Non-technical losses or commercial losses are caused by multiple factors such as: energy thefts via irregular or clandestine connections, energy meter fraud, self-connections, damaged meters, consumers default, and others. These losses entail huge financial losses for power utilities, for the distribution network and society. Namely: damages in the power quality (with an increase in blackouts, for example), in the reliability of distribution networks (with criminal changes in the network topology), an increase in the energy bill, a reduction in tax collection, and others. Numerous illicit practices that cause non-technical losses were extinguished with the development of smart grids and electronic meters. However, it is necessary that the power utilities to invest a lot of financial resources to implement these technologies. Therefore, this migration from conventional distribution networks to modern distribution networks will be slow, especially in underdeveloped countries such as Brazil. Thus, there is a need to develop new methodologies for detecting irregular consumer units in conventional energy distribution networks. In this context, this project aims to develop a methodology based on soft computing techniques with two modules: (1) automatic knowledge extraction from a database of the power utility via neural network Fuzzy ARTMAP and (2) development of a fuzzy inference system (FIS) to incorporate the empirical knowledge of specialists in inspections. The result of this methodology will be a list of consumer units suspected of irregularity and which should be inspected by the field teams. These inspections have a high financial cost; therefore, the number of false-positives should be minimized, that is, the amount of unnecessary inspections in regular consumer units.
News published in Agência FAPESP Newsletter about the scholarship: