Estudo de técnicas de previsão de consumo em sistemas de distribuição de gás natural

• Ano de defesa: 2012
• Autor(a) principal: Cruz, Gustavo Lima
• Orientador(a): Cardoso, Carlos Alberto Villacorta
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Sergipe
• Programa de Pós-Graduação: Pós-Graduação em Engenharia Elétrica
• País: BR
• Palavras-chave em Português: Previsão de consumo; Sistemas de distribuição; Gás natural
• Palavras-chave em Inglês: Forecast of consumption; Distribution systems; Natural gas
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://ri.ufs.br/handle/riufs/5018

The forecasting of gas consumption has a fundamental importance for the natural gas distribution company, since it is common for supply companies include clauses in their contracts that force the distributor companies to perform the volume programming of the natural gas to be withdrawn, these same companies are subjected to the application of penalties if the volume exceeds programmed limits previously established. Thus, in the present work has been studied the potentialities to use of predictive models based on regression, time series and artificial neural networks in forecasting gas consumption, with the intend to improved the methodologies currently used by gas distributor in the daily schedule to send to the supplier, in a scenario characterized by the predominance of industrial consumers with dissimilar characteristics. In this context, considering the potential of forecasting techniques, has been studied the gas consumption forecasting in the medium term of both the industrial consumers and the automotive segments. From these studies it was possible to identify particular types of behaviors, and the forecasting strategy most suitable approach using artificial neural networks, time series or a combination of both. To perform these studies was developed a computational tool to analyzing, parameterize and validate methods of forecasting based on historical data consumption. The results are promising because it presents boundary conditions close to actual values.