Abstract
Summary form only given as follows. Medium-term peak load forecasting is investigated and two types of applied data are investigated, i.e. weekly and monthly peak load demand of Jeddah area for nine years history. For nine years data a study the using the first eight years for forecasting the following ninth year is investigated together with another study using the first seven years data to forecast the following two years, the eighth and ninth. The Minitab statistical software package is first used for load demand prediction using the autoregressive integrated moving average (ARIMA) technique. An artificial neural network (ANN) is then utilised and several suggestions are implemented to build a suitable model of ANN. The application of a simple ANN showed poor performance. The manipulation algorithm was then enhanced to account for the trend of the peak load demand which clearly enhanced the performance of the ANN as well as an FNN (fuzzy neural network) study. The FNN provided a superior performance because it used the ANN in the training phase together with the choice of the proper membership of the fuzzy model. A comparative study is carried out to show the accuracy distinction among these techniques for our different historical two types data.