A Model for NTL Detection of Electricity Theft

Abstract

One of the major factor effecting India’s economy is because of power consumer dishonesty. Strategies used for carrying out such activity are like illegal connections, meter tampering, billing irregularities and unpaid bills. Many countries get affected with such problem. Developed model main focus is on fraud detection at Non-Technical Losses (NTL) level. Model has been carried out with the limitations on the type of datasets available. We considered a specific spatial location in Hyderabad (Telangana State).  Analysis on such data obtained has been carried out through statistical and k-means clustering approaches. Preprocessing is carried out. The proposed model uses customer consumption data on monthly basis in terms of watts. Onsite inspection rate is reduced from the outcome derived from the model. Analysis figures out abnormal consumption behavior of a consumer.  Over years a genuine customer watt usage pattern remains the same for almost all years, with this assumption model was built. Fraudulent consumer were detected if abrupt deviations were found on the consumption pattern over years when compared on the monthly basis. The result of the fraud detection model shortlists potential fraud suspects on whom onsite inspection can be carried out and this information facilitates in reducing the Non-Technical Losses incurred by the electricity board. Electricity theft can be reduced by carrying out onsite inspection or by monitoring such consumers (on the list identified through this model). Or otherwise a clear cut strategy can be followed up by restructuring existing power system distribution or by installing tamper-proof meters that paves way for clear accountability.