Detection rules for Non Technical Losses analysis in power utilities

This paper details a new procedure in the detection module of a general framework to detect and identify abnormalities which may be due to non-technical losses (NTL) in power utilities. Fraud detection techniques have been widely used in other businesses including credit card, telecommunications and insurance companies. However, there is very limited reporting on fraud detection in power utilities using customer databases. A combination of data mining tasks, including feature selection, clustering and classification techniques, have been used to test our proposed general framework and to develop detection rules to produce the most accurate benchmark to be used as a reference for individual customers. The contribution of this paper is the detection rules and the procedures using the detecting rules which have been detailed in our framework. Using real utility data, comparison results have been evaluated in order to check the classification accuracy of the proposed methods.