Abstract: In order to solve the problem of the behavior of default electricity usage increasingly serious and lack of effective detection means and early-warning system, a default electricity usage detecting and early-warning system is designed.Firstly, the linear interpolation algorithm is used to clean and correct the abnormal data of low-voltage household meter, so as to reduce the interference of the model caused by the missing measured data in the distribution station area, and the relationship among the line loss fluctuation rate, the line loss and the change-point time of current curve are analyzed to determine whether there is any behavior of default electricity usage.At the same time, the spatial dimension identification, the time dimension identification and the relation dimension identification are introduced to detect whether there is electricity stealing anomaly in low voltage users.In addition, the characteristics of user electricity theft time and electricity consumption capacity are taking into account to provide estimated quatity of electricity-stealing.Finally, the monitoring and early warning methods of power equipment are optimized to improve the system performance.The result shows that the proposed method can save 50 J energy in 8 d time, which shows that the early warning system designed in this paper has higher accuracy and robustness.