Research And Design Of Big Data Analysis System For Ahnormal Electricity Consumption In Hengshui Power Supply Company

The increasing abnormal electricity consumption and electricity theft not only damages the economic interests of power supply companies,but also jeopardizes the safe operation of the power grid,which has always been one of the problems that plague the development of the power industry.The traditional methods such as regular inspection and verification,user reporting heavily rely on manpower investigation.In that case,the efficiency is very low.Besides,the conventional abnormal electricity detection method is not practical for it has too many false positives.With the construction and development of a strong smart grid and ubiquitous internet of things in energy,the marketing measurement data generated by power users' electricity information collection and power marketing business system have grown to a large scale,and the characteristics of big data are increasingly obvious.By using the big data technology,the comprehensive analysis,abnormal screening and instant warning of marketing measurement big data can quickly improve the efficiency and accuracy of abnormal electricity analysis.Through the analysis of marketing and measurement big data,this paper gives a method of identifying electricity theft users based on marketing measurement big data.First extract raw data such as breach of electricity punitive punishment information,electric energy,working conditions,and event records for persistent storage to build a big data set for abnormal electricity analysis.Secondly,establish an evaluation index of normal electricity consumption and abnormal electricity consumption,and extract the characteristics of electricity theft behavior to construct a sample set.Finally,anomaly detection algorithms such as classification and regression tree(CART)and support vector machine(SVM)are used to train,evaluate and identify the user identification of electricity theft model.Experiments were carried out with python and scikit-learn.The results of the experiment were analyzed by using confusion matrix and ROC curve to verify the effectiveness of the method.This paper presents the physical architecture of the abnormal electricity consumption big data analysis system of Hengshui power supply company,and designs the system technology architecture,including data source layer,acquisition layer,storage processing layer,service layer,interface layer and application layer.The system meets the requirements of batch and real-time collection of marketing measurement big data,batch offline storage and processing,real-time online processing,memory calculation,and abnormal power analysis.The main application functions of the system are designed,including abnormal electricity analysis,abnormal task processing,comprehensive query statistics,abnormal electricity consumption and tampering behavior for diagnosis,screening users who have abnormal power usage and suspected power tampering,and multi-dimensional query user anomaly and task processing information.The realization of specific functions such as station selection operation,abnormal power diagnosis and statistical reports are given combined with specific cases.