Aplicando Mineração de Dados e Aprendizagem de Máquina para Planejamento Energético de Comunidades Ribeirinhas — Estudo de Caso da Reserva do Tupé no Amazonas

Abstract

The Brazilian energy sector faces challenges in ensuring the supply of electricity to all regions of the country. Some communities, due to their geographical constraints, are isolated and have difficult access. Based on this assumption, this article proposes the use of data mining and machine learning as an alternative to contribute to the improvement of rural electrification plans. Using the open source software Orange, which has algorithms that enable the development of this work, it is demonstrated that it is possible to obtain an accuracy in predicting information from a community between 52.3% and 79.9%. The higher the obtained percentage, the greater the machine’s accuracy, confirming that even with incomplete community data, this technique can be applied to obtain, for example, an estimate of the installed power in a residence and, with that, estimate the load curve. This obviates the need for additional visits to collect additional information, which would avoid extra costs and save time.