On-line analysis of suspicious electricity stealing based on level-change detection

The existing data mining methods for suspicious electricity stealing detection, as utilized based on multi-month power consumption data, are of low timeliness. In this light, this paper proposes an online method for suspicious electricity stealing analysis, which is based on level change detection (LCD) on the users' daily power consumption sequence. The power consumption data are turned into cumulative sum values so as to reduce the influence of short-time data abnormality on the detection results. In the case study section, a numerical example is offered based on the daily consumption data of the customers in a residential area in North Shanghai, where the LCD method is also compared with other detection methods. The results show that the LCD method is of high coverage and accuracy, meanwhile it can realize timely online detection by moving the data window.