Browsing by Author "Lyatuu, John Frank"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    A time series model for forecasting electricity demand in Dar es Salaam
    (University of Dar es Salaam, 2009) Lyatuu, John Frank
    This study explored matters associated with modeling, identification, estimation, diagnostic checking and forecasting the electricity demand in Dar es Salaam using box- jenkins methodology. The procedures used in evaluating models and forecasting was through the use of monthly data on quantity of electricity sold for domestic usage (TI), commercial usage (T2) and industrial usage (T3) from January 1990 to December 2007 . Plotting of the original time series data was performed, and the data were characterized by trend behavior (upward trend pattern) suggesting that the series was not stationary such that, forms of transformations were necessary. The trend behavior was removed by differencing the original monthly time series data once. Seasonal differencing was not applied as suggested by box-jenkins, implying that the monthly electricity time series data are not characterized by seasonality. In the identification, estimation and diagnostic checking stage, the study of the sample autocorrelation, and partial autocorrelations was done in order to identify the model and estimate the associated parameters. Formal test statistics such as akaike information criterion were applied to verify the model that is adequate. Four possible seasonal time series models for each of domestics usage (TI), commercial usage (T2) and industrial usage (T3) were compared to find the adequate one. The following models were identified and used to estimate the associated parameters: SARIMA (1, 1, 1) x (1, 0, 2)12 for domestic usage (TI) , SARIMA (2, 1,1) x (1,0,1)12 for commercial usage (T2) and SARIMA (0,1,1) x (1,0,0)12 for industrial usage (T3). The adequate models were then used to forecast the monthly electricity demand for the year 2008. Since the forecast values are reasonable, and are within the prediction limits, we conclude that the identified models can be taken for the respective series.