Study of non-technical losses for detecting theft of electricity using support vector machine

• Grant number: 09/09766-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: October 01, 2009
• End date: September 30, 2010
• Field of knowledge: Engineering - Electrical Engineering - Power Systems
• Principal Investigator: Pedro da Costa Junior
• Grantee: Fernando Yuji Ono
• Host Institution: Faculdade de Engenharia (FE). Universidade Estadual Paulista (UNESP). Campus de Bauru. Bauru , SP, Brazil

Abstract

The proposal of this study is to investigate and implement computational algorithms for support in identifying fraud in electricity consumer facilities. For this, it will be investigated the main strategies used by electric utilities in the mitigation of losses caused by fraud or faulty measurements. The investigated techniques should allow the characterization of electricity consumers through the construction of typical load profiles. The characterization of load profiles can be viewed as an important tool in identifying honest consumers whose measuring instruments exhibit flaws or dishonest consumers that somehow tamper the electricity registering instruments or promote the installation of illegal circuits. This research proposes to investigate a new approach for Non-Technical Loss (NTL) analysis, using an intelligence-based technique such as the Support Vector Machine (SVM) (NAGI ET AL., 2008). The proposed model preselects suspected customers to be inspected onsite for fraud based on irregularities and abnormal consumption behavior. This approach provides a method of data mining and involves feature extraction from historical customer consumption data. The SVM based approach uses customer load profile information to expose abnormal behavior that is known to be highly correlated with NTL activities.