Detecção de perdas não técnicas em redes de distribuição por meio da correlação de dados climáticos, consumo e notas de inspeção

Abstract

Non-Technical Losses (NTL), which include losses due to electricity theft and meter malfunctions, still represent a significant challenge for electricity utilities in Brazil and around the world. In Brazil, fraud and irregularities in metering cause billions of reais in losses for energy concessionaires and are the cause of a reduction in electricity quality indicators. It has been the practice of energy concessionaires in the country to carry out annual inspection campaigns to detect fraud in around 15% of the consumer units in their concession area. On the other hand, detection hit rates of less than 15% have been observed. In this scenario, it is therefore of interest to improve existing methodologies for identifying and locating NTPs in order to help increase the assertiveness of utilities, using data actually available from national and international utilities. To this end, among a variety of methodologies, there is research in the scientific literature that indicates a direct relationship between electricity consumption and fluctuations in ambient temperature. Taking this hypothesis as a premise, this work proposes an approach to fraud detection based on descriptive analysis of energy consumption and ambient temperature data from, correspondingly, consumer units, transformers and primary feeders. The case study was carried out on a medium-voltage distribution network located in central-western Brazil, with 159,939 consumer units, 4876 secondary substations and 243 primary feeders. The results show that even with a simplified methodology, it is possible to obtain an accuracy rate of over 23%.