Research On Electricity Consumption Behavior Analysis And Electricity Stealing Identification Of Power Users

In today’s fast-developing society,there is an increasing demand for electricity consumption,and related power businesses are also developing on a large scale,but the problem of electricity theft has not been covered up.The ways and means of stealing electricity are becoming more and more diversified,from the traditional way of stealing electricity to the way of high-tech electricity stealing,which is very concealed.The problem of electricity theft does not only exist in a certain period.It is a long-standing factual problem.The loss to the country cannot be underestimated.The hidden safety hazards caused by the electricity theft evolve into accidents from time to time.The accumulated user electricity data information has provided help for the anti-electricity theft work.Therefore,the use of data mining and machine learning methods to analyze and judge electric power data,and research on electricity theft has become a hot spot.This paper conducts research and analysis based on users’ historical electricity consumption data,with the theme of identifying electricity theft behavior of electricity users.Analyzed the methods of stealing electricity and common methods of stealing electricity,and integrated the characteristics of the stealing behavior to select the identification index of stealing electricity.By clustering the load curve data,the characteristic curve is obtained and the electricity consumption behavior is divided according to this;the similarity between the load curve and the characteristic curve of the user to be tested is calculated and analyzed,and the users suspected of electricity theft are initially screened;The user and the selected users who are suspected of stealing electricity together form the input user set of the model,forming and processing the data for discriminating electricity theft,and substituting it into the model that combines feature selection and Ada Boost integrated learning to carry out training and testing.The main research contents and work carried out in this paper are as follows:(1)Conduct research and analysis based on historical electricity consumption data information of users in a certain area of Yunnan Province.Starting from the principle of electricity theft,it analyzes the traditional methods of electricity theft,high-tech electricity theft methods,and the common methods of implementing electricity theft under the corresponding methods,summarizes and summarizes the changes of related electricity consumption characteristics under different methods of electricity theft,and comprehensively steals electricity.The characteristics of electrical behavior have selected 9 indicators for discriminating electricity theft.(2)Conduct research and analysis around users’ electricity consumption habits.Normal electricity usage behaviors will show a certain regularity,and when electricity theft occurs,their electricity usage trends will deviate from the previous electricity usage rules.Through the iterative process of clustering the representative daily load curve data,the characteristic curve,which is the cluster center curve,is obtained.Different clustering center curves represent the daily load characteristic curves of different types of users.Based on the clustering algorithm,the power consumption behavior of users is divided.Combining the data characteristics of the power user load curve and possible displacement changes,the DTW distance-based load curve similarity measurement method is adopted.Based on the calculation and analysis of the curve similarity,the preliminary screening of suspected users of electricity theft is realized.(3)Research on classification algorithms in data mining,analysis and modeling combined with application scenarios,further identification and judgment of users suspected of stealing electricity,and obtaining classification prediction results of stealing users or non-stealing users.The classification algorithm attributes used for the identification of electricity theft are determined,and the integrated learning and the Ada Boost algorithm used are introduced.Establish a model that combines feature selection and Ada Boost to identify users of stealing electricity,including determining user sets,forming data and processing for stealing discriminating indicators,feature selection,dividing the experimental sample set,completing the training and testing of the model,and correlating The experiment shows the rationality and effectiveness of the model.