Abstract:
This thesis presents an evaluation of forecasting methods of electrical energy demandfor various conditions and data. The five methods  smoothing, decomposition, Box-Jenkins,simple linear regression and multiple linear regression are studied in this thesis. The historicaldata used in this thesis are energy sales and free of charges from Provincial Electricity Authority(PEA) in fiscal years 1984 to 2001. Patterns of historical data are divided into two categories seasonal and non-seasonal data. The least Mean Absolute Percentage Error (MAPE) of test datais used to select the appropriate method. The results show that the appropriate forecastingmethod can be evaluated for various forecasting conditions and data, and the MAPE of seasonaldata is lower than non-seasonal data.