| The escalation of tampering means has led to frequent occurrences of hightech theft.The existing anti-stealing technology is mainly based on devices,lacking effective methods to mine features in massive historical data.Applying data mining algorithms to research anti-stealing models can solve the problems of low data mining rate of electricity theft anomalies,poor timeliness of tampering alarms,and large workloads on site and improve the efficiency of anti-stealing work,and reduce resource waste.The method this paper studies is to distinguish suspected electricity theft.Firstly,according to the characteristic analysis of power consumption data fluctuation,the anomaly law of power stealing is studied.The identification models of power stealing based on power consumption anomaly analysis and line loss anomaly analysis are constructed respectively,the electricity theft abnormal data in power consumption samples are mined by defining wave coefficient optimization and fusing outlier algorithm and clustering algorithm,what's more,it calculates the time dispersion of line loss anomaly and analyze the correlation between line loss and electric quantity by using quadratic clustering method.Then,the suspected degree of electricity theft samples is calculated by analyzing the abnormal rate of sample data and the continuity of alarm time,and the suspect level of electricity theft is classified.Finally,on the basis of identifying the suspected degree of electricity theft,this paper constructs an analysis model of electricity theft behavior by summarizing the changing rules of the characteristic parameters before and after the occurrence of electricity theft,and monitors the ways of stealing electricity from serious suspected users.This paper verifies above algorithms by experiments in anti-stealing software platform,the experimental results show that the method improves the accuracy of data mining by more than 7% and achieves the expected anti-stealing application effects. |
PDF Full Text Request