-
Vitor G. Silveira
UNESP – Universidade Estadual Paulista “Júlio de Mesquita Filho”, Câmpus de Rosana Av. dos Barrageiros, 1881, 19274-000, Rosana, São Paulo
-
Andréia S. Santos
UNESP – Universidade Estadual Paulista “Júlio de Mesquita Filho”, Câmpus de Ilha Solteira Av. Brasil, 56, 15385-000, Ilha Solteira, São Paulo
-
Mara L.M. Lopes
UNESP – Universidade Estadual Paulista “Júlio de Mesquita Filho”, Câmpus de Ilha Solteira Av. Brasil, 56, 15385-000, Ilha Solteira, São Paulo
-
José F. Resende da Silva
UNESP – Universidade Estadual Paulista “Júlio de Mesquita Filho”, Câmpus de Rosana Av. dos Barrageiros, 1881, 19274-000, Rosana, São Paulo
-
Lucas Teles Faria
UNESP – Universidade Estadual Paulista “Júlio de Mesquita Filho”, Câmpus de Rosana Av. dos Barrageiros, 1881, 19274-000, Rosana, São Paulo
Keywords:
Artificial Neural Networks (ANNs), Commercial Losses, Electrical Distribution Systems, Energy Theft, Fuzzy ARTMAP, Non-technical Losses (NTLs)
Abstract
Non-technical losses (NTLs) or commercial losses are caused by multiple factors such as: energy theft, energy meter fraud, self-reconnection, damaged energy meter, consumer default and others. NTLs cause significant financial losses to power utilities, undue changes in the distribution network and damage to consumer units (CUs) with the increase in their energy bill. In this context, we propose a methodology based on artificial neural networks (ANNs) to detect NTLs in power distribution network. The methodology has two modules: (i) extraction of statistical attributes and (ii) module for classification based on the Fuzzy ARTMAP. The proposed methodology presents high indexes for the metrics: success rate (99.0%), reliability (93.6%), and specificity (96.7%). Therefore, it has good coverage for detecting irregular CUs and a high success rate in field inspections. Thus, costly field visits by inspection teams in regular UCs are drastically reduced.
