Time Series Regression Model for Forecasting Malaysian Electricity Load Demand

The demand of electricity forms the basis for power system planning, power security and supply reliability. Forecasting electricity demand with linear methods has always been challenging tasks, as the load time series exhibit several superimposed levels of seasonality. In Malaysia, the demand for electricity has reached over 15,000 MW for the past few years and the demand is increasing. This power demand is significantly affected by many non linear factors such as temperature, holiday, special events and other seasonality. This study investigates the impact of weather variables, holidays and other type of variables on daily and monthly electricity demand in Malaysia. A multiple regression model is developed to forecast electricity demand on weather variables, holiday types, daily and monthly seasonality. Due to the nature of the time series data, a time series regression model with autoregressive errors is developed to forecast daily peak electricity demand. The empirical study shows that the Mean Absolute Percentage Error (MAPE) for model with holiday variables is approximately 1.71% in fitting the daily load model. This study also demonstrates the forecast for one month ahead using time series regression model with load reduction weights yield better accuracy. Thus it proved the suitability of the adopted time series regression method for the forecasting short-term electricity load demand.