Open AccessA Comprehensive Method of Recovering Electricity Stealing Based on Data Mining
Author(s) -
Jin Tingchao,
Wang Jian-bo,
Cheng Shuya,
Cai Hui,
Yuan Xie,
Wang Ying
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/647/1/012027
Subject(s) - electricity , computer science , artificial neural network , data mining , autoregressive integrated moving average , electricity market , power (physics) , electricity meter , time series , artificial intelligence , machine learning , engineering , electrical engineering , physics , quantum mechanics
Aiming at the current situation that the current methods of recovering the stolen electricity often use manual analysis and estimation, which leads to unsatisfactory results, this paper proposes a comprehensive electricity recovery method based on data mining. First of all, it is determined for the stealing time and other relevant data. Then, for users who can calculate the correction coefficients, the correction coefficient method is used to analyze the electricity theft electricity, and the users who meet the requirements of the prediction algorithm are analyzed using the corresponding forecast algorithm time series or neural network. Finally, the remaining users use the calibrated capacity of the electric energy meter instead of the actual load to analyze the stolen electricity. Time series forecasting uses an improved ARIMA algorithm based on white noise to predict the stolen electricity. The neural network uses an improved BP neural network based on independent variables to predict the stolen electricity. The research results show that this method takes into account the characteristics of a single power-stealing user while taking into account the user’s periodicity. It can reasonably recover the electricity of power-stealing users and provides a new idea for power recovery.