on the affecting factors of the clustered and segmented residential users.4.In the experimental part,validated the efficiency and feasibility of this method by comparing experimental results obtained from other algorithms.