Aprendizagem de máquina como ferramenta de planejamento energético: estudo de caso aplicado a comunidades não eletrificadas do baixo Rio Negro no Amazonas

• Autor(a) principal: Simões, Luís Henrique Raheem
• Data de Publicação: 2023
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFAM
• Texto Completo: http://riu.ufam.edu.br/handle/prefix/7384
• topic: Aprendizagem de máquina supervisionado; Planejamento enegético; Mineração de dados; Comunidades ribeirinhas; CIENCIAS EXATAS E DA TERRA: CIENCIA DA COMPUTACAO; Mineração de dados (Computação)

This work proposes the use of data science as an energy planning tool and can assist in the design of electrical systems in isolated communities in the Amazon region. Although it is a huge region, the Amazon basin still presents a very big challenge in providing electricity, mainly due to the lack of detailed data on communities and their energy needs. And, without correct sizing, the costs involved, whether for installation or operation of the electrification systems, can cause losses for the energy concessionaire or even dissatisfaction among users due to failure to meet their needs. The method proposed here is based on supervised machine learning, using several classifiers to predict the installed power and energy consumed of houses in a riverside community. The data used to train the method was collected from 14 non-electrified communities in the lower Rio Negro, in Amazonas. The questionnaires collected information on the socioeconomic characteristics of families, such as family structure, level of education, productive activities and energy consumption habits. The results of the method showed that it is possible to predict the installed power and energy consumed with good accuracy. The installed power was predicted with an accuracy of 79.2% and the energy consumed was predicted with an accuracy of 68,5%. This method can be used to support decision-making in the electrification of isolated communities. It can help estimate electrification costs as well as identify communities that have the greatest energetic need.