Research On Identification Method Of Stealing Electricity Based On Big Data

Stealing electricity will not only enlarge the abnormal line loss of power supply,destroy the business benefit of electric power enterprises,but also damage the normal order of social electricity consumption.At present,power supply enterprises have realized the collection and storage of users’ electricity consumption data.This thesis will study an identification method for stealing electricity through big data mining and analysis,which can lock the user area and time area of stealing electricity.The main research content of this article is the recognition and power theft Stealing electricity magnitude estimation method to study the basic theory,based on power theft of Dave random matrix identification methods are studied,based on improved LSSVR Stealing electricity magnitude estimation method of time series prediction algorithm is studied,based on improvement of time series forecasting is Stealing electricity magnitude estimate algorithm is studied,The main work contents are as follows:（1）Basic theory of behavior identification and estimation method of stolen electricity.By understanding electricity measurement principles,analyze the electric characteristics of power theft,on the current status of the power supply abnormal discriminant,and application of Dave random matrix theory,through the use of big data analysis of multidimensional random matrix to describe the user’s electricity situation,and the application of the theory of time series prediction algorithm,using the time series prediction algorithm for detecting Stealing electricity magnitude,It lays a theoretical foundation for the study of behavior identification and estimation method of stolen electricity.（2）Identification method of electric theft behavior based on large dimensional random matrix.Application line,change,the abnormal degree of user power theft suspect qualitative evaluation,quantitative criterion to determine the power of time section,focus on suspected abnormal power quickly identify the quantitative criterion of power theft recognition based on Dave random matrix theory method and process,power users of power section to determine time,realize the recognition of power theft.（3）Estimation method of stolen electricity based on improved LSSVR time series prediction algorithm.Improved semi-supervised learning LSSVR time prediction method,based on improved semi-supervised clock sequence estimation of stealing power estimation algorithm,BP neural network stealing power modeling,artificial bee swarm optimization BP neural network stealing power mode,improved the stealing power estimation method.（4）Estimation algorithm of stolen electricity based on improved semi-supervised learning time series prediction.By measuring the distribution distance between the current data training sample and the historical data training sample,a collaborative training sample was established,which made full use of the historical data,realized the multi-step prediction,and improved the rationality and accuracy of the estimation of stolen electricity.