Forecasting hourly prices in the portuguese power market with ARIMA models

• Autor(a) principal: Dias, António Vasconcellos
• Data de Publicação: 2009
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/2040
• topic: Electricity Market; Time-series analysis; Energy price; Price forecast; Mercado eléctrico; Análise de séries temporais; Preços de energia; Previsão de preços

As power markets became a recent worldwide phenomenon, electricity prices’ forecast is a new subject for investigators. Due to the electricity’s particularities, a power market has some very specific rules that must be understood before one begins its study. This empirical research presents a comparative study between two forecasting methods of the day-ahead hourly electricity prices in the Portuguese power market: a complete hourly time-series analysis and an hour-by-hour approach, each one for a Summer and an Autumn seasons. My purpose is to check if an exhaustive hourly analysis would improve significantly the energy price forecasts accuracy and, if so, would the additional computing time offsets this improvement. As it is common in energy prices empirical research, we use ARIMA models. To select the models on a first stage, the Mincer- Zarnowitz regression was considered. On a second stage, to compare the models and select the best one in terms of predictive ability, the Harvey-Newbold encompassing test was applied. Some evidence was found that, in accordance to Cuaresma et al. (2004), analysing each hour separately produced better results than considering the complete time series, although the time taken to estimate the models can be an issue for short term predictions. The ARIMA models that captured the weekly effect encompassed the others and produced more accurate forecasts.