Thesis (M.Sc.) - Cairo University - Faculty of Economics and Political Science - Department of Statistics

Electricity is an important public service for any nation. Forecasting methods are critical concerning future technical improvements. An accurate hour forecast is a vital process to balance electricity produced and electricity consumed at any time in the day. A notable feature of the electricity demand time series is the presence of both intraday and intraweek seasonal cycles. Recently, Double seasonal models and methods have been used all over the world for forecasting electricity demand. A double seasonal autoregressive integrated moving average (ARIMA) model, a double seasonal Holt-Winters method and Artificial Neural Networks (ANN) are proposed in the literature to capture the double seasonal pattern of the time series. These three forecasting methods were employed in forecasting hourly electricity demand in Egypt. The forecasts produced by these methods are accurate. Double seasonal Holt-Winters method is the best for different time horizons. Double Seasonal Holt-Winters method and Double Seasonal ARIMA model outperformed ANN in short lead times up to two weeks ahead. While for longer time horizons, double seasonal ARIMA is outperformed by ANN

Issued also as CD