Metodologia para Detecção de Possíveis Ocorrências de Perdas Não Técnicas em Consumidores Rurais Aplicando Métodos de Machine Learning

Natalia B. Sousa; Daniel P. Bernardon; Henrique S. Eichkoff; Pedro Marcolin; Daniel L. Lemes; Luciana M. Kopp; Juliano S. Andrade; Lucas M. Chiara

doi:10.20906/sbse.v2i1.3088

Natalia B. Sousa Universidade Federal de Santa Maria (UFSM), RS
Daniel P. Bernardon Universidade Federal de Santa Maria (UFSM), RS
Henrique S. Eichkoff dpbernardon@ufsm.br
Pedro Marcolin Universidade Federal de Santa Maria (UFSM), RS
Daniel L. Lemes Universidade Federal de Santa Maria (UFSM), RS
Luciana M. Kopp Universidade Federal de Pelotas (UFPEL), RS
Juliano S. Andrade Companhia Paulista Força e Luz (CPFL Energia), SP
Lucas M. Chiara Companhia Paulista Força e Luz (CPFL Energia), SP

DOI: https://doi.org/10.20906/sbse.v2i1.3088

Keywords: Non-technical losses, Electricity Consumption, Clustering, Random Forest, K-means, Rural Consumers

Abstract

The difficulty of detecting non-technical losses by electric energy concessionaires has been a great and constant challenge. Inspecting consumer units located in rural areas demands excessive time and expenses on the part of concessionaires, due to the distance from urban centers and the difficulty of access, without there being a previous technical indication of the occurrence of Non-Technical Losses. This work aims to present a methodology for estimating electricity consumption for rice crops that use flood irrigation, in the city of Uruguaiana, Rio Grande do Sul, implementing classification using artificial intelligence techniques (clustering, k- means and random forest), and with the help of indicators, report cases of possible non-technical losses.