Previsão de Queimadas nas Proximidades de Linhas de Transmissão via Redes Neurais Artificiais

Lucas Botelho da Cruz; Carlos Alexandre M. Nascimento; Fernando A. Assis; Rodolfo A. R. Moura; Marco Aurélio O. Schroeder

doi:10.20906/SBAI-SBSE-2023/3904

Lucas Botelho da Cruz Programa de Pós-Graduação em Engenharia Elétrica, PPGEL, Universidade Federal de São João del-Rei, UFSJ, São João del-Rei, MG
Carlos Alexandre M. Nascimento Companhia Energética de Minas Gerais, Cemig, Belo Horizonte, MG
Fernando A. Assis Programa de Pós-Graduação em Engenharia Elétrica, PPGEL, Universidade Federal de São João del-Rei, UFSJ, São João del-Rei, MG
Rodolfo A. R. Moura Programa de Pós-Graduação em Engenharia Elétrica, PPGEL, Universidade Federal de São João del-Rei, UFSJ, São João del-Rei, MG
Marco Aurélio O. Schroeder Programa de Pós-Graduação em Engenharia Elétrica, PPGEL, Universidade Federal de São João del-Rei, UFSJ, São João del-Rei, MG

DOI: https://doi.org/10.20906/SBAI-SBSE-2023/3904

Keywords: Transmission line shutdowns, Artificial Neural Network, Wildfires forecasting, Electric power system

Abstract

Wildfires are one of the main causes of unscheduled shutdowns in Brazil’s power grid. Therefore, tools that help predict these events in the vicinity of transmission lines (TLs) can significantly collaborate in operation planning tasks. In this context, the present work proposes a methodology that employs the artificial neural network Multi-Layer Perceptron in order to carry out wildfire forecasting in the vicinity of TLs based on meteorological data from the region of interest, representing an important indicator of aid in decision-making in relation to planning the operation of electrical power systems. A case study carried out to forecast fires in the region of the city of Uberlândia - MG, through which a 500 kV TL of the national interconnected system passes, points to a promising performance of the proposed method. Accuracy rates of 82% and 84% are obtained for forecasts in the year 2019 for the evaluated region.