Uma abordagem baseada em redes neurais artificiais e clusterização para previsão de curto prazo da demanda de energia elétrica

• Autor(a) principal: Scremin, Eric Reinoldo
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/11924
• topic: Demanda de energia elétrica; Previsão de séries temporais; Redes neurais artificiais; Clima; Electric load forecasting; Time-series forecasting; Artificial neural networks; Climate; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO

Electricity plays a crucial role in the development of a country because it directly influences many sectors of the society. In this sense, the demand forecasting is of paramount importance for the maintenance and growth of the electric power systems. Currently, there are different approaches used for such forecasting. In addition, there are many variables that can influence the performance of this process. Among these variables, one can highlight those derived from the climate. Therefore, this work proposes the application of an Artificial Neural Network of Multilayer Perceptron type with Levenberg-Marquardt training algorithm, using temperature and demand as input variables. Moreover, another objective of this work is to investigate the relationship between variables, making use of the k-means clustering method on the input data. The results show that this clustering-based approach obtains predictions with low error rates. However, slightly better results were obtained when the data were not clusterized