Interest in the applied research on short and medium-term electricity load forecasting has been remarkable during the
past few years. Forecasting electricity loads with linear methods has always been a challenging task, since the load series exhibit several superimposed levels of seasonality, together with the nonlinear effect of many important exogenous variable, such as temperature, holiday and special events. Furthermore, forecasting load profiles (the series of 24 hourly loads in the target day) as a vector forecasting problem is an order of magnitude more difficult. Yet it is precisely the forecasting of these profiles that has been the typical operational, and the market requirement for electric utilities. This paper examines the issues of forecasting using conventional regression-based methods and other methods such as neural networks and expert system. Some discussion on the practicality of using expert system and neural network for forecasting the 24 hours of daily electricity load and very much conductive to this approach. We employ the data on the daily electricity load demand from Tenaga Nasioanal Berhad (TNB). The forecast accuracy is measured based on the error statistics of forecast between the models for half an hour ahead for the short term forecast and a month ahead for the medium term are presented and behavior of data is also observed.