How Accurate is TNB's Electricity Demand Forecast?

Forecasting of electricity demand provides an essential input for decision making in power operation and development. In Malaysia, Tenaga Nasional Berhad (TNB) is using a well-established and sophisticated system to forecast the electricity demand. However, the reported predictive performance appears to be unsatisfactory if we compared to the other forecasting methods. Thus, this paper aims to evaluate and compare the predictive performance of TNB's forecast to the univariate time series methods of classical decomposition model, exponential smoothing, Holt-Winter smoothing and Box-Jenkins approach. The predictive performance of Box-Jenkins approach is found to be superior to TNB. Specifically, the Autoregressive Frictionally Integrated Moving Average (ARFIMA) is found to be able to reduce eighty five per cent of the TNB's forecast error.