Power Theft Detection Using Advanced Neural Network in Three-phase Distribution Systems
Abhilash Sen, Nien‐Che Yang
Abstract
To achieve a fully cost-effective efficient smart grid, the problem of non-technical losses (NTL) must be properly addressed. Power theft (PT), being one of the most crucial NTL, must be mitigated considerably. In this study, a power theft detection method based on parameter estimation of an unbalanced distribution line is proposed. This technique can also be utilized to locate the distribution line undergoing illegal electricity tapping. It uses the node voltages and power quantities to estimate the feeder parameters, thereby detecting instances of PT and the affected line. The proposed process does not require the phasor measurement unit (PMU) technology. A hybrid general regression neural network model equipped with multi-run optimization (GRNN-MRO) is also developed to map the line parameters and power quantities. A power theft detection (PTD) index is developed using the mean absolute percentage error (MAPE) values obtained through the GRNN-MRO estimation process. The proposed PTD algorithm along with the parameter estimation process is verified and compared with other standard methods. The results show that the PTD scheme works accurately on IEEE 13 and IEEE 37 bus systems compared with other conventional methods.