Research On Power Load Clustering Algorithms For User Interaction Power Behavior Analysis

With the deepening of information construction in the field of power system,the type and scale of power data collection is becoming more and more big data-oriented.At the same time,the new generation of smart grid is actively exploring the ubiquitous connection of multiple information,moving towards the direction of digitalization and interaction.Interactive grid is not only the basic mode of the next generation of global grid,but also the core of China’s grid modernization.As the terminal of the power system industry chain,power users are the key factor of the deep-seated integrated intelligent development of the power grid.It is of great significance to mine the multidimensional information of the user’s electricity use behavior in view of the requirement of accurately mastering the interactive electricity use characteristics in the future power construction process.However,the traditional power load clustering technology for power users has been unable to meet the increasingly urgent new requirements of ubiquitous connection,flexible innovation,open sharing,integration and interaction of continuously upgraded smart grid based on high-speed two-way communication network.Based on the above ideas,this paper proposes two improved clustering algorithms for the adjustment potential of user interaction association factors,and studies the cluster analysis model which comprehensively considers the multidimensional evaluation of interaction factors of user electricity behavior association characteristics.Main work of the thesis is as follows:(1)Based on the application of data mining technology in the clustering of user’s electric load,this paper makes a comprehensive study and analysis of the characteristics of electric load data and the associated factors of electric load,and makes a modeling analysis of the existing common clustering algorithms,and comprehensively compares the performance of each classic algorithm.(2)Considering the multi-dimensional extraction and definition of comprehensive information factors of actual power consumption and subjective and objective condition willingness of power users.For some indicators in the information,a multi-dimensional index evaluation and analysis system based on the related characteristic factors of users’ electricity consumption behavior is established.In view of multi-dimensional indexes,the method of subjective and objective comprehensive weights is used to weight them,and a refined cluster model of improved TOPSIS comprehensive evaluation oriented to multi-dimensional indexes of related characteristic factors of user electricity consumption behavior is proposed.(3)Based on the related characteristic factors of user’s electricity consumption behavior correlation characteristics,the comprehensive correlation factor adjustment potential index is developed and the clustering algorithm is studied.According to the shortcomings of the traditional algorithm,the improvement strategy is proposed and the self-organizing center K-means algorithm and the probability reconstruction Kmeans algorithm are the applied to cluster analysis and performance comparison base on the potential index of user’s interaction correlation factor adjustment.