Previsão de séries temporais com aplicações a séries de consumo de energia elétrica

• Ano de defesa: 2008
• Autor(a) principal: Roger Junio Campos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Previsão longo-prazo; ARIMA; Consumo de energia; NARIMA; Série temporal; Neuro-Fuzzy; Rede Neural; Energia elétrica Consumo Previsão; Engenharia elétrica
• Link de acesso: http://hdl.handle.net/1843/BUOS-8CTETD

The main objective of this work is to study and to apply methods oflong-term load forecasting to time series with trends and cycles just using regressors of the historical series. Two case studies were accomplished, the first using the series of load of the city ofNewEngland (USA), and the other the series of load of the state of Minas Gerais (Brazil). The work approaches many forecasting methods well know in the academic and scientific field, such as: the model ARIMA, NARIMA, the Neuro-Fuzzy Network (NFN) and the Artificial Neural Network (ANN). Using those four representations the forecast of 60 steps ahead of the load is estimated. First of all, it was defined which technique would be used to separate the components of the temporary series. The chosen methodology was proposed by Mohr (2005). The seasonal component was adjusted using the models NARIMA, RNF and RNA, and the obtained models are added to the tendency component resulting the estimates of the load. The ARIMA model is esteemed using the time series without separating their components. The models are compared using the performance indices: MPE, MAPE and RMSE. In the case of the load of the State of Minas Gerais, as the time series is composed by few observations, in the forecast of 60 steps ahead, the trend component isapproximated by a straight line, and as there is no samples to accomplish the comparison, a spectral analysis of the forecasts is accomplished with the original series of load. The results to the two case studies show that the models ARIMA, NARIMA, RNF and RNA are ecient tools that can aid in the planning and sockets of decisions in the electric sector