Characterization Of Power Consumption Behavior Of Small And Medium Load Customers Based On Spectclustering And Random Matrix Theory

The development of smart grid and demand-side management technology has made the important role and position of power consumers in the power system increasingly prominent.Research on the characteristics of customers’ electricity consumption behavior is of great significance to the refinement and efficient management of demand-side management.In order to guide customers’ electricity consumption behavior to the direction that is benefit to the stabilization of the power grid and realize the two-way interaction between the power grid and the electricity users,this paper considers combining the large amount of customers’ electricity consumption data collected by smart meters in the power grid operation and the background information of electricity users obtained from the research,using the power load curve clustering and the correlation analysis method based on the random matrix theory to analyze and study the characteristics of electricity users’ electricity consumption behavior,so as to This paper provides a reference basis for formulating detailed demand-side management plans.(1)In this paper,we propose a multi-scale similarity measure based on the mode slope distance similarity of the load curve,and improve the spectral clustering algorithm by combining the particle swarm optimization algorithm to improve the clustering effect of the algorithm in the electric load curve.This paper proposes a multi-scale similarity measure based on the slope distance of the curve.(2)Under the new situation and background of big data and demand-side management of power system,based on the massive customer load data collected and accumulated during the operation of smart grid,the main object of this paper is the load curve of residential and small and medium-sized commercial customers,which is pre-processed by data cleaning and data standardization to complete the missing data and select and discriminate the abnormal data.With the improved spectral clustering algorithm,the load curves of power users are clustered and analyzed,and the clustering effect of the proposed load clustering method is evaluated by using internal and external clustering evaluation indexes such as DBI index and ARI index.(3)Finally,based on the data of residential and small and medium-sized commercial customers obtained by clustering,under the framework of user behavior model research,integrating factors such as user psychology and economic environment influence,and use big data extraction techniques such as correlation analysis method based on random matrix theory,the correlation between the characteristics of electricity consumption behavior of residential and small and medium-sized commercial customers with different electricity consumption patterns and economic,demographic and other influencing factors is studied to explore their It provides information support for the possible future demand-side management to explore the regulable potential of customer load in a more detailed way.